CZECH TECHNICAL UNIVERSITY IN PRAGUE
FACULTY OF CIVIL ENGINEERING
MASTER 'S THESIS
2015 Jana POESOVÁ
CZECH TECHNICAL UNIVERSITY IN PRAGUE ČESKÉ VYSOKÉ UČENÍ TECHNICKÉ V PRAZE
FACULTY OF CIVIL ENGINEERING FAKULTA STAVEBNÍ
BRANCH OF GEODESY AND CARTOGRAPHY
OBOR GEODÉZIE A KARTOGRAFIE
MASTER'S THESIS DIPLOMOVÁ PRÁCE
MEASUREMENT AND SPATIAL MODEL CREATION OF THE INNER PART
OF THE HELFENBURK CASTLE NEAR ÚŠTĚK ZAMĚŘENÍ A VYTVOŘENÍ PROSTOROVÉHO MODELU VNITŘNÍHO PALÁCE
HRADU HELFENBURK U ÚŠTĚKA
Supervisor / Vedoucí práce:
Ing. Bronislav Koska, Ph.D. Department of Special Geodesy, CTU in Prague
Prof. Dr. rer. nat. Martin Oczipka Faculty of Spatial Information, HTW Dresden
Prague 2015 Bc. Jana POESOVÁ
LIST ZADANÍ
ABSTRACT
This thesis deals with the measurement and creation of documentation of castle
Helfenburk near Úštěk in northern Bohemia. Laser scanning was chosen as the most
convenient method of the data capturing. Focus was put on the inner part of the
castle; nevertheless, the entire castle was measured, including the fortification with
the castle tower.
The measurement was carried out by using the Trimble TX5 scanner. The thesis
deals briefly with the theoretical bases of laser scanning method and then with
measuring work on the object, and positioning of standpoints and check points in field.
The data processing was performed in software Geomagic Studio 2012 and Leica
Cyclone, where the registration was done and further in PoissonRecon, once more
Geomagic Studio 2012 and Agisoft Photoscan, where the triangular mesh and
its reducing was made. The final point cloud was transformed into S-JTSK coordinate
system (Datum of Uniform Trigonometric Cadastral Network) and height system Bpv
(Baltic Vertical Datum - After Adjustment).
The main aim was to produce a complete point cloud and a 3D model of the inner
part of the castle by using the triangular mesh. The results will be handed over to the
citizen association "Hrádek", which administrates the castle, and they will also serve
as as-built documentation of the castle for archaeological research and for publication
on the castle website.
KEY WORDS
Laser Scanning, 3D Model, Triangular Mesh, Point Cloud, Castle, Helfenburk near
Úštěk
ABSTRAKT
Diplomová práce se zabývá zaměřením a vytvořením dokumentace hradu
Helfenburk u Úštěka v severních Čechách. Hrad byl zaměřen metodou laserového
skenování, která pro dané podmínky a požadavky byla nejvhodnější. Největší důraz
na přesnost a úplnost naměřených dat byl kladen v oblasti vnitřního paláce hradu,
zaměřen byl ale celý hradní komplex včetně hradeb a věže.
Zaměření bylo provedeno skenovacím systémem Trimble TX5. První část práce
pojednává krátce o teoretickém základu použité metody, dále je popsán postup
při zaměření objektu, volba stanovisek a kontrolních bodů v terénu. Registrace mračen
bodů byla provedena v softwarech Geomagic Studio 2012 a Leica Cyclone.
Pro následné vyhotovení trojúhelníkové sítě a její optimalizaci byl využit opět software
Geomagic Studo 2012 spolu se softwarem PoissonRecon a Agisoft PhotoScan. Výsledné
mračno bodů bylo transformováno do souřadnicového systému S-JTSK a Bpv.
Hlavním výstupem práce je zregistrované mračno bodů a 3D trojúhelníkový model
vnitřního paláce hradu. Výsledky práce budou předány občanskému sdružení Hrádek,
který má hrad ve správě. Budou také sloužit jako dokumentace současného stavu pro
archeologický výzkum a prezentaci hradu na internetu.
KLÍČOVÁ SLOVA
Laserové skenování, 3D model, trojúhelníková síť, mračno bodů, hrad Helfenburk
u Úštěka
STATEMENT OF AFFIRMATION
I hereby confirm that I have developed and written this Master's thesis completely
by myself, and that other resources or means (including electronic media and online
sources) than those explicitly referred to, have not been used. The academic work has
not been submitted to any other examination authority.
In Prague on 28th July, 2015 …………………………………. Jana Poesová
ACKNOWLEDGEMENT
First and foremost, I want to thank to my supervisor, Ing. Bronislav Koska, Ph.D., for
his guidance, patience and support throughout my way to finish this theses. His advice
and comments were very beneficial during my elaboration of the thesis. In addition,
I would really like to thank to my German supervisor, Prof. Dr. rer. nat. Martin Oczipka,
for helpful attitude during my Erasmus stay in Dresden.
Special thanks belong to CTU in Prague for the access to a high-performance PC and
software equipment; to Hochschule für Technik und Wirtschaft in Dresden for
the access to a wide range of materials and for providing a computer; and to
Ing. Tomáš Honč from Geotronics Praha s.r.o. for lending a 3D scanner.
I share the credit of my work with my classmates Ing. Petra Dífková, Ing. Alžběta
Prokopová, Ing. Martin Toušek and Ing. Lukáš Vosyka. They helped me especially with
the measurement.
Last but not least I would like to thank to Bc. Zuzana Krotovychova and Ing. Linda
Dvořáčková for proofreading of the English language in my thesis.
And finally, I would like to express my heartfelt gratefulness and lot of thanks to my
parents and grandmother, to my sisters and numerous friends who endured this long
process with me, always offering encouragement and love.
CTU in Prague Contents
7
Contents
Contents ............................................................................................................................ 7
Index of Abbreviations ................................................................................................... 10
Introduction .................................................................................................................... 11
1 Laser Scanning ...................................................................................................... 13
1.1 The General Information about the Technology ........................................ 13
1.1.1 Basic Measurement Principle of Laser Scanning ............................ 14
1.1.2 Types of Laser Scanners ................................................................... 17
1.1.3 Factors Influencing the Measurement ............................................ 19
1.1.4 Data Processing ................................................................................. 22
1.1.5 Possibilities of Modelling ................................................................. 25
1.2 Advantages and Disadvantages of Laser Scanning ..................................... 28
2 Castle Helfenburk near Úštěk .............................................................................. 29
2.1 Location and General Information............................................................... 29
2.2 Description of the Castle Helfenburk .......................................................... 30
2.3 History of the Castle...................................................................................... 31
2.4 Cartographical Documentation of the Castle.............................................. 32
3 Measurement of the Castle ................................................................................. 34
3.1 Reconnaissance ............................................................................................. 35
3.2 Building of the Geodetic Point Field ............................................................ 35
3.3 Laser Scanning Measurement ...................................................................... 37
3.3.1 Method and Conditions.................................................................... 37
3.3.2 Using of Target during Measurements ........................................... 38
3.3.3 Layout and Work on the Standpoints.............................................. 39
CTU in Prague Contents
8
3.3.4 Laser Scanner – General Information.............................................. 41
3.3.5 Working Procedure and Scanner Settings ...................................... 43
3.4 Control Measurements ................................................................................. 45
3.4.1 Control Measurement during the Laser Scanning.......................... 45
3.4.2 Additional Control Measurement .................................................... 45
3.4.3 Computation of the Coordinates of Check Points .......................... 46
4 Data Processing .................................................................................................... 48
4.1 Export of the Measured Data ....................................................................... 49
4.2 Resampling in Geomagic Studio................................................................... 50
4.3 Registration in Cyclone ................................................................................. 52
4.3.1 Identical Points Modelling................................................................ 53
4.3.2 Workflow and Computation of Registration................................... 54
4.3.3 Transformation of a Point Cloud to S-JTSK ..................................... 61
4.4 Accuracy Assessment at Registered Model Based on the Control
Measurement .................................................................................................................. 63
4.4.1 Coordinates of Check Points ............................................................ 63
4.4.2 Control Measurement Result........................................................... 64
4.5 Completion of Registration and Preparation of Model in Geomagic ........ 66
4.5.1 Workflow in software Geomagic Studio ......................................... 66
5 3D Model............................................................................................................... 68
5.1 Poisson Surface Reconstruction ................................................................... 68
5.2 Completion of the Triangular Mesh............................................................. 70
6 Results ................................................................................................................... 73
7 Discussion.............................................................................................................. 75
8 Conclusions ........................................................................................................... 77
Reference List ................................................................................................................. 78
CTU in Prague Contents
9
List of Figures .................................................................................................................. 83
List of Tables ................................................................................................................... 84
List of Appendices .......................................................................................................... 85
CTU in Prague Index of abbreviations
10
Index of Abbreviations
ASCII American Standard Code for Information Interchange
BIM Building Information Modelling
Bpv Baltic Vertical Datum - After Adjustment
CAD Computer Aided Drafting
ČSN Czech Technical Standards
GNSS Global Navigation Satellite System
ICP Iterative Closest Point
IEC International Electrotechnical Commission
LiDAR Light Detection and Ranging
MEP Mechanical, Electrical, and Plumbing services
RGB Colour Model (Red, Green, Blue)
S-JTSK Datum of Uniform Trigonometric Cadastral Network
TFM Transcription Factor Matrix
UAV Unmanned Aerial Vehicle
2D Two-dimensional Space
3D Three-dimensional Space
CTU in Prague Introduction
11
Introduction
The idea to work out the measurement of the ruins of castle Helfenburk near Úštěk,
dated approximately from the middle of the 14th century, arose from the citizen
association "Hrádek" in need to create an as-built castle documentation. Based on the
agreement between CTU representatives and members of the association Hrádek,
a plan of measurement was adopted and it was carried out by a group of five students
during four weekends in 2014. The subsequent data processing and various types
of outcomes were divided among them.
Laser scanning was chosen as the most convenient method of data capturing. It is
a rapidly developing surveying method with wide range of use in architecture, civil
engineering, industry and cultural heritage application. Due to the capability
of capturing voluminous data by nonselective measurement method, it is suitable for
complex objects of all sizes. Nowadays, there is a wide range of output possibilities,
from viewer applications of a point cloud to a complex 3D model (spline or geometric
primitives modelling or polygonal mesh), even 2D plans like floor plans or cross
sections are possible.
This paper contains the information about the castle measurement. The first part
deals with theoretical introduction and the bases of laser scanning method; it provides
a description of how the laser scanner works and what kinds of influences can have
impact on the measurement. It also provides brief information about the data
processing, the possibilities of modelling and the advantages and disadvantages
of laser scanning.
General information about the castle, its history and the previous measurement
of the castle are described in the second chapter. Remaining parts of the thesis deal
in detail with the work and the measurement in field and with the subsequent data
processing, accuracy assessment and creation of a 3D model.
The main aim of this work is thus the creation of a detailed as-built documentation
of the castle Helfenburk near Úštěk. The results should serve as a basis for
an archaeological research with the intention of forming the original castle with all the
CTU in Prague Introduction
12
building parts which have not been preserved up to now, or for a possible
reconstruction. Considering the fact that some building parts of the castle are slightly
moving relatively to each other, the 3D model can serve for monitoring of the
movements for structural engineers. Furthermore, it could be uploaded on the castle
website so as to enable the tourist to do “virtual sightseeing”.
Since the thesis is written in English which is not my mother tongue I used the help
of the following vocabularies while writing this paper. The most helpful vocabulary was
Lingea Lexikon 5, ver. 5.1.0.0 (2010) – electronic form on DVD (Technical vocabulary
and Platinum vocabulary). Translations of technical terms from surveying field were
found in an online vocabulary developed by Research Institute of Geodesy,
Topography and Cartography, v.v.i. in Zdiby (https://www.vugtk.cz/slovnik/). In some
cases I also used Google translator (https://translate.google.com) and the online
vocabulary on Seznam.cz (http://slovnik.seznam.cz/).
CTU in Prague Laser Scanning
13
1 Laser Scanning
This chapter reviews the general information about the laser scanning technology,
its principles of measurement and the factors influencing the measurements. In brief it
is also addresses the topic of the theoretical bases of registration process and it deals
with the advantages and disadvantages of laser scanning technology.
1.1 The General Information about the Technology
Laser scanning has become well established surveying technique for the capturing
of as-built spatial information. Technological advances have led to laser scanners
capable of acquiring long range measurements at rates of tens to hundreds of
thousands of points per second, at distances of up to a few hundred meters and with
accuracy on the scale of centimetres to a millimetres. Furthermore the software tools
for processing and analysing 3D point clouds have been improving in their ability to
handle the enormous point clouds produced by laser scanners and to integrate the use
of point cloud data into CAD modelling software. [1]
In the early stages, terrestrial laser scanning was short range and mainly used in the
automotive and industrial design process to facilitate the Computer Aided Design
(CAD) process. This helped in the mass production of consumer products. However,
since technology keeps evolving, other potential fields are entered. Middle range
scanners were developed for the petrochemical industry. The continual progress and
development of laser scanning enabled the documentation of complicated industry
devices in 3D. Until then it was documented only in 2D.
The high quality 3D point clouds produced by laser scanners are nowadays used for
a diverse array of purpose including the two-dimensional drawing, the three-
dimensional modelling and analysis in wide variety of applications for the architecture,
engineering, and construction domain. [2] [3]
CTU in Prague Laser Scanning
14
1.1.1 Basic Measurement Principle of Laser Scanning
Laser scanning describes a method where a surface is sampled or scanned using
laser technology. It analyses a real-world or object environment to collect data
on its shape and possibly its appearance (for example the colour information).
Laser scanning is thus a case of a non-selective method of acquiring 3D information.
Measured values are vertical angle, horizontal angle and distance from the scanner
to the measured point. Measurements are stored as a list of polar coordinates or more
usually as the Cartesian coordinates with the origin of coordinate system in the centre
of the scanner positions. The result is called a point cloud. [3] [4]
Laser scanners usually consist of a range measurement system in combination with
a deflection of the laser beam, directing the laser beam into the direction to be
measured and all this process is driven by a software. Furthermore the system consists
of other facilities like battery, tripod etc.
Figure 1 Detection chain of a laser scanning system [2], page 11
The deflection system points the laser beam into the direction to be measured, the
laser beam is emitted and the reflected laser light is detected. The accuracy of distance
measurements depends mainly on the intensity of the reflected laser light and
therefore directly on the reflectivity of the object surface. The reflectivity depends
on the angle of incidence and surface properties. [5]
There are two basic active methods for optical measuring of a 3D surface, which are
categorised by the principle of the distance measurement system. The first method is
light transit time estimation, also known as “Time-of-Flight Measurement” or LiDAR
(Light Detection and Ranging) systems. Time-of-Flight Measurement may be also
CTU in Prague Laser Scanning
15
realised indirectly via phase measurement in continuous wave lasers (“Phase
Measurement Techniques”). In general, we can speak about the category “Time-Based
Measurement”. The second method is “Triangulation”. [2]
The subchapters cited below are adapted from [2] [3] [5] [6] [7] [8].
1.1.1.1 Time-Based Measurement
Time-of-Flight principles
Nowadays it is the most popular measurement principle. This technique allows
unambiguous measurements of distances up to several hundred metres.
The advantage of long ranges implies reasonable accuracy depending on the range and
scanner’s parameters (from the centimetres up to few millimetres). Adequate usage
is for exterior high accuracy scans, for example in an architectural reconstruction,
surveying, engineering, planning and forensics (slower data acquisition and higher
noise compare with the phase measurement techniques).
