Parametrical and Procedural Approach in the
LIDAR Data Visualisation
Complex Creation of the Photorealistic and Accurate 3D Model of the Surface
Jan Hovad and Jitka Komarkova
Faculty of Economics and Administration, University of Pardubice, Studentska 84, 532 10 Pardubice, Czech Republic
Keywords: Dtm, Dsm, Lidar, Polygon, 3D Model, Visualisation.
Abstract: Author presents a new hybrid method for LIDAR processing in the form of complex, real-world attribute-
based 3D digital model of the terrain. This model can be widely used for 3D GIS analyses, visualization for
public contracts, construction of new communications (roads), GPS navigation and generating individual
views from any location. Model is based on the LIDAR scanned values that are transferred into the
quadrilateral polygonal form. The main goal is to reconstruct large scale areas in 3D by slope based
quadrilateral grids that are interpolated from irregular raw structure. Laser scan is checked and corrected. It
is simplified by chosen interpolation technique into the quadrilateral set of grids (C++). Each grid has
different resolution to lower hardware requirements. Raw scan is analysed by slope factor and is used
further to classify grids into a groups. Classified grids are processed by 3D scripting language to form
polygonal terrain model. Vector and polygonal data are converted into 3D and utilized to reconstruct surface
features and terrain classes. Set operations are used to divide digital terrain model into the segments. Set of
high polygonal models is created for each class. These models are scattered on the top of the classified
polygonal surface.
1 INTRODUCTION
Light detection and ranging (LIDAR) is frequently
used technology that provides information about
elevation of scanned objects, their position,
classification etc.. This scan is usually obtained by a
special modified airplane or it is combined with
ground LIDAR scanners. Raw data source is formed
by individual points with basic x, y and z attributes.
Scanned area includes all captured points detected
by detection device and it is called Digital Surface
Model (DSM). These points can be segmented and
classified by chosen algorithms. For example the
laser beam reflected only from the terrain, forms
Digital Terrain Model (DTM). Other model type can
include only buildings that form, after extraction
from a DSM, a Digital Building Model (DBM).
Extracted trees form a Digital Canopy Height Model
(DCHM) and so on (Broveli at al. 2004, Omasa et al.
2008, Chen et al. 2012). All these models along with
models from other branches (engineering), can be
unified and merged together.
Creation of DTM is frequently initial activity in
the LIDAR data pre-processing. In the past 20 years
there was the whole set of algorithms to extract the
best representation of DTM (Hu 2003, Elmqvist
2002, Kraus and Pfeifer 2001).
Other categories include prediction of DTM,
accuracy analysis against the real data sets and
simplifying LIDAR point cloud. It is often
impossible to obtain data at all of the points in the
area of interest. Therefore it is necessary to compute
these missing values additionally. Spatial
interpolation calculates an unknown value from a set
of points with known values. For many applied
objectives it is necessary to create a regular square
surface formed from quadrilaterals - in this paper
called quad regular network (QRN). This procedure
can be created using the following local/global
approach of the selected algorithm - Renka-Cline,
Shepard, IDW, etc. (R. J. Renka and A. K. Cline
1984, McLain 1976, Lawson 1977, Akima 1978).
Current applied utilization of DTM creation is
very rich and intersects a wide range of scientific
disciplines. DTM and its hybrid shape forms are
used in Austria by Mandlburger et al. (2009) for
156
Hovad J. and Komarkova J..
Parametrical and Procedural Approach in the LIDAR Data Visualisation - Complex Creation of the Photorealistic and Accurate 3D Model of the Surface.
DOI: 10.5220/0005098001560161
In Proceedings of the 9th International Conference on Software Engineering and Applications (ICSOFT-EA-2014), pages 156-161
ISBN: 978-989-758-036-9
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
analysing terrain models for river flow modelling.
Mandlburger found that the data reduction is
necessary but in other hand it can bring significant
errors. DBM is created by many approaches like
slopes evaluation (Zhou et al. 2004), building
extraction from space borne imagery (Tack at al.
