现代电子技术2025，Vol.48Issue(14)：43-49,7.DOI:10.16652/j.issn.1004-373x.2025.14.008
基于改进YOLOv7-tiny的钢材表面缺陷检测
Steel surface defect detection based on improved YOLOv7-tiny
摘要
Abstract
An improved algorithm based on YOLOv7-tiny is proposed to address the issues of low detection accuracy and difficulty in detecting small targets in current steel surface defect detection algorithms.A dynamic snake-shaped efficient layer aggregation network(DSELAN)is proposed and embedded into the feature extraction network,so as to enhance the model's ability to extract key features of complex defect targets.SPDConv is introduced as a down-sampling module to avoid the loss of fine-grained information for small targets,effectively solving the problem of difficult detection of small targets.In allusion to the low efficiency of large target defect detection,a large target detection layer is added to expand the model's receptive field and improve the detection accuracy of large target defects.The experimental results show that the mean average precision(mAP)of the improved YOLOv7-tiny algorithm on the NEU-DET dataset can reach 81.4％,which is 6.7％higher than the original algorithm.The detection performance is also better than other mainstream detection algorithms,and it has fewer parameters and faster detection speed,which can meet the requirements of real-time and efficiency detection of industrial steel surface defects.关键词
YOLOv7-tiny/钢材表面缺陷检测/目标检测/细粒度/动态蛇形卷积/特征提取Key words
YOLOv7-tiny/steel surface defect detection/object detection/fine grain/dynamic snake convolution/feature extraction分类
信息技术与安全科学引用本文复制引用
张瑞雪,陈琳..基于改进YOLOv7-tiny的钢材表面缺陷检测[J].现代电子技术,2025,48(14):43-49,7.基金项目
国家科技重大专项基金(2021DJ1006) （2021DJ1006）
湖北省科技示范项目(2019ZYYD016) （2019ZYYD016）
中国高校产学研创新基金(2019ITA03004) （2019ITA03004）