These scanners have a laser diode that sends a pulsed laser beam to the scanned
object. The pulse is diffusely reflected by the surface and part of the light returns to
the receiver. The time that light needs to travel from the laser diode to the object
surface and back is measured and the distance to the object calculated using
an assumed speed of light.
Equation 1 presents basic formula for range calculation:
� =�
�
�
2
� – range (distance from source to target surface)
� – speed of light (299 792 458 m/s in a vacuum)
� – round trip (time delay)
� – index of refraction (if the light waves travel in air then the index of refraction
depends on the air temperature, pressure and humidity and must be applied
to c, n ≈ 1.00025. When we assume c= 3 x 108 m/s, then n = 1 – value for the
vacuum)
CTU in Prague Laser Scanning
16
Phase Measurement Techniques
The range for this technique is limited by the distance of one hundred metres.
Accuracy of the measured distances within some millimetres is possible. Adequate
usage is for the interior high density and high accuracy scans, for example in MEP1,
architectural and structural field, facilities management, and forensics (faster data
acquisition and lower noise compare with the time-of-flight principles).
Typical phase-shift scanners modulate their signal using sinusoidal modulation,
amplitude based or frequency based modulation, pseudo-noise or polarization
modulation. Continuous modulated signal is being sent out and compared with the
returned reflected signal. The time delay and subsequently the distance are calculated
from the phase difference between sending and receiving wave. Two wavelengths are
needed to resolve an ambiguity (shown in Figure 2 (a)).
1.1.1.2 Triangulation
Triangulation techniques are divided into passive and active methods and are based
on the optical triangulation method. Passive techniques do not use structured light,
the method is based on stereo image processing, allowing to obtain 3D reconstructions
from a set of overlapping images.
The principle of triangulation is this: a light spot or stripe is projected onto the
surface of the object and the position of the spot on the object is recorded by one
or more CCD cameras (Observation direction in Figure 2 (b)). The angle of the light
beam leaving the scanner is internally recorded and the fixed base length (B in Figure 2
(b)) between laser source and the camera is known from calibration. The distance from
the object to the instrument is geometrically determined from the recorded angle and
the base length.
These types of scanners are more suitable for use in industrial applications and
reverse engineering. Triangulation laser system allows measurements up to few
meters and accuracies up to few micrometres can be achieved with this technology.
The capability of capturing number of points per second is the smallest one within all
the methods, precisely from 200 to 10 000 points per second.
1 Mechanical, electrical, and plumbing services (MEP) is a significant component of the
construction supply chain. MEP design is critical for design decision-making, accurate documentation, performance and cost-estimating, construction planning, managing and operating the resulting facility. [39]
CTU in Prague Laser Scanning
17
Figure 2 Methods for measuring of a 3D surface: (a) Light transit time (b) Triangulation [2], page 3
1.1.2 Types of Laser Scanners
Classification of laser scanners is a little bit difficult to be done. There are several
possibilities to do it, either based on the measurement principle, as it was described
above in chapter 1.1.1 or based on the technical specifications achieved.
There is further one main classification dividing the laser scanning into the
terrestrial and airborne family of scanners.
Airborne laser scanning is a scanning technique for data capturing from the Earth’s
surface in high resolution (a digital elevation model of the landscape). It is
an important data source for example for environmental and forestry applications.
Terrestrial scanners can be divided into mobile and static scanners. Mobile mapping
is a non-invasive, state-of-the-art solution that incorporates the most advanced
ground-based LiDAR sensors, cameras, and an inertial measuring unit to collect survey-
quality point data quickly and accurately. Further in this thesis only the static
terrestrial laser scanners will be considered. There is not a unique universal laser
scanner for all conceivable applications. Depending on the application the adequate
laser scanner has to be selected. [6] [9]
The below mentioned categories are defined in [10]. Since the paper was written
in 2005, some of the limits of the categories should be updated because of the fast
development of laser scanning systems, especially the speed of acquiring points and
the range of scanners are to be highlighted.
CTU in Prague Laser Scanning
18
One of the possible categorization of scanners:
− Operating principle
o Triangulation
o Time-of-flight
− Speed of acquiring points
o Low speed systems (less than 50 000 points/sec)
o Middle speed systems (from 50 000 to 200 000 points/sec)
o High speed systems (from 200 000 to 1 000 000 points/sec)
o Very high speed systems (more than 1 000 000 points/sec)
− Accuracy
Manufacturer of the laser scanner defines accuracy of the scanner
under certain conditions (distance, ambient light, object reflexivity,
etc.)
o Very accurate systems (from 0.01 mm to 1mm) – usually
triangulation scanners for shorter distance
o Accurate systems (from 0.5 m to 2 mm) - triangulation scanners
designed for longer distances and phase-shift scanners
o Middle accurate systems (from 2 mm to 6 mm) – time-of-flight
principle designed for middle range
o Systems with low accuracy (from 10 mm to 100 mm) – time-of-flight
principle designed for long range.
− Range
o Very short range systems (from 0.1 m to 2 m)
o Short range systems (from 2 m to 10 m)
o Middle range systems (from 10 m to 100 m)
o Long range systems (more than 100 m)
− Classes of laser and safety [11]
Laser classes are based on international standard IEC 60825 (in the Czech
Republic ČSN EN 60825-1) Laser safety. Classes are defined considering the
possible damage of the human eye while looking directly to the laser source.
o Class 1: safe under all conditions of normal use.
CTU in Prague Laser Scanning
19
o Class 1M: safe under all conditions except for when the beam passes
through magnifying optics.
o Class 2: safe because they usually cause a 'blink reflex' which
protects the eye.
o Class 2M: safe because of the 'blink reflex' unless the beam passes
through magnifying optics.
o Class 3R: safe if handled carefully and with restricted beam viewing.
Class 3R lasers can be hazardous where direct beam viewing
is involved. The laser scanner Trimble TX5 that was used for
measurement of the castle is from this class (more in chapter 3.3.4).
o Class 3B: it is hazardous when direct beam viewing occurs, though
diffuse reflections of the laser are considered non-hazardous.
o Class 4: it causes eye or skin damage as a result of direct beam
exposure.
Classes 1 – 3R are considered safe for survey in Europe.
Figure 3 Pictogram of class 3R of the laser classification (http://www.lasersafetyfacts.com/3R)
1.1.3 Factors Influencing the Measurement
Laser scanning system consists of many components, where each of them has
a different specific accuracy that is called the instrumental error. The manufacturers
publish the accuracies of laser scanners to illustrate the advantages of their particular
product. However, the experience shows that sometimes these should not be taken
in account and that the accuracy varies within the scanners depending on the
individual calibration. Except these instrumental errors, the measurements are
impacted by the qualities of the measured object (called object-related errors) and it is
also swayed by an environmental and methodological errors.
CTU in Prague Laser Scanning
20
1.1.3.1 Instrumental Errors
Instrumental errors can be both random and systematic, which are caused by the
non-linearity of the measurement units.
One of these problems is the consequence of laser beam propagation, which means
the widening of laser beam with the distance travelled. The beam divergence has
a strong influence on the cloud resolution as well as on the positional uncertainty of
a measured point. It has an impact on the angular location of the point measured.
The apparent location of the observation is in a close proximity of the centreline of the
emitted beam. However, the actual point is located somewhere in the projected
footprint.
One of the most important consequences of the beam divergence is the mixed edge
problem (see Figure 4). It happens when a laser beam hits the edge of an object and it
is split into two parts. While the first one ended on the first part of object, the other
part travelled further to another surface. The final measured distance from scanner to
the point is determined based on an average of both returned signals and therefore
the point is localized in the wrong place. Mixed edge problems can cause an algorithm
to erroneous structures that do not exist, which can cause significant errors in the
dimensions of the modelled surfaces.
The other instrumental errors are the range uncertainty and the angular
uncertainty. Most middle and long range terrestrial scanners provide a range
uncertainty of about 5 mm to 50 mm within a range of 50 m. In the modelling phase
these errors are minimized by averaging or by fitting of primitive shapes to the point
cloud. Bigger problems are caused by the angular uncertainty. A small angle difference
in guidance of the laser signal in certain direction can cause a considerable coordinate
error when the distance from the scanner increases. The angular accuracy depends
on any error in the positioning of the rotating mirrors and the accuracy of the angular
measurement components.
CTU in Prague Laser Scanning
21
Figure 4 Mixed edge problem [12], page 40
1.1.3.2 Object-related errors
Laser scanners have difficulty with many types of surfaces that occur commonly
in the built environment, including low-reflectance surfaces, specular surfaces (shiny
metal and mirrors) and transparent surfaces (windows).
All laser measurement systems assume that part of the light that is emitted by the
laser will be reflected back to the sensor. The laser beam is affected by the absorption
of the signal travelling through the air, the reflection of the material being measured
and the angle of incidence between the laser beam and the surface being measured
(according to Lambert’s cosine law2). That means that for very dark surfaces, which
absorb most of the visible spectrum, the reflected signal will be very weak and
therefore the point accuracy will be corrupted by noise. The higher the percentage
of the light reflected by the material, the stronger the signal that is “bounced” back to
the receiver of the laser scanning system. Recording surfaces of different reflectivity
also leads to systematic errors in range, which are sometimes several times larger than
the standard deviation of a single range measurement.
2 Lambert's cosine law states that the intensity of radiation or luminous intensity observed from
an ideal diffusely reflecting surface or ideal diffuse radiator is directly proportional to the cosine of the angle θ between the direction of the incident light and the surface normal. [40]
CTU in Prague Laser Scanning
22
1.1.3.3 Environmental conditions
The environmental conditions such as the temperature, atmosphere, interfering
radiation and distortion from motion have considerable impact on accuracy in certain
occurrences.
The temperature inside the scanner may be increased due to the heating from
external radiation, which can cause distorting of the scanner data.
The atmosphere impact is difficult to exactly determine, but in terrestrial cases
it does not seriously affect the results of short and middle scan distances. The index
of refraction is affected by natural errors that stem primarily from atmospheric
variations of temperature, air pressure and humidity and thereby the wavelength of
electromagnetic energy which is necessary for correct measuring is modified. Fine
particles in the air, like those that compose smoke, snow, rain or fog, cause the narrow
laser beam to scatter.
1.1.3.4 Methodological errors
These errors are caused due to badly chosen survey method, incorrect choice of the
scan position and number of stand points, incorrect setting of scanner parameters
(especially the resolution) and potential errors generated during the registration
process. To put it simply, it depends actually on the users experience with this
technology.
This subchapter 1.1.3 Factors Influencing the Measurement was composed from [1]
[3] [13] [12] [14].
1.1.4 Data Processing
Usually there are multiple scan positions in scanner coordinate system which are
necessary for capturing of the entire object of interest. To be able to align different
scan positions, the exact position and orientation of these scanner coordinate systems
according to a local or global site coordinate system have to be known.
The basic principle of registration is computation of transformation elements
(rotation and translation) for all the scan positions. Usually, one scan is set as a home
scan, which means the rest of scans are transformed to the coordinate system of this
home scan. Other possibility is to transform all scans to a common reference system.
CTU in Prague Laser Scanning
23
Different approaches of solving the registration exist and are used in practice.
The most standard method is Target-to-target registration and Cloud-to-Cloud
registration.
1.1.4.1 Target-to-target Registration
The fundamental aim of this method is recognition, precise localization, and
labelling of the special shaped targets in each point cloud. These targets, whose
coordinates are known in the ground coordinate system, are used as common point
in calculation of simultaneous 3D similarity transformation to put all scans in one
common coordinate system. To perform the registration at least three target
correspondences between two scans are needed. However, it is always better to have
more than three, so that errors can be minimized by performing a least-squares
optimization.
Nowadays the aim is to achieve an automatic point cloud registration. Some
scanners have integrated the inertial measurement unit enabling the registration
in the field. But the manual registration is still more common. The effort of scanner
manufacturers is to integrate an automatic precise localization and labelling of the
target into the software. In the old version 7.0 of software Leica Cyclone, which was
used for the registration of the castle, this function has not been at disposal.
The current version of registration software such as Leica Cyclone, Faro Scene, Trimble
RealWorks have already been able to recognize the targets automatically.
The targets made from highly reflective material (during the measurement of the
castle the sphere targets were used, more in chapter 3.3.2) or printed paper targets
can be also used. When no printer or special targets are available, targets may be
improvised by using objects to which an ideal geometrical surface can be fitted.
(a) (b) (c)
Figure 5 Various types of target (a) White sphere, (b) Planar target, (c) Curved planar target (see chapter 3.3.2)
CTU in Prague Laser Scanning
24
1.1.4.2 Cloud-to-Cloud Registration
Another way of registration of two point clouds is by using the point cloud overlap.
If two point clouds have enough overlap (generally 30 – 40%), for example an Iterative
closest point algorithm (ICP) can be used for complete point cloud registration. The ICP
algorithm takes two point clouds as an input and return the rigid transformation
(rotation matrix R and translation vector T), that best aligns the point clouds. There
exist many researches how to improve the method to be the best fitted.
The methods based on surface matching can be applied only if the scans to be
aligned have a proper geometry and sufficient overlap. But it should be used with
caution. There could be a danger during the scanning of long linear structures, where
multiple setups are required. Small errors in each registration pair may multiply and
result in large global errors.
Figure 6 Example of registration algorithm (http://uk.mathworks.com/help/releases/R2015b/vision/ref/point_cloud_registration.png)
CTU in Prague Laser Scanning
25
After the registration process it is recommended to make slices of the registered
point clouds in horizontal and vertical plane to check if the scans really fit best to each
other. To achieve the best result of following processing (especially the triangulation)
it is possible to carry out the noise filtering which means removing the noisy data like
an artefact, vegetation, bad surface reflection etc. and in addition it is possible to
resample the point cloud to reduce the extensive data amount (see chapter 4.5.1).
This subchapter 1.1.4 Data Processing was drawn up of [3] [15] [16] [17] [18].
1.1.5 Possibilities of Modelling
In the moment when the scan positions are registered into one coordinate system
and the erroneous points are removed from point clouds as best as possible, there
is an option to choose from different possibilities of data output. In some cases it is
sufficient to put the scans into viewer applications enabling the basic operations and
visualization of point clouds like a measurement between the points, going through
the scan positions and viewing panoramic images from the scan positions. These
software are developed for example by Autodesk, Inc. company (Recap 360) or Faro
company (SCENE WebShare Server using the internet browser for viewing the scan
data).
Another possibility is to create 3D model using the direct modelling from point
cloud or triangulation method.
1.1.5.1 CAD Model direct from Point Cloud
The goal of direct modelling from point cloud is to create simplified representation
of object components by fitting the geometric primitives to the point cloud.
The objects can be modelled by using a spline or geometric primitives.
Direct 3D modelling from point clouds is a matter of human interpretation.
Nowadays there is an effort to create algorithms that can automate these tasks
(for example the software EdgeWise developed by ClearEdge 3D from US). But despite
the technology it is still necessary to make a detailed control of whether everything fits
into the point cloud and nothing misses there. This method has not been used so much
yet. Therefore the most expanded software packages in this domain are still plugins for
CAD packages like AutoCAD or MicroStation, whose up-to-date version supports the
work with point cloud. It is suitable for creation of spline model or 2D drawing from
CTU in Prague Laser Scanning
26
point cloud. The software to perform the geometric primitives modelling is for
example Leica Cyclone, where the registration of the castle was done (see the chapter
4.3).
