The Research And Application Of Stereo-matching Algorithm Based On Binocular Vision

With the rapid development of computer technology and the improvement of hardware performance,stereo vision has been gradually applied to various industries,such as autonomous navigation systems for mobile robots,Virtual Reality,industrial measurement,and so on.Since it has advantages in many aspects,such as simplicity,low cost and less manual intervention.Stereo vision plays an important role in the field of computer vision.Its role is to obtain 3D geometric information of scene.Stereo matching is the core of binocular stereo vision.The accuracy of stereo matching affects the accuracy of 3D geometric information directly.Therefore,whether the problem can be solved effectively is directly related to the development of stereo vision.On the basis of all that mentioned above,the research on stereo matching in binocular vision is of great significance.With uncertainty of stereo matching,there is no perfect solution to stereo matching problem.In order to effectively solve these problems,such as occlusion,depth discontinuity and weak texture matching in images,scholars have also proposed many different methods.In this thesis,an improved stereo matching algorithm was proposed based on the traditional algorithms for occlusion and depth discontinuous regions.As for the problem of depth discontinuity in matching,weighted cross window was used to aggregate the matching cost,and strict limitations for the disparity discontinuous region inthe supporting window was made.Aiming at occlusion,two interpolation methods was applied to select suitable disparity for occlusion region.Through the research and analysis of stereo matching algorithm,the thesis proposed a improved stereo matching algorithm that combining local and global stereo matching algorithm,it was tested by standard stereo image pairs provided by middlebury authoritative test platform.Experimental results showed that the proposed method are with better performance some advantages compared with other algorithms.Finally,the application of stereo matching algorithm in obstacle detection was studied,and a simple and effective obstacle detection method was proposed.The method is no need for modeling the ground.When the robot is traveling,the image pairs could be obtained by binocular cameras.According to the algorithm,the potential obstacles in the scene can be extracted,and at the same time the distance and size information of the obstacles can be obtained to achieve the purpose of early warning.