RVM-Based Classification of the Network Video Surveillance System

In this paper, a new based on RVM target recognition algorithm is proposed. The algorithm used color and texture characteristics of each frame of video in the time domain modeling, the use of the video frame sequence of images to enhance the relationship between the space division of the accuracy of the classification algorithm to achieve through the RVM MAP, in order to achieve in complex cases, the background video to identify the precise objective of the human body. In strong light case, as well as complex multi-objective circumstances, such as background on the video sequence of the human body target partition, the algorithm used in this article referred to the division better than the cumulative background subtraction as well as the segmentation approach Gaussian Mixture Model.