Target Position And Trajectory Analysis Based On Distributed Vision In Home Environment

Visual based action analysis of moving targets is a very important research topic which received lot of attention recently. It can be widely applied in many fields such as intelligent monitoring, traffic management and home service. It can detect the abnormal behavior accurately and make timely warning. The main idea of this thesis is to study the target tracking, positioning and the action of moving targets according to their move track and spatial distribution in home environment. The target tracking and positioning is the foundation of the trajectory analysis. Based on moving track and spatial distribution, we can analyze the action of targets as well as analyze if the action is normal.After comparison of the commonly used methods for motion detection, a method about background subtraction based on mixture Gaussian models is proposed. This method can build the background model quickly and remove the shadow accurately. Through the analysis of the advantages and disadvantages of the particle filter and mean shift algorithm, the mean shift algorithm is embedded into the particle filter algorithm. The experimental results show that the combination of mean shift and particle filter works well in complicated home environment.In this paper we use the binocular stereo vision to locate the target, and then a new method about camera calibration is proposed. The intrinsic parameters and the mapping matrix is calculated separately. Compared with other positioning method, this system has high positioning accuracy. A large number of experiments prove that the error of the algorithm is low.Trajectory analysis is an important direction to behavior understanding, and in special environment people’s movement often follows some laws. In this paper the environment is divided into different key points and key areas. Based on this idea, a new method for motion tracking analysis based on PCA and Mixture Gaussian models is proposed. The experimental results show that the method has good accuracy in abnormal trajectory recognition.