基于轻量卷积及跨空间学习注意力机制的安全帽佩戴检测模型

吉林大学学报(信息科学版) ›› 2025, Vol. 43 ›› Issue (5): 1014-1024.

基于轻量卷积及跨空间学习注意力机制的安全帽佩戴检测模型

吴湘宁,王梦雪,潘志鹏,方恒,蔡泽宇

中国地质大学(武汉) 计算机学院,武汉430078

收稿日期:2024-09-14 出版日期:2025-09-28 发布日期:2025-11-19
作者简介:吴湘宁(1972— ), 男, 湖南衡阳人, 中国地质大学(武汉)副教授, 硕士生导师, 主要从事机器学习研究, (Tel)86- 15342350200(E-mail)wxning@ cug. edu. cn。
基金资助:
国家自然科学基金资助项目(U21A2013); 湖北省自然科学基金资助项目 (2021CFB506)

Helmet Wearing Detection Model Based on Lightweight Convolution and Cross Spatial Learning Attention Mechanism

WU Xiangning, WANG Mengxue, PAN Zhipeng, FANG Heng, CAI Zeyu

School of Computer Science, China University of Geosciences (Wuhan), Wuhan 430078, China

Received:2024-09-14 Online:2025-09-28 Published:2025-11-19

摘要/Abstract

摘要： 为提高安全帽佩戴检测模型的效率和准确性,提出了LFE-Y8(LightConv, Focal Loss and EMA Attention- You Only Look Once version 8)模型。该模型采用 Focal Loss 损失函数, 解决了样本类别不平衡的问题; 通过 LightConv 轻量卷积优化原有模型, 提升了特征提取能力;为了更好地关注小目标,融合了跨空间学习的高效多尺度EMA(Efficient Multi scale Attention)注意力机制。实验结果表明, LFE-Y8 模型相比于改进前的 YOLOv8 模型,有效提升了安全帽佩戴检测的准确性,改进后的算法精准率提升了0.6%,召回率提升了2.1%,mAP@ 50 提升了1.2%, mAP@50-95 提升了1.5%, 证明了LFE-Y8 模型在实际应用中的有效性

关键词: YOLOv8 算法, 注意力机制, 轻量卷积, 安全帽佩戴检测

Abstract: To improve the efficiency and accuracy of the helmet wearing detection model, the LFE-Y8 (LightConv, Focal Loss and EMA Attention You Only Look Once version 8) model is proposed. This model adopts the Focal Loss function to solve the problem of imbalanced sample categories. The original model is optimized using LightConv lightweight convolution, which improves the feature extraction ability. In order to better focus on small targets, an efficient multi-scale EMA (Efficient Multi Scale Attention) attention mechanism for cross spatial learning is integrated. The experimental results show that the LFE-Y8 model effectively improves the accuracy of helmet wearing detection compared to the improved YOLOv8 model. The improved algorithm has an accuracy increase of 0. 6% and a recall increase of 2. 1%. The mAP@ 50 is improved by 1.2%, and mAP@ 50-95 is improved by 1.5%, demonstrating the effectiveness of the LFE-Y8 model in practical applications.

Key words: attention mechanism, lightweight convolution, helmet wearing test

中图分类号:

TP319

吴湘宁, 王梦雪, 潘志鹏, 方恒, 蔡泽宇. 基于轻量卷积及跨空间学习注意力机制的安全帽佩戴检测模型 [J]. 吉林大学学报(信息科学版), 2025, 43(5): 1014-1024.

WU Xiangning, WANG Mengxue, PAN Zhipeng, FANG Heng, CAI Zeyu. Helmet Wearing Detection Model Based on Lightweight Convolution and Cross Spatial Learning Attention Mechanism[J]. Journal of Jilin University (Information Science Edition), 2025, 43(5): 1014-1024.