Calibration For Welding Robot Vision System Based On PSO Algorithm And Guidance Of Initial Welding Position

Welding is one of the most important processing technologies in manufacturing industry, and now automation and intellectualization has become the trend for welding technology. Sensor technology is one of the key technique to realize this goal. As visual sensor has the characteristics of high precision andf large amount of information, it is widely used in the recognition of the welding seam, seam tracking and welding pool detection. However, all of these mentioned above are based on accurate camera calibration.Under this background, to improve the intelligent level of the welding robot, some researches of calibration for welding robot vision system and initial vision guidance of welding position have been done in this paper.In this article, firstly we analyze advantages and disadvantages of existing camera calibration method, and decide to choose the camera model of Zhang. Z. Then transform the processing of camera calibration into a nonlinear and multivariable optimization problem. As the particle swarm optimization(PSO) algorithm proves rather good performance when dealing with similar problems, in this paper, we try to apply PSO in camera calibration.And propose a new calibration method based on an improved chaos particle swarm optimization, ICPSO.However, standard PSO algorithm is inclined to be trapped into local convergence and its particle is easy to be premature. These disadvantages will affect the precision of calibration obviously. In order to solve the problem, we combine PSO with chaotic mutation, adaptive inertia coefficient and boundary conditions. Camera calibration experiments show that the improved algorithm has obvious improvement in the speed of convergence and accuracy, this proves the necessity of doing these inproments. Results of camera calibration are very close to that of Zhang. Z method, which shows the method we propsoed is feasible.At the last, a set of mage processing algorithm is proposed, which is based image reprocessing, Canny edge detection, area filtering and hough transfom, to find the initial welding position of welding seam. Then use the “one eye-double psosition” model to complete the stereo reconstructuin of feathre points. In Vision guidance we get satisfying resualts, absolute errors in X、Y derection are within 2 mm, Z 2.5 mm, and analyzes factors influencing the guidance error...