Flame detected in video based on color moments

Abstract

After analyzing the motion area of flame blocks and fluctuations of flame pixels in complex scenes video, this paper detects motion area by using blocks in spatial consecutive video frames. For the color information of the flame block, motion area is analyzed by the second order color moments in HSV color space, and then it extracts the suspected flame color motion area. For each pixel of the block area, this paper chooses spatial continuous images to do wavelet analysis and extracts pixels which meet the flame fluctuation rules as the flame pixels. Use local weighted operator to filter out non-flame discrete pixels, make the adjacent flame pixels area form connected areas and give an alarm. Experimental results show that this method can detect the video flame area taken by ordinary video cameras quickly and give an alarm, meanwhile, it also has a better robustness.