基于多特征匹配的视频图像火灾火焰检测方法研究  

Research of fire flame detection method in video image based on multi-features matching

作　　者：杨亚洁[1] 薛静[1] 乔鸿海[1] 刘宇[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《电子设计工程》2014年第3期186-189,193,共5页Electronic Design Engineering

基　　金：西北工业大学研究生创业种子基金(Z2012109)

摘　　要：相对传统的视频图像火灾检测方法,提出一种基于多特征匹配的视频图像火灾火焰检测方法。首先,算法采用改进的混合高斯分块模型对监控画面中的动态区域划分;再利用颜色概率统计模型提取动态区域的颜色特征,并分割出疑似火灾区域;最后,通过统计疑似火灾区域圆形度特征并判断火灾火焰是否存在。实验表明:算法对火灾火焰的检测识别准确率较高,并具有较好的实时性和抗干扰性。Compared the traditional method of fire detection in video image,a method based on the multi-features match for fire flame detection was presented.Firstly,this method employ improved Gaussian mixture block model to divide the dynamic area in the monitor screen; secondly,method employ color probability statistical model to extract the color feature of dynamic area and segment the suspected fire area; Finally,calculating the zone circularity characteristics of suspected fire area can judge the existence or not existence of fire.The experimental results show that this method can recognize the fire flame area accurately and has real-time monitoring ability and anti-interference.

关 键 词：火灾火焰检测 改进混合高斯分块模型 颜色概率统计模型 区域圆形度特征 

分 类 号：TN702[电子电信—电路与系统]