静止轨道干涉式毫米波大气探测仪并行仿真系统设计

发布时间：2018-03-22 09:56

本文选题：静止轨道　切入点：可见度函数　出处：《中国科学院大学(中国科学院国家空间科学中心)》2017年硕士论文　论文类型：学位论文

【摘要】：在对静止轨道大气探测需求的背景下,静止轨道干涉式毫米波大气探测仪(GIMS)应运而生。相比于太阳同步轨道卫星探测,静止轨道探测器相对地球静止,可以获得极高时间分辨率以及大观测范围,因此在天气预警和快速变化天气现象(如台风、强降雨等突发灾害性天气现象)的即时和短期天气预报上具有更好的时效性,能够发挥更大的作用。此外,静止轨道毫米波大气探测仪与红外探测器相结合,能有效弥补红外探测无法穿透云层而无法观测到云团内部温度和湿度结构以及降水参数的缺陷,从而实现在有云天气条件下大气温湿度的垂直分布探测,提供云团内部的三维结构信息,有效地监测大尺度降水结构及其变化。随着微波遥感技术的不断发展,高精度定量化遥感对微波探测技术指标的要求越来越高,为满足静止轨道对地球全圆盘观测以及高空间分辨率的需求,GIMS星载系统硬件设计上变得越来越复杂,具体表现为阵列规模增大,天线单元数增多,同时辐射灵敏度提高,亮温观测目标扩展为全球尺度范围。而探测系统运算复杂度随天线单元数的平方关系增长,由此带来的计算量巨大、仿真效率低下的问题亟待解决。为解决上述问题,本文提出了基于GPU并行处理的仿真思想,尝试利用并行计算提高系统的仿真效率。论文首先分析了静止轨道干涉式毫米波大气探测系统仿真的复杂度,找出整个仿真系统运行的瓶颈,提出了利用GPU的并行计算能力提高仿真系统时间性能的必要性。并且,进行了可行性实验,分析了实验结果,证明了基于GPU仿真系统的可行性。随后,对原始仿真系统进行了并行化处理,并在不同的仿真平台上进行了实验验证。结果表明,并行处理后的系统仿真效率较传统MATLAB仿真平台提高了6倍以上,较MATLAB-GPU仿真平台提高了5倍以上,加速效果良好。同时,为验证GPU并行处理的优势,本文随之进行了单CPU平台的串行仿真以及MATLAB多核实验。实验结果证明了并行处理在处理大规模数据集上具有高效的计算能力。最后,本文将并行处理模块与前后仿真模块间加入接口函数,使整个仿真系统形成一个完整的仿真流程,各模块之间能够互相协调,共同完成GIMS系统的仿真任务。本文所做的工作成功解决了综合孔径辐射计仿真复杂度高、仿真效率低下的问题,提供了一种新的且实用性高的仿真平台,对整个GIMS系统的仿真工作具有重大的意义。
[Abstract]:Against the background of the demand for geostationary orbit atmosphere detection, the geostationary orbit interference millimeter wave atmospheric detector (GIMS) came into being. Compared with the solar synchronous orbit satellite exploration, the geostationary orbit detector is relative to the geostationary. Very high temporal resolution and large observation range can be obtained, so it has better timeliness in both immediate and short-term weather forecasting for weather early warning and rapidly changing weather phenomena (such as typhoons, heavy rainfall and sudden severe weather phenomena). In addition, the combination of the static orbit millimeter wave atmospheric detector and the infrared detector can effectively compensate for the imperfections in the temperature and humidity structure of the cloud cluster and the precipitation parameters that the infrared detection cannot penetrate the clouds, Thus, the vertical distribution of atmospheric temperature and humidity can be detected under the condition of cloud weather, the three-dimensional structure information of cloud cluster can be provided, and the large-scale precipitation structure and its changes can be effectively monitored. With the development of microwave remote sensing technology, The requirement of high precision quantitative remote sensing for microwave detection technology is becoming higher and higher. In order to meet the demand of geostationary orbit for the whole disk observation of the earth and the high spatial resolution, the hardware design of Gims spaceborne system has become more and more complicated. In particular, the array size increases, the number of antenna elements increases, and the radiation sensitivity increases, and the bright temperature observation target expands to a global scale. The computational complexity of the detection system increases with the square relation of the antenna number. In order to solve the above problems, a simulation idea based on GPU parallel processing is proposed in this paper. This paper attempts to improve the simulation efficiency of the system by using parallel computing. Firstly, the complexity of the simulation of the static orbit interference millimeter wave atmospheric detection system is analyzed, and the bottleneck of the whole simulation system is found out. The necessity of using the parallel computing power of GPU to improve the time performance of the simulation system is put forward. Furthermore, the feasibility experiment is carried out, and the results of the experiment are analyzed to prove the feasibility of the simulation system based on GPU. The original simulation system is parallelized and verified by experiments on different simulation platforms. The results show that the system simulation efficiency after parallel processing is more than 6 times higher than that of the traditional MATLAB simulation platform. Compared with the MATLAB-GPU simulation platform, it is more than 5 times higher, and the acceleration effect is good. At the same time, in order to verify the advantages of GPU parallel processing, In this paper, serial simulation of single CPU platform and MATLAB multi-core experiment are carried out. The experimental results show that parallel processing has high efficiency in processing large scale data sets. Finally, In this paper, the interface function is added between the parallel processing module and the simulation module before and after, so that the whole simulation system can form a complete simulation flow, and the modules can coordinate with each other. This paper successfully solves the problems of high complexity and low efficiency of synthetic aperture radiometer simulation, and provides a new and practical simulation platform. It is of great significance to the simulation of the whole GIMS system.
【学位授予单位】：中国科学院大学(中国科学院国家空间科学中心)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P407.7;TP391.9

【相似文献】