Motion Trajectory Identification Under Multi-Camera Surveillance System

Intelligent video surveillance system plays an important role in our daily routine life.Multicamera network intelligent video surveillance is one of the future directions of video surveillance system.In multi-camera surveillance system several motion related techniques are integrated,including motion detection,object tracking,object matching,object modeling,object identification,and object classification.Several multi-tracking identification systems have been proposed to detect the group behavior through determining motion trajectory of the individual pedestrian.In this thesis,a matrix based temporal recursive positional identification method is proposed and extended to determine and track the trajectory of each person including the person who is newly entering or leaving the observation region during the observation time period.The surveillance area is divided into different observation zones and each zone has one camera which detects the presence of each person.Given the geometrical structure of the observation zone,the topological relation which is represented by an adjacent matrix is established.The extended matrix-based algorithm divides the topological relation into three sub-relations.Furthermore,the potential movements are represented by newly entering-into-zone,exiting-from-zone,and moving-between-zones persons.The crowd behavior analysis based on video recordings still seems challenging and time-consuming,and the observation range of the camera system is also limited.Multiple targets identification systems are implemented for crowd's behavior analysis in this paper.Moreover,the extended matrix-based algorithm is used to detect and analyze group behavior of the huge crowd.A pattern mining technique is used in the proposed method to detect the large groups and analyze the group behaviors simultaneously.Extensive experimental results conclude the effectiveness and efficiency of the proposed extended algorithm.