Point Cloud Based Three-Dimensional Reconstruction and Identification of Initial Welding Position

Abstract

Initial welding position guidance is necessary for vision-based intelligentized robotic welding. In this paper, we proposed a point cloud based approach to recognize working environment and locate welding initial position using laser stripe sensor. Calibrated laser sensor can achieve high accuracy in transforming from image coordinate system to camera coordinate system and to robot tool coordinate system with hand-eye calibration. Linear feature based image processing algorithm is developed to extract the position of laser stripe center in subpixel-level accuracy; then trajectory-queue based interpolation is implemented to convert down-sampled laser points to robot base coordinate system in real-time scanning. Identification of workpiece is implemented by segmenting workpieces from the point cloud data in the image. Before segmentation, KD-Tree based background model is constructed to filter out background points; then RANSAC fitting procedure rejects outliers and fits the correct workpiece plane model; and the welding initial position can be found along the weld seam which is the intersection of fitted planes. In verification experiment, workpiece planes and welding initial position can be correctly recognized despite the presence of abnormal noises.