基于红外图像识别技术的隧道防火预警巡检方法  

Tunnel Fire Warning Inspection Method Based on Infrared Image Recognition Technology

作　　者：母西军 铁新纳 冯博 乔涵宇 张涛 Mu Xijun;Tie Xinna;Feng Bo;Qiao Hanyu;Zhang Tao(Haitong Fire Fighting Safety Engineering Co.,Ltd.,Zhengzhou Henan 450000,China;Zhengzhou Tunnel Comprehensive Management and Maintenance Center,Zhengzhou Henan 450000,China;Henan Soling Electromechanical Equipment Installation Engineering Co.,Ltd.,Zhengzhou Henan 450000,China)

机构地区：[1]河南省海通消防安全工程有限公司,河南郑州450000 [2]郑州市城市隧道综合管理养护中心,河南郑州450000 [3]河南索凌机电设备安装工程有限公司,河南郑州450000

出　　处：《电气自动化》2023年第5期109-112,共4页Electrical Automation

基　　金：河南省发改委审批项目(郑发改投资2018【306】号)。

摘　　要：为了提高隧道火灾预防能力,基于红外图像识别技术设计了一套隧道防火预警系统。硬件部分采用红外CCD摄像头、红外光源和红外滤波光片,实现了红外图像的采集。利用树莓派4代开发板,对采集到的红外图像进行实时处理与识别,利用报警控制主机,完成隧道的防火预警。软件部分使用Python编程语言和模块化的思想,采用迁移学习和基于区域的全卷积网络算法,实现了红外图像的识别。试验结果表明,红外图像识别准确率可高达96%,系统运行12 h时,消耗的功率为100 W,红外图像识别准确率高。所设计的系统为下一步技术研究奠定了基础。In order to improve the ability of tunnel fire prevention,a tunnel fire early warning system was designed based on infrared image recognition technology.The hardware part adopts an infrared CCD camera,an infrared light source and an infrared filter to realize the acquisition of infrared images.Raspberry Pie 4 Generation Development Board was used to process and identify the collected infrared images in real time,and the alarm control host was used to complete the fire warning of tunnels.The software part used the Python programming language and the idea of modularization,adopted transfer learning and a region-based fully convolutional network algorithm,and realized the recognition of infrared images.The test results show that the accuracy of infrared image recognition can be as high as 96%.When the system runs for 12 h,the power consumption is 100 W,and the accuracy of infrared image recognition is high.The designed system lays the foundation for the next technological research.

关 键 词：红外图像识别 隧道防火 预警 迁移学习 

分 类 号：TP23[自动化与计算机技术—检测技术与自动化装置]