The Study Of Point Matching In Binocular Stereo Images

The stereo image matching is the key of Close-range photogrammetry automation. According to the practical engineering requirement on restoration of 3D image from 2D pictures, on the basis of reviewing the current status of Close-range photogrammetry especially images matching techniques, this paper focuses on the study of corresponding point automatic matching by using the theories of computer vision and digital photogrammetry. Centering on the speed, accuracy, reliability and adaptability of the corresponding point matching approach in this paper, a thoroughly theoretical analysis and a large quantity of experiments are carried out and a number of achievement are made in this thesis. Additionally, a relevant complete experimental system has been established. The main works in this paper are as follows:Firstly, the theory of binocular stereo imaging is introduced that offers theoretical foundation for the following research work. Secondly, in order to improve the quality of original picture on successive disposal, some predisposal measures of digital image processing technique are adopted to dispose of object pictures in this paper. Then the implementation process of corresponding point matching is discussed. For the sake of tradeoff between processing speed and matching accuracy the approach is presented in this paper, which bases on the geometry similarity between the binocular image-pair to reduce the search area. And at the same time, revise the position of the point by the edge and non-edge regions to enhance the measuring accuracy. Finally, a software prototype system for the disposal of such matching is implemented based on above work, the developing platform of which is Visual C++ 6.0 and which applies to Windows operating system. And in the actual measuring system, a lot of corresponding points were investigated, and the good results have been acquired by this approach.The result of image matching disposal given in this paper shows the software system is efficient and feasible in the point-matching disposal in binocular stereo images, thus it could provide foundation for the further engineering applications.