7A-1-5 
5 EXPERIMENTS 
The Bishop project 
The Bishop images cover a rural area scene in Bishop, CA, USA. 
Only the neat area from two corresponding images was extracted 
and examined. Fig. 7 shows the neat area in the Bishop images. 
The size of this area is 522x396 m 2 . The southern part is flat and 
low with a prominent road. The northern part is mountainous 
with locally steep slopes, and dirt roads. The surface is 
homogeneous with some trees. The approximated minimum and 
maximum elevations are 1230 m. and 1400 m. respectively. 
Figure 10: Examples of profiles extracted from their cost matrix 
Figure 11: Surface model of Bishop area 
Edge points were the only feature examined in this rural scene 
model. In order to prevent noise, not many distinctive features 
were extracted from this scene. Fig. 8 shows the extracted edges 
overlaid on the original images. 170 epipolar lines covering the 
whole area were investigated. Due to the small number of features 
in this model, some lines did not contain any features. There were 
approximately 0-5 matched features from each pair of epipolar 
lines. The corresponding coordinates of these feature points in 
object space were computed using Eq. 9. This information would 
be used later as constraint information in the signal matching 
process. 
Searching in the signal matching process was done in four 
different directions. There were the along-x, along-y, az45, and 
az.135 directions (see Fig. 9). The interval between each profile is 
3 m.. Fig. 10A shows the searching result of a profile along a 
road (H9). Fig. 10B shows the searching results of a profile (H84) 
in the mountainous area. The final result was obtained by 
averaging the results from four directions. The Bishop model 
could then be reconstructed. It is illustrated in Fig. 11. 
Figure 12: Some feature points along the epipolar lines on 
Washington DC images 
The Washington DC project 
Images of this project cover an urban area scene in Washington 
DC, USA. As in the previous project, only the neat area from two 
corresponding images was examined. The size of this area is 
468x535.5 m" . It includes large buildings, roads, and fields. The 
approximated minimum and maximum elevations are 0 m. and 60 
m. respectively. 
Figure 13: Effect of integrating features into 
signal matching process