摘要
智慧园区中的消防火灾系统主要包括检测火警、报警、灭火和疏散4个部分,可实现对火灾的快速响应和有效处置。传统的末端探测器存在检测精度低、安装位置和刷量不能够全覆盖及无法稳定检测区域内烟雾和火焰警情的问题。文章采用先进的目标识别框架YOLOv8,训练自定义的智慧园区室内室外烟雾和火焰目标检测及分类检测结合的模型。自训练的模型实现了数字视频系统的摄像头全覆盖区域内烟雾和火焰的目标检测,弥补了消防系统的检测死角和误报、漏报情况,也提高了智慧园区中消防系统的区域检测灵敏度、准确率和覆盖率。
The fire protection system in the smart park mainly includes four parts:fire detection,alarm,extinguishing,and evacuation,which can achieve rapid response and effective disposal of fires.Traditional end detectors have problems such as low detection accuracy,insufficient installation position and brush coverage,and unstable detection of smoke and flame alarms within the detection area.The article adopts the advanced target recognition framework YOLOv8 to train a customized smart park model that combines indoor and outdoor smoke and flame target detection and classification detection.The self trained model has achieved target detection of smoke and flames within the full coverage area of the digital video system's camera,filling in the detection blind spots,false alarms,and missed alarms of the fire protection system,and improving the sensitivity,accuracy,and coverage of the fire protection system's area detection in the smart park.
作者
谢济江
孙伟
翟剑锟
XIE Jijiang;SUN Wei;ZHAI Jiankun
出处
《今日自动化》
2024年第3期153-155,共3页
Automation Today
基金
2022年度广西高校中青年教师科研基础能力提升项目(2022KY1333)。