Roland Geibel
needed most computation time. The evaluation function used was weighted subjectively. Thereon an over-segmentation
was tolerated more than an under-segmentation. Depending on the task given and the further processing expected other
orders are possible.
In general because of the low horizontal resolution of 1m on ground not all of the details of the buildings (e.g.: dormers,
small building parts) could be segmented. Just buildings which have many such structures on their roof (e.g. dormers),
which disturb the course of a plane, cause a specific problem. Fig. 9a shows such a building in an aerial image. The
segmentation result of the FOM procedure with the parameter adjustment as used in Fig. 2d is depicted in Fig. 9c. The
segments keep so small that no connected area can be recognised. Varying the parameter dy, as shown in Fig. 9d-f ,
bigger deviations are tolerated in the plane approximation and bigger segments are produced. If however the same
parameter adjustment as used in Fig. 9f were used for all objects, then small details of buildings get lost for other
objects (see Fig. 2d). In principle also the quality of the data (noise, filtering by pre-processing) could be taken into
account during the parameter adjustment.
For the man made objects considered here it was assumed that the roof areas are plane. Curved areas which appear on
other man made objects (e.g. industrial complexes, fuel depots) were not respected. Since the investigation was done
with the data of only one flight, for a validation of the results data sets of other flights at different seasons and with
different sensors are to be examined. In order to judge the problems in the reconstruction of whole buildings caused by
faulty segmentations it is attempted to also evaluate the resulting vector descriptions comparing to a ground truth in a
test bed. Further work will also deal with the segmentation of unmasked height data.
REFERENCES
Amerinex Applied Imaging, Inc. (1996) The KBVision system task reference manual, Release 3.2,
Burns JB, Hanson AR und Riseman EM (1986) Extracting straight lines. IEEE Transactions on Pattern Analysis and
Machine Intelligence (PAMI), 8(4):425-455.
Flynn PJ, Jain AK (1988) Surface classification: Hypothesis testing and parameter estimation. Proc. IEEE CS Conf.
Computer Vision and pattern recognition (CVPR 88), 261-267,
Flynn PJ, Jain AK (1991) BONSAI: 3D object recognition using constrained search. IEEE Transactions on Pattern
Analysis and Machine Intelligence (PAMI), 13(10):1066-1075.
Goldgof DB, Huang DW, Lee H (1989) A curvature-based approach to terrain recognition. IEEE Transactions on
Pattern Analysis and Machine Intelligence (PAMI), 11(10):1213-1217.
Hoffman RL, Jain AK (1987) Segmentation and classification of range images. IEEE Transactions on Pattern Analysis
and Machine Intelligence (PAMI), 9(5):608-620.
Hoover A, Jean-Baptiste G, Jiang X, Flynn PJ, Bunke H, Goldgof DB, Bowyer K, Eggert DW, Fitzgibbon A, Fisher
RB (1996) An experimental comparison of range image segmentation algorithms. IEEE Transactions on Pattern
Analysis and Machine Intelligence (PAMI), 18(7):673-689.
Huising EJ, Gomes Pereira LM (1998) Errors and accuracy estimates of laser data acquired by various laser scanning
systems for topographic applications. ISPRS Journal of photogrammetry and remote sensing, 53:245-261
Jiang XY, Bunke H (1994), Fast segmentation of range images into planar regions by scan line grouping. Machine
vision and applications, 7(2):115-122
Klausmann P, Fries S, Willersinn D, Stilla U, Thónnessen U (1999) Application-oriented assessment of computer vision
algorithms. In: Jaehne et al. (eds) Handbook of computer vision and applications, 3:133-152
Lohr U (1998) Laserscan DEM for various applications. In: Fritsch D, Englich M, Sester M (eds) GIS - Between vision
and applications. 353-356
Stilla U, Geibel R, Repasi E (1998) Generation and update of 3D-city models from maps and elevation data. In: Ikeuchi
K (ed): Proceedings of MV A'98, 42-45.
Stilla U, Soergel U, Jaeger K (2000) Generation of 3D-city models and their utilisation in image. IAPRS Vol XXXIII
Amsterdam
334 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
CF 4 0. 0O e. me
m -L gm ep ©®
-— Un ==