Laboratory fire detection based on flame color and dynamic law

Abstract

Fire has caused great harm to human normal life, especially to human life and safety. How to effectively and reasonably prevent and control building fires and minimize or even eliminate the occurrence of such problems is currently a problem that governments at all levels and the general public attach great importance to. In this paper, the fire detection technology is studied. It is found that most of the current fire detection technology is based on temperature, smoke concentration or relative humidity and other indicators of traditional detection methods, but this kind of detection methods can not meet the needs of higher detection efficiency. In order to detect indoor fire effectively and improve the efficiency of fire detection, a method of using indoor surveillance video to realize artificial intelligence fire detection is proposed. The main principle of this method is as follows: firstly, the flame region is captured by neural network, and then the dynamic process of flame flashing color and contour in video image is studied. Then, the HSI color model of flame and the contour model of word bag are calculated and processed by HSI and word bag method, respectively. Finally, the detection results of the two models are effectively fused, and the changes of color and contour of flame in the combustion process are summarized to achieve the purpose of fire detection.