Reserch On Object Detection And Tracking System In High-definition Video

As more and more attention has been paid to urban security in recent years, video surveillance becomes closely related to peoples' daily life. And video surveillance is not only widely used in civilian areas, but also plays a big role in the field of military. The rapid development of digital image processing and digital circuit technology accelerates the change of video surveillance from traditional analog video to high-definition digital video in the meantime. Object detection and tracking are the most essential part of video surveillance, and they have received a lot of research and development both at home and abroad. In practical application, the movement of surveillance camera may occur due to various factors, such as the shake of video cameras itself or the cameras are installed on a mobile platform, which brings great challenge to video surveillance.Hence, the studying of object detection and tracking under the state of camera motion becomes extremely important, but meanwhile, it is also difficult. This paper studied and accomplished the system of real-timely object detection and tracking in high-definition video under the state of camera motion. The main work is summarized as follow:After studying of existing technologies and achievements, this paper designs the whole system based on practical application requirement and divides the system into function module. The camera array, which consists of laboratory-made camera,can expand the monitoring field. The FPGA can perform complex logic and timing design because of its high degree of integration, and deal with high-definition video data quickly owe to its powerful parallel processing capability, hence, the FPAG can guarantee real-time systems. In addition, the PC can control the system conveniently and process complex tracking algorithm. So a composite system combining the advantages of FPGA and PC is designed, which consists of hardware circuit platform based on FPGA, laboratory-made cameras and PC terminal.Focus on the moving camera, which causes changes in adjacent image background frame, this paper adopts method of global motion estimation and compensation to achieve background correction. Through the study of image block matching method,this paper proposes an improved block matching algorithm which combines the advantage of ES(exhaustive search) and TSS(three step search). With respect to ES, the new method reduces a large amount of meaningless computation, and avoids thedisadvantage of TSS which is easily failed into local optimum, and this method is suitable with hardware design. Next, takes the frame subtraction method to detect object with the corrected image, then filter the differential result, get the object information from the binary image. Finally, realizes the algorithm simulation on FPGA through Verilog programming, and extracts the object in the scene successfully.PC terminal plays the role of control center of the whole system, and it completes object tracking through the object information which sends from hardware platform.Users can interact with PC terminal through application UI, and PC terminal sends the control orders, including object detection, video display, cameras configuration and so on, to hardware platform. Meanwhile, PC terminal is required to process the object information to realize coordinate transformation, object determination, object tracking,and finally accomplish the tracking part.