PCB Electronic Component Defect Detection Method based on Improved YOLOv4 Algorithm

Abstract

In an era of information, people’s demand for electronic products is greatly increased. As an important part of electronic components, Printed Circuit Board (PCB) has a huge annual output and a variety of sizes and types. Therefore, the traditional method of manually detecting PCB defects may fail to meet the required production standards due to the high error rate. With the development of deep learning, a batch of PCB defect detection models combined with deep learning have been produced, which improves the detection efficiency. However, there are still some problems in these methods, such as low automation degree, low detection degree, and poor stability. This paper proposes an improved algorithm, based on YOLOV4, which uses PCB defect data set released by the Intelligent Robot Laboratory of Peking University, and has abundant images of different defect types, which greatly increases the reliability of the model. By analyzing the feature distribution of CSPDarkNet53 structure layer and the detection target defect size distribution in the data set used, in the data pre-processing and input stage, the image is automatically subdivided according to the average size of the callout box of the detection image, and the probability of anchor containing detection target is increased. Experimental results show that the improved YOLOV4 algorithm has a mean Average Precision (mAP) of 96.88%.