A smoke detection algorithm based on K-means segmentation

The fire accident usually causes economical and ecological damage as well as cause danger to people's lives. Therefore, its early detection is must for controlling this damage. Also smoke is considered as main constituent of fire, thus an efficient smoke detection algorithm on sequences of frame obtained from static camera is proposed. It is based on computer vision based technology. This algorithm uses color feature of smoke & is comprised of following steps: reading the image, preprocessing, classify color pixels using k-means segmentation. This paper discusses mainly the segmentation problem. It adopts L*a*b* color space and k-means clustering algorithm to isolate the smoke from video sequences. Finally the K-means algorithm used is compared with the fuzzy c-means algorithm used previously.