基于红外视频的火灾探测算法

视频火灾探测技术成为火灾探测的主要研究方向之一，该技术的难点是保证对火灾准确快速地探测。针对目前视频火灾探测识别率较低、报警时间较长的问题，为了提高火灾识别率，增强算法实时性，本文提出一种红外视频的火灾火焰探测方法，采用带有截止可见光透红外滤波片的摄像头采集现场红外影像，结合火焰的静态和动态特征，对火灾进行准确快速探测，同时，综合运用火焰圆形度和尖角个数变化判据对稳定火焰和非稳定火焰进行区别，实现对火灾的分级报警。实验结果表明，改进算法对火焰识别的准确性高，抗干扰能力较强，并且具有显著的快速性，其处理速度达到20ms/帧，并且系统平均报警时间小于200ms。

A fire flame detection algorithm based on infrared video

Video fire detection technology is one of the main research directions of the fire detection, and the technical challenge is to ensure that accurate and rapid detection of fires. The recognition rate is lower and the alarm time is longer of current video fire detection, so a novel detection algorithm in infrared video is proposed in this paper that aims to improve the fire recognition rate, and enhance the real-time nature of the algorithm, which capturing infrared images camera with a cut-off visible light through infrared filters and combining with static and dynamic characteristics of the flame. Meanwhile, it distinguishes the stable flame and unsteady flame by the integrated use of flame roundness and sharp number change criterion to achieve the classification of fire alarm. Experimental results that the proposed algorithm ensures the reliability and accuracy of fire detection, a strong anti-jamming capability and the significant rapidity. The processing speed of the algorithm is 20ms / f, and the system average alarm time is less than 200ms.