BGA Joint Defect Automatic Detection And Recognition Technology

BGA is a kind of ball grid array widely used in printed circuit board. It is so difficult totest defect, because the solder joints of the BGA are under the chip’s surface. In this paper, wedesign a set of mechanical system based on X-ray for the circuit board defect and discuss theidentification method for automatically picking up defects of the solder joints in order torealize defection of the BGA solder joints defects.According to the principle of the X-ray detection, we design a set of mechanical systembased on X-ray for the circuit board defect. We can adjust the distance between the devices ofeach part of the system, choose the best distance, so that the effect of the best imageacquisition.In the part of image preprocessing, in order to solve the noise problem in the X-ray imagethat we use the wavelet adaptive threshold method to remove the noise from the X-ray solderjoint image. This method not only can remove the noise effectively but also can preserveimage edges better. In order to improve the contrast of the X-ray solder joint image, we usethe image enhancement method based on partial differential equations to enhance the contrastof the image. This method can enhance the contrast of the image and well preserve the edgedetails at the same time.The purpose of image segmentation is the BGA solder joints and its internal defectsegmentation. In this paper we introduce the traditional segmentation method such as thesegmentation method based on threshold, the segmentation method based on region and thesegmentation method based on edge. In this paper, we mainly studied the segmentationmethod based on variation level set. Through the experiment we can found that this methodhas strong anti-noise ability and can get the edge information and contour feature. Theexperiment shows that this method has good segmentation results.Finally, we use the method of statistical pattern recognition to identify the image of theBGA solder joint defects. According to the characteristics of BGA solder joint defects, wechoice of solder joint area, perimeter, and area ratio of the circumference and othercharacteristic parameters as feature vector. Each defect should classification, learning andtraining according to a classifier which based on a decision tree method. The experimental results show that the statistical pattern recognition method can be effective and accurateidentified the defects from the BGA solder joint image.