Fire and Smoke Identification Algorithms based on Attention Mechanism
T. Luo, Nanfeng Xiao · Advances in computer science research · 2015
This paper presents several fire and smoke identification algorithms based on attention mechanism, the proposed algorithms add the attention mechanism into traditional algorithms for the fire and smoke identification, which can get the possible regions of fire and smoke quickly by saliency images, and highly accelerate the computing process.Firstly, the attention mechanism is used to get the saliency images.Secondly, the Otsu method is used to segment the saliency images.Finally, ROIs (Regions of Interest) are analyzed by the fire or smoke features, in order to judge whether the fire and smoke exists or not.To judge the fire, the fire color and intensity features are used.S imilarly, the color and circularity features are be used to judge the smoke.