Research On Stereo Matching For The Binocular Vision

Computer binocular vision uses two-dimensional images to reconstruct and perceive the three-dimensional world,and it is widely used in the field of military simulation,biomedicine and geographic mapping,Stereo matching is the most important in binocular vision research technology.Although the stereo matching has made a lot of research results,but lost depth information and the presence of noise and other interference in the course of imaging inevitably,while the image resolution is not high pixel,edge blur and other restrictions,there is no universal stereo matching method can obtain the accurate disparity map.Therefore,it is of great theoretical value and practical significance to research the stereo matching method with high accuracy in binocular vision.Therefore,the research work of this paper includes the following three aspects:(1)A local stereo matching algorithm based on the similarity measure model is proposed in this paper,aiming at the problem that the matching cost function and the aggregation method are difficult to select in the local stereo matching algorithm.The algorithm constructs the similarity measure model to calculate the matching cost,which has the gray difference absolute function based on chrominance information,the similarity function based on improved Census transform and the similarity function based on improved gradient cost.The model achieves the color difference and the measure of spatial structure similarity of pixels and has strong robustness to noise interference and illumination difference.In addition,this paper introduces a bilateral filtering method,which allocates weights for cost aggregation adaptively and improves matching efficiency.The experimental results show that the proposed algorithm is effective.(2)In order to obtain more accurate disparity map,this paper proposes a global stereo matching algorithm based on edge detection.The algorithm optimizes the traditional global algorithm and uses the Canny edge detection operator to improve the smoothing term of the energy equation.The similarity measure model is used to improve the data item,and a new energy equation including data item,smoothing term,occlusion term and feedback term is constructed.The similarity measure model is able to acquire more comprehensive and accurate reflection of pixel information,while eliminating the edge of the glitch phenomenon.Inaddition,iterative algorithm is used to optimize the energy equation according to the feedback term of the energy equation,which improves the matching precision of the edge region,disparity discontinuity region and whole region.The experimental results show that the proposed algorithm is accurate.(3)To further illustrate the effectiveness of this algorithm,this paper selects remote sensing image for experiments.Based on the disparity map acquired by the global stereo matching algorithm based on edge detection,the average absolute error of the eight buildings height is 6.05 meters(select building two as the reference building),and a more accurate matching result is obtained.