Fast classification scheme for hyperspectral imagery

Te Ming Tu, Chin-Hsing Chen

Research output: Contribution to journalArticle

Abstract

Classification for high dimensional remote sensing data generally requires a large set of data samples and enormous processing time, especially for hyperspectral image data. In this paper, a fast classification scheme is presented. The first stage of process is to develop a strategy for band selection which is designed based on the canonical analysis (CA) and the concept of loading factors to weigh bands in accordance with their energies. The suggested band selection algorithm allows one to predetermine which bands will be used for data processing so that data dimensionality is greatly reduced. It is then followed by a second stage using a maximum likelihood (ML) classifier which is recursive and designed based on Winograd's algorithm to achieve computational efficiency. The experimental results show that the proposed fast classification scheme reduces the computing time by a factor of 27 to 107 compared to the conventional one-stage ML classifier.

Original languageEnglish
Pages (from-to)456-465
Number of pages10
JournalProceedings of the National Science Council, Republic of China, Part A: Physical Science and Engineering
Volume22
Issue number4
Publication statusPublished - 1998 Jul 1

Fingerprint

Maximum likelihood
Classifiers
Computational efficiency
Remote sensing
Processing

All Science Journal Classification (ASJC) codes

  • Engineering(all)

Cite this

@article{5f8d2f016e6242628ac63ead2e599fbd,
title = "Fast classification scheme for hyperspectral imagery",
abstract = "Classification for high dimensional remote sensing data generally requires a large set of data samples and enormous processing time, especially for hyperspectral image data. In this paper, a fast classification scheme is presented. The first stage of process is to develop a strategy for band selection which is designed based on the canonical analysis (CA) and the concept of loading factors to weigh bands in accordance with their energies. The suggested band selection algorithm allows one to predetermine which bands will be used for data processing so that data dimensionality is greatly reduced. It is then followed by a second stage using a maximum likelihood (ML) classifier which is recursive and designed based on Winograd's algorithm to achieve computational efficiency. The experimental results show that the proposed fast classification scheme reduces the computing time by a factor of 27 to 107 compared to the conventional one-stage ML classifier.",
author = "Tu, {Te Ming} and Chin-Hsing Chen",
year = "1998",
month = "7",
day = "1",
language = "English",
volume = "22",
pages = "456--465",
journal = "Proceedings of the National Science Council, Republic of China, Part A: Physical Science and Engineering",
issn = "0255-6588",
publisher = "National Science Council",
number = "4",

}

TY - JOUR

T1 - Fast classification scheme for hyperspectral imagery

AU - Tu, Te Ming

AU - Chen, Chin-Hsing

PY - 1998/7/1

Y1 - 1998/7/1

N2 - Classification for high dimensional remote sensing data generally requires a large set of data samples and enormous processing time, especially for hyperspectral image data. In this paper, a fast classification scheme is presented. The first stage of process is to develop a strategy for band selection which is designed based on the canonical analysis (CA) and the concept of loading factors to weigh bands in accordance with their energies. The suggested band selection algorithm allows one to predetermine which bands will be used for data processing so that data dimensionality is greatly reduced. It is then followed by a second stage using a maximum likelihood (ML) classifier which is recursive and designed based on Winograd's algorithm to achieve computational efficiency. The experimental results show that the proposed fast classification scheme reduces the computing time by a factor of 27 to 107 compared to the conventional one-stage ML classifier.

AB - Classification for high dimensional remote sensing data generally requires a large set of data samples and enormous processing time, especially for hyperspectral image data. In this paper, a fast classification scheme is presented. The first stage of process is to develop a strategy for band selection which is designed based on the canonical analysis (CA) and the concept of loading factors to weigh bands in accordance with their energies. The suggested band selection algorithm allows one to predetermine which bands will be used for data processing so that data dimensionality is greatly reduced. It is then followed by a second stage using a maximum likelihood (ML) classifier which is recursive and designed based on Winograd's algorithm to achieve computational efficiency. The experimental results show that the proposed fast classification scheme reduces the computing time by a factor of 27 to 107 compared to the conventional one-stage ML classifier.

UR - http://www.scopus.com/inward/record.url?scp=0032119811&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0032119811&partnerID=8YFLogxK

M3 - Article

VL - 22

SP - 456

EP - 465

JO - Proceedings of the National Science Council, Republic of China, Part A: Physical Science and Engineering

JF - Proceedings of the National Science Council, Republic of China, Part A: Physical Science and Engineering

SN - 0255-6588

IS - 4

ER -