基于神经网络与加权融合的火灾火焰识别研究  

Research of fire detection algorithm based on neural network and weighted fusion

作　　者：靳琪琳[1] 段锁林[1] 

机构地区：[1]常州大学信息科学与工程学院,江苏常州213164

出　　处：《计算机工程与应用》2013年第13期156-159,共4页Computer Engineering and Applications

基　　金：常州市科技计划项目(No.CJ20110023)

摘　　要：以室内环境为应用背景,结合火灾火焰的静态和动态特征,采用了一种神经网络与加权融合的火灾火焰识别算法,对室内火灾火焰进行实时快速判决。对视频图像进行可疑运动检测,再对颜色特征进行提取,在HIS颜色空间中建立新的颜色判据,然后获取圆形度和尖峰数;研究了火焰频闪特性,将这些特征信息作为神经网络的输入端,最终利用加权融合的算法,判定区域是否为火焰。For the indoor environment as the application background,this paper designs the feature fusion flame detection algo-rithm with static and dynamic flame features, which gives the quick decision to the fire flame.Firstly, the paper makes the mo-tion detection to the video image, extracts the color feature, and creates a new color criterion in HIS space, then gets round and peak number of fire flame area.Finally, it researches on the characteristics of the flame flicker, taking all these features as neural network input, and ultimately uses of weighted fusion algorithm to determine whether it is the fire flame.

关 键 词：运动检测 火焰识别 阈值分割 特征融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]