Research On Fire Smoke Detection Based On Image Processing

Vision-based fire smoke detection technology using image processing methods is an important branch, which captured considerable attention over the past decade, in fire safety field. Due to variability and complexity of fire outcomes, yet no general theory or algorithm has formed so far in those vision-based image processing methods that have been applied in this area.Among the theories of static image processing, color image segmentation based on HIS model has excellent component independence and color authenticity better than usual RGB model image. Thus HIS model is easier to be segmented. In this article this method is designed and validated with a real fire smoke image. The result of experiment indicates its validation and good real-time characteristic in terms of announcing fire alarm.It is feasible to apply dynamic video image segmentation technology in fire detection according to characteristics of motion and texture of fire outcomes such as smoke and flame. First, moving foreground, i.e. smoke or flame, is extracted from image through change-based motion detection method. Then a consequent binary image is mapped to the original image to catch the corresponding smoke and flame area. Last, every moving objects'texture features based on gray level co-occurrence matrix (GLCM) are calculated, on the basis of texture difference between fire outcomes and other objects in the background, features with good distinguishable ability are selected to be the recognition criterion.In addition, to get a better and more precise segmentation result, morphological image processing methods are used. Morphological image processing technology is an important branch of image processing, especially in the image segmentation field. Results of experiments indicate that the approaches for color image segmentation based on HIS model and video image segmentation based on texture features recognition are effective and practicable for fire detection.