Multi-scale Visual Saliency Fusing with Spatiotemporal Features for Fire Detection

Aiming at the problem that the existing methods cannot realize the fire detection accurately, a multi-scale visual saliency fusing with spatiotemporal features for fire detection method is proposed. Firstly, the simple linear iterative clustering (SLIC) method and region contrast (RC) method are used to realize multi-scale visual saliency detection of fire images. Then the spatiotemporal features of fire are extracted in the time domain and space domain respectively. Finally, multi-scale visual saliency detection and the spatiotemporal features of fire are combined to realize the fire detection. The experimental results show the proposed method has a higher accuracy rate and a lower false rate.