The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part B3b. Beijing 2008 
678 
8 
110.0 
130.0 
121.0 
131.0 
9 
138.0 
148.0 
148.0 
149.0 
10 
162.0 
148.0 
171.0 
149.0 
11 
157.0 
148.0 
167.0 
149.0 
12 
183.5 
135.5 
194.5 
136.5 
13 
108.0 
148.0 
118.0 
149.0 
14 
187.5 
131.5 
197.5 
132.5 
15 
172.0 
128.0 
182.0 
129.0 
Table 1 Image matching coordinate at time t ] and 
© Using the matching result of step® to do stereo matching 
with corresponding images respectively, according to the 
matching result (coordinate) in the former moment to search 
corresponding feature points of the other sequence images at 
same time with the purposefully. Double matching restriction 
will be come true. Figure 5 and6 are the stereo matching result 
of corresponding images in different sequence at time /, and t 2 
apart. Table 2 and 3 are the corresponding matching coordinate 
files apart. 
Figure5 Different sequence stereo matching at time/, 
Pixel 
X 
Y 
X' 
Y' 
1 
129.0 
150.0 
158.0 
148.0 
2 
133.0 
152.0 
162.0 
150.0 
3 
127.0 
153.0 
156.0 
151.0 
4 
160.0 
126.0 
190.0 
123.0 
5 
157.0 
126.0 
187.0 
123.0 
6 
180.0 
136.0 
211.0 
132.0 
7 
153.0 
149.0 
183.0 
146.0 
8 
110.0 
130.0 
139.0 
129.0 
9 
138.0 
148.0 
168.0 
146.0 
10 
162.0 
148.0 
193.0 
144.0 
11 
157.0 
148.0 
187.0 
145.0 
12 
108.0 
148.0 
136.0 
147.0 
13 
172.0 
128.0 
203.0 
124.0 
Table2 Stereo image matching coordinate at time /, 
7 
163.0 
150.0 
193.0 
146.0 
8 
121.0 
131.0 
151.0 
129.0 
9 
148.0 
149.0 
178.0 
146.0 
10 
171.0 
149.0 
202.0 
145.0 
11 
167.0 
149.0 
198.0 
145.0 
12. 
118.0 
149.0 
147.0 
148.0 
13 
182.0 
129.0 
213.0 
125.0 
Table3 Stereo image matching coordinate at time 
©Based on the camera calibration parameters to calculate 
object-side spatial coordinates at time /, and /-> which 
correspond with table 2 and 3. The result is shown in table 4 
and 5. 
Number 
X(dm) 
Y (dm) 
Z(dm) 
1 
1.499 
1.007 
-0.128 
2 
1.435 
1.074 
-0.078 
3 
1.516 
0.957 
-0.065 
4 
1.041 
1.626 
-0.681 
5 
1.089 
1.573 
-0.682 
6 
0.613 
1.896 
-0.514 
7 
1.075 
1.418 
-0.178 
8 
1.809 
0.713 
-0.597 
9 
1.291 
1.139 
-0.209 
10 
0.867 
1.546 
-0.249 
11 
1.016 
1.493 
-0.198 
12 
1.860 
0.642 
-0.159 
13 
0.770 
1.788 
-0.688 
Table4 Spatial coordinates of 
moving object feature points at time t ] 
Number 
X (dm) 
Y (dm) 
Z(dm) 
1 
1.266 
1.146 
-0.143 
2 
1.213 
1.222 
-0.097 
3 
1.284 
1.097 
-0.079 
4 
0.886 
1.803 
-0.664 
5 
0.935 
1.752 
-0.664 
6 
0.374 
2.022 
-0.548 
7 
0.926 
1.601 
-0.156 
8 
1.587 
0.887 
-0.605 
9 
1.152 
1.328 
-0.181 
10 
0.719 
1.698 
-0.226 
11 
0.784 
1.629 
-0.226 
12 
1.639 
0.791 
-0.166 
13 
0.602 
1.953 
-0.669 
Table5 Spatial coordinates of 
moving object feature points at time t 2 
Figureó Different sequence stereo matching at time 
Pixel 
X 
Y 
X' 
Y' 
1 
139.0 
151.0 
169.0 
149.0 
2 
143.0 
153.0 
173.0 
150.0 
3 
137.0 
154.0 
167.0 
152.0 
4 
170.0 
127.0 
200.0 
123.0 
5 
167.0 
127.0 
197.0 
123.0 
6 
190.0 
137.0 
222.0 
132.0 
Feature points’ matching is get from correlation method, 
relaxation method in the process of finishing the move object 
feature points. Those methods use the elicitation knowledge 
such as the similar gray. At the same time because there are 
certain rotate angle and camera measure errors between left and 
right images, that the same point position in left image is 
different from the right image, and it can cause perspective 
projection errors. Moreover there are little difference between 
parameter in left camera and right camera, so the related points’ 
gray are not completely the same, it cause no-adjust errors etc.