Research On The Visual Flame Detection System Based On Machine Vision

To avoid the great loss caused by fire disasters, a kind of auto fire prevention sys-tem which can discover the fire, and take quick actions is needed. Mechanical vision technology with the function of obtaining spatial information can offer support for this referred system. Therefore, the research on visual fire detection system has a significant meaning. In this background, this essay is to design a visual fire detection system which is able to identify fire disasters and positioning its source. Main procedures are as be-low:1. Based on fire image analysis, how to accurately identify fire in the image on account of some potential interfere under real condition is to be discussed. The fire identifying process with comprehensive discriminated methods will then be designed.2. In the part of fire identifying, some algorithms such as the adaptive threshold segmentation method and Fractal algorithm are employed to segment image and extract fire characters. After that, the identifying process is carried out by factors of flashing degree, area change rate and center displacing rate. This identifying experiment and re-lated results analysis prove the feasibility of this calculation. The source of mistakes is given further analysis as well.3. In the part of fire positioning, different visional measuring methods are listed. After comparison, the study of passive measurement of Binocular Stereo Vision is cho-sen according to the experiment requisition. Firstly, the theory is introduced. Secondly, camera calibration, stereo rectification and stereo matching are involved to attain spatial information. To avoid the stereo matching mistakes caused by irregular appearance of fire, the algorithm of increasing the rectification’s error tolerance is used. Finally, this positioning algorithm is proved to be feasible by experiment.