CAD model is advantageous for the objects with geometric primitives (e.g. pipes,
walls), therefore it is more useful in the industrial complex. It could also be used for
creating BIM (building information modelling). Nowadays there is a big effort to
develop some method for effective automatic creation of BIMs and its application
in praxis.
Figure 7 CAD model with point cloud
(http://www.pointcloud2cad.com/wp-content/uploads/2013/07/overlay02-1030x523.png)
1.1.5.2 Triangulation
Many methods have been developed to create a regular and continuous mesh
representation from a point cloud. It is very complicated to classify all the
reconstruction methods.
The most rudimentary and easiest algorithm is the Delaunay triangulation.
The Delaunay triangulation operates on the basis of the principle that a circle through
the three points of any triangle does not include any other point of the data set.
It generates compact triangles mesh with the largest minimum angle.
More complex meshing algorithms like the ball pivoting algorithm [19] or the
marching cubes algorithm [20] have been developed and are able to triangulate huge
datasets with low memory consumption.
CTU in Prague Laser Scanning
27
When the model is created, editing operations are commonly applied to improve
and repair the polygonal surface (smoothing the surface, closing holes, reducing the
number of polygons, repairing the normals, addition of vertices etc.). If the images are
available, the texturing can be performed.
Triangle mesh is more suitable for objects with an irregular shape, for creation
of terrain model etc.
This chapter 1.1.5 Possibilities of Modelling was composed from [1] [3] [6] [7] [21]
[22] [23].
Figure 8 Smooth reconstruction of Igea and different smooth parameters (http://www.scielo.br/img/revistas/jbcos/v9n3/05f07.gif)
CTU in Prague Laser Scanning
28
1.2 Advantages and Disadvantages of Laser Scanning
Laser scanning is a rapidly developing method of survey that implies that a lot of
unresolved problems from past disappeared. [4]
The main advantages can be mentioned:
- High speed and long range of measurement
- Non-selective and non-contact method of surveying
- The elimination of errors and omissions from survey results
- Detailed 3D models can be quickly elaborated.
- Rapidly developing technology and world-wide research
- Easy to use
Between the main disadvantages can be included:
- Huge demands on data capacity of computer technology
- Time consuming data processing
- Weather requirement (measurements during the rain or snow are not
recommend because of a potential danger of noisy points from reflected
water and danger of a potential damage of the scanner)
- Relatively expensive hardware and software required to process the data
CTU in Prague Castle Helfenburk near Úštěk
29
2 Castle Helfenburk near Úštěk
This chapter deals with the general information about the surveyed castle
Helfenburk near Úštěk. An account of its location and history is provided, together
with a brief overview of the cartographical documentation of the castle.
2.1 Location and General Information
Castle Helfenburk near Ústěk, in the course of its history also known as Hrádek,
Hradec or Hradišťko, is located in the Northern Region of the Czech Republic,
in the district Litoměřice, the cadastral area named Rašovice u Kalovic (50.5792247N,
14.3838600E (http://mapy.cz)), with an approximate altitude 315 m above the sea
level. The town Úštěk, located 2.5 km to the west of the castle, is the owner of the
object. However, the citizen association "Hrádek", which has the castle in rent, takes
care of the reparation and service for the public.
The complex of walls, up to 12 meters high, is open to the public all year round.
Only the 17m high tower, which dominates the entire complex, is open only
at weekends. The castle is accessible by foot. It is also possible to get there by car, but
a special permission is required. [24] [25]
Figure 9 Location of the castle (www.mapy.cz)
CTU in Prague Castle Helfenburk near Úštěk
30
2.2 Description of the Castle Helfenburk
The castle was one of the largest castles in northern Bohemia. It consists
of an extensive complex of ruins, which are situated on a rocky sandstone ridge above
the brook called Rašovický potok, maintained here from the Middle Ages. We can find
there ruins which come from two different historical periods and which were built
in two different architecture styles. The first part consists of the outer wall and the
tower and the second one is the inner castle.
The outward fortification was built later than the inner part of the castle and it is
currently, together with the tower, one of the most well-preserved parts.
The inner castle was built on the top of a sandstone ridge, which extends over three
rocky blocks. At present, only the masonry is preserved on the middle rocky block.
The entire interior of the complex is protected by sandstone rocks.
The outer fortification consists of massive walls that are surrounded by battlements
and loopholes. Walls are up to twelve meters high at some places and the narrow
terraces line the inside wall. There was a moat around the walls. The castle was
accessible through a gate connected with the surroundings via drawbridge. A small
entrance, for pedestrians only, was situated on the right side of the gate. Another way
to the castle was through the 'down' gate which connected the outer fortification with
the inner castle. These gates and the tower are located in the east part of the castle.
Figure 10 Aerial photograph of Helfenburk near Úštěk, ©Vladimír Bärtl (http://static.panoramio.com/photos/original/66453815.jpg)
CTU in Prague Castle Helfenburk near Úštěk
31
Sandstone tower, standing on a separate rock, is a part of the fortification. A ground
plan of the tower has a square shape. The tower was completely renovated in the
19th century.
The western side of the walls is interrupted by a high rocky block, which is part
of the fortification. A small entrance is placed next to this rock. [24] [26]
2.3 History of the Castle
Based on the oldest written sources it is believed that castle Helfenburk was
founded around 1350 by Jan from Klinštejn, called also Jan von Helfenburk, of the
house of Ronovci. The Prague Archbishop Jan Očko from Vlašim bought it from Jan von
Helfenburk in 1375. The period of rule of Jan von Helfenburk was a time of prosperity
for Helfenburk, during which the castle became the new centre of archbishop manors
on the right bank of the river Elbe.
Between the years 1375 and 1379 the first great rebuilding of the castle was carried
out, with the fortification as the main part of this rebuilding. Its original length was
277metres.
The archbishop John from Jenštejn, a nephew of Jan Očko from Vlašim, continued
with an aggrandizement of the castle. Between the years 1390 and 1395 the tower and
the new fortification were built. The castle was rebuilt repeatedly in the following
years, but the base of the castle has not been changed.
The last archbishop who inhabited the castle Helfenburk was Conrad from Vechta.
Conrad was a favourite of King Václav IV. and in 1421 he converted to the Hussites.
From the second half of the 15th century the castle was owned by the aristocracy.
In 1592 Hrádek was bought by Jan Sezima from Sezimovo Ústí. However, his property
and estate were confiscated due to his participation in the anti-Habsburg rebellion
in 1622. The new owner of the castle was Jesuit Order.
The Thirty Years' War and the following decades, when the Castle was not
inhabited, caused gradual dilapidation of its buildings and fortification.
Around the year 1720 a gamekeeper's lodge was established there and in the early
19th century the remains of the castle became a destination of pilgrims.
CTU in Prague Castle Helfenburk near Úštěk
32
In 1839 the castle was bought by Ferdinand Lobkowicz. Later, the castle was owned
by a textile industrialist Josef Schroll, who bought it in 1871. During this period the
castle, and especially its tower, was renovated, but later, its dilapidation continued.
Since 1958 the castle has a status of a cultural and historical monument. In 1967
volunteers began to take care of the dilapidated castle under the supervision
of preservation specialists. The fortification and the tower were renovated. A historical
research was initiated there, which has brought interesting results; especially during
the cleaning of a 57m deep water well many interesting objects of everyday use from
different historical periods were found. [24] [27]
2.4 Cartographical Documentation of the Castle
This chapter is based on the information received from Ing. V. Kotrejch who carried
out the measurements at the castle in the second half of the 80s of the last century.
The first measurement of the castle was carried out by Mr. M. Záveský and
Mr. J. Krupka in 1983. It was measured in the local coordinate system. Survey point
number 18, to which were assigned the coordinates [100 100], was chosen for
a coordinate origin of the local coordinate system (Appendix A). A default elevation
point was located to the first step of the stairs leading to the tower, its height
was assigned the figure 100 meters. This point does not exist anymore because it was
destroyed during the rebuilding of the courtyard.
There was also a traverse measured inside and outside of the castle. A layout of
the stand points, the position of the default elevation point and the location of the
coordinate axes are displayed in Appendix A. A building of the geodetic point field was
performed using stakes with nails or nails in the rock. This geodetic point field was only
temporary and these points do not exist. Measurement was performed using
a theodolite with a distance meter. The result of the measurement was a tacheometric
plan in a scale 1: 200 (Figure 11).
Ing. Vladimir Kotrejch accomplished the measurement of the castle in the second
half of the 80s of the last century. The main idea was to use the first measurement and
to focus more not only on the terrain but also on the buildings (the shape of the walls,
chinks in the rocks etc.). The initial geodetic point field was partially destroyed by
construction changes. The initial geodetic point field must have been filled in with
CTU in Prague Castle Helfenburk near Úštěk
33
new points which had to be connected to the original local coordinate system.
Measurement was carried out using a theodolite with a distance meter, a 50 m tape
measure and a levelling instrument.
Detailed measurements were realized in the vicinity of the wall. A measuring
method of forward intersection was employed for the rock wall and its surroundings.
Thereafter, the measurement was stopped because there were not sufficient facilities
for displaying the measured data.
Figure 11 Part of tacheometrical plan made by M. Závetský and J. Krupka in 1983 (archive of Ing. V. Kotrejch)
There was further a photogrammetric mapping of the castle carried out by
Ing. Pavel Havlenka in 1988. The results were connected to the local coordinate system
through the ground control points. These points were determined during the second
measurement by Ing. Vladimir Kotrejch. The results of aerial photogrammetry were
processed between the years 1989 and 1990.
CTU in Prague Measurement of the Castle
34
3 Measurement of the Castle
The object of the measurement were the entire premises of the castle Helfenburk
(specifically the outer fortification, the tower, the ruins of the inner palace and
surroundings of the castle). Time schedule of work is displayed in the Table 1.
An accurate survey of the castle was done by the method of laser scanning and was
divided into two phases. The first one took place between 22nd and 23rd March 2014,
the second one took place between 3rd and 4th May 2014. The measurement of the
object was carried out by a group of students consisting of me and Ing. Alžběta
Prokopová, Ing. Martin Toušek and Ing. Petra Dífková.
There was also created a geodetic point field mainly in order to transfer the model
into the Czech coordinate system named S-JTSK (Datum of Uniform Trigonometric
Cadastral Network) and an altitude net named Bpv (Baltic Vertical Datum - After
Adjustment). Measurement and processing of the new geodetic net is described
in detail in the master's thesis of Ing. Lukáš Vosyka [28].
In the following sections there is described the operating procedure of precise
measurements in field including description of the scanner, preparations before the
measurement, the measurement itself and the process of check measurement.
If it is not specified otherwise, this chapter is based on the practical experiences
acquired during the data processing.
Date Carried out work
11.2.2014 Reconnaissance of the castle
28.2.2014 Building of the geodetic point field
22.-23.3.2014 Measurement of the castle by laser scanning and measurement of the geodetic point field 3.-4.5.2014
22.2.2015 Control measurement
Table 1 Time schedule of the work
CTU in Prague Measurement of the Castle
35
3.1 Reconnaissance
A reconnaissance of the castle and its surroundings was undertaken before the
beginning of the measurement of the entire castle. The reconnaissance of the entire
complex was realized on February 11, 2014. The aim of the reconnaissance was to
estimate an approximate timetable and workflow of the measurement procedure.
There was also discussed how to measure and connect a future geodetic point field.
The method of laser scanning had already been settled before the reconnaissance.
3.2 Building of the Geodetic Point Field
A geodetic point field was constructed on February 28, 2014. The main reason for
the construction of a geodetic point field of the castle was to connect the 3D model
from laser scanning with the coordinate system S-JTSK and height system Bpv.
The model was linked with a coordinate system on the basis of control points, whose
coordinates were determined from the geodetic point field.
Surveying nails and reinforcing steel were used for the building of the point field.
Length of the surveying nails was 8 cm and nails were placed in locations where it was
not possible to embed reinforcing steel, for example the rocks or the masonry of the
tower. In other cases there was used reinforcing steel of the length 40 cm, embedded
into the ground.
Figure 12 Building of the geodetic point field
CTU in Prague Measurement of the Castle
36
Two highest points (no. 601 and 602) were taken as the foundation of the point
field; first one is located on the castle tower and the second one is on a rock tower
in the north-western part of the castle. These points were determined using the GNSS
method in two periods. Geodetic point fields inside and outside of the castle were
constructed on the basis of these two points. Both geodetic point fields
are interconnected.
The construction and the method of computation of the geodetic net are described
in detail in Ing. Lukáš Vosyka's master's thesis [28].
CTU in Prague Measurement of the Castle
37
3.3 Laser Scanning Measurement
Measurement of the castle was performed during two weekends. The first weekend
was from 21th to 23th March and the second weekend was from 2nd to 4th May.
In total 91 scans were done capturing the entire area of the castle (20 scans were
done in the central part, 48 scans inside the wall in the inner part of the castle, and
23 scans of the wall from the outer side). Another 36 scans were taken inside the
tower.
The emphasis was placed on an accurate surveying of the central part of the castle.
Therefore, the density of standpoints is bigger there.
A list of equipment used for scanning:
- laser scanner Trimble TX5 (serial number LLS061203231)
- tripod with an adapter for mounting the scanner onto the tripod
- spheres and planar targets
3.3.1 Method and Conditions
The measurement of the entire castle was divided into several blocks. A different
workflow was applied for the measuring of the interior of the castle's tower and
the exterior of the castle. Selected control points used for the purpose of
transformation of the 3D model into S-JTSK were measured by using a total station.
The overview of the scanning blocks and of the weather during the measurement is
in Table 2.
Cloudy weather and nearly no rain made very good conditions for scanning. Rain
would make scanning impossible and sunnier weather would attract more tourists who
could have interfered with the measurement.
Scanning standpoints had to be chosen with regard to further processing of the
acquired data (registration of each cloud). It was necessary to secure an overlap and
common targets between two adjacent standpoints.
Scanned area of interest Date Weather
Inner palace ("interior") 22.3.2014 Cloudy
Interior of the tower 23.3.2014 Overcast, showery weather
Inner palace (courtyard) 3.5.2014 Cloudy
Area between the inner palace and walls 3.5.2014 Sunny
Outer walls 4.5.2014 Sunny
Table 2 Timesheet and weather conditions
CTU in Prague Measurement of the Castle
38
3.3.2 Using of Targets during Measurements
Before the beginning of the measurement it was necessary to place the targets in
the area of measurement so as to be visible from a currently measured scan.
The purpose of it is to facilitate later registration. These targets were white spheres
with a diameter of 200 mm and planar targets designed for laser scanning system
Trimble TX5 (shown in Figure 13). The targets were used for the registration of scans
and in addition for connecting the final 3D model to the coordinate system S-JTSK and
the height system Bpv.
Spheres were put on such places where their visibility on more scans was ensured.
Two spheres are theoretically enough to successfully execute the registration because
all the scans are already levelled (scanner is equipped with a dual axis compensator).
In case of unpredictable motion of one sphere during the measurement, it is possible
to recognize which one moved, but the subsequent registration would have to be
computed by some other ways (like ICP algorithm or natural identical points).
However, this method is very time consuming. Therefore, we endeavoured to capture
at each scan at least three spheres. Only six of them were available, so it was
necessary to choose carefully where to place them and sometimes it was difficult to
keep three spheres on each scan. The appropriate placement of the spheres also
played an important part in the subsequent merging of divided blocks of the castle. For
example, the targets connecting the inner and the outer wall were placed between
battlements. In total 71 identical spheres were captured in the scans.
