A hybrid fire detection using Hidden Markov Model and luminance map

Computer vision and pattern recognition is hot research topic recently, fire and flame recognition is an important sub-topics. Although researchers proposed many methods, there exist high false alarm and complex environment inadaptation because of fire-colored moving objects interference. This paper proposed a hybrid method using Hidden Markov Model (HMM) based on spatio-temporal feature and the variance of temporal luminance to detect fire. Here, the HMM model based spatio-temporal feature characterize the flicker feature of fire, and a series of observation points will be set on the boundary of flame to judge the fire flicker feature. While the variance of luminance character the temporal luminance feature. First, we can get the “candidate fire region”, then we concentrate our main work on analyze the “candidate fire region”. Experiment results show our method has a good result and it is robust to be used in complex environment compared with previous algorithms.