A Hybrid Clustering Algorithm for Fire Detection in Video and Analysis with Color Based Thresholding Method

Abstract

In this study an unsupervised way of fire pixel detection from video frames is depicted. A hybrid clustering algorithm is proposed, depending on color samples in video frames. A modified k-mean clustering algorithm is used here. In this algorithm hierarchical and partition clustering are used to build the hybrid. The results are analyzed with color base threshold method by considering RGB and HSI color models. The two very good color spaces which follow the mechanism, the way hardware and human being identify the color. We discuss performance based comparison by which it can be clearly shown that how our proposed algorithm perform better. Overall this algorithm does very fast processing and detects fire flame in video, which is very important because smart video surveillance need very fast response.