A novel smoke detection algorithm based on MSER tracking

The smoke detection in remote video surveillance is a very important task in the field of computer vision and forest fire prevention. In order to solve this problem, a novel smoke detection algorithm based on dual-threshold MSER tracking is proposed. The algorithm is divided into three stages: generation of candidate smoke region, confirmation of candidate smoke region and tracking of smoke region. At different stages, different thresholds are used in the MSER tracking, and then the static and dynamic judgments of smoke are proposed to identify the smoke region. Experiment results show that the algorithm is able to achieve more reliable and accurate detection performance than other state-of-art smoke detection methods in the remote video surveillance.