Research On Robot Welding Guidance System Based On 3D Structured Light Vision

At present,the majority of welding robots still use the teaching playback and offline programming mode,which has low production efficiency and poor adaptability,making it difficult to meet the requirements of complex welding tasks.Therefore,utilizing vision sensors to obtain three-dimensional point cloud data of the welding workpiece is of great significance for achieving accurate identification and positioning of weld seams,as well as welding path planning.This research focuses on the robot welding guidance system based on 3D structured light vision,and the main research contents are as follows:Firstly,hardware selection and experimental platform construction were completed,and the overall system scheme was designed.The camera and robot were chosen to adopt the Eyein-Hand configuration,where the visual sensor’s primary task is to acquire three-dimensional point cloud information of the workpiece.The industrial robot serves as the executor for capturing actions and welding actions.The control system,consisting of an industrial computer and robot control cabinet,is used as the core for data processing and robot control.Secondly,due to the complex background and the presence of numerous uncontrollable interferences in the welding field,the data obtained from the vision sensor is often contaminated with various interference data,requiring preprocessing for facilitating subsequent weld seam feature extraction.By using statistical filters,outlier noise is eliminated.The welding workpiece’s main plane normal vector is fitted using the Random Sample Consensus(RANSAC)method.Subsequently,the original point cloud is directionally corrected,and finally,the workpiece background is removed using the Euclidean clustering segmentation algorithm.Then,a feature extraction method based on point cloud slicing projection is proposed to extract feature points of the weld seam.Based on the minimum oriented bounding box of the point cloud,a set of equally spaced slicing sequences is determined,and the point cloud data in each slice are projected from 3D space to a 2D image.The feature points of the weld seam are extracted by combining coarse extraction using the maximum distance method and fine extraction using the line fitting method.Finally,the weld seam valley feature is obtained by fitting the sequence of feature points,and the intersection points of the weld seam with the edge of the workpiece point cloud are obtained as the endpoints of the weld seam.Finally,for the butt joint V-groove weld seam,based on the extracted weld groove feature points from adjacent pairs of sliced data,welding posture planning was performed.Interpolation optimization was applied to the weld point sequence to obtain a welding trajectory with good smoothness.