Research On Local Stereo Matching Algorithm In Binocular Vision Based On Improved Adaptive Support Weight

Stereo matching has always been a research hotspot and difficulty in the field of computer vision.It consists of the global stereo matching algorithm and a local stereo matching algorithm.The global algorithm has higher matching accuracy and reliability,but it has high computational complexity and low efficiency.And it is difficult to meet the requirements of practical applications.Due to its simple calculation process and fast operation,local algorithm has been widely concerned by scholars at home and abroad.However,the matching accuracy becomes the key of the local matching algorithm and it constraint for the further development of the local matching algorithm.After the adaptive support weight algorithm was proposed,the accuracy of the local algorithm was greatly improved.Based on the adaptive support weight algorithm,this paper studied the design of similarity measure function in the process of matching cost calculation and the influence of occlusion factors on the reliability of local matching results.And then,this paper proposed an improved adaptive support weight algorithm.Which have been Improved the matching accuracy of local algorithm and reduced the false matching rate.The major contributions of this article are as follow:(1)Double feature weighting local matching similarity measure function has been proposed.The weighted summation of the similarity features of the x and y directions in the grayscale image and color information features is used to calculate the matching cost.It analyzed the law of the influence of the function parameter on the performance of the function,and gived the principle of the optimal parameter selection.The performance of the method for image matching is proposed through comparative experiments.(2)For occlusion pixels in the disparity calculation will lead to a larger mis-match problem,this paper has proposed an occlusion-based partition of double weight aggregation method.It applied the left-right consistency check to achieve the division of the occlusion area and the non-occlusion area in the image and constructed a new dual-weight cost aggregation function.It can adaptively assign a smaller value to the occlusion area in the double weight calculation to reduce the impact of occlusion point on cost aggregation,which improved the matching accuracy.In order to improve the matching accuracy further,a window-adaptive disparity correction method that combined color similarity and geometric distance was proposed.The disparity value with the most occurrence in the window was selected as the correction disparity value,which effectively reduced the false matching rate.(3)An improved algorithm of adaptive support weight has been proposed.The matching cost was obtained by using the similarity measure function of double feature weighting proposed in this paper.The cost-based aggregation and the initial disparity value are obtained by the proposed method of occlusion-based partition of double weight aggregation.The initial disparity value was corrected by applying a window-adaptive disparity correction method.Experimental results erified the feasibility and accuracy of the algorithm.