Research On The Laser Vision System Based On DSP Image Processing Platform

With the higher requirements on quality of products, how to ensure the welding quality and enhance the efficiency becomes a challenge in front of us. As an effective approach, weld tracking based on laser vision technology plays an important role in welding quality control. However, finding a much better tracking method comes to be possible with the development of science and technology. Now the machine vision tracking system based on laser structured light, characterized as a large amount of information,high accuracy,wide adaptability,anti-interference etc, was concentrated on recently in the world. So far, some powerful foreign companies have already succeeded in developing a suit of machine vision system in the welding production, but domestic research still stays in the lab stage. Therefore, some studies on weld tracking and quality control using machine vision technology will have practical significance in the engineering.As is known to all, some strong noises such as welding arc,splash,smog,distributed light etc and the terrible environment, for instance high temperature,electromagnetic radiation and so forth, will exist in the welding stage. In this condition, catching and processing the welding images will be confronted with great difficulties. In this paper, a weld tracking system based on TMS320DM643 DSP was established for sake of tracking reliably. On one hand, some analyses were conducted about the feature of the welding spectrum and main factors affecting the welding quality from the origin. On the other hand, parameters were optimized according to studies on the sensor's fundamental optical path. Finally, a compact sensor with fine imaging character was designed. Besides, some filters were added to improve the image quality and reinforce the anti-interference ability of the sensor.As far as the image processing, this paper chose the typical V groove seam and the steps based on the former research were as follows: putting forward a new thinning method, improving the performance of median filtering which reduces the sorting time greatly, threshold segmenting that can segment the image automatically and has the ability to adjust to complex and variable welding environment, making full use of the relation of distance between line and point to extract feature points according to the characteristic of the image. At the same time, these algorithms could be utilized to process the butt seam after being modified a little. Thus, the application occasions were extended. In order to heighten real-time performance, the compiler optimization and software pipelining were adopted to achieve file-level and program-level optimizing.Much more experiments demonstrated that the welding tracking system was feasible and theses algorithms were provided with excellent stability and real-time performance, high reliability as well as strong anti-interference ability. All the above research will have useful value in real production and it can also offer references to future studies.