Research On Real-time Detection And Tracking System Of Objects In Complicated Static Background Based On FPGA

Nowadays, with the development of electronic technology and video image processing,intelligent video surveillance industry has developped rapidly. The detection of movingtargets is the most crucial step of the intelligent video surveillance technology. Detectingmoving objects in video, which has lots of applications in the field of machine vision,vehicle detection, human-computer interaction, objects recognition and tracking, is animportant research field of digital image and video processing. Therefore, to detect movingtargets rapidly, accurately, and effectively has very important practical value andsignificance of the research.In this circumstance, researchers propose many moving objects detection algorithms,such as frame difference algorithm, background difference algorithm, and optical flowalgorithm. At present, these algorithms have matured gradually; many of them haveimplemented in PC machine, and achieved good results. But video monitoring systemsbased on PC cost too much, and are restricted in some special occasions. In view of thissituation, researchers hope to solve this problem with a design of embedded system.FPGA (Field Programmable Gate Array) chips have rich logic resources andhigh-speed operation ability can use in video image processing, which needs large datastorage and computation. Detection and tracking systems based on the FPGA target havevery attractive future in scientific research and engineering applications. FPGA brings newdevelopment to intelligent video surveillance industry.A real-time system of moving targets detection and tracking applied to staticbackground based on FPGA is designed. While moving targets are detected， the systemcan alarm in time. The paper expounds the realization process of the hardware and softwareof this system in detail. The experimental results show that this system can detect and trackthe targets effectively in static background.The hardware platform consists of video capture module, preprocessing module,moving object detection and tracking module, SDRAM memory module, VGA displaymodule, and audio decoding module. The video capture module is responsible for the conversion from analog video to digital video; video preprocessing module is responsiblefor digital video decoding, and transforms it into RGB format data; moving target detectionand tracking module is responsible for the detection and tracking of moving objects in video;SDRAM storage module is responsible for external data cache; VGA display module isresponsible for the display of video processing results; audio decoding module isresponsible for the detection of moving targets; audio decoding module is responsible fortriggering the alarm system. By the cooperation of each module, we can construct a FPGAvideo image processing hardware platform, which is important to the further research in thisfield.The hardware platform of system, background difference method, tracking and otherkey algorithms are designed with Verilog HDL. The system have developed in Quartus II9.0, and simulated in the Modelsim SE6.4. Compared with traditional video monitoringsystem based on PC, FPGA system is not only more flexible, but also cheaper. But thissystem is only used in static background scenarios. While in the wind blowing through trees,grass and water wave dynamic scenarios, the system is not applicable, and to solve thisproblem is the goal of our research team.