Efficient semi-supervised feature selection for VHR remote sensing images

  • Xi Chen
  • , Lin Song
  • , Yuguan Hou
  • , Guofan Shao

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

9 Scopus citations

Abstract

Feature selection is often required to select a feature subset from the original feature set of objects of very high resolution (VHR) remote sensing images. However, the majority of feature selection methods is supervised, and could fail to identify the relevant features when labeled objects are scarce. To address the problem, this paper proposes a method, efficient semi-supervised feature selection (ESFS), by effectively exploiting the underlying information of the huge amount of unlabeled objects. Firstly, probability matrix of unlabeled objects is utilized in loss function to measure the relevance of features on classes, instead of using traditional graph. Secondly, construction a l2,1-norm regularization term is imposed to ensure the sparsity in rows of the selection matrix, and consequent feature selection. Experiments are carried on a VHR image demonstrate that ESFS outperforms other classical and latest methods.

Original languageEnglish
Title of host publication2016 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1500-1503
Number of pages4
ISBN (Electronic)9781509033324
DOIs
StatePublished - 1 Nov 2016
Externally publishedYes
Event36th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 - Beijing, China
Duration: 10 Jul 201615 Jul 2016

Publication series

NameInternational Geoscience and Remote Sensing Symposium (IGARSS)
Volume2016-November

Conference

Conference36th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016
Country/TerritoryChina
CityBeijing
Period10/07/1615/07/16

Keywords

  • classification
  • l-norm regularization
  • Semi-supervised features selection

Fingerprint

Dive into the research topics of 'Efficient semi-supervised feature selection for VHR remote sensing images'. Together they form a unique fingerprint.

Cite this