Research On Inspection Of Cars' Painting Defect Based On Computer Vision

There are always adverse factors in painting work caused by bad working conditions, painting materials and equipment, which bring a certain amount of defects of painting coat. After stoving varnish, many kinds of defects, which have different forms, influence the quality and appearance of cars'painting coat seriously. Thus the detection of painting coat defects is becoming an important working procedure during spray painting process. But by now the process is totally carried out by visual detection using human eyes, which is time-consuming and laborious, and some defects are always omitted which make the result of detection tends to be unreliable.Base on the characteristic of non-contact, fast, precise and intellectualized, computer vision system has been wildly applied in industry. A detecting system using computer vision technologies has been designed for the first time, to recognize painting defect automatically. The works mainly include research on illumination mode, segmentation algorithm, properties'pick-up and pattern recognition. Main productions in this paper are:1. According to the shape of a majority of cars'painting coats'defects and the quality of a material of film and the surface characteristic, illumination is proposed. Clear images are obtained so that analyses and disposal can be done much easier.2. Three different kinds of arithmetic are advanced to detect fuzzy edge of defects. An algorithm based on image centering and morphological operation is proved out by large quantity of experiments. This algorithm works well not only with fuzzy edge but also capable of detecting micro defects.3. A series of properties of painting coat defects is picked up for recognition, and then pattern recognition methods based on binary tree and BPNN are used to classify defects. Experimental results show that each method can classify most of defects effectively, for instance dirt, runs, poppings etc.The research and productions provide a good referenced method and experience to the detection of cars'painting defect.