基于视频的火灾实时检测方法  

Vision based real-time fire detecting method

作　　者：金红[1] 张荣标[2] 

机构地区：[1]江苏大学计算机中心,江苏镇江212013 [2]江苏大学电气信息工程学院,江苏镇江212013

出　　处：《计算机工程与设计》2009年第18期4330-4332,共3页Computer Engineering and Design

基　　金：江苏省高校自然科学基金项目(06KJD520037)

摘　　要：为了提高基于视频的火灾检测的准确性和实时性,提出了一种综合色彩特性分析和闪烁特性分析的火灾检测方法。首先在获取帧差视频图像的基础上,通过对HSI色彩空间中的色彩特性分析提取帧差图像中的火焰候选区域;在此基础上引入离散傅立叶变换分析火焰候选区的频率变化来确定是否存在火灾;最后通过时间累计方法给出火灾报警的可信级别。实验结果表明,该方法具有较好的可靠性、鲁棒性和运行效率。To enhance the aceuracy and efficiency ofvision based fire detecting, a novel method to detect fire in real-time by proeessing the video data is put forward, which combined fire flicker and color clues to reach a final decision. First the image difference method and color analysis technique based on the HSI color model are applied to extract the fire-like moving region between two consecutive frames. Then the spectrum of the fire-like region flickering could be got by the discrete Fourier transform of the fire-like moving region sequenee to eonfirm whether it is ftre region. Finally, a time-accumulating method is devised to estimate the fire alarm grade. Experimental results show that the fire detecting algorithm is accurate, robust and effective.

关 键 词：色彩分析 傅立叶变换 帧差 视频 火灾检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]