摘要
相较于传统单点传感器检测火灾的方法,采用视频检测火灾更快、更可靠,而且可以提供火灾大小、增速等直观的图像信息。结合图像处理技术,设计一种可以同时检测火焰和烟雾的视频火灾检测方法。首先,通过提取移动区域的方式,并利用改进的火灾颜色模型提取出疑似火灾区域;然后,选出抗干扰能力强的火灾特征并对提取各特征的方法进行设计;最后,将火焰特征输入支持向量机进行火焰判别,将烟雾各特征与阈值比较后的结果输入可调节检测灵敏度的逻辑运算器进行烟雾判别,结合火焰与烟雾判别结果决定是否进行火灾预警。大量实验结果表明,该方法不仅可以检测明火,而且可以检测阴燃火,在检测准确率、误报率及反应时间上都有更好的表现。
Compared with the traditional method of detecting fires by point sensors,video fire detection holds the promise of detecting fire incidents more quickly and reliably,and can also provide intuitive image information such as fire size and growth.Combined with image processing technology,a video fire detection method is designed that can detect flame and smoke at the same time.Firstly,the suspected fire area is extracted by the moving region extraction and improved fire color model.Then,the fire characteristics with strong anti-interference ability are selected and the methods are designed for extracting each feature.Finally,the flame feature is input into the support vector machine and the result of comparing the characteristics of the smoke with the threshold is input into a logic operator that can adjust the sensitivity of smoke detection.Results of flame and smoke detection determine whether to perform the fire warning.A large number of experiments have been completed and the results show the proposed method can not only detect open flame,but also detect smoldering fire,and has better performance in detection accuracy,false alarm rate and reaction time.
作者
何爱龙
陈美娟
HE Ai-long;CHEN Mei-juan(College of Telecommunications&Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)
出处
《软件导刊》
2020年第7期198-203,共6页
Software Guide
基金
国家自然科学基金项目(61871237)
江苏省重点研发计划项目(BE2019017)。
关键词
图像处理
特征提取
多特征融合
火焰检测
火灾预警
image processing
feature extraction
multi-feature fusion
flame detection
fire warning
