Point Cloud Target Recognition And Robot Welding Trajectory Extraction Of Vertical Plates

With the implementation of the "Made in China 2025" strategy,the replacement of manual operations with industrial robots has become the core of the intelligent upgrade of the equipment manufacturing industry.Although industrial robots have been widely used in automated welding production lines,the planning of welding trajectories is generally done by manual teaching robots.If the posture of the workpiece and the operating conditions of the robot change,the robot welding trajectory will have to be manually taught again.It is difficult to achieve real-time dynamic adjustment of welding trajectory.Therefore,equipping industrial robots with non-contact laser measuring devices to quickly measure data on the surface of workpieces,identify workpiece types,positions and postures,and automatically extract welding trajectories has become a real-time adjustment to meet industrial robot welding tasks and achieve automation welding of various components.Therefore,this thesis studies the robot welding feature recognition and trajectory extraction of multi-type,variable-size vertical plates,and proposes a welding target recognition and robot welding trajectory extraction method based on point cloud region segmentation and local feature positioning.The method first uses a random sampling consistency algorithm to delete the background noise of scattered measurement data,and then achieves the removal of point cloud outliers and the data segmentation through region growing,and uses a bilateral filter to smooth the point cloud data.On this basis,by adjusting the posture of vertical plates and combining the structural information of vertical plates,a method for quickly identifying the corners of vertical plates and extracting trajectory is proposed.Furthermore,the node-segment structural feature induction and construction method of complex vertical plates is proposed,and the edge features extracted by independent calculation methods for the straight plates and curved plates are used to complete the goal of robot machining guidance of the components.In order to verify that the feature recognition method responds to the recognition speed,accuracy and stability of the measurement point cloud of multiple types of components,a robot welding vision test platform is built.Through this platform,the identification algorithm of this thesis is used to extract the key information required for welding processing.The test results show that without using the CAD design model of vertical plates,this method can eliminate noise interference,quickly locate the structuralcharacteristics of the target from the measurement point cloud,accurately extract the welding trajectory,and ensure automatic realization of the welding process.