Research On Matching Algorithm In Binocular Stereo Vision

Three-dimensional resconstruction is performed on images which captured by twocameras,then getting3D information,this is what Binocular Stereo Vision does.Thisprocesss involves camera calibration and stereo matching. Stereo matching is the mostsignificant and difficult part,so doing some research on stereo matching is of greatesignificance for binocular stereo vision technology.This paper focused on algorithms in stereo correspondence.First,It reviews thebackground knowledge of Binocular Stereo Vision and introduces some algorithms incamera calibration.Second,we focus on feature matching and region matching algorithms,thenpresenting some modified methods. In feature matching algorithms,we have deeplystudied Harris algorithm, and have also modified it by introducing Census transformand unique constraint, which has proved that it can produce a more accurate result.Simple region matching algorithm, Rank transform-based region stereo method,feature-based region matching method and dynamic programming-based region stereomethod are researched and implemented. Inspired by Rank transform-based regionstereo method, we take advantage of another feature of Rank transform, and present amodified region matching method to save the matching time.Meanwhile it getsapproximate result and the constraint condition is the same with other algorithms.