A novel approach to hyperspectral band selection based on spectral shape similarity analysis and fast branch and bound search

Shijin Li, Jianbin Qiu, Xinxin Yang, Huan Liu, Dingsheng Wan, Yuelong Zhu

Research output: Contribution to journalArticle

30 Citations (Scopus)

Abstract

With the development of hyperspectral remote sensing technology, the spectral resolution of the hyperspectral image data becomes denser, which results in large number of bands, high correlation between neighboring bands, and high data redundancy. It is necessary to reduce these bands before further analysis, such as land cover classification and target detection. Aiming at the classification task, this paper proposes an effective band selection method from the novel perspective of spectral shape similarity analysis with key points extraction and thus retains physical information of hyperspectral remote sensing images. The proposed approach takes all the bands of hyperspectral remote sensing images as time series. Firstly, spectral clustering is utilized to cluster all the training samples, which produces the prototypical spectral curves of each cluster. Then a set of initial candidate bands are obtained based on the extraction of key points from the processed hyperspectral curves, which preserve discriminative information and narrow down the candidate band subset for the following search procedure. Finally, filtering contiguous bands according to conditional mutual information and branch and bound search are further performed sequentially to gain the optimal band combination. To verify the effectiveness of the integrated band selection method put forward in this paper, classification employing the Support Vector Machine (SVM) classifier is performed on the selected spectral bands. The experimental results on two publicly available benchmark data sets demonstrate that the presented approach can select those bands with discriminative information, usually about 10 out of 200 original bands. Compared with previous studies, the newly proposed method is competitive with far fewer bands selected and a lower computational complexity, while the classification accuracy remains comparable.

Original languageEnglish (US)
Pages (from-to)241-250
Number of pages10
JournalEngineering Applications of Artificial Intelligence
Volume27
DOIs
StatePublished - Jan 2014

Fingerprint

Remote sensing
Spectral resolution
Target tracking
Support vector machines
Redundancy
Time series
Computational complexity
Classifiers

Keywords

  • Branch and bound search
  • Conditional mutual information
  • Feature selection
  • Hyperspectral remote sensing
  • Key point of time series
  • Spectral clustering

ASJC Scopus subject areas

  • Artificial Intelligence
  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Cite this

A novel approach to hyperspectral band selection based on spectral shape similarity analysis and fast branch and bound search. / Li, Shijin; Qiu, Jianbin; Yang, Xinxin; Liu, Huan; Wan, Dingsheng; Zhu, Yuelong.

In: Engineering Applications of Artificial Intelligence, Vol. 27, 01.2014, p. 241-250.

Research output: Contribution to journalArticle

@article{80b51dc8db984504bf2b7a6a555e7f37,
title = "A novel approach to hyperspectral band selection based on spectral shape similarity analysis and fast branch and bound search",
abstract = "With the development of hyperspectral remote sensing technology, the spectral resolution of the hyperspectral image data becomes denser, which results in large number of bands, high correlation between neighboring bands, and high data redundancy. It is necessary to reduce these bands before further analysis, such as land cover classification and target detection. Aiming at the classification task, this paper proposes an effective band selection method from the novel perspective of spectral shape similarity analysis with key points extraction and thus retains physical information of hyperspectral remote sensing images. The proposed approach takes all the bands of hyperspectral remote sensing images as time series. Firstly, spectral clustering is utilized to cluster all the training samples, which produces the prototypical spectral curves of each cluster. Then a set of initial candidate bands are obtained based on the extraction of key points from the processed hyperspectral curves, which preserve discriminative information and narrow down the candidate band subset for the following search procedure. Finally, filtering contiguous bands according to conditional mutual information and branch and bound search are further performed sequentially to gain the optimal band combination. To verify the effectiveness of the integrated band selection method put forward in this paper, classification employing the Support Vector Machine (SVM) classifier is performed on the selected spectral bands. The experimental results on two publicly available benchmark data sets demonstrate that the presented approach can select those bands with discriminative information, usually about 10 out of 200 original bands. Compared with previous studies, the newly proposed method is competitive with far fewer bands selected and a lower computational complexity, while the classification accuracy remains comparable.",
keywords = "Branch and bound search, Conditional mutual information, Feature selection, Hyperspectral remote sensing, Key point of time series, Spectral clustering",
author = "Shijin Li and Jianbin Qiu and Xinxin Yang and Huan Liu and Dingsheng Wan and Yuelong Zhu",
year = "2014",
month = "1",
doi = "10.1016/j.engappai.2013.07.010",
language = "English (US)",
volume = "27",
pages = "241--250",
journal = "Engineering Applications of Artificial Intelligence",
issn = "0952-1976",
publisher = "Elsevier Limited",

