Morlet Wavelet Threshold Based Glow Warm Optimized X-Means Clustering for Forest Fire Detection

Abstract

Forest fire detection is a significant problem to be resolved for the prevention of life and property safety. Clustering performance of existing fire detection algorithm was poor. In order to resolve the above limitations, Morlet Wavelet Threshold-based Glow Warm Optimized X-Means Clustering (MWT-GWOXC) the technique is proposed. Initially, this technique takes a number of video frames as input and initializes a number of clusters. Then, the proposed approach initializes the glow warm populations with a number of video frames. Then, the technique calculates the fitness function of all clustered video frames and identifies a pre-fire stage or fire stage or critical fire stage. If any video frame not clustered, this technique employs Bayesian probability criterion which determines a higher probability of frame to become a cluster member and improves clustering accuracy. The simulation result demonstrates that the technique is able to increase fire detection accuracy and also minimize fire detection time.