Research And Implementation Of Video Target Recognition And Tracking System Based On ARM Embedded Platform

Moving target identification and track technology is regarded as core technology in intelligence video monitor system and ARM embedded technology has been widely applied in all possible fields in recent years due to its characteristics such as low cost, low power consumption, high performance and highly specialized extent. Therefore, it is the current development trend of intelligence monitor system to combine motion identification and track technology and ARM embedded technology. Besides, traditional moving target identification and track technology based on X86 main board cannot meet miniaturized and specialized requirement in both modern industry and civil fields and thus this project has much practical significance.In this thesis, a solution for video target identification and track based on ARM embedded platform is proposed and it is studied to combine ARM embedded platform with moving target identification and track system. Firstly, after analyzing the system requirement, this thesis designs general system function and demonstrates the necessity of adopting ARM embedded platform. Furthermore, this thesis sets up hardware and software platform and accomplishes the customization of Linux kernel, transfer of operation system and set-up of OpenCv computer visual database. And then, in the field of moving target identification algorithms, three target identification algorithms are studied that include frame difference method, background difference method and Surendra automatic background update method. In the field of moving tracking algorithms, CAM shift algorithm and Kalman filtering algorithm are studied and a Kalman filtering algorithm based on geometry characteristic and histogram matching is presented according to Kalman filtering algorithm. Then this thesis conducts simulation tests for the algorithms via OpenCv computer visual database code on ARM embedded platform under Linux environment and analyzes the advantages and disadvantages of each identification algorithm and tracking algorithm. By analyzing and comparing test results, this thesis finds out the identification algorithm and tracking algorithm that have the highest matching extend with ARMS3C2440 embedded development board. This thesis eventually implements the identification and track of video moving target on ARMS3C2440 embedded development board.