摘要
随着人工智能技术在生产生活中的应用,在建筑领域,安全行为检测技术成为研究热点。针对现有的安全帽佩戴检测算法精度低、模型鲁棒性差等问题,提出一种基于坐标注意力(Coordinate attention,CA)机制和YOLOv5结合的安全帽佩戴检测模型。优化后的模型显著提升了精度,在小目标检测方面具有良好表现。改进后CA机制与YOLOv5结合的算法的准确率达到了90.2%,相较于YOLOv5模型平均精度提升了1.6个百分点,且算法表现优于其他对比模型。经过对比实验验证,该模型检测效果更优,网络精度也更高,在复杂施工场景下依然可以保持稳定的性能和表现良好的检测效果。
With the application of artificial intelligence technology in production and life,safety behavior detection technology has become a research hotspot in the field of construction.Aiming at the existing helmet wearing detection algorithms with low accuracy and poor model robustness and other problems,a helmet wearing detection model based on the combination of Coordinate Attention(CA)mechanism and YOLOv5 was proposed in this paper.The optimized model in this paper significantly improves the accuracy and has good performance in small target detection.The accuracy of the algorithm combining the improved CA mechanism and YOLOv5 reaches 90.2%,which is 1.6 percentage points higher than the average accuracy of the YOLOv5 model,and the algorithm outperforms the other comparison models.After comparison experiments,it was verified that the model has better detection effect and higher network accuracy,and can still maintain stable performance and show good detection effect under complex construction scenarios.
作者
李双远
李天宇
LI Shuangyuan;LI Tianyu(Information Center,Jilin Institute of Chemical Technology,Jilin City 132022,China;School of Information and Control Engineering,Jilin Institute of Chemical Technology,Jilin City 132022,China)
出处
《吉林化工学院学报》
2024年第9期15-21,共7页
Journal of Jilin Institute of Chemical Technology
