摘要
为了快速准确地对医院内部火灾情况进行智能监测,提出一种基于多传感器融合的内控系统火灾智能监测方法。该方法一边通过YOLOv5火焰识别算法推理火焰置信度,并使用多种传感器采集建筑内部消防环境数据;另一边通过Stacking算法对BP、RF和SVM单分类器进行集成,并将处理后的消防环境数据与火焰置信度数据输入集成模型进行融合,并判定是否发生火灾。结果表明,所提基于多传感器融合的集成分类模型,在测试集上的分类准确性达到了93.46%,相较于仅使用YOLOv5火焰识别与设定阈值的识别方法提高了5.57%,误检率也有显著降低,具有较高的分类性能。与BP、RF和SVM等基分类模型相比,所提模型的F1分数也有了不同程度的提升,说明所提模型对火焰具有较好的敏感性和分类准确性。采用基于多传感器融合的集成分类方法构建的内控系统火灾智能检测模型,不仅可以实现较弱烟雾条件下的火焰识别,对所有火焰的平均响应时间仅为6.6 s,相较于常用感烟火灾探测报警器反应速度大幅提高,可以更及时地对医院内部火灾情况进行预警。
In order to monitor the fire situation in hospital quickly and accurately,an intelligent fire monitoring method of internal control system based on multi-sensor fusion is proposed.On the one hand,the YOLOv5 flame recognition algorithm was used to deduce the flame confidence,and a variety of sensors were used to collect the fire environment data inside the building.On the other side,the BP,RF,and SVM single classifiers are integrated with the Stacking algorithm,and the processed fire environment data and flame confidence data are integrated into the integrated model to determine whether a fire occurs.The results show that the proposed integrated classification model based on multi-sensor fusion has a classification accuracy of 93.46%on the test set,which is improved by 5.57%compared with the identification method using only YOLOv5 flame recognition and threshold setting.The false detection rate is also significantly reduced,and it has high classification performance.Compared with BP,RF and SVM based classification models,the F1 score of the proposed model is also improved in different degrees,indicating that the proposed model has better sensitivity to flame and classification accuracy.The intelligent fire detection model of the internal control system based on the integrated classification method based on multi-sensor fusion can not only realize the flame identification under weak smoke conditions,but also the average response time to all flames is only 6.6 s.Compared with the common smoke detection alarm,the response speed is greatly improved,and the internal fire situation of the hospital can be warned in a more timely manner.
作者
王金珠
王宇
王溢清
张明浩
黄祎
WANG Jinzhu;WANG Yu;WANG Yiqing;ZHANG Minghao;HUANG Yi(The first Affiliated Hospital of Hebei North University,zhangjiakou,Hebei 075000,China)
出处
《自动化与仪器仪表》
2024年第10期75-78,共4页
Automation & Instrumentation
关键词
火焰识别
医院内控
多传感器融合
集成学习
智能监测
flame identification
hospital internal control
multi-sensor fusion
integrated learning
intelligent monitoring
