Research on flame detection algorithm based on multi - feature fusion

Fast and accurate detection of the flame area in the surveillance video is a necessary condition to reduce the loss caused by fire. This paper combines the dynamic and static features of the flame in the video, and proposes a flame detection method combining the flame color feature and the local feature. Firstly, moving target detection method is used to extract the moving target from video streams, and the effective color segmentation threshold is analyzed and determined. The color threshold is segmented from the two color spaces of RGB and HSV respectively to obtain the suspected flame region. The local features of the target area are extracted, and the feature vector input into the Support Vector Machine classifier for flame detection. The detection effects of the two local features were compared to select the better features. The experiment result shows that the algorithm of this paper achieves the ideal detection effect. Compared with the traditional single feature flame detection method, the algorithm of this paper effectively reduces the impact of the environment on the detection results and reduces the false alarm rate of the fire flame.