Automatic Early Forest Fire Detection Based on Gaussian Mixture Model

Early forest fire detection is of great importance to avoid the huge damage of forests caused by fires. Early fire detection focuses on smoke detection. The proposed scheme of early fire detection is divided into two steps. Based on the slow spread of smoke, firstly a time delay parameter improves Gaussian mixture model for extracting candidate smoke regions. Then, two motion features of smoke, the rate of area change and motion style, are used to select smoke regions from the candidate regions. Our experiments verified the effectiveness of the proposed early fire detection scheme.