Figure 13 The inner palace with sphere targets
CTU in Prague Measurement of the Castle
39
The above mentioned second type of planar targets should be used as control
points, primarily for linking the entire model into the coordinate system S-JTSK and the
height system Bpv. The targets were also used as identical points, in the same way like
spheres, during the process of registration. The targets were placed on the wall,
the rock walls and inside the castle. Since the castle consists of sandstone and huge
rough rocks it is impossible to use any kind of tape or sticky labels to attach the planar
target to the walls for a longer period of time. So the only way to attach them was to
squeeze it between two stones. The disadvantage of this method was that the planar
target was curved instead of plane, which could cause problems with identifying the
centre of planar target. These targets were further captured using a total station from
the points of geodetic net. Totally, 20 planar targets were surveyed in this way.
An overview of the location of standpoints (blue colour) and targets (red colours)
during the measurement is in Appendix B. A drawing of the castle for this overview is
copied from the ground plan of a photogrammetric mapping (see chapter 2.4).
3.3.3 Layout and Work on the Standpoints
A layout of standpoints for scanning of the castle's exterior was chosen with
a sufficiently big overlap and, if possible, with at least two common spheres on the
acquired scans. An overlap of the scans is important for subsequent registration
without identical points, for example using ICP algorithm.
The inner part of the castle was scanned during the first weekend. Since the castle is
highly cragged and some parts are difficult to access, the work took more time than
expected. The cragged terrain is shown in Figure 14. It was often necessary to use
a ladder or a rope for a safe access of the surveyors and the equipment. In some cases,
when the tripod was built in a steep terrain, it was necessary to be careful with the
scanner in case the tripod begins to fall down. The first few standpoints were captured
also with colour information. However, since it was a highly time-consuming process,
the measurement of colour was dropped. We also took advantage of one of the
characteristics of laser scanning - the ability to make a measurement even in the dark.
Some standpoints were carried out in the dark during the first weekend because we
were not able to realize the whole plan during the day due to the hard terrain.
CTU in Prague Measurement of the Castle
40
The rest of the castle was captured during the second weekend. It consisted of the
wall from the outside and further from the inside of the castle, and of the entire tower.
Due to the limited conditions no targets could be placed in the tower. Therefore, every
scan had to have guaranteed large enough overlap with the adjacent scan. In total
36 scan positions were done in the tower. Registration of these clouds could be
executed only on the basis of the overlap. Detailed processing of the tower is
described in the master's thesis of Ing. Martin Toušek [29].
Figure 14 Laser scanning of cragged and hardly accessible terrain
CTU in Prague Measurement of the Castle
41
Capturing and measurement of walls was less time-consuming because the
surrounding terrain was not so much vertically cragged. In order to connect scans
outside and inside the walls several spheres were placed on the top of the walls
so as to be visible from both sides. Due to the steep slope around the outer walls some
standpoints must have been placed close to the wall. Therefore, it is possible that it
could have deteriorated the quality of the points which were captured under the sharp
angle.
Area of interest Numeral name of scan positions Number of scans
Inner palace ("interior") 1 - 17, 211, 201, 202 20
Inner palace (courtyard) 220 - 226, 203 - 210, 212 - 219 23
Area between the inner palace and walls 227, 256 - 265, 239 - 247, 251 - 255 25
Outer walls 228 - 238, 248 - 250, 266 - 274 23
Interior of the tower 101 - 136 36
In total 127
Table 3 Overview of the scan positions
3.3.4 Laser Scanner – General Information
The company named Geotronics Praha s.r.o. lent us a scanner Trimble TX5 for the
measurement. It is a terrestrial scanner, based on a way of distance measurement.
The scanner is ranked to the category of phase-shift scanners. According to the
manufacturer the most important features of the scanner are compact lightweight
(it easily reduces setup time), efficient and high-speed scanning, integrated sensors
(temperature, inclinometer, compass, altimeter), user-friendliness (no need for
external controllers or cables) and data management.
The scanner is able to measure up to the range of 120 m. The minimum distance is
0.6 m. These figures are valued indoors as well as outdoors if there is low ambient light
and normal incidence to a 90% reflective surface. Single point measurements can be
repeated up to 976 000 times per second. The scanner covers a 360° x 300° field
of view and has an integrated colour camera with coaxial optics for an accurate RGB's,
the resolution up to 70 megapixel colour and with automatic adaption of brightness.
As stated by the manufacturer, the ranging error is ±2 mm at 10 m (90% reflectivity)
and at 25 m (10% reflectivity). Ranging error is defined as the maximum error in the
distance measured by the scanner from the initial point to a point on a planar target.
CTU in Prague Measurement of the Castle
42
A dual axis compensator should work with levels of each scan with an accuracy
of 0.015°and a range of ±5°.
The TX5 3D laser scanner produces an invisible laser beam with a wavelength
of 905 nm. The average laser power maximum is 20 mW and the beam divergence is
typically 0.19 mrad (0.011°). TX5 3D laser scanner is classified as a Class 3R laser
system. According to the standard, direct intra beam viewing may be hazardous for the
eyes when working within an area around the Class 3R laser system where the defined
exposure limits are exceeded (for more information about the Laser Classes see
chapter 1.1.2).
Thanks to the weight (5 kg) and size of the scanner (240 mm x 200 mm x 100 mm),
it is very easy to move and can be setup in very complex environments.
Due to this characteristics the scanner is suitable for wide range of use, for example
surveying (it can measure the distance, areas and volumes), building information
modelling, industrial facilities, inspection/reverse engineering, tunnelling, crime scene
and forensic. [30] [31]
Figure 15 Laser scanner Trimble TX5
CTU in Prague Measurement of the Castle
43
3.3.5 Working Procedure and Scanner Settings
The first step on every standpoint was to set up the tripod. It is necessary to make
sure that the surface is stable, that the tripod's feet are secured and that it stands
firmly in its position. The tripod‘s plate should be levelled as horizontally as possible.
The maximum allowed inclination is 5°. For mounting the laser scanner onto the tripod
we used a standard photo camera quick release's plate.
Before switching on the scanner it is needed to insert the SD card, where the data
are stored during the measurement. The scanner was powered by a battery with life
for up to 5 hours. No electricity was available at the castle and its surroundings, only
a power generator, which was sufficient to recharge the battery one or two times
a day and to download the measured data from the scanner to a laptop.
Last thing before turning on the scanner was to check that there were no objects
that could touch the mirror unit and that the scanner was able to move freely during
the scanning.
Switching on the scanner is done by pressing the On/Off button. Before taking the
first scan the scanner setting needs to be set up. To choose the ideal parameters,
the setting menu is accessible from the home screen by pressing the Parameters
button. There are two ways how to set the scanning settings. The first one is to select
a scan profile which is a predefined set of scanning settings; or it can be set up
manually.
Figure 16 Setting the scan parameters [32], page 46
CTU in Prague Measurement of the Castle
44
There are several features to be set up:
Resolution and Quality
Resolution is set in mega points and in the middle part of the screen there is written
the number of points the final scan approximately contains (MPts). Then the resulting
scan duration, vertical and horizontal scan points (Scan Size [Pt]), as well as the point
distance are described in the middle of the view. Point distance indicates the spacing
of two scanned points which are measured at distance of 10 meters from the scanner.
Higher quality reduces the noise in the scan data and increases scan quality, but it
takes longer scanning time.
Scan Range: Scan range enables to set up vertical and horizontal scan area in
degrees.
Selection of sensors: It is possible to use an inclinometer, compass and altimeter.
The data of these sensors are always measured and attached to each scan. If the use
of sensors’ data is switched on, it is automatically used to register the scans in TX5
Scene. A temperature sensor is also integrated.
Colour settings: Capturing of coloured scans can be switched on or off. There are
more possibilities to set the device according to the current lighting conditions.
Safety Eye Distance and Advanced Settings enable to change the hardware filter
settings.
Scanning parameters were different for the inner and for the outer part of the
castle. Our conditions of demanded resolution were that the resultant points in point
cloud should not be distant more than 1 cm from each other. Considering this demand
and also the time and battery limitations, it was decided to use the parameters which
can be seen in Table 4. [33] [32]
Area of interest Resolution Quality Point distance
[mm/10m] Scan duration
[mm:ss]
Exterior 1/4 4x 6.136 7:09
Interior of the tower (rooms) 1/5 3x 7.670 2:17
Interior of the tower (stairs) 1/8 3x 12.272 0:54
Table 4 Applied scan parameters without colours
CTU in Prague Measurement of the Castle
45
3.4 Control Measurements
Control measurements were carried out together with the laser measurements
(chapter 3.4.1). The processing of measured data was supplemented with further
control measurements (chapter 3.4.2).
The main goal of control measurement is to appraise the accuracy of the final
3D model of the castle; which means whether the registered point cloud suffers from
any deformations or systematic errors. An accuracy assessment was executed by
comparing the distance between the measured points in terrain and in model.
A numerical result and conclusion of the control measurement are described in detail
in chapter 4.4. Another aim of the measurement was to determine the control points
for transformation into a global model coordinate system S-JTSK.
3.4.1 Control Measurement during the Laser Scanning
Some check points were already surveyed during the laser scanning measurement,
during the process of measuring the geodetic point field. Twenty planar targets were
placed in the castle and also around the wall from the outer side. These targets
represent the only check points from the outer side of the wall. Some of the
targets were used as control points for registration. Ten other selected check points
were measured, which were chosen and placed on the natural edges of stones and
immobile objects. The coordinates of these check points and targets are one of the
results of the thesis of Ing. Lukáš Vosyka [28].
3.4.2 Additional Control Measurement
An additional control measurement was performed on 22th of February 2015 for the
purpose of determination of spatial distances throughout the castle. Spatial distances
are calculated on the basis of the coordinates of check points. Control measurements
were done by me and other students, namely Ing. Alzběta Prokopová and Ing. Petra
Dífková.
Check points were surveyed by a polar method on the basis of the geodetic point
field of the castle. Selected points were placed on immobile and clearly identifiable
locations, for example edges of stones, corners of the windows, an arch of the gate,
edges of the battlements etc.
CTU in Prague Measurement of the Castle
46
To facilitate later identification of points during the data processing, a photographic
documentation was taken. The measured point was indicated by a laser beam of total
station on the acquired photos (Figure 17 (a)).
In total 72 check points were surveyed.
A list of equipment used for scanning:
- Trimble M3 Total Station (DR 5 ''). No. 652352
- wooden tripod
- 2x Topcon prism with rods
- 2x Leica mini Prism with stand
- tape measure
- thermometer, barometer
(a) (b)
Figure 17 Check point (a) Photograph (b) Vertex in point cloud
3.4.3 Computation of the Coordinates of Check Points
Coordinate values of geodetic check points are copied from the thesis of Ing. Lukáš
Vosyka [28]. Computation of coordinates on the basis of the additional control
measurement was calculated in software Groma 8. The check points were labelled
differently: planar targets had numbers 9xxx and other points 9xx.
For the computation of the coordinates of check points any additional distortions,
like a cartographic distortion or correction of altitude, were not taken into
consideration; and the scale factor is equal to one. Since the distances calculated from
coordinates were compared with real distances from the coordinates in the model, the
implementation of distortions is undesirable in this case. The coordinate list of
calculated points is enclosed in Appendix C. And the protocol on calculating is part
of the electronic appendix on DVD in folder “control_measurement”.
CTU in Prague Measurement of the Castle
47
In total 79 check points were surveyed, and 66 of them were identified in a point
cloud. To facilitate the identification of check points a review has been drawn up.
It contains a table with their descriptions and photographs and a ground plan with
their locations. As a base for the review we used a ground plan from photogrammetric
mapping (see chapter 2.4). The ground plan with locations of the points is in
Appendix D. Check points from the additional measurement are marked there
in orange colour and check points from the measurement performed during the laser
scanning are marked there in green colour. The table with the description and
photographs of check points is attached like an electronic appendix saved on university
computer in
c:\_data\Helfenburk_vysledne\Kontrolníměření\3_fotodokumentace\ (made by
Ing. Petra Dífková). The accuracy assessment and the results from the control
measurement are presented in chapter 4.4.
CTU in Prague Data Processing
48
4 Data Processing
Following sections gives description of the data processing. The initial part
(preparation of the data and the registration) was prepared together with the other
students (Ing. Alžběta Prokopová and Ing. Petra Dífková). The registered model of the
castle was divided among us and thereafter every student worked alone (the vector’s
model of the part of the wall was carried out by Ing. Petra Dífková and the
vector’s model of the inner part of the castle by Ing. Alžběta Prokopová). Individual
work on the castle’s tower was realized by Ing. Martin Toušek.
The first step is to convert the data from the scanner into a suitable format and cut
out the erroneous points, then process the registration and create 3D model.
The software Geomagic Studio (its current version is called Geomagic Wrap) and the
software Leica Cyclone were chosen for the process of registration and the software
Geomagic Studio and Screened Poisson Surface Reconstruction was used for modelling
of the mesh (see chapter 5).
Processing of data is very demanding, depending on the performance of the
computer. Therefore, the processing was carried out on a powerful computer that was
lent to us from the Department of Special Geodesy (Table 5). Completion of
registration, transformation of the model into the coordinate system and creation
of the 3D mesh model was carried out through a remote access to the school
computer from the computer workstation of Hochschule für Technik und Wirtschaft
in Dresden during my Erasmus study there.
If not specified otherwise, this chapter is prepared on the basis of the practical
experiences obtained during the data processing and on the basis of information
obtained from Help files of each software.
Operating system Windows 7 Enterprise
Random access memory 64 GB RAM
Processor Intel Core i7-3820 CPU @ 3,60 GHz
Graphics card AMD Radeon HD 7900 Series
Table 5 Computer configuration
CTU in Prague Data Processing
49
4.1 Export of the Measured Data
Data downloaded from the scanner can be opened only in special software called
Trimble Scene, which is delivered with the scanner by the Trimble Company.
This program is intended only for some operations with 3D data like filtering
of points, looking up specific objects in point clouds or colouring of point clouds. For
more advanced data processing it is preferable to export the measured data to some
kind of interoperable format, like PTS or PTX. In our case the data were exported to
PTX. This format allows exporting the coordinates of points X, Y, Z [m] with their
normals, colour information RGB [0,255] and intensity [0,1], which offers an easier
work and a possibility of subsequent visualization. In our case the colour information
was not exported because it was not captured on most of the scans.
Workflow for the export in Trimble Scene:
- Creation and opening of new project.
- Import the scanned data in format FLS directly into the new project.
- Export of data (Import / Export - Export scan points - window with
parameters for export like folder, where we want to save the files, format,
selected part of a scan etc.). Supported formats for export include PTX, PTC,
DXF, XYZ etc.
CTU in Prague Data Processing
50
4.2 Resampling in Geomagic Studio
Point cloud was scanned with a scheduled density according to the configured
parameters (the distance between two adjacent points was 6.1 mm at a distance
of 10 m). Density of the scanned points, however, is dependent on the distance of
scanned objects from the scanner. It means that the closest objects were scanned with
unnecessarily high density, thereby the data volume is bigger than necessary.
