Research On Automatic Seam Tracing System Based On Active Vision Sensor

Welding robots are gradually applied in many domestic manufacturing fields because manual welding has low production efficiency and bad working environment. However, the positioning error of before step and the thermal deformation of welding process may lead to a deviation between the actual seam trajectory and the welding trajectory of robot teaching programming, and thus reduce the quality of welding. Therefore, this paper develops an automatic seam tracking system based on active vision sensor and uses a weld feature point extraction algorithm based on image processing and spatio-temporal context learning to detect the weld feature point in real time and drive the robot to track welding under intense arc, spatter, smoke, dust and other interference. This study is subsidized by the national 04 major science and technology project(NO.2015ZX04005006, development and industrial engineering of five thousand robots with completely independent intellectual property rights for machine tool automatic production) and the major science and technology project of Guangdong province(NO.2014B090920001, research on integrated technology for CNC machine tools and robots).This paper establishes a mathematical model of structured light vision 3D measurement. In order to obtain the intrinsic parameters of the camera and the laser plane equation, this paper studies the calibration methods of camera and laser plane, and carries on calibration experiments. A practical calibration algorithm for the Eye-in-Hand system consisting of the laser vision sensor and the three-axis robot is presented to obtain the transformation relationship between the camera coordinate system and the base coordinate of the moving platform. After that, the transformation relation between pixel coordinate system and base coordinate system is established.An important technical indicator of seam tracking system is the distance between the measuring point and soldering point. The smaller the distance the higher the overall performance of the system, but the interference of arc and splash on the visual sensor is more serious, and thus leads to lower measurement accuracy and even measurement failure. Therefore, a weld feature point extraction algorithm based on image processing and spatio-temporal context learning is presented. Before welding, the first feature point is extracted by image processing method. The process includes filtering, threshold segmentation, morphological correction, ROI extraction, centerline extraction and feature point extraction. When welding starts, a special mode is established using the spatio-temporal context information of the first feature point, and then, the position of the feature point in current image is estimated by the model. After the feature point has been estimated, its spatio-temporal context information is used to update the model, which is used to estimate the subsequent images' feature points. In this process, the image with the lower degree of pollution is detected by evaluating the extent of noise. The real feature point of the low-noise-image is obtained by image processing method, and its spatio-temporal context information is used to correct the model. The algorithm greatly enhances the anti-jamming capability of the system.This paper designs and builds a triaxial seam tracking experiment platform based on TwinCAT real-time control software. The control system of the the platform consists of a real-time motion control module and a PC software module. The two modules run in the kernel of the TwinCAT system and in the Windows respectively. Based on the platform, a feature-point-extraction experiment and a seam tracking experiment are carried on. Experimental results show that the seam tracking system can automatically track welding under intense arc and splash interference. The minimum distance between the measuring point and soldering point is 30 mm, the maximum measurement frequency of the sensor is 30 Hz, and the mean absolute error of tracking is less than 0.4mm. The system can meet the practical requirements of welding production.