Intelligent video analysis-based forest fires smoke detection algorithms

Abstract

In order to discover forest fire as early as possible, forest fire detection should focus on the smoke in early fire. This paper focuses on three key issues: motion segmentation, feature extraction and classifier design. Background subtraction based on visual background extractor (Vibe) is chosen to divide suspected smoke area when taking into account the accuracy and time consumption. And then do some corresponding morphological processing. Later on, various static and dynamic characteristics of smoke were extracted and different tests were done based on different feature combinations in forest fire smoke detection system. Lots of smoke detecting system only think about static features which will result in a certain degree of misjudgment. Analyzing the false positive rate and recognition rate of these experiments' results, the combination of movement direction, high-frequency energy based on wavelet transformation and compactness is selected to constitute the final recognition vectors. In addition, the continuity which is not mentioned in other researches won't be ignored in this paper. The final experimental results showed that the accuracy rate of this method for smoke detection could reach 92.7%.