Video smoke detection based on color transformation and MSER

In the long-range video surveillance, the smoke is difficult to detect when the smoke is small or it moves slowly. In order to solve this problem, a video smoke detection algorithm based on smoke color enhancement transform and maximally stable extremal regions (MSER) was proposed. Firstly, a smoke color enhancement transformation was proposed to make the smoke area more salient and easier to be segmented. Secondly, in order to avoid the difficulty of smoke segmentation in the traditional color-based and motion-based methods, the MSER detection was employed to segment the smoke area. Lastly, based on the accurate segmentation, a series of static and dynamic criterions for the characteristics of smoke was proposed and the smoke was determined by the cumulative number of passing through these static and dynamic criterions. Experimental results show that the proposed algorithm can accurately and reliability detect the smoke in the far-range video surveillance.