Generating the open space 3D model based on LiDAR data.
Kalantaite, Ausra ; Parseliunas, Eimuntas Kazimieras ; Romanovas, Denis 等
1. Introduction
The usage of the LiDAR (Light Detection and Ranging) data is
growing up. One of the main motives for its usage is the high speed of
data acquisition (Schickler, Thorpe 2001; Zalnierukas, Cypas 2006;
Stankevicius 2009). The main scientific field of using the LiDAR data in
geodesy and remote sensing is to construct the digital heights models
(digital elevation models, digital terrain models, digital surface
models, digital relief models) (Arrowsmith 2006; El-Sheimy et al. 2005;
Yan et al. 2012; Sulaiman et al. 2010; Susaki 2012; Zhang, Whitman 2005;
Meng et al. 2010). But the raw LiDAR data are not only geodetic heights,
but also the information about other natural and artificial objects on
Earth's surface (for example, vegetation, buildings, etc.) (Fowler
2001; Stankevicius 2009).
An open space could be defined as a space, which is restricted by a
surface, that is generated over the physical Earth's surface,
natural and artificial objects, and in which the distances between its
objects are not less than given critical tolerance. In other words, we
have in mind the moving objects of certain dimensions, which could
freely move (fly) in such an open space. We intend to apply only 2D
restrictions caused by moving objects, so, for example, airplane could
fly over the bridge only, nor there is a free enough space to fly under
it. The open space surface will be closed to the very digital relief
model's surface in the agricultural areas and grasslands, therefore
it will be over trees in the forests, or over buildings' roofs in
the cities. In some sense the open space 3D model is similar, for
example, to the Digital Surface Model (DSM) or to the obstacles
limitation map of the airport area (Terrain 2011). Therefore, the
developers of such maps do not take into account the moving
objects' dimensions.
2. Experimental data
In 2008-2010 the LiDAR data were captured for all territory of
Lithuania. According to the technical requirements, the density of the
points is approximately 4 points in 1 sq. m. This results in a very high
resolution data set with a good spatial distribution. The accuracy of
any LiDAR data point is not worse than 15 cm in height component, and
not worse than 30 cm in plane position (Detalaus ... 2007; Lietuvos ...
2008; Zalnierukas et al. 2009). At the same time the colour
orthophotomaps were also made (Fig. 1).
The raw LiDAR data were classified into three groups: Earth's
surface data, buildings data and vegetation data. Therefore, for the
purpose of developing the open space 3D model, the data was divided into
two sets: filtered data set-Earth's surface data, and non-filtered
data set-all LiDAR data. The research territory of 1 sq. km was chosen
with a total amount of the LiDAR data points of about 1.5 m (Fig. 2).
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
3. Method of the open space 3D model construction
From Figure 2a it can be obviously seen that a density of the
points in the area of the water body is much more less than in the other
places of the territory. It is even less than it should be according to
the technical requirements. In Figure 2b we could see that the LiDAR
points of the buildings were eliminated, as well as the points captured
from the vegetation.
In the first step, the 3D models based on both data sets were
generated. They were expressed by the Triangle Irregular Networks (TIN)
(Fig. 3).
In the second step, the 3D model based on the non-filtered data set
is combined with the orthophotomap to visualise the territory (Fig. 4).
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
This combination of the 3D model and the orthophotomap will be used
for the control of the open space 3D model.
In the third step, we suggest to apply the local interpolation
algorithm (Arrowsmith 2006; El-Sheimy et al. 2005; Yan et al. 2012;
Sulaiman et al. 2010; Susaki 2012; Zhang, Whitman 2005; Meng et al.
2010) the result of which will lead to the generation of the open space
3D model. First of all, the critical dimension X of the moving object
should be defined (Fig. 5). This dimension will be the cell size of the
grid network of the open space 3D model. For example, let it be 10 m.
Later on, the LiDAR points are grouped according to the network cells,
and in each cell the maximal value of the point's height is
retrieved. This maximal height values are assigned to the central points
of each cell (Fig. 5).
Also, we should stress that density of the points in the open space
3D model is 1 point in 100 sq. m only (when X = 10 m) (Fig. 6).
[FIGURE 5 OMITTED]
[FIGURE 6 OMITTED]
Graphical representation of the created open space 3D model surface
is shown in Figure 7.
To test the open space 3D model, we could create the surface based
on all LiDAR points (LiDAR surface) (Fig. 8).
Now we can subtract the open space 3D model surface from the LiDAR
surface in order to receive the surface of two models differences (Fig.
9).
These differences should be with a sign "+". Otherwise
the open space 3D model will contain obstacles, what is against its
definition.
To analyse in more detail the quality of the open space 3D model,
we could create profiles along the created surfaces. For example, in
Figures 10 and 11 the two profiles are shown: over the building and over
the railroad.
It can be seen that in some places (for example, between points 11
and 12) the obstacles still remain in the open space 3D model. It means
that an algorithm of the open space 3D model construction should be
improved. It could be done by adding points to the maximal heights
values in the corners of each cell of the grid network. It will increase
twice the number of the points in the open space 3D model. Therefore the
open space 3D model will be free from any obstacles. Also, it should be
noted that data on some obstacles like poles, antennas, towers should be
included in the LiDAR data set additionally and, probably, manually,
because in most cases these obstacles could not be detected by the LiDAR
scanning process.
[FIGURE 7 OMITTED]
[FIGURE 8 OMITTED]
[FIGURE 9 OMITTED]
[FIGURE 10 OMITTED]
[FIGURE 11 OMITTED]
4. Conclusions
The definition of the open space 3D model was introduced. The LiDAR
full data set was suggested to use for the construction of the open
space 3D model.
