天文图像信号处理平台及其关键性技术的研究
文献类型：学位论文
| 作者 | 唐清善 |
| 学位类别 | 博士 |
| 答辩日期 | 2009-05-27 |
| 授予单位 | 中国科学院声学研究所 |
| 授予地点 | 声学研究所 |
| 关键词 | 天文图像信号处理平台 FPGA DSP LINK口 |
| 其他题名 | Study of Image Processing Platform and Key Technology in Astronomy |
| 学位专业 | 信号与信息处理 |
| 中文摘要 | 人类建立地面天文观测站对恒星和太阳黑子等天体目标进行观测时，由于大气湍流的作用，来自天体自身产生或者反射的光波，在经过地球大气以及一系列光学组件的光路传输后，所形成的图像质量将会恶化，目标分辨率会极大降低。在此背景下，基于自适应光学系统的范畴，本文对自适应光学系统中的信号处理与控制单元――天文图像信号处理平台进行了理论上的研究与实践上的探索。其中论文主要的内容可总结如下： 1）天文图像分割、复原算法的研究 图像分割与复原的是天文图像信号处理中的两个关键环节，在Young和Fried等人提出的大气传递函数解析式和Zernike图像复原算法的基础上，提出一种基于流水线技术的动态数据流分割图像的方法，可以减少处理器在图像复原过程中用于数据查找、存储的时间，从而降低系统对图像处理的整体时间。 2）系统平台的设计与实现 提出了基于大规模多DSP和FPGA的天文图像信号处理平台的设计方案，并设计和实现了一个原理系统，验证了该方案的可行性。其中系统设计按照图像数据流的方向和功能可分为：图像采集、分割板，图像复原计算板，图像计算误差校正输出板，多路（256路）数字/模拟转换输出板等部分；各个板之间的数据通信按照ADSP201的LINK口通信协议进行。 3）系统平台在自适应光学中的应用 在自适应光学系统中应用了系统平台，验证了系统的功能。最后的结果表明系统平台能够对输入的模拟测试图像进行分割，在主机上实时显示图像，并能在计算处理单元中实现图像的波前复原算法，以及通过D/A输出实现控制变形镜等功能。 在以上研究内容的基础上，论文的创新点及主要结果如下： 1）在国内，首次提出基于FPGA构建了10端口、高速（单向450MB/S，双向900MB/S）的ADSP201的LINK口数据交换体系结构。 2）提出并实现了一种基于FPGA和流水线技术对天文图像进行实时动态分割的方法，并对处理的延迟时间进行了计算与验证。 3）系统支持对高速、高帧频（2500帧/秒、数据峰值660MB/S）的天文图像的采集，并能通过PCI总线与主机互连，实时显示图像。 4）在以上的基础上，首次实现了大规模多DSP天文图像信号处理平台，该平台具备强大的图像计算处理以及输出控制能力，能满足现有天文观测方面的需要。 总之，本文对图像复原、分割算法进行了理论上的研究，以及对平台的关键技术――多端口的LINK口的设计、图像分割在FPGA上的实现以及减少图像处理延迟等方面进行了实践上的探索，设计实现了一个天文图像信号处理平台系统，得到了平台的初步测试结果。该结果表明在天文图像信号处理领域，基于具有高速并行计算处理能力的DSP以及具有灵活多变的接口形式的高性能的FPGA，建立一个具备开放性及通用性的图像信号处理平台，有助于加速太空探索的步伐。 |
| 英文摘要 | Abstract When traveling through earth's atmosphere, light waves which come from celestial body itself or reflected by others can meet the turbulence. So when we observe the macula or stars in ground astronomical observation station by a series of optical elements, we may find that both the quality and resolution of optical images are reduced. This thesis presents a novel astronomical image processing platform. For adaptive optics system, this processing platform can achieve the function of signal processing and controlling. What I have done during the PHD period are as follows: 1）The research on image segmentation and restoration algorithm. The image segmentation and restoration are the two key links in astronomical image processing. On the basis of atmospheric transfer function analytic expression proposed by Fried and Zernike image restoration algorithm, this thesis presents a method for dynamic data stream image segmentation. Based on pipelining, during the period of image restoration this method can reduce the time of data scan and storage in CPU. So the whole time of image processing in system may be reduced greatly. 2) The design and realization of system processing platform. A design for astronomical image signal processing platform based on massive multi- DSP and FPGA is proposed. According to image data stream’s direction and function, the system design can be divided into image collection and segmentation module, image restoration calculation module, the output module for image calculation error correction, output module for the converting digital to analog signals. The data interconnection among every module is in accordance with ADSP201’s LINK port protocol. 3) The application of system platform in adaptive optics. This processing platform can be applied in adaptive optics system and its function is achieved successfully. The experimental results indicate that in calculation and processing module, for input analog image this processing platform can achieve image segmentation, then display the real-time image in computer. In addition, this platform also can realize the algorithm of image’s wave-front restoration, and control the deform mirror by output of D/A. The main innovations of this thesis are as follows. 1) In China, for the first time, ten ports, high-speed (unidirectional speed is 450MB/s and bidirectional speed is 900 MB/s) ADSP201’s LINK port data exchange system structure based on FPGA is proposed. 2) On the basis of FPGA and pipelining, a novel method for astronomical image’s real-time and dynamic segmentation is presented and realized. In addition, the delay time derived from the image segmentation processing is calculated and verified. 3) This system can collect the high speed astronomical image data whose frame frequency is 2500 frame/s and peak speed is 660MB/s. The system also can be connected to computer by PCI bus and display the real-time, dynamic image. 4) For the first time a processing platform for astronomical image signal based on massive multi-DSP is realized in China. This platform has powerful ability of calculation and extensibility, and can satisfy the requirement of astronomical observation. In conclusion, this thesis does some theoretical research on image segmentation and restoration algorithm. The thesis also carries out a lot of practical research in some key technology. For example, these technologies includes the design of multi-port LINK port, the realization of image segmentation in FPGA and the reduction of delay time in image processing. An astronomical image signal processing platform is realized and the testing results are given. These experimental results prove that in the field of astronomical image signal processing, with the help of high-speed DSP that has the ability of parallel calculation and high-powered FPGA which has flexible ports, the establishment of an open and versatile platform for image signal processing will help to accelerate the pace of space exploration greatly. |
| 语种 | 中文 |
| 公开日期 | 2011-05-07 |
| 页码 | 124 |
| 源URL | [http://159.226.59.140/handle/311008/474]  |
| 专题 | 声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 唐清善. 天文图像信号处理平台及其关键性技术的研究[D]. 声学研究所. 中国科学院声学研究所. 2009. |
入库方式： OAI收割
来源：声学研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。