Printed Circuit Board Optical Detection Algorithm

With the rapid development of Surface Mount Technology (SMT), it changes the traditional concept of the electronic assembly, and Creats conditions for the miniaturization of electronic products. At present, most of SMT equipment has entered a more advanced stage.However; the inspection of electronic products is still in a low level. Automated Optical Inspection (AOI) is a highly integrated visual system which integrates image processing and pattern recognition technology; it has an important practical and research value.Now, most of detection algorithms have been studied. On the basis of previous studies, the paper is focused on the detection algorithm (such as short-circuit, open-circuit, burr, etc).Firstly, a series of image pre-processing algorithms are anaylized in the paper. Lots of improvements mainly on gray-scale stretch, smoothing, sharpening, binarization, and registration etc are propsed.The adaptive algorithm of gray-scale stretch is proposed in the paper.The paper improves the contrast enhancement based on the mask smooth and gives a new algorithm of sharping.Genetic algorithm is used in this paper for the improvement of maximum between-cluster variance.Secondly, the paper uses Hough Transformation to calibrate the images and analizes the template matching, and detects the line width. On the basis of above, the paper extracts the features.Two methods are used to improve real-time: On one hand, it reduces the calculation of the template.On another hand, it reduces the times of moving.The paper improves a lot in the case of real-time in the use of genetic algorithm.To remove the the isolated points, the paper carries out a series of processing of morphology by further. The paper extracts the features of image. Last we inspect the wire space is whether to meet the requirements through the thinning.Lastly, the paper analyzes two pattern recognition algorithms for PCB defect detection. Reference method uses tree classifier to recognize the defects (such as short- circuit, open-circuit, burr, etc), while no reference method uses BP neural network to classify the defects. The paper proposes improved BP algorithm which can escape from local minimum points through Cauchy algorithm.On the basis of above research, a software system with good accuracy, real-time capability and anti-jamming performance is implemented in this paper.