The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bib. Beijing 2008 
become more robust, which makes the boundaries with weak 
signals in image can be robustly detected. 
To demonstrate the effectiveness of the proposed algorithm for 
boundary segments selection, it is compared with the result of 
line segments extraction. Figure 3(d) are the results of boundary 
segments selected from Figure 3 (c). Comparing Figure 3(c) and 
(d), the number of line segments reduces from 4141 to 779, 
3362 line segments (81%) have been removed by our selection 
algorithm. The final boundaries in Figure 3(d) are overlain with 
the original image, which shows that all important segments of 
building boundaries are kept; most of irrelative line segments 
(mainly line segments of the rooftops) have been removed. And 
the determined boundaries are detailed and have a highly 
geometric precision. 
To evaluate the quality of the boundaries quantitatively, the 
correctness of the boundaries are estimated. We check the 
distance and angle between a boundary segment and its 
corresponding segment in the original image. If the angle is 
smaller than 3 degrees and the distance is smaller than 5 pixels, 
then the boundary segment is considered as a true one; 
otherwise, it is considered as a wrong one. There are 779 
boundary segments in Figure 3(d), 709 true boundaries are 
found according to the evaluation criterion. To evaluate the 
quality of the boundaries quantitatively, the correctness of the 
boundaries are estimated. We check the distance and angle 
between a boundary segment and its corresponding segment in 
the original image. If the angle is smaller than 3 degrees and the 
distance is smaller than 5 pixels, then the boundary segment is 
considered as a true one; otherwise, it is considered as a wrong 
one. There are 779 boundary segments in Figure 3(d), 709 true 
boundaries are found according to the evaluation criterion. Only 
70 boundaries are determined wrongly by our approach. The 
correctness of the determined boundaries is 91%. By 
overlapping the final boundaries, the original image, and Lidar 
data, it can be found that almost all wrong boundaries are kept 
wrongly because of a local jump of density of Lidar data, and 
most of the wrong boundaries lie in the rooftop of building. 
4. CONCLUSIONS 
To automatically obtain detailed building boundaries with 
precise geometric position, a new approach integrated very high 
resolution imagery and Lidar data is proposed in this study. The 
approach consists of a sequence of four steps: pre-processing, 
building image generation, line segments extraction, and 
boundary segments selection. Firstly, the segmented building 
points need to be determined from raw Lidar data. Then, a 
building image is generated by processing an original image 
using a bounding rectangle and a buffering zone, which are 
derived from the segmented building points. Based on the 
building image and rough principal orientations constraints, an 
algorithm is proposed for estimating the principal orientations 
of a building, which ensures the accuracy and robustness of the 
subsequent line segments extraction. Finally, an algorithm 
based on Lidar point density analysis and Kmeans clustering is 
proposed to identify accurate boundary segments from the 
extracted line segments dynamically. All these strategies ensure 
a high correctness (91%) of the determined boundaries. And the 
boundaries are detailed and have a highly geometric precision. 
REFERENCES 
Brenner, C., 2005. Building reconstruction from images and 
laser scanning. International Journal of Applied Earth 
Observation and Geoinformation, 6(3/4), pp. 187-198. 
Chaudhuri, D. and Samal, A., 2007. A simple method for fitting 
of bounding rectangle to closed regions. Pattern Recognition, 
40(7), pp. 1981-1989. 
Haala, N. and Brenner, C., 1999. Virtual city models from laser 
altimeter and 2D map data. Photogrammetric Engineering and 
Remote Sensing, 65(79), pp. 787-795. 
Kim, T. and Muller, J. P., 1999. Development of a graph-based 
approach for building detection. Image and Vision Computing, 
17(1), pp. 3-14. 
Kraus, K. and Pfeifer, N., 1998. Determination of terrain 
models in wooded areas with airborne laser scanner data. ISPRS 
Journal of Photogrammetry and Remote Sensing, 53(4), pp. 
193-203. 
Lin, C. and Nevada, R., 1998. Building detection and 
description from a single intensity image. Computer Vision 
Image Understanding, 72(2), pp. 101-121. 
Mayer, H., 1999. Automatic object extraction from aerial 
imagery - a survey focusing on buildings. Computer Vision and 
Image Understanding, 74(2), pp. 1 
Mayer, S., 2001. Constrained optimization of building contours 
from high resolution ortho-images. In: Proceedings of 
International Conference on Image Processing. 
38-149. 
Rottensteiner, F., Trinder, J., Clode, S. and Kubik, K., 2005. 
Using the Dempster-Shafer method for the fusion of Lidar data 
and multi-spectral images for building detection. Information 
Fusion, 6(4), pp. 283-300. 
Sampath, A. and Shan, J., 2007. Building boundary tracing and 
regularization from airborne Lidar point clouds. 
photogrammetric Engineering and Remote Sensing, 73(7), pp. 
805-812. 
Sohn, G. and Dowman, I., 2004. Extraction of buildings from 
high resolution satellite data and LIDAR. In: ISPRS 20th 
Congress WGIII/4, Istanbul, Turkey. 
Suveg, I. and Vosselman, G., 2004. Reconstruction of 3D 
building models from aerial images and maps. ISPRS Journal of 
Photogrammetry and Remote Sensing, 58(3/4), pp. 202-224. 
Vosselman, G. and Kessels, P., 2005. The utilisation of airborne 
laser scanning for mapping. International Journal of Applied 
Earth Observations and Geoinformation, 6(3/4), pp. 177-186. 
Weidner, U. and Forstner, W., 1995. Towards automatic 
building extraction from high resolution digital elevation 
models. ISPRS Journal of Photogrammetry and Remote Sensing, 
50(4), pp, 38-49. 
Rau, J. Y. and Chen, L. C., 2003. Fast straight lines detection 
using hough transform with principal axis analysis. Journal of 
Photogrammetry and Remote Sensing, 8(1), pp. 15-34.Xu, F., 
Niu, X. and Li, R., 2002. Automatic recognition of civil 
infrastructure objects using hopfield neural networks. In: 
ASPRS annual conference. 
Zhang, K., Yan, J. and Chen, S. C., 2006. Automatic 
construction of building footprints from airborne Lidar data. 
IEEE Transaction on Geoscience and Remote Sensing, 44(9), 
pp. 2523-2533. 
ACKNOWLEDGEMENTS 
This work was supported by 973 Project (Grant No. 
2006CB701300) and the China / Ireland science and technology 
collaboration research fund (ICT, 2006-2007). 
697