中文摘要：

基于彩色监控视频图像,利用火焰影像的时间运动特性,采用混合高斯背景建模方法从监控视频图像序列中提取出运动前景像素;在RGB色彩空间模型中通过火焰颜色的加权判别算法提取出具有火焰颜色的运动像素区域;最后,提出一种基于统计频率计数的火焰频闪特征识别方法,用于将视频图像中真实的火焰区域从具有火焰像素颜色的运动区域中区分出来。Matlab仿真结果表明,该算法具有很高的有效性和鲁棒性,能够应用实际的视觉火灾检测系统之中。

英文摘要：

Based on the colour surveillance video images,firstly makes use of temporal motion feature of flame image,adopts Gaussian mixture model to extract motion foreground pixels from surveillance video image sequence;then flame color weighted distinguishing algorithm is adopted to extract the motion pixel regions which have flame-like color from RGB colour space model;Finally,a kind of flame time-frequency flicker feature recognition method based on statistical frequency counting is proposed to distinguish the true flame region from the motion regions which have flame-like color.The Matlab simulation results show that the proposed algorithm is perfectly effective and robust,and it is able to be used in actual vision fire detection systems.