Processing of such voluminous data would be highly ineffective. Due to the demands
on the computing operations it would be almost impossible to compute and process
the data in terms of computer technology. Therefore, resampling was made before the
registration, in the software Geomagic Studio using a function called Uniform. This
function allows to resample the data to a specified density. The setting of the “Curvate
Priority” tool of this function means that the places with large curvature retain high
density of points and, at the same time, the places which are straight or only a little bit
curved are reduced. Thanks to these points, we can have sufficiently accurate
modelling from point cloud after the resampling (Figure 18 (a)).
(a) (b)
Figure 18 Dialogue window (a) Uniform sample (b) Batch processing
The resampling run automatically for each of 127 scans using the function Macro
and Batch processing, into which was uploaded the function for reducing a point cloud.
Batch processing facilitates future work on multiple files. The size of the data was
thereby reduced from 157 GB to 25 GB.
CTU in Prague Data Processing
51
Workflow for the resampling:
- Recording of the macro (Tools – Macro – Record). Since this moment every
operation is stored in the file, which is written in Phyton programming
language.
- Resampling process was executed using the function Uniform (Points –
Sample – Uniform). Required spacing was defined to 10 mm and maximum
curvature priority was set up (Figure 18 (a)).
- Points – Shading – Repair Normals – Recompute Normals. It can repair
a normal for each point in point cloud, so all normals are perpendicular
to their surface.
- Bring Macro function to a stop (Tools – Macros – Stop). Python script is
stored in the directory which can be selected in the Geomagic start menu –
Options – General – Directories – Macros – Browse.
- Beginning of the batch processing Tools – Advanced – Batch Processing
(Figure 18 (b)). This operation is divided into three parts. At first, we have to
choose the input directory and a loading method (Open or Import).
Another step is to choose the Actions (Run Macros and Save Files).
And finally, we have to choose the Output Files and the directory in which
the outputs are saved. In our case VTX ASCI format was chosen because it is
suitable for the following step. It contains the coordinates X Y Z and
the normals I J K.
Batch processing of all scans took almost a day.
CTU in Prague Data Processing
52
4.3 Registration in Cyclone
First of all, it is necessary to rewrite the suffix of the scans from VTX to TXT, which is
very simple: in Total Commander the suffix has to be rewritten.
A new database called "hel_palac" was made after a running software Leica
Cyclone. Seventeen scan stations representing only the inner castle had been imported
in these databases. The remaining 74 scan stations, displaying the fortification,
the exterior of the towers and the courtyard, were imported into the second database
called "helfenburg".
The import was carried out by right-clicking on the database and selecting the
Import possibility. The import dialog is showed in Figure 19. In the Options tab there
must be set up the correct value range of colours and intensities so as the imported
point cloud could be properly displayed. Furthermore, the number of columns of the
imported file has to be set up, and the number of lines in heading which must be
skipped. For each imported column it is necessary to specify which type of data format
the column contains (X, Y, Z, etc.). Importing rules are stored as a_xzy_xynz.afr and
applied for each scan.
For each imported point cloud the ScanWorld is created; for more information and
for the structure of the software see my bachelor thesis, pages 26 and 27 [34].
Figure 19 Import file dialog
CTU in Prague Data Processing
53
4.3.1 Identical Points Modelling
Identical points had to be found one by one in all scan positions. These points are
represented by scanned spherical targets which were placed in the course of scanning.
Only the visible part of the surface of a spherical target was captured from every scan
station. A different part of the surface is scanned from each position, which means
that the scanned points are not identical among the standpoints for one particular
target. Therefore, it was necessary to interpose a spherical object among the points
of a target. Sphere diameter was fixed to the value of 200 mm, which is specified
by the producer. That is carried out using a function Edit - Preferences Object - Fit
Diameter.
Modelling of spheres was executed by choosing one point of sphere which has to be
fitted, and then the function Create Object - Region Grow - Sphere was applied. This
function automatically selects the other points of the cloud which represent the target
of the sphere. Using a setting of Region Size it is possible to adjust the area of the
selected sphere; the options of setting are shown in Figure 20.
Figure 20 Application of Region grow sphere
The centre of this sphere was determined as an identical point. The function Create
Object - Insert - Vertex was used to insert the identical point. Each created identical
point was labelled by a function Tools - Registration - Add / Edit Registration Label,
thence each vertex was assigned a unique number (Figure 21). A vertex thus created
CTU in Prague Data Processing
54
and labelled in software Leica Cyclone can be itself recognized as an identical point.
An identical point represented by the centre of one particular target has the same
number in all scans.
(a) (b)
Figure 21 (a) Modelled sphere target (b) Vertex with Registration label
All the modelling was carried out in the window of Model Space. Therefore, it is
necessary to copy the final point cloud into the window of ControlSpace. Identical
points and their labels are inserted automatically into ControlSpace. Since the program
works only with the data saved in ControlSpace during the registration, this step must
not be forgotten.
4.3.2 Workflow and Computation of Registration
Each point cloud of one scan position is captured in a local coordinate system.
Registration is a process of transforming coordinates of all the points in a point cloud
into a common coordinate system. As a result we get a registered model of the
complete scanned object.
The entire process of registration of the castle was the most time consuming part
of data processing. There are two possibilities how to approach the process of
registration (for more details see chapter 1.1.4). The first one is registration by means
of identical points, which we preferred. The second possibility is to compute an ICP
algorithm. The first problem with computing the registration using only the identical
points was the fact that we did not have enough spheres to represent identical points.
Therefore, we could not leave the spheres positioned on one place as connecting
identical points for later use. We also needed to connect the scans from the
measurement of the first weekend with the second weekend. And, moreover, since
CTU in Prague Data Processing
55
the scanner Trimble TX5 is equipped with a dual axis compensator, we assumed that
two identical points should be enough to join two adjacent scan stations. However,
in that case, there are no supernumerary conditions if one of the sphere moves.
We could analyse which sphere moved, but in that case it would not be possible to use
it for the following registration and there would arise the problem of insufficient
number of constraints. Therefore, ICP was added as a third constraint. For all these
reasons we had to combine both types of registration. The entire process was more
complicated than we expected.
Due to the considerably high number of processed scan positions, the registration
process was divided into three successive sections. In the first block we tried to find
as much scans as possible which could be joined via identical points into several blocks.
Then the adjacent scans were added using spheres in combination with ICP algorithm.
The last step was to interconnect these sections of scans. The advantage of this
approach is that it is easier than creating one big complex registration. A little
drawback is that registrations which have been already frozen cannot be adjusted with
the new scans.
4.3.2.1 General Description of Workflow for Registration
The processing of individual registrations was drawn up in a similar way. At first, the
dialog box of registration was created in the corresponding database by selecting
Create - Registration. All other steps were performed in the window of registration.
- Choice of scans which are transformed in one common coordinate system
using ScanWorld - Add ScanWorld.
- Setting of one selected scan station as a Home ScanWorld, which means that
the other scan positions are transformed into this coordinate system.
The first point cloud in the list is automatically determined as the Home
ScanWorld and this is written in bold.
- Checking of clouds that are marked as levelled. Since the scanner is
equipped with a dual axis compensator, the scans entering in the
computation have to be levelled.
- Looking up for the identical points and their interconnection is performed
automatically by function Constraint - Auto Add Constraints.
CTU in Prague Data Processing
56
- Checking, troubleshooting and potential editing of newly created conditions
is conducted in the tab Constraint List. It is possible also to set the weight
of constraints with which the constraint of identical point enters into
computation. In case of insufficient number of constraints, it is
supplemented with adding of ICP (more detail in chapter 1.1.4.2).
- The computation of registration (by the function Registration - Register).
- Checking of the numerical results of registration (Registration – Show
diagnostics). There is displayed the attained deviation of each scan and
the mean absolute error for enabled constraints together with the
transformation key for each scan.
- Checking of merged control clouds. By using the function Registration -
Interim Result View it is possible to preview the interim result and a visual
control can be done. For more details about the checking see chapter
4.3.2.5.
- Completion of registration. By using the function Registration - Create Scan
World / Freeze Registration the new ModelSpace containing registered
clouds is created. Scans are not combined into one cloud, but still they
represented by individual clouds.
- Saving the final report from registration (Registration - Show Diagnostics).
4.3.2.2 First Section of Registered Scans
Ten groups, which were joined mostly by identical spheres, were created in this
section. The registration groups were saved in databases "Helfenburg" and
"helf_palac". Processing of registration in Leica Cyclone software allows to weight set
for the constraints. In our case the weight was set to value 1 for the constraints
containing the identical spheres.
Table of the registered clouds from this section and of attained mean absolute
errors are in the Table 6.
CTU in Prague Data Processing
57
Registration Registered point clouds
Mean absolute error [m] Mark Name
A reg1-10 1 - 10 0.0023
B reg13-17 13 - 17 0.0020
C reg205-210 205 - 210 0.0032
D reg214-217 214, 215, 217 0.0024
E reg219-223 219 -223 0.0012
F reg251-254 251 - 254 0.0028
G reg236-238 236 - 238 0.0026
H reg256-258 256 - 258 0.0013
I reg229-233 229 -233 0.0083
J reg269-274 269 - 274 0.0118
Table 6 Overview of point clouds in the first section of registration
In this section there were two significant problems with registrations marked in my
theses as letter "I" and "J". Constraints constituted only on the basis of available
identical spheres were not sufficient; therefore, it was necessary to increase the
amount of constraints of registration with other naturally signalized identical points.
The first problem was with registration "I", which consists of the clouds in the east
part of the walls and the surroundings of the tower. It has been found during the
registration process that the distance between two adjacent scans is up to 6 cm
at some places. Therefore, little planes with labels were added into the point cloud and
these points were used as an identical point as well. After that, the check distance
in plane section dropped to the value around 1.5 cm between registered clouds.
This problem was not solved more elaborately because the complete model of the
tower was processed independently by Ing. Martin Toušek as part of his thesis.
Deviation, however, could be caused due to the fact that the tower was scanned from
many standpoints under sharp angles and from a long distance.
Registration "J" consists of the scans located in north part of the walls. In the first
trial registration we found out that the registered scans did not fit. Erroneous part
of point clouds was interpreted as incorrectly scanned due to a sharp angle of
incidence. The problem was solved by trimming each incorrect part of the clouds.
Points in distance bigger than 10 metres from the scanner position were deleted.
There was probably also a slight movement of the sphere number 320 during the
measurement, as it was only laid on the ground in the leaves. This target was removed
and not included in the registration.
CTU in Prague Data Processing
58
Since the ICP algorithm is used in the subsequent process, the trimming
of vegetation in point cloud was performed, and therefore the ICP algorithm is
computed only from the real estate of the castle. The trimming was carried out for all
scans and all moving objects like aforementioned vegetation and also people,
equipment etc. Such objects could be captured in different position on different scan,
which is unacceptable. Unnecessary points were selected by the function Polygonal
Fence Mode and deleted by the function Fence – Delete inside.
The results in reports (the mean absolute error of targets and transformational
matrix with translation and rotation) were stored with 10-digits precision because
it would be necessary for future work with the software Geomagic. The reports are
attached in DVD in folder “reports_cyclone_registration”.
4.3.2.3 Second Section of Registered Scans
Eight groups of registration were created and named "doregistraceXX" in the
second section. Variable XX says to which groups from the first section it is related.
These "doregistraceXX" groups were made by joining the remnant clouds to existing
groups from the first section of registration. The constraints are a combination
of overlap (ICP algorithm) and identical points. A new working database named
"helf_vse" was established for these registrations.
Before the beginning of the registration process a schedule with the overview was
established for the matching of still unused scans to the already registered groups.
Constraints using overlaps were used in these registrations most frequently. The
weight for different types of constraints was adopted in two ways. When constraints
were only from an overlap, the value of weight is 1. In the case that the constraints
were a combination of overlaps and identical points, the value of weight for identical
points was 1 and for overlap 0.7. Overview of the registered clouds and attained mean
absolute errors is given in the Table 7.
CTU in Prague Data Processing
59
Second section Registered point clouds and scans
Mean absolute error [m] Mark Name
K doregistrace13-17 B, 201, 202 0.0021
L doregistrace205-210 C, 204, 211 - 213, 218 0.0031
M doregistrace214-217 D, 216 0.0003
N doregistrace219-223 E, 225 - 227, 203 0.0024
O doregistrace251-254 F, 224, 239 - 244, 255 0.0022
P doregistrace236-238 G, 235, 247 - 250, 266 - 268 0.0017
Q doregistrace256-258 H, 245, 246, 261 - 265, 259, 260 0.0028
R doregistrace229-233 I, 228, 234 0.0036
Table 7 Overview of point clouds and scans in the second section of registration
The constraints consisting of an overlap are possible to be input by the function
Cloud Constraint - Cloud Constraints Wizard in the registration window. A window with
table is opened when the function is activated. It is necessary to choose a pair of scans
between which the overlap will be made. Then, the two working windows containing
the clouds are displayed. Among these clouds we must choose at least two pairs
of points representing the approximate coordinates for the start of the computation.
Selection of a point is confirmed and the possibility Preview is applied. If we are
satisfied with the preview of the registration, it is confirmed and saved by the function
Constraint. A new constraint from overlap is displayed in the Constraint list. There
is possibility of optimizing this constraint in the right-click menu (Cloud Constraint -
Optimize Cloud Alignment).
Figure 22 Registration window with Cloud constraints wizard
CTU in Prague Data Processing
60
4.3.2.4 Final Section of Registered Scans
At this stage of registration it is necessary to merge all groups into one unit.
This registration was saved in a database named "helf_vse".
Weight of constraint for identical points and overlaps was chosen so as to be
congruent with the registrations in the second block. An overview of the registered
cloud and attained mean absolute errors are given in the Table 8.
Final section Registered point clouds
Mean absolute error [m] Mark Name
S registrace_fin A, J, K, L, M, N, O, P, Q, R 0.0107
Table 8 Overview of point clouds in the final section of registration
4.3.2.5 Checking of the Registered Blocks of Scans
After finishing each group of registration the visual checkout was always done in
a window of ModelSpace that is created from Registration window by the function
Registration - Interim Result View. The checkout consisted of creating a cross and plane
section of the point cloud (using either Limit Box, or a new Copy Fenced's ModelSpace).
In each section it was checked whether the point clouds are joined correctly and
without gaps between them. A mutual distance between two mismatch point clouds
can be measured by picking up two corresponding points from each scan and apply the
function Distance - Point to point from Menu - Tools - Measure.
In Figure 23 there can be seen a section with two mismatch point clouds and
the distance between them.
Figure 23 Mismatched point clouds
CTU in Prague Data Processing
61
4.3.3 Transformation of a Point Cloud to S-JTSK
The registered point clouds were transformed into the coordinate system S-JTSK
and the altitude system Bpv using control points represented by vertexes in the
centres of planar targets. The rigid 3D transformation was used in the software Leica
Cyclone by way of creation a new registration with the coordinates of planar targets
ascertained from the measurement of Lukáš Vosyka (it is described in detail in his
theses [28]). The rigid 3D transformation was chosen in our case to preserve real
distances in a point cloud. Distortions of a map projection in the castle area do not
exceed 3 mm which is less than the accuracy of the point cloud and therefore the map
projection can be ignored.
Furthermore, two forgotten point clouds (number 11 and 12) were added to the
registered point clouds.
The software Leica Cyclone works with the classical mathematical coordinate
system. But the standard S-JTSK axes have a different direction in comparison with the
classical system. In S-JTSK X axis goes to the south and Y axis goes to the west.
Therefore, the transformation equations between these two systems had to be applied
as follows: YS-JTSK/EN = -XS-JTSK, XS-JTSK/EN = -YS-JTSK, ZS-JTSK/EN = ZS-JTSK. Using this system the
whole measurement is moved to the third quadrant. The orientation does not change.
