Extraction of fire region from forest fire images using color rules and texture analysis

Abstract

Day by day the rate of forest fire is increasing and the losses suffered are huge. There are many methods available for detecting the fire flames. However most of them fail in discriminating fire and fire colored local objects. In this paper we propose a novel and robust method that segments fire flame regions from an image. The proposed method combines the unique color and texture features of the fire flames. The RGB images obtained from digital cameras are used as inputs. The mean value estimation is used to assume the presence of fire flames in an image. Its presence is confirmed by performing the background elimination. Color rules defined for the RGB, HSV, and YCbCr color spaces are used to segment the fire regions. The HSV and YCbCr color spaces are used because of their capability to separate intensity (luminance) and color characteristics (chrominance). One of the unique characteristics of the fire flame pixels is its high intensity value. These color spaces helps to take this characteristic into account thereby reducing a lot of errors. Finally the result is made more accurate by using the texture analysis. The Gray-Level Co-occurrence Matrix (GLCM) is used to calculate five texture features of the fire flames. The proposed method is tested on a set of fire and non-fire images and it achieves a high detection rate and low false rate.