2011) or they were extruded from the DEM
(Priestnall et al. 2000). DBM can be used for GPS
signal prediction. Li et al. (2007) explained a ray
tracing method, which was applied to prediction and
visualization of GPS signal in dense urban areas.
Implementation of a real-time satellite visibility was
also researched by Taylor et al. (2006). Besides GPS
the DBM can be used directly in evaluating the Sun
exposure time in case of constructing the solar
panels on the top of the building roofs. This study
was explained by Nguyen et al. (2012). Scientific
research and applied usage of LIDAR is also
directed into agriculture and forestry.
LIDAR topic is widely discussed but its
applications can be still extended. Each algorithm or
method is usually used for certain purposes like in
urban areas to create DBM. Different results can be
obtained when creating DTM, topographic position
can shift the algorithm precision and so on.
Approach of combining different steps, algorithms
and methods is usually needed to solve a specific
problem. Proposed solution allows direct
interconnection of the geography, architecture,
engineering and transport outputs (simulation,
analyses) together. For example, authors are able to
calculate GIS analyses, import 3D model of the
bridge with calculated stress and tension analysis
from the engineer, integrate building models from
the architect and print everything by the 3D printer
or create photorealistic animations/visualisation of
the output. This kind of approach is described in this
article.
2 GOALS
The main goal is: LIDAR data utilization in case of
reconstructing large scale areas in the 3D by slope
based quadrilateral grids to form photo-realistic
model of the surface, which is based on the real-
world values scanned by laser.
Partial goals are: Laser scan is carefully checked
and corrected - registered, merged, de-noised etc.
Laser scan is simplified by chosen interpolation
technique into the form of quadrilateral set of grids
(C++/Python). Each grid has different resolution to
lower hardware requirements - adaptive approach.
Laser scan is analysed by slopes and this analysis is
used further to classify interpolated grids into
groups. Classified grids are processed by 3D
scripting language to form polygonal quadrilateral
digital terrain model. Points are intersected by
splines predictably and they form a regular shaped
quadrilaterals that can be further divided and
adjusted to match the imported object properties
(Figure 1).
Figure 1: Surface division.
Vector data are converted into 3D and utilized to
reconstruct rivers, roads, buildings and terrain
classes. All other objects can be placed into the
model. Quadrilateral structure allows unlimited
smoothing by intersecting additive splines through
the polygon. Set operations are used to divide digital
terrain model into the classes (forests, fields, urban
areas, water surfaces etc.). A set of high polygonal
models is created for each class. Models are
scattered on the top of the classified surface and the
static outputs are formed (Figure 2).
2.1 Used Data, Software and Hardware
Authors have two data inputs available. LIDAR scan
for the area size roughly 10×20 km. Data set is
scanned by aircraft Turbolet L-410 FG. Outputs
create two types of LIDAR models. First, the Digital
Surface Model 1
st
Generation (DSM 1G) which
includes all objects on the surface (terrain +
vegetation, buildings…). The second is the Digital
Model of the Relief 5
th
Generation (DMR 5G)
(Belka 2012). Scanned values are stored as XYZ
coordinates (Figure 3).
Each scanned value is represented by separate
line with information about X/Y coordinate (WGS
1984) and Z elevation attribute (meters above sea
level).
The second data type is vector layer from the
database ZABAGED
®
, which contains the basic
surface features like rivers, roads, terrain types etc.
ParametricalandProceduralApproachintheLIDARDataVisualisation-ComplexCreationofthePhotorealisticand
Accurate3DModeloftheSurface
157
Figure 2: Proposed solution.
Figure 3: LIDAR point cloud.
Hardware requirements are quite moderate
because authors process the big data volumes.
Authors used two computers to perform individual
operations. AMD Opteron and Intel i7 2600K CPU,
with minimum of 16 GB memory, fast SATA III
SSD (500 MB/s) and GeForce 460 GTX. The most
important is high amount of memory to work with a
large number of points. Render farm in Germany is
used as a remote cloud station to get quick static
outputs from the 3D model.
