一种基于视频图像相关性的火灾火焰识别方法  

Renovated method for identifying fire plume based on image correlation

作　　者：沈诗林[1] 于春雨[1] 袁非牛[1,2] 陈志斌[1] 张永明[1] 

机构地区：[1]中国科学技术大学火灾科学国家重点实验室,合肥230027 [2]江西财经大学信息学院

出　　处：《安全与环境学报》2007年第6期96-99,共4页Journal of Safety and Environment

基　　金：国家科技支撑计划项目(2006BAK06B07)

摘　　要：由于火焰燃烧时气体羽流的卷吸特性和空气流动的影响,燃烧的火焰表现出不停的振荡特性,基于这种振荡性提出利用图像的相关性来探测火灾的发生。火灾视觉特征的提取是视觉火灾探测的重要问题,首先确定运动区域,再利用火焰振荡特性提出了相关性算法,并分析了R、G、B3种分量对相关性的影响。实验结果表明利用相关性来分析时,火焰的相关性系数图是剧烈振荡的,而非火焰的相关性系数图是相对比较平坦的,为可视化火灾探测系统提供了一种有效的判据;R、G、B对相关性的影响是一致的,在实际应用中只需利用其中的1个分量即可,可有效提高火灾探测系统的响应时间。This article is inclined to present its authors＇ renovated method for identiicing fire plume based on the plume image correlation. As in the fire-fighting practice in recent years, it has become a latest developmental trend of identifying the fire burning strength by judging its burning color, texture and time features, for a fire flame is characteristic of oscillation of an air flow, which itself is different from other rigid matters. Thus, we have brought forward the so-called oscillation-identifying method whose main feature is to identify and measure the correlation coefficients between the time of burning and the appearance of the flame. In our experiments, location was presented for surveying ordinary CCD between the first frame and the corresponding subsequent frames. First of all, it is to determine the flame- moving area based on picking-up the commix background. Secondly, it is to make use of the image correlation for fire detection based on its moving region. In the meantime, we have to analyze the fireplume image correlation of its intensity titled as R, G and B for the identification based on the fire oscillation. The present method is inclined to be used for reducing the fire disaster likely to cause casualty and material damage due to the failure to get reported so as to boost up the robust of the detection system. Experimental results indicate that the algorithm is effective and useful. The method can also be used for non-fire detection purposes. All in all, the image of the correlation coefficients of the flames is characteristic of oscillation, whereas those of the non-flame matters are rather smooth, whereas the mutual influence of the R, G or B intensities proves to be fairly stable. When the method is put for practical application, it is recommended to choose a sort of intensity among the three intensities of R, G or B to reduce the operation time and enhance the system＇s capability.

关 键 词：安全工程 视觉特征 相关性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] X932[自动化与计算机技术—计算机科学与技术]