}

TY - JOUR

T1 - A novel approach to hyperspectral band selection based on spectral shape similarity analysis and fast branch and bound search

AU - Li, Shijin

AU - Qiu, Jianbin

AU - Yang, Xinxin

AU - Liu, Huan

AU - Wan, Dingsheng

AU - Zhu, Yuelong

PY - 2014/1

Y1 - 2014/1

N2 - With the development of hyperspectral remote sensing technology, the spectral resolution of the hyperspectral image data becomes denser, which results in large number of bands, high correlation between neighboring bands, and high data redundancy. It is necessary to reduce these bands before further analysis, such as land cover classification and target detection. Aiming at the classification task, this paper proposes an effective band selection method from the novel perspective of spectral shape similarity analysis with key points extraction and thus retains physical information of hyperspectral remote sensing images. The proposed approach takes all the bands of hyperspectral remote sensing images as time series. Firstly, spectral clustering is utilized to cluster all the training samples, which produces the prototypical spectral curves of each cluster. Then a set of initial candidate bands are obtained based on the extraction of key points from the processed hyperspectral curves, which preserve discriminative information and narrow down the candidate band subset for the following search procedure. Finally, filtering contiguous bands according to conditional mutual information and branch and bound search are further performed sequentially to gain the optimal band combination. To verify the effectiveness of the integrated band selection method put forward in this paper, classification employing the Support Vector Machine (SVM) classifier is performed on the selected spectral bands. The experimental results on two publicly available benchmark data sets demonstrate that the presented approach can select those bands with discriminative information, usually about 10 out of 200 original bands. Compared with previous studies, the newly proposed method is competitive with far fewer bands selected and a lower computational complexity, while the classification accuracy remains comparable.

AB - With the development of hyperspectral remote sensing technology, the spectral resolution of the hyperspectral image data becomes denser, which results in large number of bands, high correlation between neighboring bands, and high data redundancy. It is necessary to reduce these bands before further analysis, such as land cover classification and target detection. Aiming at the classification task, this paper proposes an effective band selection method from the novel perspective of spectral shape similarity analysis with key points extraction and thus retains physical information of hyperspectral remote sensing images. The proposed approach takes all the bands of hyperspectral remote sensing images as time series. Firstly, spectral clustering is utilized to cluster all the training samples, which produces the prototypical spectral curves of each cluster. Then a set of initial candidate bands are obtained based on the extraction of key points from the processed hyperspectral curves, which preserve discriminative information and narrow down the candidate band subset for the following search procedure. Finally, filtering contiguous bands according to conditional mutual information and branch and bound search are further performed sequentially to gain the optimal band combination. To verify the effectiveness of the integrated band selection method put forward in this paper, classification employing the Support Vector Machine (SVM) classifier is performed on the selected spectral bands. The experimental results on two publicly available benchmark data sets demonstrate that the presented approach can select those bands with discriminative information, usually about 10 out of 200 original bands. Compared with previous studies, the newly proposed method is competitive with far fewer bands selected and a lower computational complexity, while the classification accuracy remains comparable.

KW - Branch and bound search

KW - Conditional mutual information

KW - Feature selection

KW - Hyperspectral remote sensing

KW - Key point of time series

KW - Spectral clustering

UR - http://www.scopus.com/inward/record.url?scp=84888291350&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84888291350&partnerID=8YFLogxK

U2 - 10.1016/j.engappai.2013.07.010

DO - 10.1016/j.engappai.2013.07.010

M3 - Article

AN - SCOPUS:84888291350

VL - 27

SP - 241

EP - 250

JO - Engineering Applications of Artificial Intelligence

JF - Engineering Applications of Artificial Intelligence

SN - 0952-1976

ER -