Conference Paper

Fire Smoke Detection in Video Images Using Kalman Filter and Gaussian Mixture Color Model

DOI: 10.1109/AICI.2010.107 Conference: Artificial Intelligence and Computational Intelligence (AICI), 2010 International Conference on, Volume: 1
Source: IEEE Xplore

ABSTRACT Fire smoke detections are crucial for forest resource protections and public security in surveillance systems. A novel approach for smoke detections with combined Kalman filter and a Gaussian color model is proposed in the paper in open areas. Moving objects are firstly generated by image subtractions from adaptive background of a scene through Kalman filter and MHI(Moving History Image) analysis. Then a Gaussian color model, trained from samples offline by an EM algorithm, is performed to detect candidate fire smoke regions. Final validation is carried out by temporal analysis of dynamic features of suspected smoke areas where higher frequency energies in wavelet domains and color blending coefficients are utilized as smoke features. Experimental results show the proposed method is capable of detecting fire smoke reliably.

  • [Show abstract] [Hide abstract]
    ABSTRACT: It has long been a big challenge to extract dense smoke regions by motion detection. As a result, there are too few suspected smoke regions being recognized for an early fire alarm. In this paper, an early smoke detecting system that can efficiently extract dense smoke regions is proposed. Firstly, since the brightness in the areas that have dynamic texture is not constant, the residuals of optical flow are calculated to locate suspected smoke regions. A certain threshold of the increment of optical flow residuals is also used to distinguish smoke from other dynamic texture. Secondly, five features that can jointly represent a smoke area, including grayish color, chrominance decrease, edge energy decrease, optical flow orientation diffusion and circularity, are chosen by thorough experiments. Experimental results show that the proposed system can detect the smoke in early time and is robust to most kinds of interferences, especially other dynamic textures.
    2012 International Conference on Machine Learning and Cybernetics (ICMLC); 07/2012
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Fire accident remains a problem in modern society. This leads great efforts in finding ways to prevent, detect and control it. Conventional fire detection systems are mostly point detectors, which have limitation for early smoke detection, especially in a high-ceiling atrium. A video-based smoke detection system is an interesting alternative approach. It has better area coverage and detecting smoke faster. In this work, a video-based smoke detection system was developed with two main processes, i.e. moving objects segmentation with Gaussian Mixture Models (GMM) and smoke classifications with Mathematical Model of Meaning (MMM). In the MMM model, the interpretation of dangerous smoke is based on the context provided. Then the classification results are compared with conventional smoke detector. The results show that MMM can recognize the dangerous smoke faster than conventional smoke detectors.
    Procedia Engineering 12/2013; 62:963-971. DOI:10.1016/j.proeng.2013.08.149
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Standard wildfire smoke detection systems detect fires using remote cameras located at observation posts. Images from the cameras are analyzed using standard computer vision techniques, and human intervention is required only in situations in which the system raises an alarm. The number of alarms depends largely on manually set detection sensitivity parameters. One of the primary drawbacks of this approach is the false alarm rate, which impairs the usability of the system. In this paper, we present a novel approach using GIS and augmented reality to include the spatial and fire risk data of the observed scene. This information is used to improve the reliability of the existing systems through automatic parameter adjustment. For evaluation, three smoke detection methods were improved using this approach and compared to the standard versions. The results demonstrated significant improvement in different smoke detection aspects, including detection range, rate of correct detections and decrease in the false alarm rate.
    Computer Vision and Image Understanding 01/2013; DOI:10.1016/j.cviu.2013.10.003 · 1.36 Impact Factor