Workflow for the transformation:
- Creation of a new ScanWorld named "SJTSK_Lukas" in database "helf_final"
Import of coordinates of 18 control points into a ModelSpace named
"registrace_fin". The number markers of these points were increased
by 9000 so as not to be mixed up with identical sphere targets.
- Creation of registration named "SJTSK" with input scans "SJTSK_Lukas"
as a Home ScanWorld and "registrace_fin".
- Following the process mentioned above in chapter 4.3.2.1.
The standard deviation of the transformation was 13 mm. Considering the error in
identifying the centres of planar targets in a point cloud, this is a good result.
The attained deviations at all the scans, the mean absolute error and the
transformation matrix are in the protocol of registration in electronical appendix
in folder “reports_cyclone_registration”.
CTU in Prague Data Processing
62
Point S-JTSK Leica Cyclone (S-JTSK/EN)
Y [m] X [m] Z [m] x [m] y [m] z [m]
9001 737990.478 988443.537 327.681 -737990.478 -988443.537 327.681
9003 737990.084 988438.428 325.873 -737990.084 -988438.428 325.873
9101 737985.791 988443.514 320.747 -737985.791 -988443.514 320.747
9102 737992.972 988429.258 319.284 -737992.972 -988429.258 319.284
9103 737976.591 988437.188 318.461 -737976.591 -988437.188 318.461
9106 737983.646 988467.783 316.868 -737983.646 -988467.783 316.868
9107 737989.635 988482.621 314.607 -737989.635 -988482.621 314.607
9108 737967.281 988482.76 314.709 -737967.281 -988482.760 314.709
9111 737952.549 988492.464 319.156 -737952.549 -988492.464 319.156
9112 737982.424 988490.538 310.867 -737982.424 -988490.538 310.867
9114 737998.989 988468.007 310.685 -737998.989 -988468.007 310.685
9116 738014.235 988436.192 310.159 -738014.235 -988436.192 310.159
9117 738016.279 988432.669 309.872 -738016.279 -988432.669 309.872
9118 738020.213 988418.384 309.388 -738020.213 -988418.384 309.388
9120 738020.863 988407.093 309.055 -738020.863 -988407.093 309.055
9122 738015.617 988404.628 308.568 -738015.617 -988404.628 308.568
9125 738000.992 988405.432 306.069 -738000.992 -988405.432 306.069
9127 737936.849 988450.977 309.748 -737936.849 -988450.977 309.748
Table 9 Coordinates of control points used for the transformation
Since some software have problems with high coordinates used in S-JTSK
a reduced system had to be used. Shift of coordinates was done using formulas:
Yreduced = YS-JTSK/EN + 988 000, Xreduced = XS-JTSK/EN + 730 000, Zreduced = ZS-JTSK/EN. The shift
could have been performed in this simple way because there is no rotation or scale
involved. The transformation procedure was done in Leica Cyclone as another section
of the registration, in the same way as in chapter 4.3.3, only with the different input
coordinates of control points which were reduced according to the above mentioned
formula (see Table 10).
Point x [m] y [m] z [m] Point x [m] y [m] z [m]
9001 -7990.478 -443.537 327.681 9112 -7982.424 -490.538 310.867
9003 -7990.084 -438.428 325.873 9114 -7998.989 -468.007 310.685
9101 -7985.791 -443.514 320.747 9116 -8014.235 -436.192 310.159
9102 -7992.972 -429.258 319.284 9117 -8016.279 -432.669 309.872
9103 -7976.591 -437.188 318.461 9118 -8020.213 -418.384 309.388
9106 -7983.646 -467.783 316.868 9120 -8020.863 -407.093 309.055
9107 -7989.635 -482.621 314.607 9122 -8015.617 -404.628 308.568
9108 -7967.281 -482.760 314.709 9125 -8000.992 -405.432 306.069
9111 -7952.549 -492.464 319.156 9127 -7936.849 -450.977 309.748
Table 10 Reduced coordinates of control points
CTU in Prague Data Processing
63
4.4 Accuracy Assessment at Registered Model Based on the
Control Measurement
Assessment of the accuracy of the registered 3D model (point cloud) was performed
by comparing the check distances obtained from the check coordinates measured
in field and from model coordinates of these check points read in the final model
(point clouds). Detailed information about the way of measurement is described
in chapter 3.4. Particulars of the method of computation are described in detail in the
theses of Ing. Alžběta Prokopová [35].
4.4.1 Coordinates of Check Points
Check points surveyed in field are available from two measurement periods (two
types of check points). The coordinates of centres of planar targets, which were also
used for the transformation to S-JTSK as control points, are taken from the
measurement of Ing. Lukáš Vosyka [28]. In total, 18 targets were placed and measured.
Figure 24 Planar target in a point cloud
In the second period there were measured the points of edges of stones and
immobile objects. Since the castle is built from sandstone it turned up to be quite
difficult to find sharp edges and easily identifiable object. For this reason these points
are not so accurate. Expected accuracy could be around 1 cm or worse. In total
73 points were measured in the whole castle, 66 of them were identified in point
clouds. The numbers of these points were raised by an invariable 900 not to be mixed
up with identical sphere targets.
The software Cyclone was used for finding check points in a registered point cloud.
Check points have model coordinates and are saved in a separate layer. The model
coordinates of planar targets are shown in Table 11.
CTU in Prague Data Processing
64
Point x [m] y [m] z [m] Point x [m] y [m] z [m]
9001 -5.225 -4.949 267.58 9112 13.444 -48.876 250.763
9003 -5.989 0.094 265.782 9114 -7.869 -30.732 250.569
9101 -0.645 -3.874 260.649 9116 -30.010 -3.279 250.060
9102 -10.913 8.353 259.185 9117 -32.814 -0.313 249.770
9103 6.857 4.402 258.361 9118 -39.923 12.684 249.290
9106 7.033 -26.992 256.758 9120 -43.168 23.527 248.951
9107 4.605 -42.829 254.501 9122 -38.620 27.136 248.464
9108 26.375 -37.811 254.622 9125 -24.191 29.712 245.977
9111 42.963 -43.878 259.053 9127 48.700 0.087 249.651
Table 11 Model coordinates of control points
4.4.2 Control Measurement Result
Coordinates with numbered points were exported into TXT file and further
processed in Matlab to obtain 3D distances. The result is a matrix of distances among
all measured points and second matrix with distances from model (point clouds).
The difference between the real and the model distances is obtained by subtracting
those two matrixes.
Calculated distance differences were compared with permissible deviation of
distance differences that were specified on the basis of the accuracy analysis
of the measurement [35]. The values of maximum permissible difference of distances
are as follows:
For planar targets: δΔd1 = 2.0 cm
For planar targets and stone edges: δΔd2 = 3.2 cm
For stone edges: δΔd3 = 4 cm
Since the table with the results is very large it is enclosed on DVD in the folder
“control_measurement” (made by Ing. Alžběta Prokopová). Based on the gained
computed distance differences and their comparison with the maximum permissible
difference we can conclude that the final model of the castle is not deformed.
The longest independent distances in longitudinal, transverse and Z axis direction
were chosen to evaluate the model accuracy. Only several chosen distances are shown
in the following tables (Table 12 (a), (b), (c)). Then, the standard deviation of the
sample was calculated from distance differences between selected points by
the formula σ� = √ (ΣΔ�2/�), where � is a distance difference and � is a number of
distances. The value of standard deviation is 1.5 cm. [10] [35] [41]
CTU in Prague Data Processing
65
Point A Point B d [m] Δd [m]
9112 9118 81.485 -0.021
9111 9125 100.479 -0.018
9108 9120 92.902 -0.012
Point A Point B d [m] Δd [m]
9107 9103 47.442 -0.017
9116 9102 24.151 0.004
9118 9125 23.414 -0.004
(a) (b)
Point A Point B d [m] Δd [m] |ΔZ| [m]
9101 9001 8.370 -0.007 6.934
9111 9125 100.479 -0.018 8.713
9108 9120 92.902 -0.012 17.305
(c)
Table 12 Distances in (a) longitudinal direction, (b) transverse direction, (c) Z axis direction
CTU in Prague Data Processing
66
4.5 Completion of Registration and Preparation of Model
in Geomagic
The Leica Cyclone software was chosen for the processing of registration for several
reasons. One of the main reasons is the possibility of setting the weight for each
constraint and furthermore, although time-consuming, clearly arranged entering and
combining of constraints for using the ICP algorithm. One of the drawbacks, however,
is the fact that we are not able to export the point clouds from Leica Cyclone in any
format with information about the values of the normals. For the processing and
computation of the mesh, which is the aim of this work, the normals are necessary.
This problem can be solved by procedure of extracting the TFM file from the
registration reports by using the program Cyclone2gm.exe developed by Ing. Branislav
Koska, Phd. The Transcription Factor Matrix TFM file can be applied on the raw scans
(after resampling from Geomagic Studio, see chapter 4.2) in the software Geomagic
Studio.
4.5.1 Workflow in software Geomagic Studio
From the registration window in the Leica Cyclone software the reports of each
registration group of scans are stored with 10-digits precision (Registration – View
Interim Result-Show Diagnostics…) and named in accordance with column the “name”
from Table 6 and Table 7.
Then the Cyclone2gm.exe program is applied. That runs from the command line by
the statement for example "Cyclone2gm.exe name_of_registration_group.txt" in the
directory where the reports are saved. The program extracted information about
translation and rotation of each scan from the report and calculates and creates a file
in TFM format “name_of_each_scan_from_registration_group.tfm” for each scan. This
procedure is applied successively to all reports. TFM files are attached on the DVD that
is appendant to this thesis.
Another step is sequential import of all groups of scans (exported from Geomagic
Studio in format VTX, see chapter 4.2) into Geomagic Studio working window. If the
directory where the scans are saved contains also the TFM files named identically with
the scans, the point clouds are automatically transformed after that according to the
information in TFM files (saved in electronical appendix in folder “tfm_matrix”).
CTU in Prague Data Processing
67
This method represents the easiest way. There is also possibility to import a cloud
which is not saved in the same directory with the identically named TFM file.
Consequently, it is then necessary to make the transformation manually by the
function Capture - Manage Transform and load the TFM file.
If the group of scans is imported and transformed in the right way, the function
Combine is applied. It is also possible to use the function Merge which offers more
settings to control the overall process and provides some additional post-processing
features such as global registration; however, it is not necessary in our case because
we need to maintain the point cloud as well as the position of points without any
changes. Due to the extensive amount of the data the function Uniform (with setting
of 1 cm) is used and the data are exported to VTX format. The same procedure
is applied for each group in the first section of registration and accordingly to the
above procedure until the final transformation of point clouds into the coordinate
system S-JTSK. The final point cloud is resampled again at 1 cm and is saved in WRP
format (the native format of Geomagic Studio). Since we used the raw uncleaned
scans, the point cloud had to be cleared of all outlying points, vegetation, movable and
other undesirable objects.
The final model of the castle is represented by one point cloud that captured the
entire area of our interest. The model consists of 172 million of points and it is placed
in a reduced coordinate system S-JTSK and Bpv.
Finally, the model is cut into four parts so that the huge amount of data is
acceptable for further work. It is necessary also to rewrite the suffix of the scans from
VTX to TXT and two rows at the beginning of the file has to be deleted. The overview of
the split parts of the point cloud is displayed in Table 13.
Number of points
[million] Size of file
[GB]
part1.vtx 45.5 2.7
part2.vtx 41.3 2.5
walls.vtx 64.4 3.9
walls_tower.vtx 21.0 1.3
Table 13 Overview of parts of the point cloud
CTU in Prague 3D Model
68
5 3D Model
The basic principle of creation of the triangular mesh is described above in chapter
1.1.5.2. The Poisson Surface Reconstruction was chosen for the point cloud of the
castle as one of the best solutions we know for the surface reconstruction.
5.1 Poisson Surface Reconstruction
Poisson surface reconstruction creates watertight surfaces from point clouds
acquired with 3D scanners, this technique is resistant to noisy data. That is caused by
the fact that the Poisson formulation considers all the points at once, without resorting
to spatial partitioning or blending.
Unlike radial basis function schemes, the Poisson approach allows a hierarchy of
locally supported basis functions, and therefore the solution is reduced to a well-
conditioned sparse linear system. This approach enables faster, higher-quality surface
reconstructions. [36] [37]
The solution of surface reconstruction using the Poisson approach was developed
by Michal Kazhdan from Johns Hopkins University and by Hugues Hoppe from
Microsoft Research. The program is available on the website [38] and the easiest way
is to run it from the command line. Screened Poisson Surface Reconstruction consists
of two programs – from PoissonRecon, which generates the mesh, and from
SurfaceTrimmer. The computation of mesh sets off from the command line in the corresponding
directory, where the input point cloud and EXE file with Poisson Recon or Surface
Trimmer must also occur.
The most important input parameters and their clarification are adduced
in Appendix E [38].
CTU in Prague 3D Model
69
The examples of the commands:
PoissonRecon.x64 --in cast5.vtx --out part1_d11_pw8.ply --depth 11 --pointWeight 8 --
polygonMesh --density ––verbose
SurfaceTrimmer.x64 --in part1_d11_pw8.ply --out part1_d11_pw8_aR0_polM.ply --
trim 8 --aRatio 0 –polygonMesh
Figure 25 Triangular mesh from Poisson Surface reconstruction with many holes
Triangular mesh was created for each part of the point cloud from Table 13.
An overview of bounding dimension (B), the selected parameter of depth (d) for
computation of mesh, the approximate length of edge in the final mesh (l), the used
trimming value (t) and the number of triangles of each part after running the process
in PoissonRecon (nP) and in SurfaceTrimmer (nT) are in Table 14 and the computation
reports are available in electronical appendix in folder “reports_poisson”.
B [m] d l [cm] nP [million] t nT [million]
part1.ply 51 x 48 11 2.5 14.65 9 14.52
part2.ply 51 x 49 11 2.5 14.54 9 14.38
walls.ply 106 x 107 12 2.6 26.75 10 26.30
walls_tower.ply 31 x 42 11 2.1 13.16 9 13.00
Table 14 Information about meshes
CTU in Prague 3D Model
70
5.2 Completion of the Triangular Mesh
Since the created mesh models contain many noise mistakes, it is necessary to use
some editing operations. First of all, the function Relax was applied in software
Geomagic Studio 2012. This command smooths the polygon mesh by minimizing
crease angles between individual triangles. In menu Polygons – Smooth – Relax the
parameters like scale of smoothness, level strength and curved priority must be set
(in our case it was 4, 2, 8 in the same order). In the mesh there also exist a lot of
isolated triangles. It would be very time-consuming to delete one by one; therefore,
the components of polygons whose area is less than or equal to the elected
percentage of the whole object (for example 50%) are automatically selected using the
function Select – Data – Select By – Area.
The most difficult stage of mesh optimization is filling up holes or, as the case may
be, the shape modelling of a missing part. When using the function Polygons – Fill
Holes – Fill Single there are three possibilities of the filling technique. The Curvature
possibility specifies that the new mesh that fills up the selected holes has to match
the curvature of the surrounding mesh. The Tangent possibility specifies that the new
mesh that fills up the selected holes also has to match the curvature of the
surrounding mesh, but with more tapering than Curvature. And the Flat possibility
specifies that the new mesh that fills up the selected holes is generally flat. In Figure 26
we can see the differences among them. For the inner part of the castle the
combinations of the filling technique were applied.
