DefinePK

Abstract

CC cams are everywhere and in this work, we explore these sensors ability and the corresponding algorithms to detect fire. The two drawbacks that raises questions on the performance of fire detection algorithms are: (1) Ambient lighting that masks the fire for color feature based detection and (2) Sizeable object movement near the fire for dynamic change based algorithms. This paper addresses these problems with CC camera footages of fire in indoor and outdoor environments under the two problematic conditions. We test models on color, frame subtraction, back ground modelling with Gaussian Mixture Models, Independent component analysis, Geometric - Independent component analysis (GICA). A 4-parameter statistical model checks the quality of the proposed algorithm. Results show the potentiality of the proposed algorithm is solving the above two problems for fire detection.