Based On The Target Trajectories And The Spatial Distribution Of Behavior Analysis

Visual based action analysis of moving targets is a very important area which is of much concern in computer vision recently. Its research object is the image sequences which containing the moving targets captured by the camera, and it aims at analyzing and understanding the action of moving targets according to detection, tracking and recognition of the moving targets in the image sequence.The main idea of this thesis is to study the action of moving targets according to their move track and spatial distribution. Move track and spatial distribution are macroscopical features of motion which is valuable for understanding the action of targets. Relying on move track and spatial distribution, we can analyze the action of targets accurately as well as analyze if the action is normal or abnormal.Detection and tracking is indispensable for obtaining target's move track. Although target detection and tracking are very complex and difficult problems, there are still several practical methods in use by the advance of the research. These methods could provide a favorable foundation of obtaining the track of moving targets.Several important methods for detection and tracking of moving targets are introduced in the paper. After detailed analysis of the features of the digital environment, several reasons for interpreting why entity construction is needed is put forward, further more, a dynamic entity construction method is studied based on the principal of statistical validity of movement relativity; Based on detailed and comprehensively analysis of the features of move tracks, a new method for move track analysis ground on rotation angle histogram and rotation angle difference histogram is proposed, and also, a new method for track classification based on PCA is proposed, in succession, variety analysis and condensation arithmetic is analyzed and based on which a new method for measuring the similarity of movements is proposed, its of much advantage by comparing with some existing methods; the practical value of the action analysis method based on move track and spatial distribution is discussed, several important action analysis methods is advanced, furthermore, we show up the relevant arithmetic and also the stimulation experiment results which is preferable.