基于目标跟踪和多特征融合的火焰检测算法

针对目前视频火焰检测算法无法达到高检测率、低误检率和实时的工业需求,提出一种基于目标跟踪和多特征融合的火焰检测算法.首先利用混合高斯背景建模获取前景图像,在HSV色彩空间中根据火焰的颜色特性分离出疑似火焰区域,对火焰疑似区域采用卡尔曼滤波器实现运动目标的跟踪,再结合火焰的相似度、区域增长率和跳动频率特征用加权求和得到的值与报警阈值相比,最后根据判断比较确定真实火焰区域,并且实现对火焰的持续跟踪.实验结果证明,该算法能够对火焰区域进行有效的检测与跟踪并且具有良好的实时性和抗干扰能力.

Fire Detection Algorithm Based on Object Tracking and Multi-feature Fusion

Current video flame detection algorithm can hardly meet the requirements of high detection rate,low false-alarm rate and real time in industrial application, this paper proposed a method that combined the object tracking algorithm and flame characteristics to detect the fire region. First, Gaussian mixture background model was utilized to derive foreground image, and then in the HSV color space separating the suspected flame area according to the characteristics of fire. After that using the Kalman filter for object tracking, and combining with the similarity of flame, region growing rate and the beating frequency features obtained an weighted sum to compare with the alarm threshold. According to judgment to determine the true flame region, at the same time flame would be continuous tracking. The experimental results show that the algorithm can effectively detect and track the real flame region. In addition it has a good real-time performance and anti-jamming capability.