Intelligent fire-detection model using statistical color models data fusion with Dezert–Smarandache method

Abstract

The aim of this article is to present a novel and efficient method for early fire detection in video sequences. Fire detection in video sequences is the need of the hour for surveillance applications and de-noising processes used in target detection in defense applications. The proposed fire-detection model fuses evidences from statistical models using Dezert–Smarandache Theory (DSmT) of plausible and paradoxical reasoning for evidence fusion. The conflicting masses are redistributed using the proportional conflict redistribution (PCR5) rule of combination for calculation of final belief of evidence. The results of the proposed model are compared with some of the earlier methods by taking large data sets, which contain both fire and fire-like regions created under different environmental conditions. Results show that color models data fusion using DSmT resolve the conflict in earlier models and increase belief level i.e. high fire-detection rate and low false-alarm rate.