In order to improve the measurement accuracy of 3D reconstruction of welds under different poses of binocular cameras,a 3D reconstruction measurement method of pipeline welds based on stereo vision image error compensation was proposed.The improved gray wolf algorithm(IGWO)was used to optimize the generalized regression neural network(GRNN)to compensate the coordinate er-ror of the 3D reconstructed image points of the weld.The GWO algorithm was improved by chaotic mapping,nonlinear convergence factor and optimal memory preservation idea,and the simulation verification was carried out through 8 standard test functions.The IGWO was used to construct the point cloud of the weld,and the weld width,height and length were measured in three dimensions.The experimental results show that the model can accurately achieve the 3D reconstruction of the weld seam under different poses of the binocular camera.The relative error of the 3D measurement of the weld seam is within 0.9％.
关键词
立体视觉/图像误差补偿/改进灰狼优化/广义回归神经网络/焊缝三维重构测量
Key words
stereo vision/image error compensation/improved gray wolf algorithm(IGWO)/generalized regression neural network(GRNN)/3D reconstruction measurement of pipeline weld
维普
万方数据