Indian Journal of Science and Technology
Year: 2018, Volume: 11, Issue: 19, Pages: 1-17
Aditya Gupta*, Neeraj Bokde, Dushyant Marathe and Kishore
Electronics and Communication, Visvesvaraya National Institute of Technology Nagpur - 440010, Maharashtra, India; [email protected], [email protected], [email protected], [email protected] ece.vnit.ac.in
*Author for correspondence
Electronics and Communication, Visvesvaraya National Institute of Technology Nagpur - 440010, Maharashtra, India; [email protected]
Objectives: A novel video based fire detection algorithm based on rule base technique using RGB and HSV color space and spatial analysis based on wavelet analysis is proposed. Methods/Statistical Analysis: Rule base technique utilizing RGB and HSV color space for extraction of fire pixels in the frame was used. Threshold based spatial energy methodology is used for differentiating fire and fire like objects. Wavelet analysis is performed for calculation of spatial energy. Texture analysis using Local Binary Pattern (LBP) is also performed when fire or fire like candidate having spatial energy near to threshold of fire pixel. Findings: The usage of RGB color space alone for identification of fire in the video frames is not sufficient as they suffer from false detection. Two novel rules based on HSV plane is proposed which have improved the detection ability of system when compare to previous studies. But still suffers from false detection. Spatial energy methodology based on differentiating fire and fire like objects performs well and has achieved greater efficiency with low false detection rate of 4% on standard datasets. Texture analysis using Local Binary Pattern (LBP) is also performed in rare case when fire candidate is having spatial energy near to that of fire like object category. This has helped in reducing the computational complexity of the system. The system shows 100% accurate results. Improvements: The results obtained for different standard datasets using the proposed hybrid spatial and texture base analysis shows 100% accuracy with zero false positive and false negative rates which is not observed in any of the present articles.
Keywords: Fire Detection, Local Binary Pattern (LBP), Spatial Analysis, Texture Analysis, Wavelet Analysis
Subscribe now for latest articles and news.