Research Of The Stereo Matching Algorithm Based On Binocular Vision

Along with the development of computer technology,the computer is widely used in our life,and the computer technology in computer vision is gradually influence our lives.As the developing of Computer Vision,Binocular vision is widely used in areas like UAV,robot and 3D measurement.Binocular vision is the most important part in stereo matching,the whole process of stereo matching is accurate or not ultimately affect the results of 3d reconstruction.However,in the process of the stereo matching are influenced by many factors,such as the object of shade,the pictures of parallax discontinuous zone,the image texture difference is not abundant,matching accuracy is not high,and so on.This article mainly aimed at the weak texture matching problem of area and the accuracy of the matching problem,and the article proposes a kind of the global matching algorithm based on the segmentation information,ultimately through the draw parallax figure compared with real parallax verify whether it can improve the effects of matching.First of all the article introduces the principle of stereo matching and the the process of stereo matching disusing the inherent defects in the process of stereo matching.After compared the methods and the focus which scholars proposed in study over years,the article proposes a new global stereo matching algorithm based on segmentation.Secondly,the article rebuilds the global matching energy function of the algorithm,using the meanshift segmentation algorithm.The image segmentation based on segmentation block instead of the original matching primitives,using as a new matching primitives,one of the advantage is it can improve the accuracy of parallax figure,another is for the weak texture region due to the similar or similar areas in the divided blocks will be consolidated,the texture information is abundant area will be split into an area can improve the matching rate of the area;This article does some researches and discussion about the camera calibration process because the calibration accuracy affects the accuracy of the matching results.And through the Visual Studio2012 based on MFC graphical user interface design,can be operated by a user interface to realize the whole process of the stereo matching,and obtain the parallax figure.The article uses Visual Studio2012 for the experiment platform to explore the software development platform,the maximum resolution of 640*640 an external USB interface of the binocular cameras,using OpenCV standard libraries for the library function call,using the Matlab toolbox auxiliary calibration function and OpenCV function calibrating camera,programming the image acquisition,camera calibration,stereo calibration,image segmentation,stereo matching,final parallax image output,the graphical user interface performance block.Through the experimental process,finally realized the improvement of parallax,and the results show that the matching accuracy is improved,and the weak texture area also has a better matching effect.