Research On Seam Recognition And Tracking Technology Of Mobile Welding Robot Based On Stereo Vision

In view of the overall situation, most of the current welding robot systems belong to the1st generation’s teach and play back type or the2nd generation’s off-line programming welding robot. Those welding robots cannot adapt to the changes of the environment and working conditions. It is necessary for welding robot to realize self-regulation according to the environment changes and autonomous welding, which plays an important role in improving the intelligent level of welding robot.The essay mainly researches seam recognition and real-time tracking technologies, including stereo vision, three-dimensional reconstruction and multi-sensor data fusion technologies. Hardware platform of seam detection system of mobile welding robot based on monocular stereo vision is built. Theoretical and experimental researches are conducted around seam data extraction, three-dimensional information computing and seam tracking control technologies. Specific works are as follows:(1) Hardware platform of seam detection and tracking for mobile welding robot is built.(2) After designing the two-step experimental procedure based on Kalman filtering algorithm for camera calibration, the vision system is calibrated by five-parameter model. And combining with the hand-eye calibration model, calibration for the welding and camera coordinate system is accomplished. Experimental result shows that the calibration accuracy of this method is very high.(3) Three-dimensional information computing method based on monocular stereo vision is studied. Firstly, the weld feature points are obtained by pre-processing of the collected weld pictures, and then welding feature points are transformed into three-dimensional coordinate. This method can meet the requirements of the welding process. In addition, a new weld feature point extraction method is proposed to convert the feature point detection into finding a local extremum in the nodal accumulation space, which performances better in avoiding false feature points compared with traditional methods.(4) In order to solve the torch deviation adjustment problem in seam tracking process, multi-sensor data fusion technology is use to detect the weld location. PID control method based on double weight neural network is proposed to tracking of the weld, in which the motor used for driving Oldham motion to move the torch is the control object. The simulation results show that the controller for the weld tracking control is of high response speed and accuracy.Seam recognition and seam tracking technologies base on stereo vision are researched from theoretical and experimental aspects respectively in this essay. The experiment platform and method proposed are useful to improve the tracking accuracy, stability and intelligence of the mobile welding robot control system, which provide reliable experiment platform and theoretical basis for further research on mobile welding robot.