粟毅,
陈思伟,
国防科技大学电子科学学院电子信息系统复杂电磁环境效应国家重点实验室 长沙 410073
基金项目:国家自然科学基金(61771480)，湖南省自然科学基金(2020JJ2034)，湖湘青年英才项目(2019RS2025)，装备预研基金项目(61404160109)，国防科技大学科研计划重点项目(ZK18-02-14)
详细信息
作者简介:崔兴超(1994–)，男，山东人，国防科技大学博士研究生。主要研究方向为极化SAR和目标检测。E-mail: nudt_cui@163.com
粟毅：粟 毅(1961–)，男，山东人，博士，国防科技大学教授，博士生导师。主要研究方向为微波成像、遥感应用、探地雷达、超宽带雷达系统等。E-mail: y.su@yeah.net
陈思伟(1984–)，男，四川人，博士，国防科技大学电子科学学院****，硕士生导师。主要研究方向为极化雷达成像、目标识别、电子对抗等。E-mail: chenswnudt@163.com
通讯作者:陈思伟 chenswnudt@163.com
责任主编：杨健 Corresponding Editor: YANG Jian中图分类号:TN958
计量
文章访问数:569
HTML全文浏览量:343
PDF下载量:156
被引次数:0
出版历程
收稿日期:2020-12-19
修回日期:2021-02-02
网络出版日期:2021-02-22
Polarimetric SAR Ship Detection Based on Polarimetric Rotation Domain Features and Superpixel Technique
CUI Xingchao,SU Yi,
CHEN Siwei,
The State Key Laboratory of Complex Electromagnetic Environment Effects on Electronics and Information, Changsha 410073, China
Funds:The National Natural Science Foundation of China (61771480), The Natural Science Foundation of Hunan Province (2020JJ2034), The Youth Talents Project of Hunan Province (2019RS2025), The Equipment Pre-Research Foundation (61404160109), The Key Research Projects of National University of Defense Technology (ZK18-02-14)
More Information
Corresponding author:CHEN Siwei, chenswnudt@163.com
摘要
摘要:对海监视是极化SAR的重要应用，密集区域的舰船目标检测是当前面临的主要挑战之一。舰船密集区域受多目标串扰，传统的恒虚警率(CFAR)检测滑窗难以选取纯净的海杂波样本用于确定检测门限，将导致检测性能下降。针对这一问题，该文从特征提取和检测器设计两方面出发，提出一种融合极化旋转域特征和超像素技术的极化SAR舰船检测方法。在特征提取方面，雷达目标的后向散射敏感于目标姿态与雷达视线的相对几何关系，由此带来的散射多样性隐含信息可通过极化旋转域分析进行挖掘。该文利用极化相关方向图及导出的一系列极化旋转域特征，根据目标杂波比(TCR)分析，优选TCR最高的3个极化特征量用于构建目标检测器。在此基础上，该文在检测器设计方面提出了一种基于K均值聚类的杂波超像素筛选方法，有效避免了密集区域舰船目标对邻近杂波的影响，基于筛选的杂波像素点得到舰船目标CFAR检测结果。基于Radarsat-2和高分三号星载全极化SAR数据的对比实验表明，所提方法能有效实现密集区域舰船目标检测，检测品质因数达到95%。
关键词:极化合成孔径雷达/
极化旋转域/
超像素/
特征融合/
舰船检测
Abstract:Sea surveillance is an important application of polarimetric Synthetic Aperture Radar (SAR), but ship detection in dense areas remains a major challenge. Due to the crosstalk of multiple targets in dense ship areas, it can be difficult to collect pure sea clutter samples for threshold determination when using the traditional Constant False Alarm Rate (CFAR) moving window, which decreases the detection performance. To address this issue, in this paper, a polarimetric SAR ship detection method is proposed based on polarimetric rotation domain features and superpixel technique, with consideration of both feature selection and detector design. For feature selection, the backscattering of radar targets is sensitive to the relative geometry between the target orientations and the radar line of sight. The information hidden in this scattering diversity can be mined using polarimetric rotation domain analysis, from which the polarimetric correlation pattern and a set of polarimetric rotation domain features are obtained. Target-to-Clutter Ratio (TCR) analysis is conducted, and the three polarimetric features with the highest TCR values are selected for successive target detection. On this basis, a clutter superpixel selection method is developed for detector design based on K-means clustering, which effectively circumvents the influence of dense ship targets on near sea clutter. CFAR ship detection results can be obtained based on the selected clutter samples. Experimental studies on spaceborne Radarsat-2 and GaoFen-3 full polarimetric SAR datasets indicate that, the proposed method can effectively detect dense ship targets with 95% higher figures of merit.
Key words:Polarimetric Synthetic Aperture Radar (SAR)/
Polarimetric rotation domain/
Superpixel/
Feature fusion/
Ship detection
PDF全文下载地址:
https://plugin.sowise.cn/viewpdf/198_ed3033d5-51fb-4011-b7d9-a35592949e61_R20147
