基于FPGA在轨数据处理技术研究

发布时间：2018-05-10 11:06

本文选题：多核流水 + 在轨数据处理　；参考：《北京邮电大学》2015年硕士论文

【摘要】：目前现有的星载计算机处理能力不足,尤其是在进行遥感图像数据处理时,通常将获取到的遥感图像通过压缩,然后下传到地面站来进行处理。由于当前遥感卫星获取到的图像分辨率越来越高,而下传数据的带宽有限,导致现有遥感卫星的高数据获取量和低下传带宽之间的矛盾越来越严重。迫切需要提高星载计算机的处理能力,让遥感卫星具有在轨数据处理能力,提升现有遥感卫星的自主性与灵活性。多核处理技术是通过采用多个处理器来共同完成一项任务,来加快任务的执行效率,根据实现方式的不同,可分为多核并行和多核流水。现有商用FPGA,比如Xilinx公司所提供的部分芯片,具有很高的逻辑容量和处理速度,并且具有一定的抗辐射能力,能够适应严苛的空间环境,可以应用于在轨数据的处理。本文针对在轨数据处理技术的研究现状,提出了一种基于FPGA的异构多核在轨数据处理方法,来进行遥感图像特征点的提取。图像特征点的提取选用尺度、旋转不变SURF算法,并在Xilinx Virtex-4芯片内部实现了该系统,进行相应的实验验证。本文的主要研究内容有：基于FPGA的片上系统及异构多核技术的研究；尺度、旋转不变特征点提取SURF算法及在异构多核系统上的实现；异构多核在轨数据处理系统软硬件的总体设计与具体实现；在轨数据处理系统的实验与数据分析。本文预期达到的目标：综合分析现有星载计算机的处理能力,以及多核处理技术的特点,提出基于FPGA异构多核在轨数据流水处理的方法。通过应用异构多核数据处理技术,解决现有星载计算机处理能力不足的问题。对SURF算法进行研究,对其进行合理的任务划分,将划分好的任务映射到异构多核系统的每个处理器上进行实现,从而在多核间形成流水。进行实验和数据分析,测试异构多核系统的执行效率。
[Abstract]:At present, the existing spaceborne computer processing ability is insufficient, especially in the remote sensing image data processing, the obtained remote sensing image is usually compressed and then sent down to the earth station for processing. Because the resolution of image acquired by remote sensing satellite is getting higher and higher, and the bandwidth of data transmitted down is limited, the contradiction between high data acquisition and low transmission bandwidth of existing remote sensing satellite is becoming more and more serious. There is an urgent need to improve the processing capacity of spaceborne computers, to enable remote sensing satellites to process data in orbit, and to enhance the autonomy and flexibility of existing remote sensing satellites. Multi-core processing technology can be divided into multi-core parallelism and multi-core pipelining according to the different implementation methods by using multiple processors to complete a task together to speed up the task execution efficiency. The existing commercial FPGAs, such as some chips provided by Xilinx, have very high logic capacity and processing speed, and have the ability to resist radiation, which can adapt to the harsh space environment and can be applied to the processing of in-orbit data. In this paper, a heterogeneous multi-core on-orbit data processing method based on FPGA is proposed to extract the feature points of remote sensing image. The image feature points are extracted by scale, rotation invariant SURF algorithm, and the system is implemented in the Xilinx Virtex-4 chip, and the corresponding experimental verification is carried out. The main research contents of this paper are as follows: the research of the system on chip and heterogeneous multi-core technology based on FPGA, the SURF algorithm of extracting the invariant feature points from scale and rotation, and the realization of the algorithm on the heterogeneous multi-core system. The hardware and software design and implementation of heterogeneous multi-core on-orbit data processing system, the experiment and data analysis of in-orbit data processing system. The purpose of this paper is to comprehensively analyze the processing capability of existing spaceborne computers and the characteristics of multi-core processing technology, and propose a water treatment method based on FPGA heterogeneous multi-core on-orbit data flow. By applying heterogeneous multi-core data processing technology, the problem of insufficient processing capacity of existing spaceborne computers is solved. The SURF algorithm is studied, and the task is divided reasonably, and the partitioned tasks are mapped to each processor in the heterogeneous multi-core system for implementation, so as to form a pipeline between the multi-cores. Experiments and data analysis are carried out to test the execution efficiency of heterogeneous multicore systems.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN791;TP274

【参考文献】