Research On Optimization Of Real-time Target Tracking Algorithm Based On Miniaturized DSP Platform

Real-time video target tracking technology is a hot and difficult problem in the field of computer vision.Real-time target tracking systems based on the miniaturized embedded platform and using real-time video target tracking algorithm as the core of software have important prospects and application value in many fields such as virtual reality,intelligent transportation system and weapon guidance.In this paper,according to the requirements of practical applications,a miniaturized,low-cost,low-power embedded DSP image processing hardware platform and a target tracking algorithm software development framework are constructed.Then,three typical real-time target tracking algorithms are studied in depth,and optimized based on their shortcomings towards respective applications and DSP system,and then the algorithm programs are further optimized based on the features and advantages of the DSP platform.Experimental results show that the tracking system in this paper can track the target object in real time and has good theoretical and practical value.The main work of this paper is summarized as follows:1.Through the reading of relevant documents and materials,understand and analyze the development and research status of real-time video target tracking technology,and expound the unique advantages of TI-DSP in video image processing.2.According to the demand of the system for miniaturization,low cost and low power consumption,TI DM6437 chip is taken as the core processor and the system hardware design is completed based on its hardware peripheral modules.Based on TI's DVSDK and DSP / BIOS embedded real-time operating system,the software development framework of the target tracking algorithm is constructed according to the actual system functional requirements.3.The target tracking algorithm based on template correlation,compressed target tracking algorithm based on compressed sensing and target tracking algorithm based on particle filter are studied.According to their respective tracking robustness,accuracy and real-time performance of the DSP platform,the paper proposes several improvement measures in terms of algorithm strategy and calculation,and then implements and validates on the DSP platform.The experimental results show that the improved algorithm can get better performance on the DSP platform than the original algorithm.4.After the completion and ensure of the correctness and accuracy of the algorithms,in order to achieve better real-time tracking performance,taking the target tracking algorithm based on particle filter as an example,three aspects of optimization measures:the optimization settings of the compiler,DSP processing hardware-layer optimization and TI's official support library and the corresponding optimization results are introduced.The results show that after these optimization measures of program,the algorithm in the low-cost DSP platform can achieve better processing speed.