The Douglas-Peucker-Algorithm was used to get the feature
points of the original point set by recursively eliminating
points. Then the solution of least squares with
perpendicularity was applied to determine regularized
building boundaries. Finally the building outlines were
converted in DXF format for ISPRS test evaluation. Fig.(5)
showed the regularized building boundaries.
3.3 Result Analysis
Our methods detected 8 buildings in Vaihingen A2 area and
6 buildings in suburban area. Pixel-based evaluation and
object-based evaluation on Point-in-Polygon were used as
evaluation based on the method described in Rutzinger et al.,
(2009). The evaluation of Vaihingen A2 area was provided
by Rottensteiner et al. as result of ISPRS Benchmark on
Urban Object Classificaton and 3D Building Reconstruction,
while the evaluation of suburban area was calculated
manually by ourselves.
In pixel-based evaluation, completeness represented
Producer’s Accuracy, correctness represented User’s
Accuracy, and quality represented balances of completeness
and correctness. Fig.(6) showed the pixels classified as True
Positive (TP), False Positive (FP) and False Negative (FN)
in the pixel-based evaluation, which indicated that the
algorithm had detected the majority of the buildings, without
too many FPs performance. Tab.(2) showed the result of
pixel-based evaluation.
Table 2 Results of a pixel-based evaluation
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B3, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
Pixel Pixel Pixel
completeness correctness quality
A2 88.5% 98.9% 87.6%
Suburban 87.9% 98.3% 86.6%
Since object-based evaluation techniques are less sensitive to
errors at the building outlines and can be related to building
parameters (Rottensteiner et al., 2005), object-based evaluation
based on Point-in-Polygon Tests was also used to evaluate.
Tab.(3) showed the results of the performance evaluation based
on the PIP test.
Table 3 Results of an object-based evaluation using PIP test
Object Object Object
completeness correctness quality
A2 71.4% 100.0% 71.4%
Suburban — 70.396 100.0% 70.3%
The results of evaluation indicated that building extraction
from DSM in a complex urban area by using extended
LEGION can achieve considerable accuracy.
4. Conclusion
The study presented in this paper has used the neural oscillator
network LEGION approach and applied it to building
extraction from DSM. The extended LEGION method needs no
assumption about the underlying structures in DSM data and no
prior knowledge regarding the number of regions. This method
successfully segmented real DSM data, which shows that it
may represent a generic DSM segmentation method. Building
objects in urban and suburban areas DSM can be detected
efficiently and effectively. The boundary regularization method
takes rectangularity constraints and arc constraints into account,
and thus produces promising results.
Acknowledgements
This work is supported by National Basic Research Program of
China (2012CB957702) and the Innovation Program of
Shanghai Municipal Education Commission (No.
11YZ290) .We acknowledge the data set is a subset of the data
used for the test of digital aerial cameras carried out by the
German Association of Photogrammetry and Remote Sensing
(DGPF) ‚which referenced from M. Cramer. And we would
like to thank for The ISPRS Benchmark on Urban Object
Classification and 3D Building Reconstruction, which provided
by F. Rottensteiner, G. Sohn, M. Gerke, C. Baillard, S. Benitez,
U. Breitkopf.
References
Al-Harthy, A., and Bethel J., 2002. Heuristic filtering and 3D
feature extraction, Proceedings of ISPRS Commission III
Symposium, Graz, Austria, unpaginated CD-ROM.
Chen K., Wang D L., Liu X W., 2000. Weight Adaptation and ,
Oscillatory Correlation for Image Segmentation.IEEE
Transactions on Neural Networks,11(5):1106-1123
Cramer, M., 2010. The DGPF test on digital aerial camera
evaluation — overview and test design. Photogrammetrie
Fernerkundung Geoinformation, 2:73-82.
Li, Y., Zhu, L., et al. 2010. A refined marker controlled
watershed for building extraction from DSM and imagery.
International Journal of Remote Sensing 31(6), 1441-1452.
Lu, Y.H., Trinder, J.C., et al. 2006. Automatic building
detection using the Dempster-Shafer algorithm.
Photogrammetric Engineering and Remote Sensing 72 (4), 395-
403.
Mario, I., Chacon, M., et al. 2005. Image complexity measure :
A human criterion free app roach Proceedings of IEEE Annual
Meeting of the North American Fuzzy Information Processing
Society. Detroit, 28 June, 2412246.
Miliaresis, G. and Kokkas, N., 2007. Segmentation and object-
based classification for the extraction of the building class from
LIDAR DEMs. Computers & Geosciences, 33(8): 1076-1087.
Ren, M.W., Yangl.Y., Sun, H.,2002. Tracing boundary
contours in a binary image, Image and Vision Computing,
20(2),125-131
Rottensteiner, F., Trinder, J., et al.2005. Using the Dempster-
Shafer method for the fusion of LIDAR data and multi-spectral
images for building detection. Information Fusion 6(4),283-300.
Rutzinger, M., Rottensteiner, F., et al. 2009. A comparison of
evaluation techniques for building extraction from airborne
laser scanning. IEEE Journal of Selected Topics in Applied
Earth Observations and Remote Sensing 2(1),11-20.
Tullis, J. A., Jensen, J.R., 2003. Expert system house detection
in high spatial resolution imagery using size, shape, and context.
Geocarto International, 18(1), 5-15.
Walter, V., 2004. Object-based classification of remote sensing
data for change detection. ISPRS Journal of Photogrammetry
and Remote Sensing 58(3-4), 225-238.
Wang, D.L., Terman, D., 1995. Locally excitatory globally
inhibitory oscillator networks. IEEE Transactionson Neural
Networks, 6, 283-286.
Wan
oscil
for (
Spri
