ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision", Graz, 2002
Figure 5: Clustering results for vegetated area in the
Stuttgart data overlaid on mesh of the data. Bright points —
smooth or planar surface, gray points — vegetation points.
Dataset std. range [m] number
of clusters [%]
0«s«.05 61
Stuttgart | 0.05 < s « .10 38
0d0«s«.12 1
0«s«.05 60
Vahingen | 0.05 « s « .10 34
0.10< s<.12 6
Table 2: Accuracy estimate of the surfaces clusters
to seven points, which offers redundancy of four point in
plane fitting, and also refers to the point density and the
size of objects in the Vahingen dataset (in particular roof
faces). Results are summarized in Table 2. The quality of
the results is an indication to the potential quality of infor-
mation that can be achieved by LIDAR data. As can be
seen from Table 2 in both cases the majority of the clus-
ters had a std. smaller than 5 cm, which was the minimum
threshold that was set. In both cases a small fraction of
clusters had a std. larger then 10 cm but did not exceed 13
cm even though the upper limit was set to 15 cm. The re-
sults indicate that the cluster proposals manage to propose
natural clusters. The surface fitting accuracy of the large
clusters within all three datasets was below 5 cm. The size
of the large clusters was on the order of several hundred of
points per cluster. The majority of the clusters in the high-
accuracy category had a relatively large number of points
per cluster. There is a high similarity between the number
of points per cluster and surface quality, so in addition to
the data density the number of points has an effect on the
ability to determine the surface parameters accurately. This
realization was very evident in the Vahingen dataset, where
few of the roof faces clusters had their fitting accuracy in
the third category (10 cm < std. < 12 cm), without much
place for improvement by removing points. It was evi-
dent that these points represent a structure, as they all were
part of one roof face, so dismissing them seemed a wrong
decision. As these objects are very likely to represent a
structure in the data that due to low point density cannot
be defined more precisely, these points are considered as a
coarse representation of these objects. The std. value that
is attached to these clusters serves as an indication for that.
A- 124
5 CONCLUSIONS
The paper presented a methodology for clustering laser
data surfaces. As a first step surface categories were de-
fined; the categories present one way to interpretation of
the surface. Features that enable distinguishing among these
categories and among surfaces within each category were
defined and a way to measure them was developed. Fol-
lowing the definition of the features a method for modeling
surface texture in the data was derived, and the clustering
algorithm was established. The approach that is taken does
not require defining windows to identify surface texture in
the data and does not require limiting the data volume that
is processed. The interaction between the parameter space
and object space, and the validation phase relaxes the de-
pendency of parameters that are determined within the al-
gorithm, and makes the process more robust to the exis-
tence of errors. The results show that even with relatively
sparse datasets, structure can be identified alluding to the
generality of the algorithm.
References
Axelsson, P., 1999. Processing of laser scanner data — algo-
rithms and applications. ISPRS Journal of Photogram-
metry and Remote Sensing, 54(2-3): 138-147.
Besl, P.J., 1988. Surface in Range Image Understanding.
Perception Engeeniring. Springer-Verlag, New-York,
339 pp.
Hoover, A., Jean-Baptiste, G., Jiang, X., Flynn, PJ.,
Bunke, H., Goldgof, D., Bowyer, K., Eggert, D.,
Fitzgibbon, A.W. And Fisher, R.B., 1996. An experi-
mental comparison of range image segmentation algo-
rithms. IEEE Trans. on PAMI, 18(7): 673 -689.
Kóster K. and Spann M., 2000. MIR: an approach to ro-
bust clustering - application to range image segmenta-
tion. IEEE Trans. on PAMI, 22(5): 430-444.
Lee, I. and Schenk, T., 2001. 3D perceptual organization
of laser altimetry data. International archives of pho-
togrammetry and remote sensing, 34(3/W4): 57-65.
Maas, H.G., 1999, The potential of height texture mea-
sures for the segmentation of airborne laserscanner data.
In: Proceedings of the Fourth International Airborne
Remote Sensing Conference, Ottawa, Canada. pp. 154-
161.
Oude Elberink, S. and Maas, H.G., 2000. The use of
anisotropic height texture measures for the segment of
airborne laser scanner data. International archives of
photogrammetry and remote sensing, 33(B3/2): 678-
684.
Rousseeuw P.J. and Leroy, A. M., 1987, Robust regression
and outlier detection, Wiley & Sons. New York. 329 pp.
Vosselman, G., 2001. 3D building model reconstruc-
tion from point clouds and ground plans. Interna-
tional archives of photogrammetry and remote sensing,
34(3/W4): 37-43.
