| 摘 要: 聚焦以智能安防系统为基础的吸烟行为自动检测问题,解决复杂背景影响下吸烟检测的误检漏报缺陷,提出了基于YOLOv7的人体关联实时吸烟目标检测方法。通过定位香烟和场地内人员,利用目标关联的方式解决吸烟行为的实时检测问题。实验测得该方法的mAP 值为90%,帧率为130.1 FPS,结果说明基于YOLOv7的人体关联实时吸烟目标检测方法可适用于高精度的实时目标检测系统。 |
| 关键词: 吸烟检测;目标关联;YOLOv7;目标检测 |
|
中图分类号: TP391.4
文献标识码: A
|
| 基金项目: 国家级大学生创新创业训练项目(202211258001) |
|
| Real-Time Smoking Target Detection Method for Human Body Based on YOLOv7 |
|
SUN Bing1, LI Hao1, HUANG Xinkai1, REN Changning2, ZOU Qijie1
|
(1.Dalian University, Dalian 116622, China;
2.Dalian Neusof t University of Inf ormation, Dalian 116023, China)
15762467351@163.com; lihao@s.dlu.edu.cn; 1432796737@qq.com; renchangning@neusoft.edu.cn; 55440414@qq.com
|
| Abstract: Focusing on the problem of automatic detection of smoking behavior based on intelligent security systems, this paper proposes a real-time smoking target detection method based on YOLOv7 to solve the problems of false detection and missed reporting defects of smoking detection under complex circumstances. By locating cigarettes and people in the venue, real-time detection of smoking behavior can be solved through target association. The experimental results show that the mAP value of the proposed method is 90% and the frame rate is 130. 1FPS, indicating that the real-time smoking target detection method for human body based on YOLOv7 is more suitable for high-precision real-time target detection systems. |
| Keywords: smoking detection; target association; YOLOv7; target detection |
