Research On Stitching And Multi-View Reconstruction Of Cable Surface Image

In recent years,the non-manual detection method of bridge cables has developed rapidly.The method based on image acquisition of climbing robots has attracted attention because of its low cost,convenient maintenance,simple assembly and higher safety factor than manual detection.Problems such as related image processing are directly related to the effect of detection.In this paper,the related processing of the cable surface image is studied.The computer vision-based correlation algorithm analyzes and synthesizes the data information used for detection,and improves the effectiveness and availability of the information collected by the robot.An algorithm for image preprocessing of the cable surface PE layer acquired by the robot is designed.Combined with the specific application scenarios,three conventional filtering algorithms are combined,and reasonable parameters are set to denoise the corrected image,which improves the signal-to-noise ratio of the initial data.According to the applicable conditions of the homography matrix,based on the single-frame bicubic interpolation algorithm,the weight function is modified for the specific neighborhood,and the non-scaled super-resolution reconstruction is performed on the image of the cable region with the background removed.Remove blockiness and ensure clarity.An image mosaic algorithm based on improved Laplacian fusion is proposed.Based on the characteristics of inter-frame positional relationship during cable image acquisition,combined with the analysis of six algorithms,A-SIFT(Affine-SIFT)algorithm is used to extract the characteristics of anti-affine deformation.The algorithm can guarantee the low registration degree in the image registration stage.Stability and accuracy in such cases.Aiming at the local luminosity abrupt problem caused by the conventional fusion rule of the Laplacian pyramid algorithm in the fusion stage,the bidirectional transition distance weight map is introduced to improve the fusion effect and ensure the clarity of the edge feature information.A multi-view reconstruction algorithm combining cable geometry features was designed.For the problem that the spatial structure and the actual difference between the camera parameters and the sparse point cloud are greatly different for the beam adjustment,based on the triangulation measurement error analysis,the depth initial estimation combined with the geometry is used as the optimization initial value,which is reduced.The initial data error for beamadjustment optimization improves the accuracy of sparse point cloud position estimation.The patch diffusion algorithm and Poisson surface reconstruction algorithm are used to complete the structural compact reconstruction,which improves the information dimension compared with the original data.