Research On Target Location Algorithm Of Binocular Fisheye System In Simple Distributing Station

Computer vision technology has been a hot topic since the rise of the 20 th century.People can get the objective information of things through the visual system,and get the accurate and complete information through the fine processing of computer technology.After half a century of development and improvement,binocular stereo vision in computer vision has been widely used in unmanned vehicle,medical imaging,underwater machine vision system and other fields,and has become a very important branch of computer vision.However,the traditional binocular system adopts the binocular parallel mode,which also limits the size of the public field of view angle in the binocular system to a certain extent.Moreover,in some environments requiring the acquisition of target information of wide field of view angle,fisheye camera or panoramic camera should be used instead of the traditional camera to obtain a wider field of view.In the environment of complex background,there are some problems such as occlusion of target,unclear feature and unclear fringe,which make binocular view hard to match.Aiming at the problems of traditional binocular positioning system,such as small field angle and difficult stereo matching,this paper studies stereo matching and target positioning algorithm under the condition of simple binocular positioning station.Firstly,the coordinate system and transformation relationship in computer vision are introduced.This paper introduces the basic model of binocular stereo vision,the analysis of several binocular model,based on binocular convergence model proposed binocular relative pose more flexible simple cloth station model,this paper derived the positioning principle of binocular simple cloth station model,the system is given under the 2 d image coordinates to 3d space coordinates formula of location and boundary conditions are discussed under simple cloth station model feasibility of target.Secondly,the single target determination method is introduced.The fisheye lens with large distortion is selected in this paper,which leads to the failure of corner detection of partially tilted image.The corner detection algorithm based on growth is adopted,and the distortion of fisheye image is corrected by bilinear interpolation method.Next,binocular calibration method based on singular value decomposition of intrinsic matrix and stereo correction of fish-eye image pair are introduced.In view of the nature of the simple station system in this paper to restore the 3d information of the target based on binocular angle attitude and internal parameter matrix,a biaxial stereo correction method is proposed to preserve the inclination attitude in this paper.The above algorithms are verified by experiments,and the experimental results are given.Thirdly,the classification of stereo matching algorithm and matching constraints are introduced.According to the characteristics of target positioning in the binocular simple layout station system in this paper,the pixel coordinates of target points in double view and the existence of polar constraints in stereo matching are only needed,the region-based matching algorithm is selected.Area matching method is introduced in the NCC algorithm and Census similarity measure algorithm,and the problem of center of Census algorithm relies on pixel is improved,in view of the uneven illumination and noise in the process of matching two obvious experiment environment the applicability of the algorithm to some extent this characteristic,the two kinds of weighted fusion algorithm,the simulation and experiment have been carried out to verify the above algorithm.Finally,the selection of equipment and the interface of software control for binocular fisheye simple cloth station are introduced.In the indoor environment,the positioning accuracy of the dip angle attitude transformation of the simple distribution station system is experimentally verified,and then the positioning accuracy of the outdoor environment in the wide field of view is experimentally verified,the target positioning accuracy meets the engineering requirements.