Research Of Visual Seam Tracking System For Welding Robots

The application of welding in industry is more and more extensive.In the past,manual welding and semi-automatic welding methods commonly used in industry have being gradually replaced by welding robots due to the difficulty in ensuring welding quality or the lack of versatility.The demand of welding robots is growing because of the breakthrough of manual welding defects,improvement of production efficiency and reduction of labor costs.In this paper,the principle of structured light vision is applied to the welding robot.The welding robot is guided by non-contact vision measurements to follow the weld.A visual seam tracking system for welding robots which can realize automatic identification,accurate locating and tracking of welding seams is designed in this paper.Based on the accuracy requirements of weld tracking and the principle of structured light vision,the mechanical structure of the seam tracking sensor is designed.The seam tracking sensor can be fixed at the welding torch flexibly and the Eye-in-Hand system is formed to collect the image of the welding seam with structured light.The six-axis robot is used as the moving part of the system.The welding torch is fixed on the end of the robot and the six-axis robot can realize the flexible movement of the welding torch and the weld tracking sensor.The industrial computer is connected to the weld seam tracking sensor and the six-axis robot via Ethernet.It is the control center of the system.It can collect and process the image of the welding seam with structured light through the weld seam tracking sensor,and calculates the weld seam position in combination with the information of the six-axis robot to control the robot to achieve the tracking of the weld.In this paper,the welding spot extraction algorithm for the image of the welding seam with structured light is studied.By performing edge cut,image filtering,threshold segmentation,light bar center extraction and the algorithm for finding the intersection point of straight lines on both sides of the approximate position of welding spot by least-squares-fitting,the pixel coordinates of welding spot of the image can reach the sub-pixel level.In this paper,the calibration method of camera internal parameters of the seam tracking sensor is studied.For the problem that the 650 nm filter makes it difficult to capture the calibration image,a method of using the backlight source to enhance the brightness of the calibration grid image is proposed to realize the calibration of structured light plane equation and the hand-eye parameters of the sensor.The accuracy of hand-eye parameters of the sensor is improved by parameters justification through experimental data acquisition.The software of the visual seam tracking system is programmed,and the communication process between robot,industrial computer and camera is designed.Within the working range of the welding robot,the weld seam tracking accuracy is tested.Weld seam tracking results show that the system has high tracking accuracy on the X-axis and Y-axis and the average absolute error is within 0.5mm.The tracking error on the Z-axis is relatively large and the average error is 0.95 mm.The system can meet the requirement of V fillet weld tracking for medium plate.