The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3h. Beijing 2008 
677 
goal of these two matches are the same from the aspects of 
image processing. During the process of object-side 3D feature 
point’s correspondence, both stereo-sequence match and 
sequence-stereo match can be applied. However, the different 
matching order will has different matching effect to the final 
moving object-side 3D feature point correspondence. In actual 
operation, the adjacent images took by one camera are similar, 
because the time interval between two adjacent images is very 
short. Thus, the sequence image match from same camera is 
easy, but because its base line between viewpoints is relatively 
short, 3D reconstruction is difficult. Therefore, the estimated 
depth is not precise in the situation when noise exists (it is even 
impossible when baseline is fairly short). With this 
correspondence, there is usually a certain baseline between 
different cameras, the object-side 3D reconstruction precision is 
fairly high among stereovision because the distance between 
viewpoints is large, but stereo matching is difficult, especially 
when huge disparity and image distortion exist. The double 
match restraint namely first extracts random feature points on 
moving objects from different-time but same-sequence images, 
determining the corresponding relationship of feature points 
among the same sequence images, establish the sequence match 
of binocular image sequences. If the image sequence sample 
density is appropriate, the reliability of the feature match in 
sequence image can be guaranteed. Then according to match 
corresponding point coordinates which obtained from the 
sequence match result matching with corresponding images of 
same-time different sequence (left and right images stereo 
match). Therefore, the difficulty of random stereo match can be 
decreased to a great level through this double match restraint. 
As a result, the whole correspondence of moving object random 
three-dimensional feature points can be obtained preferably. 
The process of double matching restriction as figure 2 shows. 
The left and right images at time /,■ are denoted by /, and A 
corresponds point in images plane 7, and /’, are denoted by m, 
and m The point Mi in 3D space expressed in the coordinate 
system attached to m, and m 7? u+1 ,6, +1 is the rotation matrix 
and the translation vector describing movement object from at 
time ti to t i+l .R LR ,,t LR is the rotation matrix and the translation 
vector between the left and the right camera. The images 
matching between points m, and m ) is stereo matching, while 
the images matching between points m, and m, +1 (or between 
points m and m +1 ) is motion matching. The correspondences 
between points Mi and M i+X are features correspondences. Here 
/=3. 
Figure2 motion-stereo double matching restriction 
confirming corresponding connection of feature points in 
sequence image and finishing move matching of double 
sequence image. As long as sampling consistency of sequence 
image is suitable, we can obtain the credible feature matching 
of moving sequence image; Secondly, carry on stereo matching 
between different sequence images and the corresponding 
feature point is searched on the other image for matching 
according to the preceding match result (coordinates). In order 
to advance calculation speed and matching precision, the 
strategy which examine the dynamic moving object, and limit 
the matched object on moving object is used before moving 
matching (screen the static background of moving object). 
3. EXPERIMENTAL RESULTS OF REAL DATA 
Using the established binocular stereo vision system in doors 
for this research, obtains binocular sequence motion images of 
model car in a certain way to move (as figure3). According the 
request of 3D movement object’s location and tacking, the 
feature point’s correspondence of movement object is 
performed by motion-stereo double matching restriction of 
binocular sequence image. Combining the results of the camera 
calibration, using the triangulation process for reconstruction 
feature points of moving object 3D coordinate from binocular 
sequence images, the method guarantees the correspondence of 
arbitrary object-side feature point of the move object at 
different time. The concrete steps as follows: 
Figure3. Binocular vision system 
(T)Doing motion matching for movement object binocular 
sequence images can obtain motion matching result which in 
the same sequence at time t x and t 2 (as figured).The coordinate 
matching file as table 1. Where X, Y and X', Y' is the matching 
coordinate of the same feature point at time /j and t 2 
respectively. 
Figure4 Same sequence motion matching at time /[ and t 2 
In order to finish double matching restriction, firstly the feature 
points are extracted from moving images at different the time 
(front and rear) simultaneously, and original matching table is 
created between the features of two images, the possible 
candidate match points of a feature are found in the other image. 
Based on the above matching methods and some hypothesizes, 
considering compatibility of feature matching in a certain range, 
finding out the best feature as the final matching result, 
Pixel 
X 
Y 
X' 
Y' 
1 
129.0 
150.0 
139.0 
151.0 
2 
133.0 
152.0 
143.0 
153.0 
3 
127.0 
153.0 
137.0 
154.0 
4 
160.0 
126.0 
170.0 
127.0 
5 
157.0 
126.0 
167.0 
127.0 
6 
180.0 
136.0 
190.0 
137.0 
7 
153.0 
149.0 
163.0 
150.0