Fire Smoke Detection Based On Video Surveillance In Indoor Large Space

With the unceasing enhancement of the ability of microprocessors and widespreadapplication of security video monitoring using computer vision technology, fire smokedetection methods attract increasingly more attention. This paper focuses on thealgorithms of smoke detection based on video surveillance in indoor large spaceconsidering its characteristics.Comparing the experimental results between Kalman filter based and mixed Gaussianbased background modeling, in this paper, we choose the latter one as the backgroundupdating method which provides the background image for following foregroundextracting and analyzing the smoke characters.In this paper, the Otsu algorithm is used to extract foreground image from thereal-time image via using background image. Through implementing median filter andmorphology filter on the foreground image, a denoised black-white image is got for thefurther calculation of the circularity. The recognition of fire smoke is based on two majorcriteria, namely circularity and structural similarity. Firstly, if the region meets the criteriaof circularity, the structural similarity is calculated. Then, if the region still meets thecriteria, we count the numbers of frames which meet the criteria in the following frames.At last, a judgment of whether setting an alarm is made by previous tests.The algorithm is practical by testing three groups of video clips with fire smoke. Atlast four conclusions and algorithms improvement are established by analyzing theexperimental data.