Memetic three-dimensional gabor feature extraction for hyperspectral imagery classification

Zexuan Zhu, Linlin Shen, Yiwen Sun, Shan He, Zhen Ji

Research output: Chapter in Book/Conference proceedingConference contributionpeer-review

4 Citations (Scopus)

Abstract

This paper proposes a three-dimensional Gabor feature extraction for pixel-based hyperspectral imagery classification using a memetic algorithm. The proposed algorithm named MGFE combines 3-D Gabor wavelet feature generation and feature selection together to capture the signal variances of hyperspectral imagery, thereby extracting the discriminative 3-D Gabor features for accurate classification. MGFE is characterized with a novel fitness evaluation function based on independent feature relevance and a pruning local search for eliminating redundant features. The experimental results on two real-world hyperspectral imagery datasets show that MGFE succeeds in obtaining significantly improved classification accuracy with parsimonious feature selection.

Original languageEnglish
Title of host publicationAdvances in Swarm Intelligence - Third International Conference, ICSI 2012, Proceedings
Pages479-488
Number of pages10
EditionPART 1
DOIs
Publication statusPublished - 2012
Externally publishedYes
Event3rd International Conference on Swarm Intelligence, ICSI 2012 - Shenzhen, China
Duration: 17 Jun 201220 Jun 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume7331 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference3rd International Conference on Swarm Intelligence, ICSI 2012
Country/TerritoryChina
CityShenzhen
Period17/06/1220/06/12

Keywords

  • Gabor Feature Extraction
  • Hyperspectral Imagery Classification
  • Memetic Algorithm

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'Memetic three-dimensional gabor feature extraction for hyperspectral imagery classification'. Together they form a unique fingerprint.

Cite this