Dissertation > Excellent graduate degree dissertation topics show

Research on Real-time Image Processing Platform Based on Multi-core DSP

Author: SunKeLin
Tutor: WuQinZhang; PengZhenMing
School: University of Electronic Science and Technology
Course: Signal and Information Processing
Keywords: real-time image processing multi-core DSP CPCI Express fabric real-timeimage fiber-optic transceiver image fiber-optic transmission protocol
CLC: TP391.41
Type: PhD thesis
Year: 2012
Downloads: 1011
Quote: 3
Read: Download Dissertation

Abstract


To meet the requirements of high frame rate, high-resolution real-time imageprocessing in the photoelectric imaging tracking and measuring system, as well as therequirements of complex multi-model image tracking algorithm under complexbackground, a high-performance multi-core DSP real-time image processing platform isstudied. Firstly, the history and current status of the development of real-time imageprocessing platform architecture, and the core advantage of multi-core DSP technologyin the field of high performance computing are briefly introduced, and then theperformance characteristics of the real-time image processing platform in thephotoelectric imaging tracking and measuring system are analyzed.The limitations of architecture and bottlenecks of processing performance ofexisting real-time image processing platform are studied firstly, and then the internetand storage structure of the processing platform are analyzed in this dissertation. Bycomparing to its architecture inheritance of the image processing platform, a newgeneration of compact coupled CPCI Express bus standard is selected as the researchplatform architecture. By introducing standards and electrical characteristics of CPCIExpress module board, a design of CPCI Express backplane and switching architectureare discussed in detail. The master processing module and switching module of CPCIExpress are analyzed, as well as the compact coupled CPCI Express real-time imageprocessing platform constituted by CPCI Express peripherals. In addition, the internalstructure principle and the PCI Express interface technology of the processing platformare described in detail.Combined with the performance requirements of processing platform for imageprocessing tasks in different application environment, the requirements of high-speedreal-time image processing and multiple image parallel processing for data flowstructure, and the requirements of multi-model image tracking algorithm for platformhardware architecture are analyzed, as well as the requirements of multi-source imagedata fusion processing for the data interaction between different modules and chips.Besides, the theoretical analysis of the data bandwidth required by the processing algorithms is conducted.For the data input/output bandwidth bottleneck problem of real-time imageprocessing platform in photoelectric imaging tracking and measuring system, areal-time image fiber-optic transceiver is designed. In the design, the high-speed cameralink image data can be transmitted directly to the real-time image processing platform.By using fiber-optic high-speed serial transmission interconnection technology andfiber-optic slip ring technology, multi-channel real-time image data can be transmitteddown to the machine with high bandwidth, high reliability and high stability. On thisbasis, a custom image fiber-optic transmission protocol is developed. The protocolmakes the image data interconnected among the processing subsystem, displaysubsystem and recording subsystem, and then the real-time transmission, processing,displaying and recording of the image data can be achieved.On the basis of analysis of multi-model image tracking algorithm, and combinedwith the FPGA flexible parallel water structure and the storage resource sharingcharacteristics of multi-core DSP, a real-time image processing module based on FPGAand multi-core DSP architecture is designed. For data stream processing in the imageprocessing module, the FPGA is used as a co-processor for image parallel processing,and multi-channel parallel structure can be constituted combined with multi-core DSP.Therefore, by using SIMD technology of FPGA and VLIW technology of multi-coreDSP, a higher processing performance can be obtained.After the design of compact coupled internet and fiber-optic high-speed serialtransmission interconnection network between multi-core DSP image processingmodules, and interconnection structure between chips, the implementation ideas andmethods for image parallel processing technology used on the real time imageprocessing platform are described. From the aspects of the chip memory resourcesharing structure of multi-core DSP, and the resource sharing structure between FPGAand multi-core DSP, the parallel processing performance of the image processingmodule are illustrated. Furthermore, to excavate parallel processing performance in theimage processing platform, the parallel structure of many multi-core DSP imageprocessing module is expanded in this dissertation.At last, from the aspects of computing performance, data throughput, storagebandwidth and power consumption, the performance of a single block multi-core DSP real-time image processing module and the whole multi-core DSP real-time imageprocessing platform are concluded. The applications of the real time image processingplatform in photoelectric imaging tracking and measuring system are analyzed. In viewof the development trend of multi-core DSP will act as an important role in the field ofhigh performance computing in the future, the technology foresight of softwarearchitecture based on multi-core parallel processing in inner-chip, multi-core parallelprocessing between chips and modules is proposed.

Related Dissertations

  1. Research on the Image Real-Time Acquisition, Storage and Image Processing System,TP391.41
  2. Design and Implementation of Parallel Image Matching Algorithm Based on MDSP,TP391.41
  3. Multi-Core DSP Implementation of Femtocell Receiver of WCDMA System,TN929.533
  4. Multi-Core DSP Design of TD-LTE Femto-Cell Downlink Transmitter Processing,TP368.12
  5. LTE downlink channel estimation and DSP Implementation,TN929.5
  6. LTE Channel Coding System Research and Implementation,TN929.5
  7. LTE downlink adaptive technology research and implementation,TN929.5
  8. Software Implementation of Rafar Imaging Algorithm,TN957.52
  9. Design and Implementation of TD-LTE Random Access Processing System,TN929.5
  10. TD-SCDMA uplink synchronization algorithms and DSP Implementation,TN929.533
  11. TD-SCDMA system channel estimation technology research and DSP,TN929.533
  12. Design and Implementation of Parallel MPEG-4 Decoder Based on Multi-core DSP,TN764
  13. CoStar Ⅱ based heterogeneous multi-core DSP Design and Implementation,TP368.12
  14. Design and Implementation of Video Codec System on the Embedded Multi-Processor,TN919.81
  15. Optimization Techniques of Cache in Chip MultiThreading,TP332
  16. Key Techniques on Data Stream Speculation for Heterogeneous Multi-Core Digital Signal Processors,TP368.1
  17. Key Techniques of Network-on-Chip Design for Multi-Core System-on-Chip,TN47
  18. The Hardware Design of Multi-core DSP Image Processing System,TP391.41
  19. Design and Implementation of Multi-core Digital Signal Processing Platform,TN911.7
  20. Study on the restoration of real-time wavefront based on multi core DSP,O436
  21. Research and Implementation on Language Recognition System Based on GSV-SVM,TN912.34

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile