Research On Fast Stereo Matching Method Of Stereo Vision Measurement

Stereo vision is one important branch of machine vision technology and it is applied in medical diagnosis, aeronautic scientific technology, practical measurement and other application areas. Stereo vision with its advantages plays an important role especially in the area of non-contact industrial measurement, and stereo matching which directly affects the speed and precision of measurement provides the powerful technical support for stereo vision measurement. The fast stereo matching method is the key technology for the efficient measurement. To further improve the speed of stereo vision measurement, it is necessary to carry on deeper researches on stereo matching method.On the basis of studying mathematical model of stereo vision measurement, a fast stereo matching method of stereo vision measurement is presented. Comparing with the traditional methods, it is a stereo matching method used to stereo vision measurement. When the matching precision is guaranteed, the matching speed is also improved. It is more suitable for online real-time measuring.At the purpose of simplifying the searching, stereo rectification as the preprocessing method of stereo matching is achieved by using the geometric relation of binocular stereo vision mathematical model to transform the positions of the two poles. Stereo rectification is to reduce the search range and improve the matching speeds. It can guarantee the successful implementation of stereo matching algorithm.A stereo matching algorithm based on high-quality edge feature points is presented for the application fields of stereo vision measurement. Canny operator was choosen to extract the edge feature points in left and right images according to the stereo matching algorithm of this paper. The edge of a single pixel-width line helps to ensure tbe disparity continuity. High-quality edge feature points are obtained by high reliability matching of feature points and original images. The self-adaptive window based on high-quality edge feature points is built to compute the matching costs of nor high-quality edge feature points.These are to reduce false matching rate. Then, to complete the first matching progress of nor high-quality edge feature points, the first matching disparity values are optimized by the region growing theory and credibility to obtain the better disparity map.Finally, standard test images from Middlebury image library and the image pair of actual scene were used to achieve the ideal disparity maps by the algorithm that was effective to improve the velocity and precision of matching, which verified the validity of the algorithm.This paper provided a fast stereo matching method of stereo vision measurement, which not only improves the speed of stereo matching, but also improves the precision. Therefore, this method has certain value for actual applications.