Software that is used is mainly a 64 bit to fully
utilize RAM capacity. It is ArcGIS 10 SP3,
SagaGIS, OriginLab. These applications fully
sufficed in all GIS analyses and allowed
programming the batch interpolation tasks.
Programming can be made with the help of
integrated Python or C++. Authors utilized national
algorithm library (NAG) for C++. Graphical tasks
are completed in the Autodesk 3D Studio Max 2012
and its Maxxscript module that allows object
oriented scripting and processing of wide range of
data inputs.
3 METHODS
This chapter characterizes details of the individual
steps as described in the Figure 2.
3.1 Spatial Interpolation and Adaptive
Approach
This text uses local interpolation of Inverse Distance
Weighting (IDW) and global version of Renka-Cline
interpolation. The whole irregular structure is batch
processed by a C++. The code saves all interpolated
grid outputs into specified folder. Each output is
saved in the different resolution defined by a target
size of required polygon. Target size of polygons is
set to array of {5 10 20 40 60 100} meters.
Computation results in two input arrays that form a
raw empty grid that is filled by interp. values, COLS
{100,167,251,502,1005,2010}, ROWS {199,332,49
8,997,1995,3990}. C++ code solves the batch tasks
of performing Renka-Cline within an application
OriginLab. Source code shows only a few lines of
procedure code due to the limits of the article size
(Figure 4).
Figure 4: Main part of the source code for Renka-Cline
interpolation.
Files with grids are stored in the created
directory. Each of them is compared against slope
analysed DEM and clipped into 3 major groups
based on the National Slope Classification system.
This process creates detailed grids for areas with
sharp elevation changes and low resolution grids for
flat areas. Minor overlaps are apparent and they help
to connect all regions. Hardware requirements are
lowered (Figure 5) (Svobodová 2011).
ICSOFT-EA2014-9thInternationalConferenceonSoftwareEngineeringandApplications
158
Figure 5: Spatial and adaptive grids.
3.2 3D Scripting of Chosen Point
Inputs
According to the previous step, the set of grids is
chosen. That minimizes the amount of spatial
deviation in the area of interest. This structure is still
represented by a points or optionally raster coded
values. Each set has a different resolution and point
offset. Each row or column (points are interpolated
into predefined grid size already) must be sorted and
intersected by a spline to fit each point in the good
order. This step forms a quadrilateral polygon
structure, where the polygons are equal in the area
size, which means that they can be easily
subdivided. The following procedure can be done in
any 3D environment that supports work with
vectors/polygons and allows the use of scripting or
programming techniques to access the individual
objects within the source code. This requirement is
met for example by the Maxxscript, 3D scripting
language, which is available as the part of the 3D
Studio Max. Full implementation of the principle
can be expressed in the graphical form (Figure 6).
Figure 6: Author´s approach to create point to polygon
surface.
Selected data are loaded from the text files line
by line (point by point). Each array is then sorted
and the line is intersected vertically and horizontally.
Sorting process can be time consuming task (in case
of higher inputs). Authors recommend to pre-
compute this step by distributed approach (Hadoop).
3.3 Parametrical and Procedural
Modelling
In this phase of work, the terrain is formed from
continuous quadrilateral polygonal structure with
houses on the top of the surface. What needs to be
done is to clip the surface into groups, classes. There
are two ways to do this. The first one utilizes
modern LIDAR scanners that handle classification
automatically. This is not the author’s case. The
second option is to use set operations (intersect) and
digitized vectors from publicly available databases.
The process of creating surface ID for each class is
shown in the following picture (Figure 7).
Figure 7: Set operations to create unique terrain segments.
The terrain model is ready to be fitted with
models. Each terrain ID receives a set of adequate
models that are scattered on the top of the surface.
The model for the given area after applied
intersection is shown in Figure 8.
Figure 8: Segmented terrain 3D model.
