基于记忆矩阵和跳动频率的视频火焰识别方法

Video flame detection using motion memory matrix and flame jumping frequency

针对大空间环境下的早期火灾的探测问题,提出一种基于记忆矩阵的运动目标检测方法,通过火焰质心的跳动直方图计算火焰的跳动频率,利用HSV颜色空间中火焰像素的特征结合火焰的跳动频率作为logistic回归模型的特征向量,构建火焰识别模型,实现视频火灾火焰探测。实验结果表明提出的算法在复杂的视频场景中都有较高的识别精度。

In order to solve the problem of early fire detection in the large space, a moving object detecting method using motion memory matrix is proposed. By the histogram of flame centroid healing, the flame beating fre(|ucncy was calculated. Combining the flamepixel feature in HSV color space “nd lhe flame beating frequency as the feature vector of the logistic regression model, the flame identi" fication model was built to achieve flame detection. Test results showed that lhe algoriihm has high recognition accuracy in coni pl ex video scene.