Round-the-clock Video Fire Detection System

Fire is an important symbol of human civilization, but fire is a double-edged sword, on the one hand, it has brought great impetus to the progress of human civilization and social development. On the other hand, fire is a disaster that damages widely and happens frequently and brings harm to human's safety of life and property. In addition, forest fires, building fires, mine fire which happen frequently cause great destruction on the ecological environment. So the prevention and control of fire always get extreme care and more research.Most fire detection systems are mostly based on sensor alarm system now, such as feeling smoke, feeling temperature and photosensitive detector. They made full use of temperature and light and other physical properties of flame and smoke to detect the existing of flame. But due to the velocity of air, temperature, humidity, dust and many others factors, the sensor monitoring has lots of shortage such as limited extent, and easily failure, the detection effect are also often not satisfied. In recent years, as video monitoring system with the popularization and the increasingly mature of computer vision technology ,the flame detection technology based on image processing has been widely concerned. the fire detect researches from home and abroad are mainly based on the flow combined aspects, the color features, morphological characteristics, dynamic characteristic and the combination of dynamic and static features of flame. The traditional visualization fire detection technology are basically based on analyzing images from infrared cameras or visible camera to detect flame, These methods cann't meet the requirements of round-the-clock monitoring flame in engineering application. A multi-feature fusion flame detection system using video-type judgment algorithm(VTJA) is proposed.In order to choose corresponding types of suspected flame region extraction algorithm and to improve the accuracy of detection system. The system firstly use two video type judgment algorithm RGB-VTJA and HSV-H-VTJA to determine the image types of each frame which acquired from camera. The main function of these algorithm is to determine the video type belongs to optical image or infrared image.Improving the accuracy of suspected flame area can largely improve the detection rate of the follow-up flame detection algorithm ,the processing and extraction of foreground and background is the basis of flame detection based on video , three kinds of visible light conditions extraction algorithm of flame area are employed which are base on YCbCr, RGB, HSI, HSV four kinds of color space, the system used the flame detection algorithm based on YCbCr color space to extract the suspected flame area under the visible light condition. The algorithm can accurately extract the fire area and directly exclude the interferences of the objects with flame color. After analysis the flame characteristics of video images, the system use several aspects of the flame, color features, morphological characteristics, dynamic characteristic and combination of dynamic and static characteristic features, and the algorithm based on flame sharp-angled features, flame centroid detection algorithm, flame wavelet detection algorithm, flame frequency based on DCT and frame correlation between the flame detection algorithm is analyzed in the system, these five features algorithm different characteristics ensuring high detection rate and reduce false examining rate.Flame combust procession can be followed, but these rules is always broken by the environmental factors and other various uncertainties of interference. so the system use a more features fusion flame detector which can achieved in maintaining high detection rates, reducing the mistake examining rate. The system has a lot of advantages: real-time, can detect and warn flame within 4 seconds in the early period of fire. Robustness, the algorithms of system has good robustness and also applied invisible light conditions and infrared conditions, Stability, code correctly, timely release system memory, clean up the useless memory space, Portability, this system adopted through the optimized, algorithm can be applied to multiple platform, which can save cost, the system can be use for existing monitoring system, which can save a large part of the money.Experiments show that the system can detect flame fast and reduce the error rate. And can make full use of existing video monitoring equipments with good application prospects.