All work in the visualisation is planned to be
purely parametrical and procedural without any
needs of the subjective interventions. The word
parametrical means that authors prepared for each
ID a set of high-polygonal objects that are suitable
for individual terrain sections. These objects are
modelled separately, which take some time, but they
can be used in the infinitely many projects
repeatedly. Authors are able to exclude this step in
ParametricalandProceduralApproachintheLIDARDataVisualisation-ComplexCreationofthePhotorealisticand
Accurate3DModeloftheSurface
159
the future. Everything can be performed almost
automatically. More models for each ID section
forms better outputs because of variability.
Word procedural is here crucial because every
object material is very simple. Materials are created
only in the procedural way of combining colours.
All the details are stored in the high-polygonal
models that are stored as proxies. Proxies are not
hardware challenging but they are unpacked when
making final render. Memory is saved and GPU,
which influences the speed while working with the
model, is not so busy. FPS (Frame Per Second) is
higher and total operability is increased. The
different colour of models is made by colour map,
which is applied to each section. Each section has a
scattered class assigned, that has multiple properties
used to vary each object from the other one. Main
attributes are distribution (uniform, random),
random/fixed scale/move/rotation, collisions,
normals and many other parameters. These objects
that contain all the scattering information and
reference models can be transferred from one project
to any other. This makes the visualization of ArcGIS
analysis very flexible and time effective. Next very
simplified image shows a few model clusters ready
to be placed and modified on the basis of the
assigned ID. Attributes can be based on the exact
value from LIDAR.
The point structure is very reduced, all the
surface objects are stored as Proxy objects, terrain is
quadrilateral, each polygon contains only 4 points
while the surface is adaptive (higher slopes have
smaller polygons, flatter areas have bigger
polygons). This situation can be illustrated by the
following 3D model output. This is completely
computer generated imagery, not a photograph
(CGI), which is based on the LIDAR point cloud
values. The values are derived into the 3D model ()
The camera can be placed anywhere in the model
or it can be animated. The general level of detail is
very high even in case of close-ups because the
proxy objects are very detailed. This can be
demonstrated by placing the secondary camera into
the altitude of 550 meters and the third camera into
the middle of the forest below (Figure 9).
4 DISCUSSION AND
CONCLUSION
This section summarizes the results and outlines
possible ways forward in the future research.
Figure 9: Level of detail, random camera placement.
4.1 Discussion
Authors completed all stated goals with success.
However, some tasks can be adjusted and improved.
Therefore, the future work is going to be directed
into the area of a raster driven scattering. Every
model on the surface will be scattered on the basis of
the black-white raster map. This map can store the
LIDAR values like tree height or tree width in much
lower hardware demanding form. There is no need
to use statistical approach while scattering the proxy
objects. Every tree can have an exact position as in
the real world in the time of the scanning.
The second task that can be enhanced is the
digital weather model. In this case, clouds are
generated randomly. In the future work, the NOAA
images are connected with the 3D model. Particle
system is created and clouds are generated in 15
minutes intervals.
4.2 Conclusion
Authors propose a method, which encapsulates
outputs from the GIS analyses, LIDAR data
processing, DBM acquisition and terrain creation to
utilize all these steps in parametrical and procedural
scientific visualization. The resulting 3D model can
be widely used for 3D GIS analyses, visualizations
for public contracts, planning and construction of
new communications (roads, bridges, etc.),
simulations, 3D printing, GPS navigation and
generating individual static/dynamic views from any
location. The result connects GIS with other
sciences like architecture, agriculture, forestry or
road design and can be further used as a GIS server
with large dimensioned 3D terrain model (C# .NET).
Client side can provide user interface to fill
parameters like camera position, target position,
camera settings, weather condition, file formats etc.
The rendered output can be sent to the client.
Studios that must remodel each environment for the
ICSOFT-EA2014-9thInternationalConferenceonSoftwareEngineeringandApplications
160
client manually can lower the costs significantly by
use of already created large scaled model. There are
so many application possibilities that it is hard to
mention all of them.
ACKNOWLEDGEMENTS
This work was supported by the project No.
CZ.1.07/2.2.00/28.032 Innovation and support of
doctoral study program (INDOP), financed from EU
and Czech Republic funds.
