The robotic welding technology has been more and more popular due to the high demand of improving welding quality and efficiency in the industrial field. However, as for the complex seam path, the programming process of teaching playback robot which is widely applied is of low efficiency and complicated. Meanwhile, the accuracy mainly depends on operator’s experience. Moreover, the accuracy of off line programming robot is not ideal for the lack of detecting the real environment which results in difficult description of the real 3D space. Aiming at these problems, the technology of seam recognition and path planning of robot based on binocular vision has been researched which is helpful to improve robotic welding accuracy and automatic level.The robotic welding system has been designed. It includes binocular stereo vision platform which combines line structure-light to aid identification, image acquisition cards, IPC, 6 axis welding robot, welding machine, welding torch and MCU communication module. The system model is built to measure the 3D target point and locate the welding torch of robot according to camera calibration, distorted image correction, binocular vision calibration, hand-eye calibration, realization of system measurement, robotic TCP calibration, etc.The system model is built to measure the 3D target point and locate the welding torch of robot. The main steps are as follows: camera calibration, distorted image correction, realization of binocular vision calibration and measurement, hand-eye calibration, realization of system measurement and robotic TCP calibration.The recognition algorithm of the V-groove curve seam has been designed. Gauss image Pyramid, FFT and Integral graph theory are used to improve the NCC template matching algorithm, so that it can quickly find the reference map of the maximum correlation matching position. Then, the geometric classifier for two types of targets is designed to define the existence of V-groove curve seam according to the feature of NCC coefficient and the absolute difference of reference image entropy and template image entropy.Robotic path planning algorithm based on binocular vision has been designed. The maximum correlation matching position is used as the approximate initial seam position. Then the line structure-light is guided to this position. Three turning points of the line structure-light which are inflected on the V-groove curve seam are extracted. The chord direction is used to scan the seam path. The smooth curve of seam path is formed according to data processing. Moving coordinate system is established. The robot path planning is completed while the pose of welding torch coincides with the coordinate system point by point.The system software of seam recognition and path planning of robot based on binocular vision has been designed. Meanwhile, real seam tracking experiment has been done by using this system and the welding accuracy has been analyzed.The algorithm of seam recognition and path planning of robot based on binocular vision can effectively achieve identification and tracking of curve seam. The average deviation of central line between pool width and groove is 0.5mm. It meets the demand of robotic welding for curve seam. |