Algorithm Of Forest Smoke Video Detection Based On Codebook And Dynamic Characteristic

The frequent forest firing will make considerable damage for the economic development or ecological environment. So far, some methods have been used to prevent forest firing and have got a lot of achievements, but there are still much space to improve at timeliness and accuracy. Therefore, this paper proposed a method aimed at detecting smoke in forest video. And, codebook model and dynamic characteristics of smoke are put forward to improving the timeliness and accuracy of fire warning method so that the fire can be stopped at early time.Analyzing whether there is smoke in video produced by firing is major work in this paper. Firstly, the background subtraction is selected as moving objects detection method, then three modeling methods such as moving average method, Gaussian mixture background modeling and codebook algorithm are introduced, Owing to the contour's irregularities and the moving disorder of shaking leaves are similar with the dynamic characteristics of smoke in forest smoke video, the experiments comparing foreground result and progressing time among this three methods are made. At last the codebook is determined to be the background updating method. It mainly rules out shaking leaves interference, can extracted the motion area accurately.Then, motion vector direction features is selected as dynamic characteristics of smoke to analyze the main directions of motion and adjust motion detection area whether is smoke, which was based on features of forest smoke video in this paper. Meanwhile, irregularity of contours as the other characteristic is used to detect smoke. Through comparing the effect of two characteristics detecting, the irregularity as the first characteristic, and motion vector as the second is to determine whether there are smoke or not with alternate testing in this article.Finally, in order to verify the effectiveness of this proposed algorithm, there are a number of different scenarios videos which are tested, including smoke or non-smoke, and smoke and disruptors coexistence. The results show that the video smoke detection algorithm in paper can identify smoke accurately, improve real-time and effectively in video and has some anti-jamming capability.