Research On Matching Technique In Binocular Microscopic Vision

Micro stereo vision is an important branch of machine vision, and it is the key point of micro operation system, because of it can improve the accuracy of positioning. In the stereo vision, the stereo matching is the most important and difficult part.In this thesis, a micro stereo vision system was built based on the research of micro stereo vision technology in domestic and foreign. This thesis focuses on the research of stereo matching technology in stereo vision. A stereo matching method is proposed combining the stereo matching algorithm based on feature and the matching algorithm based on region. First, the SIFT stereo matching method is adopted to matching the stereo image pair to obtain the accurate disparity of feature points. And the SIFT matching algorithm is improved. In the matching process, the constraints of the polar line and the constraint of the Homography matrix are increased to reduce the search space and improve matching accuracy. Then the stereo matching algorithm based on region is adopted for non-feature points. And the search scope and the window size are adjusted according to the disparity gradient to reduce the amount of computation, and the constraint of Homography matrix is adopted to eliminate the false matching points and improve the matching accuracy.The main work of this research is as follows. Firstly, build the hardware configuration system consisted of stereo microscope, electron microscope eyepiece, calibration board, PC and so on. Secondly, based on the optical path calibration problem and the practical requirements, a complete micro stereo vision imaging model is established. The method is used to calibrate the system. After calibration, the acquisition of the stereo images is carried out, and the images are pre-processed. Finally, the disparity map can be obtained by the proposed stereo matching algorithm.