Smoke Detection Method Based on LBP and SVM from Surveillance Camera

摘要

Wildfire is a regular incident worldwide today. It destroys forests and also the living areas of wild animals. So, to reduce the harmful effects of such disasters this paper describes a method of smoke detection for surveillance cameras. This smoke detection will ease the fire detection. Proposed method is based on Local Binary Pattern (LBP) and Support Vector Machine (SVM). Initially, Approximate Median Filtering Algorithm was applied to subtract the background from input frame. Then, shape based filtering method was applied to get the region of interest. Thirdly, LBP values and histograms were calculated from the pixels of region of interest to form a feature vector. The proposed method also applied Bhattacharyya coefficients to verify the smoke region for accurate result. Finally, SVM classified the region of interest as smoke image. Results using real scene data show that the proposed method can give accurate results in different conditions of real world situations.