Ahmed Elaksher
The inclusion of the feature constraints within the signal matching showed favorable results in many parts of this
experiment. Figure 7 shows a cost matrix with the extracted surface profile without and with the feature constraints.
(4) (B)
Figure 7. Cost Matrices and Extracted Profiles Without and With Features
After integrating the feature information the building shapes were substantially improved and several structures that had
been completely missed were found. This example indicates the strength and the benefits of using both feature and
signal matching. This process was carried out over the entire overlap area. The object surface reconstruction is shown
in Figure 8.
Figure 8. Perspective Rendering of Extracted Surface of the Purdue Model
A preliminary evaluation of the results was carried out by comparing the surface model to that produced by a
commercial system using an automated terrain extraction. The error histograms from the two results are shown in
Figure 9.
272 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.