Binocular Vision-Based Calibration Algorithm For Welding Robot

This paper introduced the welding trajectory tracking visual servo control system based on the binocular vision. Including hardware and software system consists; aluminum alloy welding pool binocular vision sensors manufacture; binocular vision system calibration; welding trajectory matching method based on the linear characteristics and the visual control of seam tracking method and experimental results.The system base on SAF company's automatic welding trolley mobile platforms, two dimensional Sliders as Bias adjustment and the binocular vision sensor. After binocular vision sensor calibration the system can follow the welding trajectory.Firstly, articles consult other calibration methods; propose the calibration method in this paper. The method only need planar pattern target includes radial distortion and tangential distortion. Algorithm auto select near the center point which image distortion small to solve the initial value, and then select all the points nonlinear optimization to improve the calculation accuracy. After my formula derivation, the objective function reduced to standard least square method which has formula can be directly calculated. The Algorithm only contains plus; subtract; multiply; divide prescription. Greatly simplified the calculation process, and reduction of the calculation rounding error.Secondly, we match the two pictures after welding trajectory Image Processing and edge detection. Articles proposed a line characteristics matching method which was based on a broad inflection point invariants principle and same cross-ratio principle. Compared with traditional methods, the method has the advantages of welding trajectory match. Seam plane average matching error is 1.4mm and height average matching error is 0.9mm.Finally, The binocular vision and control system for welding seam tracking were validated by experiment. the feasibility also been proved.