Research Of Auto Parts Defects Detection Technology Based On Machine Vision

Machine vision technology has been used widely at home and abroad due to its high efficiency, high precision, easy integration and some other significant advantages. However, although this technology is very popular in industry, agriculture, military and medical fields abroad, it is still in the stage of exploration in China, and great efforts have been made in industrial inspection, medical analysis and traffic monitoring mainly by colleges and universities, research institutes and enterprises. The applications of these products are grow rapidly. On bases of the understanding of the principle of machine vision technology, this paper accesses to a large number of machine vision application materials on automotive manufacturing and testing aspects, and designs a complete set of auto parts defect detection system based on this technology, which achieves on-line flaw detection of auto-parts. The main contents of this paper are summarized as below:The technology of machine vision theory, development situation at home and abroad, application situation in automotive manufacturing industry, as well as the software platform of this system (LabVIEW software and Vision module) are introduced.According to the preset parameters and technical requirements, a complete set of auto parts defect detection system is designed, which includes the selection and arrangements of the lighting source, optical lens, camera and photoelectric switches. The camera parameters are determined based on the relationship between the platform transmission speed, detection part rotating speed and the camera shooting speed.Intensively studied the image processing technology, which includes image smoothing, image segmentation and image analyzing. The appropriate image processing methods are determined through analyzing and comparing the effects of these methods.On bases of the camera calibration principle, calibration procedure is written applying the spatial intersection method, while the coordinates of the conversion procedure is written using spatial forward intersection method. By using the geometrical relations of the detecting parts, pixel-physical coordinate conversion is completed, which could only be achieved by binocular calibration otherwise and thus greatly reduces the cost.A set of parts defect inspection procedure is written in accordance with the modular programming ideas based on LabVIEW. The procedures are divided into image acquisition module, image processing module, results storage module, data management module and interface designation module. Taking the shaft parts for example, the effectiveness of this system is verified.By carrying simple experiments, results of defect detection, detection accuracy and speed were calculated, which verified the system feasibility. Finally, the error reasons were full analyzed.