监督SNPE：高光谱图像分类的多层流形特征提取方法

131 浏览量更新于2024-08-26 收藏 296KB PDF 举报

本文主要探讨了"监督线性流形学习特征提取在高光谱图像分类中的应用"。作者Jinhuan Wen、Weidong Yan和Wei Lin来自西北工业大学科学学院，西安，陕西，针对高光谱图像的复杂特性，提出了一个新颖的监督邻域保持嵌入（Supervised Neighborhood Preserving Embedding, SNPE）方法，用于有效提升图像分类的性能。高光谱图像具有高维度、大量数据以及频带间的高度冗余，这些特点使得传统的分类方法面临挑战。SNPE方法的核心在于引入了一个新的距离度量，这个距离是基于先验的类别标签信息设计的。通过这种方法，同一类别的样本点之间的距离被拉得更近，而不同类别的样本则被刻意拉开，从而突破了单一流形假设的限制。这使得算法能够处理那些分布在多个潜在低维流形上的数据集。相比于无监督的邻域保持嵌入（NPE），SNPE在利用局部邻域结构的同时，充分利用了类别信息，提高了特征提取的精度。在实验部分，作者采用AVIRIS高光谱图像数据集进行了验证，结果显示，该方法在保持信息完整性的前提下，显著提高了高光谱图像的分类准确性和效率。关键词包括：特征提取、降维、邻域保持嵌入、流形学习、高光谱图像分类。这篇研究论文的意义在于，它提供了一种有效的策略，解决了高光谱图像分类中的复杂问题，并可能在未来推动相关领域的技术发展。这篇论文展示了监督线性流形学习作为一种强大的工具，如何在处理高光谱数据时提高图像分类任务的性能。

SUPERVISED LINEAR MANIFOLD LEARNING FEATURE EXTRACTION FOR

HYPERSPECTRAL IMAGE CLASSIFICATION

Jinhuan Wen, Weidong Yan, Wei Lin

School of Science, Northwestern Polytechnical University, Xi’an, Shaanxi, 710072, P.R. China

ABSTRACT

A supervised neighborhood preserving embedding (SNPE)

linear manifold learning feature extraction method for

hyperspectral image classification is presented in this paper.

A point’s k nearest neighbors is found by using new

distance which is proposed according to prior class-label

information. The new distance makes intra-class more

tightly and inter-class more separately. SNPE overcomes the

single manifold assumption of NPE. Data sets lay on (or

near) multiple manifolds can be processed. Experimental

results on AVIRIS hyperspectral data set demonstrate the

effectiveness of our method.

Index Terms—feature extraction, dimensionality

reduction, neighborhood preserving embedding, manifold

learning, hyperspectral image classification

1. INTRODUCTION

Comparable to multispectral image data, hyperspectral

image has high spectral dimension, vast data and altitudinal

interband redundancy which present a challenge to

traditional data processing techniques. In many cases, it is

unnecessary to process all the spectral bands of a

hyperspectral image, since most materials have specific

characteristics only at certain bands, which makes the

remaining spectral bands somewhat redundant. Jimenez [1]

pointed out that the hyperspectral data is centralized in low-

dimensional space because of the high-dimensional space of

hyperspectral image is relatively empty. Therefore, reducing

the dimensionality of hyperspectral data without loosing

important information about objects of interest is a very

important issue for the remote sensing community. The

dimension reduction of hyperspectral image is mainly

divided into two categories, feature extraction and feature

selection. Feature extraction transforms the original data

from high dimension into low dimension with the most of the

desired information content preserved. A number of methods

have been developed to mitigate the effects of dimensionality

on information extraction from hyperspectral data, such as

principal component analysis (PCA) [2] and segmented

principal components transform (SPCT) [3]. Because of not

taking the class information of the input data into account,

PCA may probably discard much useful information and

weaken the recognition accuracy, especially when the

number of sample points is very large [4].

Recently, nonlinear manifold learning dimension

reduction techniques, such as locally linear embedding

(LLE) [5], Isomap [6] and laplacian eigenmaps (LE) [7]

have been successfully applied to hyperspectral data [8]-

[17]. These methods are appropriate for representation of

nonlinear data, but only defined on the training data.

Another major drawback is that these methods are much

more intensive in computation and memory consumption,

which result incapability to process image larger than 70×70

[8-11]. Because it remains a difficult issue to map a new test

sample to the low dimensional space, the above manifold

learning algorithms cannot be easily extended for

classification problems. To address this problem, locally

preserving projection (LPP) [18] and neighborhood

preserving embedding (NPE) [19] are

proposed to be

directly applicable for dealing with new test data.

When the class information is available, NPE becomes

supervised NPE (in order to distinguish with our proposed

algorithm, we denote it as S-NPE). In S-NPE, the Euclidean

distance is used to compute a data point’s k nearest

neighbors. The main assumption behind S-NPE is that data

set is sampled from a single manifold. But in real world,

data set is always sampled from multiple manifolds. To

address this problem, a supervised neighborhood preserving

embedding (SNPE) linear manifold learning feature

extraction method for hyperspectral image classification is

presented in this paper. According to prior class-label

information, a new distance which makes intra-class more

tightly and inter-class more separately is proposed. The

intrinsic structure of original data in low-dimensional space

can be better described by using the new distance to

compute a data point’s k nearest neighbors. Data sets lay on

(or near) multiple manifolds can be processed.

The rest of the paper is organized as follows. In section

2, we briefly review the NPE algorithm. Section 3

introduces our SNPE algorithm, followed with new distance.

Experimental results on AVIRIS hyperspectral data set are

presented in Section 4. Section 5 ends the paper by

presenting some conclusions.

2. NEIGHBORHOOD PRESERVING EMBEDDING

下载后可阅读完整内容，剩余3页未读，立即下载

weixin_38695452

粉丝: 3
资源: 899

监督SNPE：高光谱图像分类的多层流形特征提取方法

空谱协同竞争保持图嵌入高光谱图像特征提取.docx

使用Hypergraph学习以提取高光谱图像特征

基于流形光谱降维和深度学习的高光谱影像分类.pdf

基于多特征和改进自编码的高光谱图像分类

特征提取的多线性无监督判别投影方法

空间一致性邻域保留嵌入的高光谱数据特征提取

线性嵌入与张量流形提升高光谱数据分类精度

非线性流形学习算法分析与应用

监督空间正则化流形鉴别分析在高光谱影像分类中的应用

小样本高光谱图像分类：NPLDE特征提取的创新方法

最新资源