Hiroshi Masaharu
threshold will increase the number of small regions and is not suited for modeling purposes. Therefore it is necessary
to develop a method to cope with this omission. We plan to try developing combination of other techniques such as
edge detection or contour lines of horizontal section.
4.3 Obtaining Ground Surface DEM
From the result of region segmentation, ground surface region is obtained as a complementary set of building regions.
Then it is possible to derive ground surface DEM from laser scanner data by interpolating the regions of buildings from
neighbor pixels on ground surface region using TIN (triangular irregular network) and averaging. This result is shown
as Figure 8.
This is another application of laser scanner data and region segmentation. But in order to get correct ground surface,
all the buildings have to be extracted. It can be seen that Figure 8 is not the case. Some buildings were not extracted
and are shown to be unusual rise of the ground surface. This can be used, on the other hand, to detect the omission of
building extraction.
4.4 Distinguishing Trees from Buildings
The region segmentation method cannot distinguish trees from buildings. To include trees in the 3-D model may not
be bad for scenery reproduction but it is necessary to distinguish trees from buildings to have meaningful 3-D city model.
We have not yet got results on this. We plan to use optical CCD sensor data obtained simultaneously. A new laser
sensor can measure also the backscatter signal intensity. These data will be able to be used together for classifying
trees and buildings.
5 FUTURE STUDY
As written in the previous section, the region segmentation method is still incomplete as a method to extract buildings
from laser scanner data. One of the most important study items is to develop a method to prevent omission in building
extraction. Another important study item is to distinguish trees from buildings.
We have used grid type data derived from laser scanner data up to now. Using original data expressed as a set of three-
dimensional coordinates of reflection points should be explored.
We have described a method to make a 3-D city model only from laser scanner data in this paper. But it is necessary to
combine various data sources in order to utilize the 3-D model in GIS environment. In this regard, it should be pursued
to develop matching techniques of the laser scanner data with other data such as 2-D map data. There are problems of
horizontal displacement of the scanner data due to measuring errors and changes of buildings due to time difference.
To develop methods to cope with these problems is an important issue in order to expand the usefulness of the laser
scanner data.
6 CONCLUSION
A method to extract boundary polygon data of buildings from laser scanner data was developed. The merit of this
region segmentation method compared to other methods is that polygon data can be obtained in a natural way. By
using the resulting polygon data, 3-D city model was generated. It demonstrates 3-D modeling capability of the laser
scanner data. Although there still remains problems to be solved, this shows high potential of the laser scanner data
and of this method for the 3-D data capture of urban area.
ACKNOWLEDGEMENTS
We would like to thank Mr. Izumi Kamiya for providing the 3-D models by applying his program originally made for 3-
D model generation using laser scanner data and 2-D digital map data (Kamiya and Hasegawa, 1999).
REFERENCES
Ackermann, Friedrich, 1999. Airborne laser scanning - present status and future expectations. ISPRS Journal of
Photogrammetry & Remote Sensing, Vol.54, pp.64-67.
Douglas, D. and T. Peucker, 1973. Algorithms for the reduction of the number of points required to represent a digitized
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 561