基于轻量化YOLOv5煤矿人员不安全行为识别研究
RESEARCH ON THE IDENTIFICATION OF UNSAFE BEHAVIOUR OF COAL MINE PERSONNEL BASED ON THE LIGHTWEIGHT YOLOV5
方成焰 1杨超宇1
作者信息
- 1. 安徽理工大学计算机科学与工程学院 安徽 淮南 232001
- 折叠
摘要
煤矿的安全事故大多由矿工的不安全行为导致,YOLOv5 目标检测模型可通过视频数据检测矿工的违规行为,以降低事故发生的概率.但是 YOLOv5 模型存在训练速度慢、召回率低的问题,针对该问题对模型进行修改.改进后的模型相较于原模型大大提高了召回率及准确率,测试结果优秀,可成功应用在煤矿违规行为识别过程中.
Abstract
Most safety accidents in coal mines are caused by unsafe behaviors of miners.The YOLOv5 object detection model can de-tect miners'violations through video data to reduce the probability of accidents occurring.However,YOLOv5 has issues with slow training speed and low recall rate,and the model needs to be modified to address these issues.The improved model significantly im-proves the recall and accuracy compared to the original model,and the test results are excellent,which can be successfully applied in the process of identifying violations in coal mines.
关键词
轻量化网络/煤矿安全/行为识别Key words
lightweight/coal mine safety/action recognition引用本文复制引用
出版年
2024
NETL
NSTL
万方数据