George Vosselman
was shown that a function which minimises the probability of a classification error produces a DEM with smaller errors
than a function which tries to preserve shape characteristics in the training data.
As expected, the filter results deteriorate with an decreasing point density. The filtering of reflections on low vegetation
can not be perfect, and will always cause errors in the derived digital elevation model. Although the mean errors in the
computed DEM's were relatively small (4-6 cm for 1 point per 16 m^), the RMS error values were found to be in the
order of 20-30 cm. The true precision of the DEM is expected to be slightly worse, since the effects of the measurement
noise was not included in the computed RMS values.
In this paper points were classified solely by comparing height differences between two points. Better classifications
can be expected if other features, like height textures derived from multiple points are also used [Maas, 1999, Oude
Elberink and Maas, 2000]. In that case it should become easier to make a distinction between a ground point on a sloped
surface and a vegetation point on a horizontal surface, even though the maximum height differences between these
points and their surrounding points are the same.
Another way to improve the classification is to introduce support from an image analyst. If the morphological
characteristics of a terrain vary within an area to be processed, one could (by roughly drawing polygons on an image)
indicate areas which are more or less homogeneous. For each type of terrain one could then use a different training set
to derive the optimal filter for that terrain type. The amount of required interaction is quite low, but the filter results
could improve considerably.
ACKNOWLEDGMENTS
The laser altimetry data was kindly provided by the Survey Department of the Ministry of Transport, Public Works and
Water Management.
REFERENCES
Baltsavias, E.P., 1999. Airborne laser scanning: existing systems and firms and other resources. ISPRS Journal of
Photogrammetry and Remote Sensing, Vol. 54, No. 2-3, pp. 164-198.
Haralick, R.M., Shapiro, L.G., 1992. Computer and robot vision. Vol. 1, Addison-Wesley, Reading, MA.
Huising, E.J., Gomes Pereira, L.M., 1998. Errors and accuracy estimates of laser data acquired by various laser
scanning systems for topographic applications. ISPRS Journal of Photogrammetry and Remote Sensing, Vol. 53, No. 5,
pp. 245-261.
Kilian, J., Haala, N., Englich, M., 1996. Capture and evaluation of airborne laser scanner data. In: International
Archives of Photogrammetry and Remote Sensing, Vol. XXXI, Part B3, Vienna, pp. 383-388.
Lindenberger, J., 1993. Laser-Profilmessungen zur topographischen Geländeaufnahme. Deutsche Geodätische
Kommission, Series C, No. 400, Munich.
Maas, H.-G., 1999. The potential of height texture measures for the segmentation of airborne laser scanner data. Fourth
International Airborne Remote Sensing Conference and Exhibition / 21st Canadian Symposium on Remote Sensing,
Ottawa, Ontario, Canada, 21-24 June 1999, pp. 1-154-161.
Pfeifer, N. Kostli, A., Kraus, K., 1998. Interpolation and filtering of laser scanner data - implementation and first
results. In: International Archives of Photogrammetry and Remote Sensing, Vol. XXXII, Part 3/1, Columbus, pp 153-
150.
Oude Elberink, S., Maas, H-G., 2000. The use of anisotropic height texture measures for the segmentation of airborne
laserscanner data. In: International Archives of Photogrammetry and Remote Sensing, Vol. XXXIII, Commission III,
Amsterdam.
942 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
A]
gr
foi
Fi
pa
of
m