Research On Machine Vision For Arc Welding Seam Auto-tracking

With the rapid development of computer science, imagine processing, optoelectric technology and pattern recognition technology, the research and application of machine vision have been getting more and more concern and many primary achievements have been done. In this thesis, welding automation is taken as studying purpose and automatic seam tracking in the help of machine vision is attached great importance. This thesis presents a novel method for seam tracking. This method with simple structure can achieve high precision and intelligence.First, this thesis introduces the structure of system for welding. This seam tracking system includes a vision sensing system, an image processing and seam recognizing system, a fuzzy controller and an executive unit. In this system, a line structure-light is used to lighting the groove in the front of the arc and a suitable filter is selected. Good seam image can be captured by this vision sensing system. This thesis establishes the tracking model of the line structure-light sensor and its calibration technology.In the processing of image for welding seam structure-light, this thesis presents a method for quick process and recognition of image. In filtering methods for image, we use median filter to improve velocity of the system and withhold the edge of line structure-light and presents an improvement method for rapid median filter. The binarization of seam image are studied and an adaptive threshold adjusting method are worked out. A seam recognition principle based on the cross-correlation is showed in this thesis. The result of welding seam tracking validate the reliability and accuracy of this method. It's better than the methods based on edge detection. And the technology is adaptive to welding seam tracking system.In the analysis of the whole process of system, fuzzy control is introduced in automatic seam tracking, since welding process is a process with high nonlionliar, strong coupling and time-variation. This thesis presents a Fuzzy-P controller for vision tracking. Linear and curved seams are tracked by this system, and finally the errors and it's reason are analyzed through the experiment of the whole system.