Research Of Multi-core Image Processing System And Moving Target Detection And Tracking Techniques

With the increasing demand for visual experience and the wide application of image acquisition equipment with high resolution and high frame rate,the image processing systems are confronting more and more challenges,including the increasing of the sampling rate of processed images,the complexity of processing algorithms and the enhancement of the real-time requirements of processing,which also bringing higher performance requirements for the image processing system.The fast development of on-chip multi-core processor system which integrates multiple Digital Signal Processors(DSP)cores in a single chip makes the image processing system able to deal with the performance requirements of high-speed and real-time image processing.This thesis designs a multi-core image processing system based on DSP+FPGA architecture.It has the powerful computing capability of multi-core DSP and excellent logic control capability of Field-Programmable Gate Array(FPGA)at the same time,and can perform high-speed and real-time image processing tasks outstandingly.In addition,based on the image processing system,the target detection and tracking techniques have been studied,and the target real-time detection and tracking is achieved on the designed multi-core image processing system.The main research contents of this thesis are the design and implementation of multi-core image processing system,the study of target detection and tracking algorithm and its realization and optimization on multi-core image processing system.Specific work is as follows:Firstly,this thesis describes the task source and task requirements of this thesis,and proposes a multi-core image processing system based on DSP+FPGA architecture.At the same time,according to the multi-core image processing system,this thesis describes the hardware design of the multi-core image processing system such as hardware structure building,DSP and FPGA chip selection,the determination of data interaction scheme,the power supply design and related peripheral circuit design;Secondly,this thesis studies the bootloader of multi-core processor,and eventually designs to start the DSP from Nand Flash through I2 C to realize an offline self-loading of the multi-core image processing system;Then,the related target detection and tracking technology has been studied,and thetarget detection algorithm based on Principal Component Analysis and the target tracking algorithm based on Spatio-Temporal Context are selected to be implemented on the multi-core image processing system.The simulation results show that the algorithm is feasible and effective;Finally,the multi-core processing techniques related to multi-core DSP includes storage management,inter-core communication and multi-core application development architecture has been studied.The target detection and tracking algorithm is programmed and transplanted into the multi-core image processing system,and the feasibility of parallelization of modules is analyzed according to the algorithm structure.Efficient parallel processing of algorithms and good code optimization are achieved.The experimental results show that the multi-core image parallel processing system and the optimized codes work well and efficiently.