REFERENCES
Akima, H., 1978. A method of bivariate interpolation and
smooth surface fitting for irregularly distributed data
points, ACM TOMS, 4, 148-64.
Belka, L., 2012. Airborne laser scanning and production of
the new elevation model in the Czech Republic,
Vojenský geografický obzor, 55 (1), 19-25.
Broveli, M. A., Canata, M. and Longoni, U.M, 2004.
LIDAR data filtering and DTM interpolation within
GRASS, Transactions in GIS, 8(2), 155-174.
Chen, Z., Devereux, B., Gao, B. and Amable, G., 2012.
Upward-fusion urban DTM generating method using
airborne LIDAR data, Journal of Photogrammetry and
Remote Sensing, 72, 121-130.
Elmquist, M., 2002. Ground surface estimation from
airborne laser scanning data with morphological
methods, Photogrammetric Engineering & Remote
Sensing, 73 (2), 175-185.
Hu, Y., 2003, Automated Extraction of Digital Terrain
Models, PhD Thesis. University of Calgary, Canada
Kraus, K. and Pfeifer, N., 2001. Advanced DTM
generation from LIDAR data, International Archives
of Photogrammetry and Remote Sensing, 34(3), 23-30.
Lawson, C. L., 1977. Software for C surface interpolation,
Mathematical Software, 3, 161-194.
Li, J., Taylor G., Kidner, D. and Ware, M., 2008.
Prediction and visualization of GPS multipath signals
in urban areas using LIDAR Digital Surface Models
and building footprints, International Journal of
Geographical Information Science, 22 (11-12), 1197-
1218.
Mandburger, G., Hauer, C., Hofle, B., Habersack, H. and
Pfeifer, N., 2009. Optimization of LIDAR derived
terrain models for river flow modeling, Hydrology and
Earth System Sciences, 1453-1466.
Mclain, D.H., 1976. Two dimensional interpolation from
Random Data, Computer J., 384, 179-181.
Nguyen, H. T., Pearce, J. M., Harrap, R. and Barber, G.,
2012. The Application of LIDAR to Assessment of
Rooftop Solar Photovoltaic Deployment Potential in a
Municipal District Unit, Sensors, 12, 4534-4559.
Omasa, K., Hosoi, F., Uenishi, T. M., Shimizu, Y. and
Akiyama, Y., 2008. Three-Dimensional Modeling of
an Urban Park and Trees by Combined Airborne and
Portable On-Ground Scanning LIDAR, Remote
Sensing, Environmental Modeling & Assessment,
13(4), 473-481.
Priestnall, G., Jaafar, J. and Duncan, A., 2000. Extracting
urban features from LIDAR digital surface models,
Computers, Environmental and Urban Systems, 24,
65-78.
Renka, R. J. and Cline, A. K., 1984. A triangle-based C
Interpolation Method. Journal of Mathematics, 14,
223-237.
Svobodova, J., 2011. Quality assessment of digital
elevation models for environmental applications.,
Thesis, University of Ostrava, 174p, Available in
Czech only: Title: Hodnocení kvality digitálních
výškových modelů pro environmentální aplikace
Tack, F., Buyuksalih, G. and Goossens, R., 2012. 3D
building reconstruction based on given ground plan
information and surface models extracted from
spaceborne imagery, ISPRS Journal of
Photogrammetry and Remote Sensing, 67, 52-64.
Taylor, G., Li, J., Kidner, D., Brunsdon, CH. and Ware,
M., 2007. Modeling and prediction of GPS availability
with digital photogrammetry and LIDAR,
International Journal of Geographical Information
Science, 21 (1), 1-20.
Zhou, G., Song, C., Simmers, J. and CHeng, P., 2004.
Urban 3D GIS From LIDAR and digital aerial images,
Computers & Geosciences, 30, 345-353.
ParametricalandProceduralApproachintheLIDARDataVisualisation-ComplexCreationofthePhotorealisticand
Accurate3DModeloftheSurface
161