基于对数正态ρ-度量的SAR图像舰船检测提升方法

151 浏览量更新于2024-08-26 收藏 876KB PDF 举报

本文主要探讨了"基于对数正态rho-metric的SAR图像舰船检测"这一主题，发表在2018年9月的IEEE GEOSCIENCE AND REMOTE SENSING LETTERS上。作者Meng Yang和Chunsheng Guo都是IEEE成员，他们将信息几何学的研究成果应用于舰船检测问题，特别是针对合成孔径雷达（SAR）图像中的复杂背景。信息几何学源于对概率分布构成的流形的几何结构的研究，其中引入了一种独特的黎曼度量——Fisher信息度量。然而，在SAR图像中，传统的Fisher度量方法在处理非均匀和非稳定杂波背景时存在局限性，这对舰船检测的精度构成了挑战。为解决这个问题，论文提出了一种改进的Fisher度量，即对数正态rho-metric，它结合了对数正态模型与黎曼几何的优势。对数正态rho-metric旨在增强SAR图像中目标与背景之间的对比度，减少假警报的发生，从而提高舰船检测的性能。通过这种方法，研究人员能够有效地应对SAR图像中的复杂环境，使得船舶在杂乱的雷达回波中更加容易被识别。论文的关键内容可能包括理论框架的建立、对数正态rho-metric的具体计算公式和算法设计、实验数据的处理与分析，以及对比传统方法的性能提升效果。此外，研究可能还涉及对不同类型的SAR图像（如极化、多通道等）的应用测试，以及对实际应用中可能遇到的挑战和优化策略的讨论。总结来说，这篇研究论文是一项创新性的尝试，将信息几何学的理论与SAR图像处理相结合，为舰船检测提供了一种更有效的方法，对于提升SAR图像分析的精确性和鲁棒性具有重要意义。

1372 IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 15, NO. 9, SEPTEMBER 2018

Ship Detection in SAR Images Based

on Lognormal ρ-Metric

Meng Yang , Member, IEEE, and Chunsheng Guo, Member, IEEE

Abstract— Information geometry emerged from the study of

the geometrical structure of a manifold of probability distribu-

tions. It deﬁnes a Riemannian metric uniquely, which is Fisher

information metric. However, the classical Fisher metric method

has a limitation that it does not overcome the problem of

inhomogeneous and nonstationary clutter for ship detection in

synthetic aperture radar (SAR) images. By combing lognormal

model and the Riemannian geometry, this letter presents a mod-

iﬁed Fisher metric (lognormal ρ-metric) based on information

geometry. Experiments show that lognormal metric for ρ in

the ship detection from SAR images can be tuned to increase

the performance of improving contrast between the object and

background, and reducing false alarms.

Index Terms— Information geometry, lognormal ρ-metric,

synthetic aperture radar (SAR) images, ship detection.

I. INTRODUCTION

YNTHETIC aperture radar (SAR) is an active remote

sensor, which makes it functional in all weather and day-

and-night operating conditions. SAR sensors have become an

important technology for the observation and surveillance of

the sea surface. In particular, the improved r esolution of SAR

images and the large amounts of open data being available

nowadays promote the development of new automatic ship

detection too ls in time. This is also evident from a large

number of research articles published on this subject. Marino

and Hajnsekand [1] tend to approach this topic from various

perspectives.

Among these, there are two fundamentally different means

of ship detection in SAR images: adaptive threshold and

feature representation. For the adaptive threshold algorithms,

constant false alarm rate (CFAR) detection of targets in heavy

sea clutter background is an important unit in the ﬁeld of

automatic target recognitio n of SAR images [2]. However,

the difﬁculty of CFAR methods is to detect the ships appearing

on sea surfaces of varying degree of homogeneity and, at the

same time, to keep the false-alarm rate (FAR) down to a prede-

ﬁned level. And, there are also many possible variations of the

shape and size of the target, guard, and background windows.

Manuscript received February 12, 2018; revised May 1, 2018; accepted

May 14, 2018. Date of publication May 30, 2018; date of current version

August 27, 2018. This work was supported by the Chinese National Natural

Science Foundation under Grant 61501152. (Corresponding author:

Meng Yang.)

The authors are with the College of Communication Engineering, Hangzhou

Dianzi University, Hangzhou 310018, China (e-mail: yangmeng@hdu.edu.cn).

Color versions of one or more of the ﬁgures in this letter are available

online at http://ieeexplore.ieee.org.

Digital Object Identiﬁer 10.1109/LGRS.2018.2838043

Therefore, the ﬁnal choice will be determined by the detection

problems at hand. Due to its simplicity and effectiveness, it is

expected that the distribution-based detection performance will

be improved by incorporating some novel analysis techniques.

Another approach to detect targets in a radar clutter

background is to extract the features of ship images [3].

Wang et al. [3] focus on the feature analysis of ships in high-

resolution SAR images and p ropose an optimizing algorithm

for ship detection. However, each feature representation for

ship detection has its own strengths and weaknesses, which

should be evaluated in relation to the practical application

scenarios. In addition, the resolution of most satellite SAR

images is often not high enough to extract detailed ship

inform ation. Therefore, some difﬁculty exists in employing

the ship features for d etection.

Information geometry has emerged from investigating the

relationship between differential geometry and information

theory. Theory of information geometry has experienced a very

fast development in the past few d ecades [4]. It has penetrated

into many other mathematical disciplines and has been making

great progress in theory and application in recent y ears [5].

Apart from classical Riemannian descendants, there are many

special metrics associated with various geometrical structures

in the theory of information geometry, sometimes in a rather

unexpected and subtle way. It provides the information sci-

ences with a more efﬁcient method for constructing signal/data

abstract model. It has been usefully applied to various areas

and provides them with a new perspective to view the structure

of the investigated systems.

Our principal tool is the statistical manifold and metric-

topological constructions based on information geometry. The

aim of this letter is to show the beneﬁts of statistical manifolds

suitable for feature analysis in SAR imagery and based on

information geometry theory. Statistical manifolds are the

representations of smooth families of probability density func-

tions. It allows differential geometry to be applied to problems

in ship target detection of SAR images.

The main contributions in the proposed method are as

follows: 1) ρ-metric construction for lognormal manifold with

high performance is exploited based on information geometry

and 2) a salience-based d escriptor method that exploits the

microstructure feature of statistical manifold.

II. P

ROPOSED METHODS

For this letter, we draw motivation from the modern math-

ematics thought and method, many of which are reﬂected in

See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

下载后可阅读完整内容，剩余4页未读，立即下载

weixin_38693506

粉丝: 5
资源: 966

基于对数正态ρ-度量的SAR图像舰船检测提升方法

PyPI 官网下载 | rho-django-user-1.0.0.tar.gz

一种基于对数极坐标变换的图像配准方法

Rho-Bot for Half-Life-开源

pollard-rho-factoriser:使用 Miller-Rabin 检验的 Pollard-Rho 数因子分解器

rho-concept：Rho Concept是一些概念示例项目的集合，这些项目示例了机器固有的一些概念

rho - 频率稳定裕度：[P,C] 对的频率稳定裕度-matlab开发

双变量累积正态概率：该函数给出双变量累积正态概率，也包括正态概率。-matlab开发

PyPI 官网发布 rho-django-user 包的 1.0.0 版本

Rho-Bot：Half-Life开源计算机对手开发

Matlab代码实现MRI T1rho值计算及图像显示

最新资源