PCB defect detection method based on YOLOv5

Abstract

The invention discloses a PCB defect detection method based on YOLOv5. The method comprises: image acquisition, image selection and cutting, image preprocessing and marking, image training and image testing. The method has the beneficial effects that: through testing, the accuracy rate of PCB defect detection and identification reaches 97.27%, the recall rate reaches 97.88%, the mAP (at) 0.5 value reaches 0.9877, the mAP (at) [. 5:. 95] value reaches 0.6588, and various properties are extremely excellent. Compared with the traditional image processing, the method provided by the invention has higher accuracy and generalization ability in the aspect of PCB defect surface detection. Compared with other deep learning algorithms such as Faster-RCNN for PCB surface defect detection, the method provided by the invention has higher accuracy.

Description

technical field [0001] The invention relates to a PCB defect detection method, in particular to a PCB defect detection method based on YOLOv5, and belongs to the technical field of PCB defect detection in industrial production. Background technique [0002] With the rapid development of high-tech such as big data, artificial intelligence, 5G communication, and the Internet of Things, the development of printed circuit boards (hereinafter referred to as PCB boards) has been promoted. As the mother of electronic products, the PCB board is as important to electronic products as the human heart. PCB boards are essential components of various electronic products, and the quality of electronic products depends largely on the quality of PCB boards. Therefore, PCB boards also have a great impact on the competitiveness of various brands and businesses. The quality requirements and detection efficiency requirements for PCB bare boards are also getting higher and higher. [0003] At ...

AI Technical Summary

Problems solved by technology

[0003] At present, the detection of bare PCB boards mainly relies on traditional manual visual inspection, but with the continuous increase of circuit density, the difficulty of manual detection increases

In order to solve this problem, some enterprises have used the online testing method of bed of needles detection to detect it. Although this method has made significant progress compared with the traditional manual method, it also brings high cost of the template and large detection size. Restricted, the contact detection method is easy to cause damage to the bare PCB, and still cannot meet the requirements of mass production, high precision and high efficiency detection

Although the use of more advanced automatic optical inspection systems (Automatic Optical Inspection, AOI) with machine vision as the core can solve the above problems, but because most of the companies with complete functions and complete industries are foreign companies, the continuous introduction of foreign AOI equipment will undoubtedly increase the domestic market. The cost of the enterprise is not conducive to the development of the domestic industry, so solving this problem is a huge challenge for the PCB industry

Images