(a) (b)
(c) (d)
Figure 26 Different types of the filling technique (a) Hole (b) Flat (c) Tangent (d) Curvate
CTU in Prague 3D Model
71
In the case of larger holes the entire filling up of the holes in one step would be
insufficient because of a higher danger of distorting the reality. This danger is present
in every case of filling up holes; nevertheless, to achieve the waterproof triangular
mesh necessary for the potential 3D print it has to be performed. Using the tool Bridge
implemented in the function Fill Single a bridge of mesh is built across a hole. So the
huge complex hole is divided into smaller ones that can be filled up more correctly.
(a) (b)
Figure 27 Filling up of a large hole (a) Bridges (b) Completed filling
This tool was applied for combining of part1.ply and part2.ply together. A little gap
was created between them and thereafter the “Bridges” of mesh, placed depending
on the surrounding terrain situation, were created and the gaps were consecutively
filled up using the Flat possibility.
Figure 28 Joining of two parts of mesh
Since the number of polygons is too high for correctly displaying the model in most
of software, reducing of this high number is necessary. Polygons could be decimated
in the software Geomagic Studio, but the output and the final quality is better in
CTU in Prague 3D Model
72
software Agisoft PhotoScan. A simple procedure was performed there for the purpose
of mesh decimation.
Workflow of mesh decimation in Agisoft PhotoScan:
- Importing the mesh Tools – Import-Mesh and the choice of possible shift
of coordinates.
- Decimation process (Tools – Mesh – Decimate) and the parameter “Target
face count”, which means how many triangles the new model contains.
In our case, the model was reduced to 20 percent of its original size. For the
placement into Sketchfab it had to be reduced to the size of 50 MB
(see chapter 6).
- Export of the model in format PLY (File – Export Model – Export
OBJ/FBX/KMZ…).
As a matter of interest the analysis of deviation of a point cloud from a triangular
mesh was done in the software Geomagic Studio using the tool Analysis – Compare -
Deviation. The result is shown in Figure 29. The deviations between the point cloud
and the created triangular mesh are very low. This finding is consistent with the
expected conclusion that there was no reason for prediction of high deviation.
Figure 29 3D Deviation analysis (top view)
CTU in Prague Results
73
6 Results
The main purpose of the thesis was to create a 3D model of the castle Helfenburk
near Úštěk. It could be divided into two main results: a point cloud and a triangular
mesh.
To capture the entire complex of the castle 91 scan positions were measured. After
completing the registration process the point cloud was further transformed by rigid
3D transformation into S-JTSK and Bpv using 18 planar targets as control points. The
standard deviation of the transformation into S-JTSK was 1.4 cm. The registration
process was checked by a control measurement and no systematic distortion was
found. The final model coordinates were reduced so as they could be displayed
in a wide range of CAD software and other programs. The reduction formulas
Yreduced = YSJTSK/EN + 988 000, Xreduced = XS-JTSK/EN + 730 000, Zreduced = ZS-JTSK/EN were used.
The triangular mesh was created on the basis of the final point cloud. An overview
of the created parts is shown in Table 15.
Trimmed mesh [million triangles]
Size of file [MB]
Filled mesh [million triangles] Size of file
[MB]
part1 14.5 228 14.5 265
part2 14.5 226 14.6 264
inner_palace 29.1 530 29.0 530
Filled decimated mesh [million triangles]
Size of file [MB]
Decimated mesh for Sketchfab [million triangles]
Size of file [MB]
part1 3.0 71.5 1.8 42.9
part2 3.0 71.5 1.9 46.5
inner_palace 5.9 140 1.9 45.2
Table 15 Overview of created PLY files of modelled parts
The final results are saved in external appendices in the university workstation as
TXT files for the point cloud and PLY files for the triangular mesh.
CTU in Prague Results
74
All files are saved in coordinate system reduced S-JTSK and height system Bpv. It
could be opened in many software, for example an open source software like
CloudCompare3 or MeshLab4 where each model can be rotated, zoomed or moved.
The files are also attached on the DVD that is appendant to this thesis (in folder
“meshes”).
The final results can be seen in the universal online 3D viewer Sketchfab. Due to
the Sketchfab rule that the maximum file size of freely published models is 50 MB
the models had to be reduced. It can be found at the following links:
inner_part https://skfb.ly/JQPU
part2 https://skfb.ly/JQPt
part1 https://skfb.ly/JQPu
Figure 30 Final point cloud
Figure 31 Final triangular mesh of inner palace
More pictures of the final triangular mesh are shown in Appendix F.
3 http://www.danielgm.net/cc/ 4http://meshlab.sourceforge.net/
CTU in Prague Discussion
75
7 Discussion
Due to its many obvious advantages the laser scanning was chosen as a suitable
technique for detailed documentation of the castle Helfenburk near Úštěk. The main
advantage of scanning is that it can capture large numbers of surface points very
quickly and directly. However, acquiring data and processing it in order to obtain
a complete 3D model is a very complex work which includes, for instance, the time
managing of measurement and data processing, or the necessity of learning new
technologies, which can cause some troubles.
During the measurement in field only six sphere targets were available, so it was
necessary to constantly reconsider their location. Nevertheless, it was not sufficient
and in following registration procedure the other constraints (overlap of scans using
ICP algorithm) had to be added. It prolonged the already long time of data processing.
Another factor that made the data processing so time consuming was the few year old
software used for the work with data. For example, the new version of Leica Cyclone
automatically recognizes and models the sphere targets; we did this part of work
manually. And finally, the long duration of the data processing was also influenced by
the management of the cooperation of three students. How can be seen above, the
process and the time management were highly time consuming. After this experience
it can be said that a skilled person should be able to manage the data processing more
effectively and in a considerably shorter time.
One of the resulting drawbacks of this thesis could be the fact that the laser
scanning was carried out without capturing the true colour information. It was due to
the necessity of saving the battery (because there was no possibility of charging) and
also of saving time. For the purpose of an archaeological research and of gaining
knowledge of the geometrical shape of the castle it is not so important. Nevertheless,
for other possible applications such as visualization it would be recommendable to
reconsider it despite the bad conditions in the castle.
Now the question arises if the other measurement techniques would not be
a better solution. Traditional geodetic methods like tachymetric surveying can be ruled
out directly because of too cragged and hardly accessible terrain. Another option is
photogrammetry. The terrestrial analogical and analytical photogrammetry is already
CTU in Prague Discussion
76
past its prime, the aerial photogrammetry from aeroplane would be absolutely
insufficient and expensive. However, a brand-new type of gaining pictures for
photogrammetry is available now, a photogrammetry from UAV which has become
very popular. Thanks to the rapid development of UAV, a computer technology and
software that are able to process huge amount of data and to count the model more
precisely, this technique appears to be suitable for processing of a detailed
documentation of the cultural heritage as well.
In my opinion the best solution is to combine the aerial photogrammetry from UAVs
with the laser scanning. Laser scanners provide data which can be transformed to
a highly accurate reconstruction of the surface. It allows us to keep and present the
geometry of the object very precisely. And photogrammetry introduces the possibility
of keeping the realistic colours of the objects, acquiring the data from inaccessible
places like roof, chimney, dormer window etc. and so it highly improves the final
products.
This area offers a lot of opportunities for further work and possibly for another
master's thesis – measurement of the castle by photogrammetry from UAV and its
comparison with the laser scanning method.
CTU in Prague Conclusions
77
8 Conclusions
The aim of this master's thesis was the measurement and documentation of castle
Helfenburk near Úštěk. First of all a geodetic point field was built so as the final
documentation could have been transformed into coordinate system S-JTSK and height
system Bpv. The entire complex of the castle was scanned by using the scanner
Trimble TX5. To capture the entire complex of the castle 91 scan positions were done.
Proportions of the point cloud were compared with points obtained from the control
measurement. No systematic distortion was found.
The registration of point clouds and the modelling procedure were realized by
a combination of several software packages and the workflow consisted of many
exporting and importing operations with different tools. Major part of the registration
work was carried out in Geomagic Studio and Leica Cyclone, the reconstruction of the
castle surface in PoissonRecon and completion of modelling again in Geomagic Studio.
Final outputs of the thesis are the point cloud of the entire castle and the triangular
meshes of the inner palace. These results will be handed over to the citizen association
"Hrádek".
It can be concluded that the process of measurement and data processing, if made
by an experienced person, may not be so time-consuming while maintaining the same
or higher level of accuracy.
CTU in Prague Reference List
78
Reference List
[1] HUBER, Daniel, et al. Using Laser Scanners for Modeling and Analysis in
Architecture, Engineering, and Construction. Carbegie Mellon University, The
Robotics Institue. [Online] 2010. [Cited: 15 07 2015.]
http://ri.cmu.edu/pub_files/2010/3/Huber%202010%20-
%20Using%20Laser%20Scanners%20for%20Modeling%20and%20Analysis%20in
%20Architecture,%20Engineering,%20and%20Construction.pdf.
[2] VOSSELMAN, George and MAAS, Hans-Gerd. Airborne and Terrestrial Laser
Scanning. Dunbeath, Scotland : Whittles Publishing, 2010. 978-190-4445-87-6.
[3] QUINTERO, M.S., et al. 3D RiskMapping, Theory and practice on Terrestrial Laser
Scanning. KU Leuven Nederlands, European Leonardo Da Vinci programme.
[Online] 2008. [Cited: 13 07 2015.]
https://lirias.kuleuven.be/bitstream/123456789/201130/2/leonardo_tutorial_fin
al_vers5_english.pdf.
[4] ŠTRONER, Martin, et al. 3D skenovací systémy. Praha : České vysoké učení
technické, 2013. 978-80-01-05371-3.
[5] FRÖHLICH, C. and METTENLEITER, M. Terrestrial Laser Scanning - New
Perspectives in 3D Surveying. Institute of Computer Graphics and Algorithms.
[Online] [Cited: 12 07 2015.]
http://oldwww.prip.tuwien.ac.at/cvch07/download/download/lectures/FROEHLI
CH.pdf.
[6] BOEHLER, W., HEIZ, G. and MARBS, A. The Potencitial of Non-contact Close range
Laser Scanners for Cultural Heritage Recording. Institute for Spatial Information
and Surveying Technology, FH Mainz. [Online] 2011. [Cited: 13 07 2015.]
[7] SETKOWICZ, Joanna Alina. Evaluation of algorithms and tools for 3D modeling of
laser scanning data. Norwegian University of Science and Technology,
Department of Civil and Transport Engineering. [Online] 2014. [Cited: 13 07
2015.] http://brage.bibsys.no/xmlui/handle/11250/233080.
CTU in Prague Reference List
79
[8] Bianco, G., et al. A Comparison Between Active ans Passive techniques for
Underwater 3D Applications. University of Calabria, Department of Mechanical
Engineering, Rende - Italy. [Online] 2011. [Cited: 13 07 2015.] http://www.int-
arch-photogramm-remote-sens-spatial-inf-sci.net/XXXVIII-5-
W16/357/2011/isprsarchives-XXXVIII-5-W16-357-2011.pdf.
[9] O’Day, Eugene. 3D Laser Scanning: Different Type of Scanners. Ideate, Inc.
[Online] 2013. [Cited: 16 07 2015.]
http://ideatesolutions.blogspot.cz/2013/07/3d-laser-scanning-different-type-
of.html.
[10] ŠTRONER, Martin and POSPÍŠIL, Jiří. Moderní geodetické technologie a přístroje
pro laserové skenování. Stavební obzor. 2005, Vol. 14, 8.
[11] PAYNE, Angie. Laser Scanning for Archaeology: Introduction to the Laser Scanning
Guide. Archaeology Data Service / Digital Antiquity. [Online] 2009. [Cited: 14 07
2015.]
http://guides.archaeologydataservice.ac.uk/g2gp/LaserScan_1-2.
[12] ŠTRONER, Martin and POSPÍŠIL, Jiří. Terestrické skenovací systémy. Praha : České
vysoké učení technické, katedra speciální geodézie, 2008. 978-80-01-04141-3.
[13] ANGELOPOULOU, Elli and WRIGHT, John. Laser Scanner Technology. University of
Pennsylvania, Department of Computer & Information Science. [Online] 1999.
[Cited: 13 07 2015.]
http://repository.upenn.edu/cgi/viewcontent.cgi?article=1083&context=cis_rep
orts.
[14] BOEHLER, Wolfgang and MARBS, Andreas. Investigating Laser Scanner Accuracy.
Institute for Spatial Information and Surveying Technology, University of Applied
Sciences, Mainz, Germany. [Online] 2003. [Cited: 17 07 2015.]
http://dev.cyark.org/temp/i3mainzresults300305.pdf.
[15] AKCA, Devrim. Full Automatic Registration of Lase Scanner. Institute of Geodesy
and Photogrammetry, Swiss Federal Institute of Technology, ETH Hoenggerberg.
[Online] [Cited: 23 07 2015.]
http://e-collection.library.ethz.ch/eserv/eth:26938/eth-26938-01.pdf.
CTU in Prague Reference List
80
[16] PREVITALI, M., et al. Laser Scan Registration Using Planar Features. The
International Archives of the Photogrammetry, Remote Sensing and Spatial
Information Sciences. [Online] 2014. [Cited: 23 07 2015.]
[17] AXELSSON, Peter. Processing of laser scanner data—algorithms and applications.
ISPRS Journal of Photogrammetry & Remote Sensing. 1999, 54.
[18] WILM, Jakob. Iterative Closest Point. MATLAB Central Contests. [Online] 2013.
[Cited: 24 07 2015.]
http://www.mathworks.com/matlabcentral/fileexchange/27804-iterative-
closest-point.
[19] BERNARDINI, Fausto, et al. The Ball-Pivoting Algorithm for Surface
Reconstruction. IEE Transactions on Visualization and Computer Graphics.
[Online] 1999. [Cited: 27 07 2015.]
http://vgc.poly.edu/~csilva/papers/tvcg99.pdf.
[20] LORENSEN, William and CLINE, Harvey. Marching Cubes: A high resolution 3D
surface construction algorithm. General Electric Company, Corporate Research
and Development, Schenectady, New York. [Online] [Cited: 27 07 2015.]
http://www.eecs.berkeley.edu/~jrs/meshpapers/LorensenCline.pdf
[21] MENCL, Robert. Reconstruction of Surfaces from Unorganized Three-Dimensional
Point Clouds. Universität Dortmund. [Online] 2001. [Cited: 27 07 2015.]
https://archive.org/stream/ReconstructionOfSurfacesFromUnorganizedThree-
dimensionalPointClouds/Surface_Reconstruction__Robert_Mencl_PhD_thesis#p
age/n0/mode/2up.
[22] REMONDINO, Fabio. From Point Cloud to Surface: the Modeling and Visualization
Problem. Institute of Geodesy and Photogrammetry, ETH Zurich. [Online] 2003.
[Cited: 27 07 2015.] http://www.isprs.org/proceedings/xxxiv/5-
W10/papers/remondin.pdf.
[23] HOPPE, Hugues, et al. Mesh Optimization. University of Washington, Seattle.
[Online] 1992. [Cited: 27 07 2015.]
http://research.microsoft.com/en-us/um/people/hoppe/meshopt.pdf.
[24] KUKAL, Zdeněk, et al. Hrady Čech a Moravy z čeho jsou a na čem stojí. 2010 :
Grada Publishing, a.s. a Česká geologická společnost, 2010. 978-80-7075-740-6.
