Research On Laser Vision Weld Seam Tracking System Of Complex Component Welding Robot

Due to the disadvantages of manual welding,welding robots based on traditional teaching methods are gradually applied in the field of modern welding.In the unstructured welding environment,especially in the welding process of complex components such as ships,containers and pressure tanks,the components are prone to uncertain factors such as clamping errors and thermal deformation,which leads to a large error between the actual trajectory and teaching trajectory,and the teaching welding robot cannot correct the welding trajectory according to the change of the welding seam position,which ultimately causes problems such as low welding efficiency and poor welding quality.Therefore,thesis studies the laser vision-based welding seam detection and tracking algorithm to achieve the trajectory correction of the welding robot of complex components,and improve the welding efficiency and quality.The main contents of thesis are as follows:(1)In order to convert the two-dimensional detection results of the weld seam in pixel coordinates into the three-dimensional coordinates of the robot's base coordinate system to guide the robot to correct the trajectory,thesis analyzes the measurement principle of laser vision and the calibration algorithms such as camera calibration,hand-eye calibration and laser plane calibration involved in the tracking system are studied,and finally use the Halcon vision software package to develop the corresponding program to complete the accurate calibration of each part.(2)Aiming at the problem that it is difficult to accurately track the weld seam in real time under strong disturbances such as arc and splash,thesis proposes a deep siamese region proposal network tracking algorithm that combines the siamese network and region proposal network.In the initial frame,the traditional image processing algorithm is used to extract the weld feature points,and the tracking target template is established.In subsequent frames,the siamese sub-network extracts the features of the template frame and the frame to be detected and inputs it into the region proposal sub-network;the latter calculates the similarity between the template frame and the frame to be detected to accurately predict the position of the weld.In the online tracking phase,it is executed as a one-shot detection task.Furthermore,according to the characteristics of the small target offset during the seamtracking process,the inputted image size is reduced,and the calculation amount is further reduced.(3)In view of the safety,efficiency and welding direction constraints involved in the welding process of complex components,thesis proposes a hierarchical tracking path planning method.On the basis of using the fast expanding random tree algorithm to achieve collision-free local path planning,the improved adaptive ant colony algorithm is then used to perform global path planning to obtain the shortest length of the weld seam tracking path while meeting the direction constraints.(4)Thesis designs and builds an actual welding hardware and software platform to perform visual tracking experiments on fillet joints and V-shaped welds.Furthermore,a simulation environment is built based on the robot D-H parameters and component model parameters to conduct welding seam tracking path planning experiments.The experimental results show that the average accuracy of the weld tracking in this paper is 0.5mm,and the average frame rate is 125 fps,which can achieve robust and accurate real-time tracking of the weld under strong interference,and the planned tracking path is the shortest without collision and meeting the direction constraint Path,can effectively improve the welding efficiency and quality of complex components.