基于支持向量机(SVM)算法的雷达图像分类技术

 2022-05-17 21:31:39

论文总字数:24833字

摘 要

合成孔径雷达(SAR)是一种具有较高分辨率的对地观测成像雷达,因为其具有地物穿透能力、不受天候干扰等特点,因而受到各国重视,如何快速准确的完成SAR图像的分类也一直是国内外的研究重点。主成分分析(PCA)是模式识别中的一种经典算法,在人脸识别,图像分类等领域被广泛用来进行特征提取。支持向量机(SVM)是一种监督分类方法,在小样本,非线性数据分类上表现出色。

本文通过SVM对SAR图像的分类进行了研究,旨在完成导弹,F-117,P51,Su-27战机的SAR图像的分类。首先,采用PCA方法对原始SAR图像进行降维完成特征提取,因为PCA中协方差矩阵的维数较大,求解特征值时耗费时间多,空间大,采用了核主成分分析(KPCA)方法进行SAR图像的特征提取,之后将从SAR图像中提取的特征值通过SVM进行训练,通过一对一的方法实现多分类。

通过实验表明,基于PCA和KPCA完成特征提取,再通过SVM进行训练是一种有效的SAR图像分类方法,能够完美分类导弹,F-117,P51和Su-27战机,实验中同时进行了SVM训练集数量选取对分类结果影响的研究以及SVM参数的选取对分类结果影响的研究,并采用粒子群优化算法(PSO)完成SVM最优参数的选择。

关键词:SAR图像,主成分分析,核主成分分析,支持向量机

ABSTRACT

Synthetic Aperture Radar (SAR) is a geodetic imaging radar with high resolution. Due to its characteristics of ground penetrating ability and undisturbed interference, it is highly valued by all countries. How to classify SAR images quickly and accurately has always been the focus of research at home and abroad. Principal Component Analysis (PCA) is a classical algorithm in pattern recognition. It is widely used in feature recognition in the fields of face recognition and image classification. Support Vector Machine (SVM) is a supervised classification method that excels in small sample and nonlinear data classification.

This paper studies the classification of SAR images based on SVM, aiming at the classification of SAR images of missiles, F-117, P51 and Su-27 fighters. Firstly, the original SAR image is extracted by the PCA method. Because the dimension of the covariance matrix in PCA is large, it takes time and space to solve the eigenvalues, the Kernel Principal Component Analysis (KPCA) method is used to extract features of SAR images. Then the features ​​extracted from the images are trained by SVM, and multi-classification is realized by a one-virus-one method.

Experiments show that the feature extraction based on PCA and KPCA, and training data through SVM is an effective classification method. It can classify missiles, F-117, P51 and Su-27 fighters perfectly. The experiment also studies the influence of the number of SVM training sets and the selection of SVM parameters on the classification results. Particle Swarm Optimization (PSO) algorithm is used to optimize the selection of SVM parameters.

KEY WORDS: SAR image, Principal Component Analysis, Kernel Principal Component Analysis, Support Vector Machine

目 录

摘 要 Ⅰ

ABSTRACT Ⅱ

第一章 绪论 1

1.1研究背景 1

1.2国内外研究现状 2

1.3论文主要工作及结构 2

第二章 基于主成分分析(PCA)的SAR图像特征提取 4

2.1引言 4

2.2 PCA原理及流程 4

2.3实验与分析 6

2.4本章总结 8

第三章 基于核主成分分析(KPCA)的SAR图像特征提取 9

3.1引言 9

3.2 KPCA原理及流程 9

3.3实验与分析 10

3.4本章总结 12

第四章 基于支持向量机(SVM)的SAR图像分类 13

4.1引言 13

4.2 SVM原理 13

4.2.1 SVM线性分类 13

4.2.2 SVM非线性分类 15

4.2.3 SVM多分类 16

4.3 SVM分类结果 16

4.3.1特征数对分类结果的影响 16

4.3.2 训练集数量对SVM分类结果的影响 17

4.3.3 SVM参数影响 19

4.4基于粒子群优化(PSO)的SVM参数选择 20

4.4.1 PSO算法原理 20

4.4.2 PSO-SVM实验与分析 22

4.5本章总结 24

第五章 总结与展望 25

5.1本文的主要工作与不足 25

5.2展望 25

参考文献 27

致谢 29

第一章 绪论

剩余内容已隐藏,请支付后下载全文,论文总字数:24833字

您需要先支付 80元 才能查看全部内容!立即支付

该课题毕业论文、开题报告、外文翻译、程序设计、图纸设计等资料可联系客服协助查找;