The method for the development of the open space 3D model was
presented. The method uses the local interpolation algorithm and the
critical dimensions of the moving objects in the open space to create
the grid network of the open space 3D model surface.
To verify the correctness of the open space 3D model it was
suggested to investigate the differences between LiDAR surface and open
space 3D model surface. It is stated that these differences should be
with sign "+", otherwise the obstacles will still remain in
the open space 3D model.
doi:10.3846/20296991.2012.758438
References
Arrowsmith, J. R. 2006. Notes on LiDAR Interpolation. Arizona State
University. 12 p. (draft).
Detalaus erdvinio modelio sudarymas. LiDAR skrydziu ir duo-menu
kaupimo kokybes kontroles ataskaita. 2007. FIT Conseil, UAB InfoERA. 19
p.
El-Sheimy, N.; Valeo, C.; Habib, A. 2005. Digital Terrain Modeling:
Acquisition, Manipulation, and Applications. Boston, MA: Artech House.
257 p.
Fowler, R. 2001. Topographic LiDAR, in D. Maune (Ed.). Digital
Elevation Model Technologies and Applications. Maryland: American
Society for Photogrammetry and Remote Sensing, 207-236.
Lietuvos Respublikos teritorijos detalaus erdvinio modelio sudarymo
kontroles ataskaita. 2008. UAB "Aerogeodezijos institutas". 58
p.
Meng, X.; Currit, N.; Zhao, K. 2010. Ground filtering algorithms
for airborne LiDAR data: a review of critical issues, Remote Sensing 2:
833-860. http://dx.doi.org/10.3390/rs2030833
Schickler, W.; Thorpe, A. 2001. Surface estimations based on LiDAR,
in Proceedings of the ASPRS Annual Conference, April 23-27, St. Louis,
Missouri. 11 p.
Stankevicius, Z.; Kalantaite, A. 2009. LiDAR zemes pavirsiaus tasku
masyvo supaprastinimo algoritmu parametru parinkimas, Geodezija ir
kartografija [Geodesy and Cartography] 35(2): 44-49.
Sulaiman, N. S.; Majid, Z.; Setan, H. 2010. DTM Generation from
LiDAR data by using different filters in open-source software,
Geoinformation Science Journal 10(2): 89-109.
Susaki, J. 2012. Adaptive slope filtering of airborne LiDAR data in
urban areas for digital terrain model (DTM) generation, Remote Sensing
4(6): 1804-1819. http://dx.doi.org/10.3390/rs4061804
Terrain and Obstacle Data Manual, Edition 2.0. 2011. Eurocontrol.
247 p.
Yan, M.; Blaschke, T.; Liu, Y.; Wu, L. 2012. An object-based
analysis filtering algorithm for airborne laser scanning, International
Journal of Remote Sensing 33(22): 7099-7116.
http://dx.doi.org/10.1080/01431161.2012.699694
Zhang, K.; Whitman, D. 2005. Comparison of three algorithms for
filtering airborne LiDAR data, Photogrammetric Engineering & Remote
Sensing 71(3): 313-324.
Zalnierukas, A.; Cypas, K. 2006. Zemes skenavimo lazeriu is
orlaivio technologijos analize, Geodezija ir kartografija [Geodesy and
Cartography] 32(4): 101-105.
Zalnierukas, A.; Ruzgiene, B.; Kalantaite, A.; Valaitiene, R. 2009.
Miestu skenavimo LiDAR metodu tikslumo analize, Geodezija ir
kartografija [Geodesy and Cartography] 35(2): 55-60.
Ausra KALANTAITE. Head of GIS and cartography division at National
Land service under the Ministry of Agriculture Ph + 370 5 239 8446,
doctoral student of Vilnius Gediminas Technical University, Ph +370 5
274 4703, e-mail: gkk@vgtu.lt.
A graduate of Vilnius Gediminas Technical University (MSc, 1997).
Participation in projects: Land Parcel Identification System creation in
Lithuania (2002-2003), Land Parcel Identification System and Block
Database update in Lithuania (2004-2006).
Research interests: digital mapping, GIS, LiDAR.
Eimuntas Kazimieras PARSELIUNAS. Professor, Doctor. Vilnius
Gediminas Technical University. Dept of Geodesy and Cadastre, Ph +370 5
274 4703, Fax +370 5 274 4705, e-mail: eimis@vgtu.lt.
Doctor (1992). Habilitation procedure in 2008. Author of two
educational books and more than 50 scientific papers. Participated in
many intern conferences.
Research interests: graphs theory in geodesy, adjustment of
geodetic networks, geoinformation systems.
Denis ROMANOVAS. MSc student, Vilnius Gediminas Technical
University. GIS programmer at the National Centre of Remote Sensing and
Geoinformatics "GIS-Centras", Seliu g. 66, LT-08109 Vilnius,
Lithuania, Ph +370 5 2724 741, Fax +370 5 3737 723, e-mail:
d.romanovas@gis-centras.lt.
Research interests: GIS analysis and application programming.
Dominykas SLIKAS. Doctoral student. Vilnius Gediminas Technical
University. Dept of Geodesy and Cadastre, Sauletekio al. 11, LT-10223
Vilnius, Lithuania, Ph +370 5 274 4703, Fax +370 5 274 4705, e-mail:
dominykas.slikas@vgtu.lt.
MSc at VGTU (2007).
Research interests: calibration of geodetic instruments,
engineering geodesy, airborne and terrestrial laser scanning.
Ausra Kalantaite (1), Eimuntas Kazimieras Parseliunas (2), Denis
Romanovas (3), Dominykas Slikas (4)
Vilnius Gediminas Technical University, Sauletekio al. 11, LT-10223
Vilnius, Lithuania
E-mail: gi@vgtu.lt (corresponding author)
Received 4 November 2012; accepted 12 December 2012