CTU in Prague Reference List
81
[25] Nemovité památky - hrad Helfenburk, zřícenina. MonumNet. Národní památkový
ústav. [Online] 2013-15 [Cited: 22 06 2015.]
http://monumnet.npu.cz/pamfond/list.php?IdReg=154930.
[26] Kol. autorů. Hrady, zámky a tvrze v Čechách, na Moravě a ve Slezsku III - Severní
Čechy. Praha : Svoboda, 1984. 25-089-84.
[27] ČÍŽEK, Jiří. Helfenburk (Hrádek) u Úštěka. HRADY.CZ Cestujte s přehledem,
Atthero s.r.o. [Online] 2002 [Cited: 27 06 2015.] http://www.hrady.cz/?OID=356
[28] VOSYKA, Lukáš. Vybudování vztažné sítě pro detailní zaměření hradu Helfenburk.
Diplomová práce. ČVUT v Praze. [Online] 2015. [Cited: 04 07 2015.]
http://geo.fsv.cvut.cz/user/cepek/proj/dp/2015/lukas-vosyka-dp-2015.pdf.
[29] TOUŠEK, Martin. Zaměření a vytvoření prostorového modelu hlavní věže hradu
Helfenburk u Úštěka. Diplomová práce. ČVUT v Praze. [Online] 2015. [Cited: 28
05 2015.]
http://geo.fsv.cvut.cz/user/cepek/proj/dp/2015/martin-tousek-dp-2015.pdf.
[30] Datasheet TRIMBLE TX5 scanner. [Online] 2012. [Cited: 04 07 2015.]
http://trl.trimble.com/docushare/dsweb/Get/Document-628869/022504-
122_Trimble_TX5_DS_1012_LR.pdf.
[31] Trimble TX5. Trimble. [Online] 2015. [Cited: 04 07 2015.]
http://www.trimble.com/3d-laser-scanning/tx5.aspx.
[32] User Guide Trimble TX5 3D Laser Scanner version 2.00. Trimble.[Online] 2013.
[Cited: 17 05 2015.] http://trl.trimble.com/docushare/dsweb/Get/Document-
633127/Trimble%20TX5%20%20User%20Guide%20V2%20-%20English.pdf.
[33] TRIMBLE® TX5 3D Laser Scanner Quick Start Guide. Trimble. [Online] 2012.
[Cited: 17 05 2015.]http://trl.trimble.com/docushare/dsweb/Get/Document-
633125/87655-10_TX5_QuickStartGuide_EN_Trimble_LR.pdf.
[34] POESOVÁ, Jana. Laserové skenování pro potřeby geometrické analýzy žebrové
klenby z doby Lucemburků na Pražském hradě. Bakalářská práce. ČVUT v Praze.
[Online] 2013. http://gama.fsv.cvut.cz/~cepek/proj/bp/2013/jana-poesova-bp-
2013.pdf.
[35] PROKOPOVÁ, Alžběta. Zaměření hradu Helfenburk u Úštěka a vytvoření části jeho
prostorového modelu. s.l. : Diplomová práce, 2015. ČVUT v Praze.
CTU in Prague Reference List
82
[36] KAZHDAN, Michael, BOLITHO, Matthew and HOPPE, Hugues. Poisson Surface
Reconstruction. Eurographics Symposium on Geometry Proccessing. [Online]
2006. [Cited: 10 07 2015.]
http://www.cs.jhu.edu/~misha/MyPapers/SGP06.pdf.
[37] KAZHDAN, Michael and HOPPE, Hugues. Screened Poisson Surface
Reconstruction. Johns Hopkins University and Microsoft Research. [Online] 2012.
[Cited: 10 07 2015.] http://www.cs.jhu.edu/~misha/MyPapers/ToG13.pdf.
[38] KAZHDAN, Michael. Screened Poisson Surface Reconstruction (Version 8.0).
[Online] [Cited: 10 07 2015.]
http://www.cs.jhu.edu/~misha/Code/PoissonRecon/Version8.0.
[39] Mechanical, electrical, and plumbing. Wikipedia, the free encyclopedia. [Online]
2015. [Cited: 20 07 2015.]
https://en.wikipedia.org/wiki/Mechanical,_electrical,_and_plumbing.
[40] Lambert's cosine law. Wikipedia, the free encyclopedia. [Online] 2015. [Cited: 21
07 2015.] https://en.wikipedia.org/wiki/Lambert%27s_cosine_law.
[41] ŠTRONER, Martin and HAMPACHER, Miroslav. Zpracování a analýza měření
v inženýrské geodézii. Praha : České vysoké učení technické v Praze, 2011. 978-
80-01-04900-6.
CTU in Prague List of Figures
83
List of Figures
Figure 1 Detection chain of a laser scanning system ...................................................................................... 14
Figure 2 Methods for measuring of a 3D surface: (a) Light transit time (b) Triangulation.......................... 17
Figure 4 Mixed edge problem .......................................................................................................................... 21
Figure 5 Various types of target (a) White sphere, (b) Planar target, (c) Curved planar target.................. 23
Figure 6 Example of registration algorithm..................................................................................................... 24
Figure 7 CAD model with point cloud............................................................................................................... 26
Figure 8 Smooth reconstruction of Igea and different smooth parameters ................................................. 27
Figure 9 Location of the castle ......................................................................................................................... 29
Figure 10 Aerial photograph of Helfenburk near Úštěk.................................................................................. 30
Figure 11 Part of tacheometrical plan made by M. Závetský and J. Krupka in 1983 ................................... 33
Figure 12 Building of the geodetic point field ................................................................................................. 35
Figure 13 The inner palace with sphere targets .............................................................................................. 38
Figure 14 Laser scanning of cragged and hardly accessible terrain .............................................................. 40
Figure 15 Laser scanner Trimble TX5 ............................................................................................................... 42
Figure 16 Setting the scan parameters ............................................................................................................ 43
Figure 17 Check point (a) Photograph (b) Vertex in point cloud .................................................................... 46
Figure 18 Dialogue window (a) Uniform sample (b) Batch processing ......................................................... 50
Figure 19 Import file dialog .............................................................................................................................. 52
Figure 20 Application of Region grow sphere ................................................................................................. 53
Figure 21 (a) Modelled sphere target (b) Vertex with Registration label...................................................... 54
Figure 22 Registration window with Cloud constraints wizard...................................................................... 59
Figure 23 Mismatched point clouds ................................................................................................................. 60
Figure 24 Planar target in a point cloud .......................................................................................................... 63
Figure 25 Triangular mesh from Poisson Surface reconstruction with many holes...................................... 69
Figure 26 Different types of the filling technique (a) Hole (b) Flat (c) Tangent (d) Curvate......................... 70
Figure 27 Filling up of a large hole (a) Bridges (b) Completed filling............................................................. 71
Figure 28 Joining of two parts of mesh ............................................................................................................ 71
Figure 29 3D Deviation analysis (top view) ..................................................................................................... 72
Figure 30 Final point cloud ............................................................................................................................... 74
Figure 31 Final triangular mesh of inner palace ............................................................................................. 74
CTU in Prague List of Tables
84
List of Tables
Table 1 Time schedule of the work................................................................................................................... 34
Table 2 Timesheet and weather conditions..................................................................................................... 37
Table 3 Overview of the scan positions ........................................................................................................... 41
Table 4 Applied scan parameters without colours.......................................................................................... 44
Table 5 Computer configuration ...................................................................................................................... 48
Table 6 Overview of point clouds in the first section of registration ............................................................. 57
Table 7 Overview of point clouds and scans in the second section of registration ...................................... 59
Table 8 Overview of point clouds in the final section of registration ............................................................ 60
Table 9 Coordinates of control points used for the transformation .............................................................. 62
Table 10 Reduced coordinates of control points ............................................................................................. 62
Table 11 Model coordinates of control points ................................................................................................ 64
Table 12 Distances in (a) longitudinal direction, (b) transverse direction, (c) Z axis direction .................... 65
Table 13 Overview of parts of the point cloud ................................................................................................ 67
Table 14 Information about meshes ................................................................................................................ 69
Table 15 Overview of created PLY files of modelled parts ............................................................................. 73
CTU in Prague List of Appendices
85
List of Appendices
Appendix A Situation of the Geodetic Net and Points of the Traverse from the Measurement in 1983 ...... I
Appendix B Situation of the Scan Positions and Targets .................................................................................II
Appendix C List of Coordinates of the Check Points .......................................................................................III
Appendix D Situation of the Check Points ....................................................................................................... V
Appendix E Input Parameters for Poisson Surface Reconstruction .............................................................. VI
Appendix F Final Triangular Mesh ................................................................................................................ VIII
Electronical appendices attached on DVD:
- Reports of control measurement computation, differences of check distances among the
coordinates from control measurement and model coordinates
Directory “control_measurement”
- Reports of registration computation in Leica Cyclone
Directory “reports_cyclone_registration”
- Reports from mesh triangulation in PoissonRecon
Directory “reports_poisson”
- TFM files
Directory “tfm_matrix”
- Triangular mesh
Directory “meshes”
� Meshes made directly from Poisson Surface “ply_trimmer”
� Meshes repaired in Geomagic Studio “ply_filled”
� Decimated meshes “ply_decimate”
- Papers of the thesis
jana-poesova-dp-2015.pdf
Electronical appendices saved on university workstation:
Electronical appendices can be found on university workstation in directory
“Helfenburk_vysledne” (c:\_data\Helfenburk_vysledne\)
- Measured data from laser scanning
- Databases of Leica Cyclone with the registered final point cloud and check points
- Overview of check points with photo documentation
- Final point cloud in ASCII TXT format
- The same appendices as on DVD
I
Appendix A Situation of the Geodetic Net and Points of The Traverse
from the Measurement in 1983
II
Appendix B Situation of the Scan Positions and Targets
III
Appendix C List of Coordinates of the Check Points
Point S-JTSK
Y [m] X [m] Z [m]
1 737947.915 988466.529 321.187
2 737948.266 988467.839 322.564
3 737947.192 988465.357 322.550
4 737948.530 988466.224 318.464
5 737942.011 988455.686 315.477
6 737941.198 988454.970 316.830
7 737943.074 988457.408 316.086
8 737937.861 988452.185 313.813
9 737937.425 988452.110 314.055
10 737936.910 988451.893 312.975
11 737937.006 988451.905 311.752
12 737939.126 988452.647 312.198
13 737938.029 988452.253 311.674
14 737959.789 988480.101 315.914
16 737948.548 988470.777 313.860
17 737946.554 988469.134 314.064
18 737936.372 988473.172 313.754
19 737983.453 988417.410 315.311
20 737981.906 988418.601 315.504
21 737981.693 988418.784 315.488
22 737979.325 988420.661 317.372
23 737975.811 988423.210 315.650
24 737975.671 988423.311 317.278
25 737973.600 988424.808 317.347
26 737983.853 988416.768 317.193
27 737984.959 988416.247 316.279
29 737983.561 988418.650 313.224
30 737980.985 988420.605 312.800
33 737976.725 988423.834 313.521
34 737986.590 988416.294 313.443
35 737989.751 988435.323 318.462
37 737992.863 988429.308 318.544
IV
Point S-JTSK
Y [m] X [m] Z [m]
38 737992.343 988432.372 325.748
39 737991.547 988432.958 324.014
40 737965.308 988442.863 327.982
41 737977.136 988437.755 319.963
42 737978.386 988439.466 319.506
43 738006.391 988418.768 321.826
44 738005.724 988418.089 322.497
45 738005.460 988417.910 323.829
46 738000.106 988462.169 314.632
47 737999.792 988462.673 315.399
48 738001.772 988458.830 314.270
49 738002.517 988457.392 314.547
50 738004.156 988454.177 314.553
51 737994.356 988442.110 331.110
52 737992.392 988447.671 320.469
53 737987.268 988462.939 320.556
54 737987.778 988464.754 316.383
55 737973.318 988487.635 316.847
56 737984.032 988489.109 316.709
57 738018.617 988416.627 311.229
58 738017.679 988410.574 311.679
59 738018.795 988412.424 312.975
60 738018.556 988419.012 312.880
61 737942.299 988453.506 315.479
62 737944.505 988456.379 315.532
63 737944.386 988456.331 319.370
64 737945.567 988457.840 319.379
65 737948.402 988463.716 318.704
66 737948.406 988463.714 319.549
67 737955.340 988459.433 318.478
68 737960.797 988455.635 318.503
70 737956.497 988449.450 317.301
71 737956.444 988469.525 328.991
72 737955.331 988470.561 329.212
V
Appendix D Situation of the Check Points
VI
Appendix E Input Parameters for Poisson Surface Reconstruction
The information about the parameters mentioned below is adopted from [38].
POISSON RECON
--in <input points>
This string is the name of the file from which the point set will be read. In our case it
has to be an ascii file with groups of 6 columns with white space delimited and the first
three places are the x-, y-, and z-coordinates of the position of the point, followed by
the x-, y- and z-coordinates of the normal of the point. There should not be any
specified information about the number of oriented point samples. The other
possibilities of input file are in [38].
--out <output triangle mesh>
This string is the name of the file to which the triangle mesh will be written. The file is
written in PLY format.
--depth <reconstruction depth>
This integer is the maximum depth of the tree that will be used for surface
reconstruction. Running at depth corresponds to solving of a voxel grid whose
resolution is no larger than 2^d x 2^d x 2^d. Since the reconstructor adapts the octree
to the sampling density, the specified reconstruction depth is only an upper bound.
It means by these parameters we assess the final definition of the triangular mesh.
The approximate length of triangular edge is following:
l = B / (2^d) [m], where
B is the maximal bounding dimension of a point cloud [m]
d is the depth parameter, the default value is set up to 8.
--pointWeight <interpolation weight>
This floating point value specifies the importance of the fact that the interpolation of
the point samples is given in the formulation of the screened Poisson equation. The
results of the original (unscreened) Poisson Reconstruction can be obtained by setting
this value to 0. The most applicable value for the parts of the castle was empirically
found out and set up to 8.
VII
--polygonMesh
Enabling this flag tells the reconstructor that output is a polygon mesh (rather than
triangulating the results of Marching Cubes).
--density
Enabling this flag tells the reconstructor that output has the estimated depth values of
the iso-surface vertices. It is necessary for the subsequent computation with the
SurfaceTrimmer, because the reconstructor appends the field with the mesh vertices
for the output.
--verbose
Enabling this flag provides a more verbose description of the running times and
memory usages of individual components of the surface reconstructor.
SURFACE TRIMMER
--in <input triangle mesh>
This string carries the name of the file from which the triangle mesh will be read. The
file is read in PLY format.
--trim <trimming value>
This floating point values specifies the value for mesh trimming. The subset of the
mesh with signal value less than the trim value is discarded. Usually, this value was
adjusted to 1-2 less than the depth parameter.
--aRatio <island area ratio>
This floating point value specifies the area ratio that defines a disconnected
component as an "island". Connected components whose area, relatively to the total
area of the mesh, is smaller than this value, will be merged into the output surface to
close small holes, and will be discarded from the output surface to remove small
disconnected components. The default value is 0.001. By the empirical ascertainment
it was found out that the most appropriate solution is aRatioseting to the value 0 and
afterwards the isolated undesirable connected components are deleted in the
software Geomagic Studio using the function Select by Area.
VIII
Appendix F Final Triangular Mesh
Top View
IX
South View
X
Northwest View
XI
Detailed View
It can be compared with Figure 14 on page 40.