Research on PCB Defect Detection Using Deep Convolutional Nerual Network

In view of the low robustness of the existing traditional PCB defect detection algorithms, this paper applies a PCB defect detection and recognition algorithm based on deep convolutional nerual network framework SSD(Single Shot Detector). This algorithm structure utilizes multi-scale feature maps to customise boundary boxes with different scales, and applies small convolution kernel (3*3)to predict the classification results and boundary box information. Then the detection results gracefully optimize by non-maximum suppression (NMS). Finally, in order to prove the superiority of this algorithm, this paper conducts comparative experiments. The experimental results show that the algorithm has a significant improvement in the accuracy of PCB defect detection, and the identification accuracy of PCB nodules can be as high as 94.69%. It has good applicability in the application of PCB defect detection.