Chen J, He YP, Wang J (2008) 'Study on Method of Multi-Feature Fusion Based Video Flame Detection.' (Science Press Beijing: Beijing) 863-868
motion detection, Gaussian mixture model, color mapping, flickering,
Abstract: A real-time video flame detection method based on the multi-feature fusion is described in this paper. In this method, dynamic characteristic and color clues are utilized to detect fire in color video sequences. Moving object, in complex scenes are detected by the Gaussian mixture model method to first model the background and then extract moving foreground of detection scenes, secondly, detected moving objects were then categorized into flame and non-flame regions by using flame color decision-making algorithm. As a result of the violent oscillation of fire boundaries, flame flickering is also discussed as a potential criterion to improve the robustness of our detection algorithm in the future research. Testing results show that the algorithm is effective and adaptable, meaning that the proposed method can fulfill the real-time processing requirements.