Application Study On Computer Vision In Welding Mesh Robot

This paper roots in "Key technology research on welding mesh robot" which issupported by Science Program Foundation of Education Department of ShandongProvince. Steel bar 3D positioning based on computer vision in the welding meshrobot system is mainly studied to solve the problem of inaccurate steel bar loading inthe existing welding mesh equipment.Binocular vision technology is employed to position the steel bars in a 3D space.Firstly, the hardware of the binocular vision cameras is introduced. The intrinsicparameters and the radial distortion coefficients are calibrated by using plane board.Secondly, the improved Canny edge detection algorithm based on bilateralfiltering has been proposed. Bilateral filtering combines gray levels or colors based onboth their geometric closeness and their photometric similarity, and prefers near valuesto distant values in both domain and range so as to smooth images while preservingedges. 8-neighborhood gradient algorithm in the traditional Canny edge detection isreplaced by 8-neighborhood bilinear interpolation non-maximal suppression toenhance the accuracy for the edge detection.Thirdly, the improved straight line extraction algorithm based on endpointspre-positioning and the steel bar positioning algorithm based on straight line extractionhave been proposed in this paper. Harris algorithm, which omits the Gaussian filtering,is adopted to obtain pairs of feature points on the edge of the steel bar. Throughexamining the pairs of feather points, the steel bars are positioned. The experimentresults show that the above method proposed by this paper can run faster than straightline extraction algorithm based on Hough transform, and be more effective to avoid thefake straight lines.Last but not least, the epipolar geometry constraint between two images andnormalized cross-correlation algorithm are employed to determine coincidence relation between the key points' projections in two images and reconstruct the key points in the3D space. Through the reconstruction of the key points, the straight line equation,which represents steel bar under the welding mesh robot coordinate system, isobtained.The experiments show that the method proposed by this paper can obtain steel bar3D positioning information more quickly and accurately, which, makes a foundation forwelding mesh robot to capture steel bars.