The Research And Application Of Flame Video Feature Detection And Analysis

According to statistics, fire is the first kind of threat to economic development and people's life and property safety. Fire detection method based on video feature detection and anlysis makes use of the characteristic of flame image when the fire just happened, and it recognize the fire information correctly from image sequence with background noice to detect the fire. Compared to traditional method, it has the advantage of large detection and usability range, high accuracy, high reflection, easy to find cause of the fire.In this paper, our objective is to build up a new fire image recognition system with the real requirement, which is based on the research of flame video feature detection in tunnel. It focuses on the flame features, such as moving, color, frequency, shape characteristic, and these are used as theoretical basis for implementing related fire detection method. Then we set the thresholds and give out the effect of video testing. The main constructure and content is:First, from chapter one to chapter three, we introduce the background of flame recognition system, give the general expatiation of video monitoring and flame detection research status, and anlysis for logical constructure of system and the kernel part of image recognition.Then, we divided the system into five functional models, including moving detection, color detection, temporal frequency detection, spatial frequency detection and shape detection. From chapter four to chapter six, we give detailed description of the design and implementation process of these models, and focus on the process of set detection methods thresholds from large number of testing data. We also give out the testing effect. At the sametime, we introduce the interface of input and the image recognition kernel part.At last, we summarize the result that we have gained, and also analyze the existed problems as well as the research contents further more.