The Research On Key Algorithms Of AOI System For PCB Inspection

During the complex production processes of PCB, a variety of defects may emerge. Compared to the failure detection during the PCB assembly process, the cost of failure detection is much higher after the product is put into the market. Thus, as a quality control method, the failure inspection of PCB is very important. As the PCB industry is developing towards ultra-thin, high density, fine pitch, low-defect, the traditional manual inspection of PCB defect cannot meet the needs of actual production and has been replaced by AOI. Therefore, the research in the improvement of AOI system has a good application prospect.In this paper, two important core algorithms of AOI system are mainly studied, and some improvements are proposed which are based on the detailed analysis of many existing deficiency of AOI systems in practical application, such as the serious dislocation in process of Image mosaic, and in the process of detection, the system has the problem of slow response, short precision and so on. The PCB image mosaic algorithm and defect detection algorithm which are studied in this paper are all verified in VC++and OpenCV platform based on AOI hardware system.The PCB image mosaic algorithm which is mainly researched in this paper is improved in two aspects. First of all, a method of decreasing every entire line image is put forward which uses the line image as the reference of registration. The experimental results indicate that this mosaic method has less data quantity and faster speed; it can also solve the serious Dislocation deviation caused by the accumulation of error in the traditional way. Secondly, the advanced surf registration algorithm is brought forward in this paper to optimize the traditional image mosaic algorithm, which can solve the problem of obviously splicing aperture caused by the opposite trend of image gray scale on edge with uneven illumination.Traditional detection algorithm is difficult to accurately identify defects if it is difficult to extract the components'characteristic information due to the impact of uneven illumination or noise factors. Also, traditional detection algorithm can not identify the defects for specific components with complex structure. In order to break through the limits of traditional detection algorithm, a new effective weighted algorithm for defect detection is proposed. Experimental results show that the proposed algorithm is robust in improving the accuracy of the system and just needs very few templates. It can also save system memory, and make detection much faster.