DSP-based Image Processing Platform Design And Algorithm Implementation

In recent decades,digital video image processing technology has developed rapidly.It has been widely used in the military field,traffic and security field,medical field,and the latest artificial intelligence technology.The traditional image processing system based on general CPU can not meet the needs of practical application for various reasons,but the DSP as a dedicated digital signal processing chip,with high integration,fast operation,the user can be customized and other features are widely used.TI's C6000 series DSP is designed for video processing design of high-performance processor.This thesis selected TMS320DM642 chip is one of the outstanding performance of a fixed-point chip.Based on this chip,the main work of this thesis is as follows:(1)Completed the image processing hardware platform design and construction.Based on TMS320DM642 processor,this thesis designs a set of basic platform for image processing.The platform is mainly composed of three modules: image acquisition,image processing and image display module.The image processing module is mainly used in industrial CCD and video decoder chip;image processing module based on TMS320DM642 core chip,including EMIF memory interface,EDMA and other modules;image display module by the image coding chip and LCD display.This thesis focuses on the principle of real-time processing of processor chip in image processing module and data processing by EDMA.(2)Based on the deep research on the image processing algorithm,the algorithm simulation experiment is carried out to verify the correctness of the algorithm.Then the algorithm is transplanted to the DSP processing platform and do a lot of debugging for the characteristics of DSP processors.In this thesis,image preprocessing,image feature extraction,image segmentation,image moving target detection and tracking algorithms in image processing are studied in depth.These algorithms were implemented in C language based on the x86 architecture and the VS2010 platform.Finally,the results of the simulation experiment are analyzed in detail.(3)The system is optimized based on the goal of optimizing the system's real time and accuracy.According to the characteristics of C64 x DSP chip,the thesis focuses on the optimization of DSP software development.By analyzing the time-consuming situation in the process of executing the program,different optimization strategies are used to optimize the key codes that affect the running speed of the system.Finally,a better real-time processing effect is achieved.(4)Encapsulates the algorithm and integrates the system.At the end of this thesis,the algorithm is encapsulated into an algorithm library.A moving object tracking system is designed based on the image processing platform and algorithm library to Evaluate the performance of the hardware platform and algorithm library.The results show that the designed platform has good real-time performance and robustness,can meet the needs of real-time video image processing.