Recognition And Classification Of Coating Film Defects On Autobody Based On Image Processing

Paint film can protect and decorate the automobile. The quality of paint film which directly affects competitiveness in the car market is the intuitive evaluation from customer. At the same time, the paint film can improve corrosion resistance and prolong automobile life. Therefore, it is important to ensure the quality of paint film on the autobody. However, it is inevitable to produce paint film defects in the automobile painting process due to coating environment and equipments. It is very important to detect defects of paint film of the automobile.The based on computer vision technology is advanced technology in many engineering fields as a detection technique. Computer vision technology gets more and more attention. In this work, the computer vision technology is used to recognize and classify paint film defects. The purpose of the paper is to change the traditional manual detection method, research effective detection methods, improve the detection accuracy and speed, and match the need of online detection.In this paper, paint film defects of the automobile are the detection objects, and the detection process is mainly divided into: image collection, image preprocessing, image contrast enhancement, characteristic parameters acquisition, and paint film defects recognition and classification. The paper established detection scheme of paint defects of the automobile, designed the related test device, and analyzed the detection experiments on common paint film defects. The top hat transform technology and bottom hat transform technology can be used to eliminate uneven illumination, which effects image quality. The application effects of top hat transform technology and bottom hat transform technology are analyzed by experiment. The based on morphology enhancement method is proposed by analyzing the applications of common image enhancement methods. The comparative experiment between the new enhancement method and the common image enhancement methods is carried out. In order to improve segmentation effect of the image, the original segmentation method based on graph theory is improved and verified by test. In order to implement defect detection automatically, the corresponding detection plug-in is developed on MATLAB software platform. The common paint film defects, including sparkling, particles, sagging, pinhole, scratches and orange, are used to verify the detection methodology. The main contents are as follows:1. The common types and the causes of the paint film defects are introduced and analyzed; the visual detection scheme based on image processing is proposed and the structure of test device and the hardware materials are determined.2. The major issues of image preprocessing are analyzed, including image cropping, uneven illumination eliminating, image denoising and image enhancement. The based on morphology image enhancement method is proposed and compared with common image enhancement methods, and the method can obviously improve the image contrast. The based on graph theory image segmentation method is improved and the application effect shows that the improved method can not only avoid the phenomenon of excessive segmentation, but also reduce the phenomenon of non-closed curves.3. The principal component analysis is used to reduce characteristic parameter dimension of defect image and improve the efficiency of operation. The parameters of classifier kernel function are determined by using genetic algorithm, particle swarm optimization algorithm(pso) and the grid division algorithm.4. The corresponding detection plug-in is developed on MATLAB software platform. In order to improve the applicable scope of the GUI programs, the GUI programs are conversed into exe executable file.5. The common paint film defects, including sparkling, particles, sagging, pinhole, scratches and orange, are used to verify the detection scheme of the paint film defects.