Defect detection of flexible circuit board based on convolutional neural network

Abstract

Flexible printed circuit (FPCs) is one of the important links in the manufacture of electronic products. A minor flaw in the FPC can lead to a major flaw in the final product. Therefore, it is critical to detect and locate all defects on a FPC. Although great progress has been made in FPC defect detection, traditional detection methods are still difficult to deal with complex and diverse FPC. Therefore, this paper designs a depth model that can accurately detect FPC defects from non-detection templates and defect detection image input pairs. This method uses the multi-scale pyramid hierarchy structure inherent in deep neural network (DNN) to construct multi-scale characteristics. First of all, k-means clustering is used to design reasonable anchor points. Then, the network strengthens the relationship between the feature mapping at different levels and the advantages of the underlying structure information and is suitable for the detection of minor defects. The experimental results show that the accuracy of defect detection is improved effectively.