6 CONCLUSIONS 
Figure 14:Location of some profiles on Washington DC image 
An integrated stereo matching technique for object surface 
reconstruction was described in this paper. It combines feature 
matching and signal matching in one algorithm. Two innovative 
features, i.e. the plateau and the spike point were introduced. The 
classical dynamic programming method for stereo matching was 
modified and employed in the feature matching process. In the 
signal matching part, dynamic programming for line following 
was applied by searching for the optimal elevation profiles. The 
object coordinates of matched features were used as constraints. 
The promising results in both an urban area and a rural area 
demonstrate the promising capabilities of such an approach. 
Three types of features i.e., straight lines, plateaus, and spikes 
were extracted along epipolar lines in the image. There were 120 
epipolar lines investigated. In each line, about 35-40 matched 
features were found. Points in Fig. 12 represent matched features 
in some epipolar lines. The corresponding coordinates of these 
feature points in object space were computed. 
Figure 15: Examples of some profiles from Washington DC project 
Signal matching was processed directly in the object space. 
Searching was done in the along-x and along-y directions. It 
started from profile VO to profile VI19 in the along-x search and 
profile HO to profile HI04 in the along-y search. The interval 
between each profile is 4.5 m. Fig. 13A shows the search result of 
profile H90 (Fig. 14) without using constraints. The result was 
improved after including constraints from feature points 
(Fig. 13B). Other examples of profiles are also illustrated in 
Fig. 15. The searching results of profiles H20, V60, and V84 are 
shown in Fig. 15 A, B, and C respectively. The final result was 
obtained by averaging the results from two directions. The 
extracted DC model is illustrated in Fig. 16. 
Figure 16: Surface model of Washington D.C. area 
REFERENCES 
Ayache, N. and Faverjon, B., 1987. Efficient Registration of 
Stereo Images by Matching Graph Description. International 
Journal of Computer Vision, pp. 107-131. 
Benard, M., 1984. Automatic Stereophotogrammetry: A method 
based on feature detection and dynamic programming. 
Photogrammetria, Vol. 39, pp. 169-181. 
Bethel, J.S., Mikhail, E.M., Kim, K., Marshal, J., 1998. 
Intelligent Map Understanding - Automated Feature Recognition 
and delineation. Technical Report, School of Civil Engineering, 
Purdue University, IN. 
Boldt, M., Weissa R. and Riseman E., 1989. Token-based 
Extraction of Straight lines. IEEE Transactions on Systems, Man, 
and Cybernetics, Vol. 19, November, pp. 1581-1594. 
Bums, J. B., Hanson, A. R., and Riseman, E.M., 1986. Extracting 
Straight lines. IEEE Transactions on Pattern Analysis and 
Machine Intelligence, PAM(l-8), July, pp.429-455.