The Research Of Background Extraction Based On The Event Monitoring

First,This article illustrates current situation about algorithm of background extraction and update at home and abroad, and states several commons theory on moving target detection algorithm,what is more, summarizes the advantages and disadvantages.Which is commonly used in the event monitoring for background extraction methods, such as:statistical histogram method, statistical method, more frames between average method, frame difference method and the method of the gaussian mixture model has carried on the empirical analysis. Based on the analysis results show that the statistical histogram method and the median method to get the initial background is very good, other methods in both the initialization and background update, all appeared the trailing, ghost phenomenon, the background image is poorer.Based on the advantage of statistical histogram method on the background initialized, this paper proposes an improved algorithm. First,statistics the area that former30frames video images per point falls into, looking for the most and more area.According to the difference of two area corresponding to the mean and setting threshold value to judge, to determine the initial background image. Then, calculate the current frame image mean and the standard deviation, their sum as an adaptive threshold whether update background, and according to the setting weight update.background on real-time.Through the empirical analysis, we come to the conclusion that the improved method can not only get a clear initial background image, and in the real-time system update, the background image is superior to the commonly used background updating method, when the frame number reaches a certain number, the background is in stable state, little ghost phenomenon.