| In recent years,the traditional equipment manufacturing industry in China is developing towards the direction of intelligence,information and data.Welding is a very important means of processing in equipment manufacturing,welding robots gradually replace welders and become the main force of welding operation.But now the most of traditional welding robots use the teaching method,of machining error,clamping and welding in the process of the thermal deformation caused by weld path deviation occur,will not be able to real-time correction,make the welding efficiency and welding quality cannot be guaranteed.With the vigorous development of emerging technologies such as robotics,visual processing technology and information big data,it is the main direction for the future development of the welding industry to combine visual processing technology with industrial robots to realize automation,intelligence and informatization of welding.Therefore,the research of welding tracking technology combining vision sensing and robot is of great significance.The robot visual seam tracking system is built,and the host computer program with functions of data interaction,robot motion control,image processing and data calculation is designed.Using the principle of laser triangulation,combining the line structured light with the industrial camera,the three-dimensional coordinates of the weld seam are obtained through camera calibration and image processing,and the angle information of the weld is described by constructing the vector method,and the position and pose of the weld were obtained.The realtime tracking algorithm is designed,and the welding seam tracking results were verified and analyzed by experiments.First of all,the robot vision system to analyze the relationship between the coordinate system,using Zhang’s calibration method to calculate the camera parameters,using the least squares fitting out of light plane structure,using the rotating flat semi-analytical coupling method to solve the hand-eye relation matrix,determine the pixel coordinates with the mapping relationship between 3D spatial coordinates,design experiments to validate the calibration results,the calibration error is compensated.Secondly,the weld image collected by the camera is processed,through mean filtering and gray scale transformation,highlighted the key characteristics of the weld,choose interested area decrease the range of image processing,improve the processing speed,use based on the between-cluster variance method of threshold segmentation method to extract the laser stripe region,gaussian fitting method is used to extract the center line of laser stripe on both sides of weld,The intersection point of two center lines was taken as the characteristic point of the weld and its pixel coordinates were extracted.Then,for the V-shaped weld,the projection of the laser stripes on both sides of the weld is used to establish a space vector to determine the angle information of the weld,and the angle information of the weld is determined,according to the actual welding conditions,the rotation angle of the welding torch to the optimal welding posture is calculated,the Euler angle method is used to transform the angle into the coordinate axis rotation vector,the coordinates of weld feature points including three translations and three rotations are constructed.Finally,based on the off-line tracking method,the real-time tracking dynamic fitting algorithm is designed,and the off-line tracking and real-time tracking experiments are carried out respectively.The error analysis of the weld trajectory was carried out through the experimental data.The results show that the real-time tracking algorithm can meet the requirements of welding accuracy and real-time performance.The maximum error of the weld real-time tracking trajectory is 0.3324 mm,and the real-time tracking trajectory has a high accuracy. |