There are three important approaches in human motion analysis, include moving objects detecting, tracking, and recognizing the human activities, among them moving objects detecting is foundation of tracking and behaviors recognition. So this paper introduces the first step briefly and emphasizes on the research of pedestrians tracking and behaviors recognition. In the respect of pedestrians tracking, first we suppose an algorithm called mantle ratio which can automatically select the largest characteristic region. Secondly we choose the weighted color histogram in this region as the target characteristic of human body. Finally Bhattacharyya distance is employed to estimate the similarity between the target'color model and the particles. It can be regard as the powerful fact to measure the weight. Finally we can track pedestrians in the framework of particle filtering theory automatically and real-timely. For human activities recognition this paper adopts HMM(hidden Markov model) to analyze the temporal-spatial feature of human contour. While extracting feature, first we construct an integrative contour and make its width Single-pixel. Then the single-pixel-width contour is unfolded to its corresponding one-dimensional signal of distance. And this normalized one-dimensional signal is selected as feature vector. At last in this foundation focusing on the application of intelligent security monitoring, this paper designs and implements a relevant network human motion analysis system. |