基于YOLOv12的PCB缺陷检测系统研究

Research on PCB Defect Detection System Based on YOLOv12

摘要: 针对传统PCB缺陷检测方法主观性强、漏检率高、泛化能力弱的问题，本文设计并实现了一种基于YOLOv12的PCB缺陷自动化检测系统。系统采用数据层、业务层、表现层三级架构，以DeepPCB数据集为训练基础，通过YOLOv12的注意力机制与高效特征聚合网络，实现对开路、短路、鼠咬等6类典型PCB缺陷的精准识别与定位。基于PyQt5构建可视化交互界面，集成缺陷检测、台账管理、数据导出等核心功能，单张图像检测耗时 ≤ 20 ms，满足工业场景实时检测需求，为电子制造业质检智能化升级提供技术支撑。

Abstract: Aiming at the problems of strong subjectivity, high missing detection rate and weak generalization ability of traditional PCB defect detection methods, this paper designs and implements an automated PCB defect detection system based on YOLOv12. The system adopts a three-level architecture consisting of data layer, business layer and presentation layer. Based on the DeepPCB dataset for training, it realizes accurate identification and localization of 6 types of typical PCB defects such as open circuit, short circuit and rat-bite through the attention mechanism and efficient feature aggregation network of YOLOv12. A visual interactive interface is built based on PyQt5, which integrates core functions including defect detection, account management and data export. The detection time of a single image is ≤20 ms, which meets the real-time detection requirements in industrial scenarios, and provides technical support for the intelligent upgrading of quality inspection in the electronic manufacturing industry.