Aoi System Parallel Algorithm In The Application Research

In recent years, automated optical inspection systems have been widely used in theproduction process of printed circuit boards, and with the development of electronicinformation industry, People put forward higher requirements on product quality andproduction of printed circuit boards. In this trend, many manufacturers and researchershave put a lot of effort dedicated to improving the detection results of the automatedoptical inspection systems. In the actual production process, automatic opticalinspection equipment are expected to be accurate, stable, fast detection of the printedcircuit board in the printed circuit board production lines. Usually, improve imageprocessing module processing speed to improve system detection effect plays a vitalrole.In order to solve the problem of speed in the AOI, parallel processing methods toaccelerate the image processing of the automatic optical inspection system is realized inthis dissertation, and strive to significantly improve the system detection speed, and atthe same time control system production costs.The Parallel processing is realized in both Computer multiple processors andgraphics processors in this thesis. In computer Multiple processors platform, We use thetwo ways of Windows API and OpenMP on image parallel processing, which includefiltering, edge detection, binary image, morphological operation, to accelerate thesystem inspect speed. In graphics processor platform, GPU Parallel mathematicalmorphology operations based on CUDA is realized to improve the speed of imageprocessing. The experimental results show that both the computer and multi-coreprocessors platform, the processing speed of the system can be significantly improvedwhen process larger image. The results also showed that the acceleration effect of theOpenMP better than the Windows API. Image processing in parallel on the graphicsprocessor to accelerate the effect is more obvious, can reach one to two orders ofmagnitude speed upgrade. At present, there are not graphics processor be applied to theautomatic optical inspection system of parallel acceleration reports in China.