MOTION—STEREO DOUBLE MATCHING 
RESTRICTION IN 3D MOVEMENT ANALYSIS 
ZHANG Chun-sen 
Dept of Survey, Xi’an University of Science and Technology, No.58 Yantazhonglu, Xi’an 710054,China 
-zhchunsen@yahoo.com.cn 
Commission III 
KEY WORDS: Image matching, Stereo-motion, Sequence image, 3D reconstruction, 3D motion analysis 
ABSTRACT: 
In order to realize the correspondence of arbitrariness object-side feature points of movement object in the different time in 3D 
motion analysis based on binocular sequence images, Aiming at the fact that motion and stereo matching exists simultaneously, 
motion and stereo matching of images in the process is studied. The algorithm of double restriction matching combining motion and 
stereo image matching is presented. The basic image matching based on point feature is completed by correlation coefficients and 
relaxation algorithm, and 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. And using these object-side sequence 
“image” (coordinate) accomplishes 3-D object tracking location. A set of experimental results of real data are presented, it shows 
that the accuracy of the final correspondence is about 76.5%,which can meet the requirements of the 3D motion object tracking 
location based on point feature. 
1. INTRODUCTION 
The problems of 3D (three-dimension) motion estimation from 
visual data are among the most challenging problems in 
computer vision. The way of 3D motion estimation includes the 
method of monocular sequence image and stereo (or binocular) 
sequence image, but there are distinctions in the complexity of 
computing and accuracy of computing result. When monocular 
sequence image estimation is being used, only relative moving 
information can be gotten, and it’s exist a scale factor related to 
the structure information. In order to get the structure 
information, the traditional approach is having the aid of the 
equipment of range finder. But it is difficulty to integrate the 
different equipment or eliminate the system error. This paper 
presents an approach of integrating the two cues of motion and 
stereo when two cameras that takes pictures of moving object 
repeatedly. Multi-ocular cues or binocular sequence images 
approach, compared with single sequence images, not only the 
computation is simple, but also the absolute translation in space 
(structure information) can be acquired. But it requires solutions 
of these sub-problems: the images matching problem, referred 
to stereo matching in different sequence which at the same time 
and motion (sequence) matching in same sequence which at the 
different time, the reconstruction problem, in which 3D 
information is to be reconstructed from the correspondences, 
and the features correspondence problem, referred to different 
time correspondence of feature points in object side sequence 
“image”. All of these are the most important and difficult things 
in computer vision. 
Aiming at the fact that motion and stereo matching exists 
simultaneously in 3D movement estimation based on binocular 
sequence images, the author studies the motion and stereo 
matching of images in the process. The algorithm of double 
matching restriction combining motion and stereo image 
matching is presented. Matching of images based on point 
feature is completed by correlation coefficients and relaxation 
algorithm. 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. And using these object-side sequence “image” 
(coordinate) accomplishes 3-D object tracking. An experimental 
result of real data by means of this algorithm is presented in the 
article. The result indicates that the accuracy of the final 
correspondence is about 76.5%. It can satisfy the 3D motion 
object tracking location based on point feature. 
2. DOUBLE MATCHING RESTRICTION 
2.1 Object feature points extracting and initial matching 
Extracting feature points from image is the first step of feature 
image matching. This paper uses Harris operator to extract 
arbitrariness feature points on sequence image. The experiment 
shows that this operator is simple, stable, and insensitive to 
noise, illumination and so on. It can also extract as ration. And 
the distribution of feature points extracted is rational. The aim 
of initial matching is gotten a matching candidate set T. The 
correlation coefficient method is used in this paper [2] . Namely, 
for each feature point which m, G image a, m 2 G image b. Their 
image coordinates are supposed to be (wi,vi), (w 2 ,v 2 ).If the 
difference between the coordinate of nq and the coordinate of 
m 2 dose not exceed a certain threshold, calculates correlation 
coefficient of (2n+l)x(2n+l) window which is irq, m 2 as centre. 
A pair of points is presented, if they are considered as matching 
points candidate, the correlation coefficient must be greater 
than a certain threshold. The matching candidate relation 
between a certain feature point in image and some feature