基于双目视觉的局部立体匹配算法研究-电子与通信工程专业论文.docxVIP

优秀毕业论文 精品参考文献资料 Il AbstractStereo Abstract Stereo matching method is a very important part in computer vision，and due to its simple configuration and strong representative，binocular stereo matching has become a target for many researchers．Similar to the person’S eyes，binocular stereo matching use two camera which is arranged in parallel rows to film the scene around， and then compute three-dimensional structure of the scene by the images．The whole stereo matching method should include six steps：image acquisition，camera calibration， feature extraction， image matching，depth calculation and three·dimensional reconstruction．Image matching is the core content in stereo matching method．The image matching can be divided into local stereo matching algorithms and global optimization algorithms according to its optimization．Although local stereo matching algorithm is slightly lower than global optimization algorithm in accuracy,its still be widely researched because of its advantages such as simple theory，easy to implement and SO on．And with some of the algorithms proposed in recent years，some local matching algorithms Can achieve the leval of the global optimization algorithm in accuracy． In this paper,firstly,aimed at the problem that matching points’S searching area is too large，proposed a method using matching feature points to constraint the searching area．And then aimed at the feature points hard to match in the binocular images’S repeating textures region，proposed an improved algorithm，which can get precise matching points．Secondly,on the basis of AD—Census algorithm，an improved algorithm is proposed which can reflect the pixel’S speciality in image． The characteristics and innovations in this thesis are as follows： (1)aimed at the feature points hard to match in the binocular images’S repeating textures region，proposed an improved algorithm．Improved algorithm is based on SURF algorithm and neighborhood disparity constraint，then matching again．Because of the acculate