Efficiency of HSV over RGB Gaussian Mixture Model for fire detection

Computer Vision based systems have already been proposed to detect fire automatically. To increase the reliability of such systems, Gaussian Mixture Models of fire need to be developed in order to decide if an object in a scene is a fire or no. The models are trained using color images in RGB color model. We, however, believe HSV model is more suitable to present the statistics of a set of colored images precisely. To vindicate this claim, the paper includes a rigorous comparative evaluation of the two aforementioned color models in a fire detection system to demonstrate the superiority of the HSV color model. It makes hence the recommendation of using the latter model in vision based fire detection systems.