Video flame detection based on color and contour features

In this paper, a video flame detection method combining fire color and contour features is proposed. Firstly, fire color areas are obtained using fire color detection rules based on RGB color space and HSI color space. Secondly, the whole continuous and close flame contours are extracted by applying the Chan-Vese model in fire color areas. Then the moving properties of flame contours are analyzed and those with characteristic displacement and vibration features are accepted as flame objects. Experimental results show the proposed approach can remove the static or dynamic non-flame objects and segment the flame area from the complex background, and also has a high time effectiveness.