倪嘉成1,2,,
张群1,2,3,
1.空军工程大学信息与导航学院 西安 710077
2.信息感知技术协同创新中心 西安 710077
3.复旦大学电磁波信息科学教育部重点实验室 上海 200433
基金项目:国家自然科学基金(61631019, 61971434)
详细信息
作者简介:罗迎：罗 迎(1984–)，男，湖南益阳人，空军工程大学信息与导航学院副教授，博士生导师，主要研究方向为雷达成像与目标识别。E-mail: luoying2002521@163.com
倪嘉成(1990–)，男，陕西西安人，空军工程大学信息与导航学院讲师，主要研究方向为SAR成像与目标识别。E-mail: littlenjc@sina.com
张群：张 群(1964–)，男，陕西合阳人，空军工程大学信息与导航学院教授，博士生导师，主要研究方向为雷达成像与目标识别。E-mail: zhangqunnus@gmail.com
通讯作者:罗迎 luoying2002521@163.com
中图分类号:TN957.5计量
文章访问数:3439
HTML全文浏览量:1153
PDF下载量:526
被引次数:0
出版历程
收稿日期:2019-11-27
修回日期:2020-02-26
网络出版日期:2020-03-10
Synthetic Aperture Radar Learning-imaging Method Based onData-driven Technique and Artificial Intelligence
LUO Ying1,2,3,,,NI Jiacheng1,2,,
ZHANG Qun1,2,3,
1. College of Information and Navigation, Air Force Engineering University, Xi’an 710077, China
2. Collaborative Innovation Center of Information Sensing and Understanding, Xi’an 710077, China
3. The Key Laboratory for Information Science of Electromagnetic Waves (Ministry of Education), Fudan University, Shanghai 200433, China
Funds:The National Natural Science Foundation of China (61631019, 61971434)
More Information
Corresponding author:LUO Ying, luoying2002521@163.com
摘要
摘要:对感兴趣目标的数量、位置、型号等参数信息的精确获取一直是合成孔径雷达(SAR)技术中最为重要的研究内容之一。现阶段的SAR信息处理主要分为成像和解译两大部分，两者的研究相对独立。SAR成像和解译各自开发了大量算法，复杂度越来越高，但SAR解译并未因成像分辨率提升而变得简单，特别是对重点目标识别率低的问题并未从本质上得以解决。针对上述问题，该文从SAR成像解译一体化角度出发，尝试利用“数据驱动+智能学习”的方法提升机载SAR的信息处理能力。首先分析了基于“数据驱动+智能学习”方法的SAR成像解译一体化的可行性及现阶段存在的主要问题；在此基础上，提出一种“数据驱动+智能学习”的SAR学习成像方法，给出了学习成像框架、网络参数选取方法、网络训练方法和初步的仿真结果，并分析了需要解决的关键性技术问题。
关键词:合成孔径雷达(SAR)/
SAR成像解译一体化/
SAR学习成像/
数据驱动/
深度学习
Abstract:One of the most important research fields in Synthetic Aperture Radar (SAR) technology is to improve the accuracies of the number, location, classification, and other parameters of targets of interest. SAR information processing can be mainly divided into two tasks: imaging and interpretation. At present, research efforts on these two tasks are relatively independent. Many algorithms have been developed for SAR imaging and interpretation, and they have become increasingly complex. However, SAR interpretation has not been made simpler by improvements in the imaging resolution. The low recognition rate of key targets, in particular, has yet to be adequately resolved. In this paper, we use a “data driven + intelligence learning” method to improve the information processing ability of airborne SAR based on SAR imaging & interpretation integration. First, we analyze the feasibility and main problems of SAR imaging & interpretation integration using a “data driven + intelligence learning” method. Based on the results, we propose a SAR learning-imaging method based on “data driven + intelligence learning” with the goal of producing an imaging network. The proposed learning-imaging framework, parameter selection method, network training method, and preliminary simulation results are presented, and the key technical problems to be solved are identified and analyzed.
Key words:Synthetic Aperture Radar (SAR)/
SAR imaging & interpretation integration/
SAR learning-imaging/
Data driven/
Deep learning
PDF全文下载地址:
https://plugin.sowise.cn/viewpdf/198_357b9b4f-6be5-49c0-b690-dc5cce8cc8d5_R19103
