基于YOLOv5的地铁站场景下烟火检测系统研究

Research on a Smoke and Fire Detection System for Subway Station Scenes Based on YOLOv5

针对当前地铁车站安防监控智能化欠缺的问题,提出一种基于YOLOv5的烟火检测系统,以提高火灾检测效率。本系统以YOLOv5算法实现异常情况检测,支持对烟雾和火焰的动态检测;可通过多路视频并行推理监测,提高系统监测效率。在主干网络中加入CBAM注意力机制,基于火焰动态特性,提出基于帧差法的虚检抑制算法ODF-LOOK,降低静态虚检对输出结果的影响。经过实验测试,改进后算法识别精度提高到98.9%,语音警报和云端短信报警功能效果良好,对烟火异常情况检测有效而便利。该系统对于预防火灾事故、提高安全水平具有较高的实用意义。

In view of the lack of intelligence in the current security monitoring of subway stations,a smoke and fire detection system based on YOLOv5 is proposed to improve the efficiency of fire detection.The system uses YOLOv5 algorithm to detect the abnormal situation and supports the dynamic detection of smoke and flame.Multi-channel video parallel reasoning monitoring can improve the monitoring effici-ency of the system.CBAM attention mechanism is added to the backbone network,and based on the dynamic characteristics of the flame,the false detection suppression algorithm ODF-LOOK based on frame difference method is proposed to reduce the influence of static false detection on the output results.The experimental results show that the recognition accuracy of the improved algorithm is improved to 98.9%,and the functions of voice alarm and cloud short message alarm work fine,which is effective and convenient for detecting abnormal smoke and fire.The system has high practical significance in preventing fire accidents and im-proving safety level.