Research And Implementation Of Infrared Binocular Camera Calibration Method

With the development of infrared detectors in recent years,binocular stereo vision technology in the infrared band has been widely used in military reconnaissance,target tracking,video surveillance,automatic driving,field search and rescue,aerial photography and other fields,playing an increasingly important role.For building an infrared binocular stereo vision system,solving the intrinsic parameters of each camera and the extrinsic parameters between two cameras by calibration is an important step,which affects the accuracy of the subsequent tasks of the entire vision system.This paper has carried out in-depth research on corner detection and camera calibration in infrared binocular stereo vision system:The contributions in corner detection are as follows:(1)Aiming at the problems of poor imaging contrast,complex structure,cumbersome operation,low calibration accuracy and other common problems of current infrared calibration board,this paper presents a design scheme of a pure passive infrared calibration board.The infrared calibration board has the advantages of high imaging contrast,simple structure,convenient operation and high calibration accuracy;(2)Aiming at the problem that the current method of acquiring checkerboard images cannot extract feature of the edge area of lens,this paper proposes an infrared corner detection algorithm based on homography matrix,which can effectively extract the feature of edge area and make the intrinsic and extrinsic parameters of infrared binocular camera more accurate;(3)Aiming at the problem that some wrong corners and pseudo-corners detected by current corner detection algorithms under complex imaging conditions and the problem that unsuitable size of the iterative window for sub-pixel corner detection will cause the corner detection error,this paper proposes an infrared corner robustness verification algorithm,which realizes the elimination of wrong corners and adaptive infrared corners detection.The contributions in camera calibration are as follows:(1)Aiming at the problem that the corresponding points of left and right images still cannot meet the parallel equivalence requirements after epipolar correction,this paper proposes a binocular extrinsic parameters optimization algorithm based on essential matrix,which optimizes the extrinsic parameters between two cameras;(2)In order to solve the problem of unstable Z-direction values of the translation vector after the essential matrix decomposition,this paper scales and translates the right image,which optimizes parallel equivalence relationship of the corresponding corners and makes subsequent stereo matching and other computer vision tasks more accurate.The Experimental result shows that the proposed method makes the intrinsic parameters and extrinsic parameters between two cameras more accurate compared with the traditional method,which can significantly reduce the parallel equivalence error between left and right images and improve the accuracy of subsequent vision tasks.