Video-based fire detection with spatio-temporal SURF and color features

This paper proposes a new video-based method for fire detection, which uses color, spatial and temporal information. Our method divides videos into spatio-temporal blocks and extracts a novel volume SURF features from three orthogonal planes of these blocks. Combining spatio-temporal SURF features and the global color histogram of Lab color space on blocks, we achieve a new final features to detect fire. And we use a linear SVM in classification stage. Experiments show that the classification performance of our proposed method is better than a previous method and can achieve better true positive rate in various scenes.