Binocular Stereo Matching Algorithm Combining BT And Improved Brief Propagation

Stereo vision,as an important aspect of computer vision,is a hot topic of research for many scholars and is widely used in robot navigation,industrial measurement and military fields.Binocular stereo vision mimic the stereo perception process of human visual,obtain three dimensional space scene information.In a stereo vision system,the accuracy of the stereo matching directly affects the effect of3 d reconstruction,and since the scene illumination and noise interference,stereo matching becomes the most important and most difficult part of the stereo vision.There are a lot of stereo matching algorithms and some of them have achieved remarkable results.This paper divides stereo matching algorithm into different types,and discuss the advantages and disadvantages of various algorithms,and select the brief propagation algorithm of the global stereo matching algorithm for research.The advantages of the brief propagation algorithm are high matching accuracy and robustness,but at the same time there are still some problems,mainly including: poor real-time,mis-matching during the acquisition of the data items in the energy function,and smoothing items that can cause disparity holes.To address these problems,this paper investigates the data term and the smoothing term of the energy function in the brief propagation algorithm,the global energy function is optimized by combining the BT algorithm and bilateral filtering;At the same time,the traditional brief propagation algorithm node iterative update strategy is improved to reduce the time complexity The main work of this paper includes the following four aspects.(1)The BT algorithm instead of AD transform can reduce the mis-match and obtain sub-pixel level disparity to improve the matching accuracy while ensuring the continuity of subsequent disparity image to complete the depth recovery.Then the left-right consistency check method is used to remove the mis-matching points and obtain the initial reliable disparity image to provide a solid foundation for the subsequent node iterations;(2)The smoothing constraint by bilateral filtering can solve the problem that the Potts model in the traditional brief propagation algorithm is poorly matched in the weak texture region and causes the disparity hole,and an improved energy function is constructed to obtain data items by BT algorithm and smoothing items by bilateral filtering;(3)A fast convergence confidence brief propagation algorithm is adopted,which only accepts the information passed by the pixel in the adaptive size region with the pixel as the center pixel when calculating the information of any node in the image,reducing redundant calculations and improving the matching speed of the algorithm;(4)The relevant parameters in the experiment are determined through experiments,and the stereo image pairs provided in the Middle-bury evaluation platform are used for experimental tests to compare and analyze the experimental results of this paper's algorithm with other stereo matching algorithms,and the effectiveness of this paper's algorithm is analyzed in terms of both accuracy and real-time performance.The paper has 22 figures,4 tables,and 66 references.