Research Of Fire Detection Algorithm Based On Video-images

The fire detection system based on image-processing detects the fire by using the video images which were taken by the watching cameras automatically. The sequence images were processed and analyzed by using some kinds of image-processing technologies. Then the result is judged according to the changing characters of the video fire images in the early period.Firstly, the paper introduced the theory used by the fire detection algorithm based on video-image, such as color-model theory, wavelet and BP neural networking theory.And then a new fire detection method was brought forward based on those theories introduced before. The method detected the moving region in the video firstly and segmented the fire-color region in it using HIS model. The advantage of the segmentation method is it can fast the processing speed, which is good for the processing of the real-time system. Then the roundness, area change rate, fire flicker frequency, area change tendency, etc, those fire characters in the segment regions were analyzed. At the same time, the relevant computation and analysis methods were brought forward. The fire flicker frequency and area change tendency characters were obtained by using the part analysis ability of wavelet technology. The curve which was consisted of the area of segmentation regions in every frame of the videos was processed using wavelet. Low-frequency heft was used to analyze the area changing current, and the high-frequency heft was used to analyze the flicker frequency. At last the construction, input and output unit of BP Neural Network were designed to classify and recognize the fire images.The simulation test was done at the last of the paper. The results show that this fire detection method can recognize the fire effectively and has good automatic recognition ability.