Welding Anomaly Detecting Based On PCB Images

In order to evaluate manual-welding printed Circuit Board(PCB)quality automatically,this paper study the imagery anomaly detection method for PCB welding quality without using the design pattern.The main contents and results are as follows:(1)An imagery anomaly detection method was proposed to evaluate PCB welding quality.A statistic model for the image feature was built,and the position of the welding anomaly was found when the corresponding image feature didn't obey this model.(2)A PCB welding quality evaluation system using image was designed.The lighting source,the camera,the lens,the image acquisition card and the bracket of system was selected to build the hardware system.After analyzing the system characteristic and examination requirement,the software module was designed.(3)A contour-based PCB image registration algorithm was designed,which used the centroid and the minimum inertial principal axis' s orientation of contours to calculate translation and rotation parameters accurately.A device image segmentation algorithm based on quantized color space was designed,which utilized the color difference between devices and PCB.(4)A hierarchical device welding anomaly detection algorithm based on feature fusion was proposed.Firstly,all device images were divided into normal images and suspected abnormal images using perpetual hash method.Secondly,all normal images were used to build a statistic model based on kernel density estimation method,where color feature and gradient feature were fused.Thirdly,suspected abnormal images were further tested using this statistical model.Experiment results show that the proposed algorithm has both lower false alarm rate and missed detection rate.