The Research Of Binocular Stereo Area Based Matching Algorithms

Stereo vision is an important research field of computer vision inhuman-computer interaction, intelligent control, the three dimensional measurement,video surveillance, robot navigation, and unmanned aerial vehicle (uav)，and otherfields has been widely used. Stereo matching is through the image of thecorresponding point matching, it is a key step to obtain image depth information.Stereo matching is an important research direction in the ifeld of computer vision, isalso a bottleneck problem in stereo vision. Therefore the study of the stereo matchingof the binocular vision not only has important theoretical significance, also has animportant practical value.Many experts and scholars have the deep research in stereo matching problemfor many years, so far still did not find a suitable algorithm can well solve theproblems that exist in the match. For discontinuous barrier existing in stereo matching,depth and weak texture and other technical problems, solutions emerge in endlessly,but how to guarantee the precision of matching cases improve the speed of stereomatching algorithm is still a challenging research direction. Stereo matchingaccording to the local stereo matching and global optimization method can be dividedinto different stereo matching two kinds big, global algorithm accuracy is high, butthe large amount of calculation, it is difficult to meet the actual needs. Partialmatching algorithm because of the advantages of simple and easy to implement andbeen adopted by many scholars.This paper introduces the horizontal binocular stereo system in stereo matching related theory, focusing on regional stereo matching algorithm are studied, and putforward the improved algorithm, is advantageous to the late3-d reconstruction.Stereo matching algorithm based on the analysis of area ifrst, on the basis ofaiming at the shortcomings of the fixed window，and the characteristics of weakrobustness, introduced a concept of disparity gradient to choose the window size.Stereo matching algorithm is given based on the characteristics of the area, using theHarris corner has stronger noise resistance and robustness, and has strong real-timecharacteristic at the same time, in the case of weighing the real-time, make up for thedeficiency of the area algorithm.Then based on regional matching repetitive texture, the depth of discontinuousarea precision is not high question，ifrst using the variable window, according to thedifferent types of pixel adaptive to adjust the window size, other areas using theconnectivity and color similarity through scan lines produce dense disparity map.At the end of the paper according to regional matching shade problem, on thebasis of the study of adaptive weights, weight calculation according to shade the areainformation improvement. The algorithm improves the matching precision of the keepout area.