Skip to main navigation Skip to search Skip to main content

Using CNN to classify hyperspectral data based on spatial-spectral information

  • Harbin Institute of Technology

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

Abstract

Currently, the dimensionality of hyperspectral images is increasing, and the images have the characteristics of nonlinearity and spatial correlation, making it more and more difficult to classify these data. In this study, convolutional neural network (CNN) which has been successfully applied in image recognition and language detection is introduced. The spectral and spatial information is combined and used for hyperspectral image classification. According to the character of CNN that its input is two-dimensional image data, two methods are proposed converting the spectral and spatial information of hyperspectral images into two dimensional images. One of them converts the spatial-spectral information into gray level images and uses the varying texture features between spectral bands. The other converts the spatial-spectral information into waveforms and uses the wave characteristics of the spectral bands. Experiments on KSC and Pavia U data sets demonstrate the feasibility and efficacy of CNN in hyperspectral image classification.

Original languageEnglish
Title of host publicationAdvances in Intelligent Information Hiding and Multimedia Signal Processing - Proceeding of the 12th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, 2016
EditorsJeng-Shyang Pan, Pei-Wei Tsai, Hsiang-Cheh Huang
PublisherSpringer Science and Business Media Deutschland GmbH
Pages61-68
Number of pages8
ISBN (Print)9783319502113
DOIs
StatePublished - 2017
Event12th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, IIH-MSP 2016 - Kaohsiung, Taiwan, Province of China
Duration: 21 Nov 201623 Nov 2016

Publication series

NameSmart Innovation, Systems and Technologies
Volume64
ISSN (Print)2190-3018
ISSN (Electronic)2190-3026

Conference

Conference12th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, IIH-MSP 2016
Country/TerritoryTaiwan, Province of China
CityKaohsiung
Period21/11/1623/11/16

Keywords

  • Convolutional neural network (CNN)
  • Hyperspectral image classification
  • Joint spatial and spectral feature

Fingerprint

Dive into the research topics of 'Using CNN to classify hyperspectral data based on spatial-spectral information'. Together they form a unique fingerprint.

Cite this