Curve Weld Extraction And Trajectory Planning Based On Visual Sensing

Welding robots are gradually becoming a crucial component of welding production technology as the Chinese equipment sector moves toward intelligent production.Most welding robots were formerly operated using the teach-and-train replication method,with set actions and no efficient way to adjust trajectory mistakes brought on by clamping,design flaws,and thermal deformation.This decreased the efficacy and quality of the weld seam.A major development in future technology is the integration of industrial robots,adaptive trajectory planning,and image processing for automated and intelligent welding operations.To extract weld seams and design trajectories,we thus suggest in this research a mix of visual perception technology and industrial robots,concentrating on the following:The establishment of the robot vision system and completion of the upper computer’s software development enabled the realization of the vision system calibration,image processing,capability for visualizing information exchange and controlling robot movements.The calibration approach implements the transformation matrix from the pixel locations of the feature points to the robot base coordinates using the vision system of the welding robot.In order to improve the sensitivity of the recognition,a modified Canny threshold extraction approach was implemented.The weld picture was fitted using the ellipse equation,and the center line was extracted using the mean method.To validate and analyze the weld seam extraction and robot trajectory planning findings,experiments were created and carried out.Initially,the coordinate system of each welding robot component and the internal and external camera parameters were connected through transformation.The light plane coefficients were determined in conjunction with the structural light plane calibration approach based on external factors and image processing techniques.A non-linear optimisation is employed in the hand-eye calibration section to transform the pixel coordinates to base coordinates,and an experimental demonstration is finished.The main information of the curved weld is then highlighted by reducing noise to make the feature details more distinct after the weld picture has been pre-processed using the contrast approach and different graphic pre-processing techniques.An adaptive evolutionary algorithm optimizes the joint motion and finds the best joint motion trajectory.By simulation and comparative analysis,the improved joint trajectory is smoother and results in less invalid motion.In order to calculate the robot position information at the feature spots for S-curve welds,the spatial position of the weld centerline is extracted and a spatial vector is established.The rotation angle of the tool to the welding posture at the weld feature is measured under simulations of the real welding settings.On the basis of the offline welding technique,comprehensive experiments are created,and the fitted weld trajectory is subjected to error analysis.The results of the trials demonstrate that the adaptive genetic algorithm-based trajectory optimization approach is capable of achieving the welding criteria.It contributes to the automation of the welding process.