关键词: 背景建模, 时空熵, 高斯分布, 组合高斯模型, 前景分割

Abstract:

In order to increase the update speed of Gaussian models for the background in video surveillance systems,the concept and computation method of the scene moving complexity are devised,according to which a combinational Gaussian background modeling method is devised. In this method, according to the spatio-temporal model of pixels,the scene moving complexity is analyzed and the entropy image of the scene is calculated. Then,this image is segmented into the stable region and the dynamic region by means of the maximum entropy threshold. In the two different regions,two different Gaussian models and corresponding updating algorithms are respectively adopted. Finally,the devised method is used to implement the foreground segmentation of the video sequences with a size of 384 × 288 pixels. The results show that the devised method is of a greater update speed for the background model and can effectively segment moving objects.

Key words: background modeling, spatio-temporal entropy, Gaussian distribution, combinational Gaussian model, foreground segmentation