燕山大学学报2025，Vol.49Issue(2)：167-176,10.DOI:10.3969/j.issn.1007-791X.2025.02.009
基于改进YOLOv7-tiny的PCB缺陷检测算法
PCB defect detection algorithm based on improved YOLOv7-tiny
摘要
Abstract
In response to the issues of low detection efficiency,a large number of parameters,and the complex structure of existing PCB defect detection algorithms,an improved YOLOv7-tiny algorithm is proposed.A multi-scale capture module is designed to enhance the algorithm's ability to extract image features through multi-scale feature capture,context information fusion,and feature enhancement,addressing the problem of a single pooling operation at the CSPSPP layer masking effective information within the feature map.A global-local gated perception module is introduced,reducing the parameter count of the neck network through selective feature fusion and the combination of local and global information.Experimental results on the DeepPCB dataset show that the improved model achieves a 1.5%increase in accuracy compared to traditional models,with a reduction of 66%in parameters and 20.6%in computational workload,and a 66.3%decrease in model size.The improved algorithm demonstrates high recognition accuracy,a reduced parameter count,and lower computational requirements,providing favorable conditions for fast and accurate identification of PCB defects.关键词
PCB表面缺陷检测/YOLOv7-tiny/多尺度捕获模块/全局局部门控感知模块/轻量化Key words
PCB surface defect detection/YOLOv7-tiny/multi-scale capture module/global-local gated perception module/lightweight分类
信息技术与安全科学引用本文复制引用
侯培国,韩超明,李宁,宋涛..基于改进YOLOv7-tiny的PCB缺陷检测算法[J].燕山大学学报,2025,49(2):167-176,10.基金项目
河北省自然科学基金资助项目(F2023203005) （F2023203005）
河北省教育厅科学研究资助项目(CXY2024024) （CXY2024024）
