A modified algorithm for multi-target detection in hyperspectral image

  • Jihao Yin*
  • , Yan Wang
  • , Yisong Wang
  • , Zhanjie Zhao
  • *Corresponding author for this work

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

Abstract

In the area of hyperspectral image detection, Constrained Energy Minimization (CEM) algorithm has received considerable interest recently. However, it is very sensitive to noise and can only detect one target at one time. In order to solve this problem, Winner-Take-All Constrained Energy Minimization (WTACEM) was proposed. Based on it, we propose a modified multi-target detection approach. We study the sample correlation matrix and find if we exclude the desired target pixel vectors from the correlation matrix, we can achieve better results. The simulation results indicate that the errors caused by the modified WTACEM have decreased obviously and the output is more close to the true abundance fraction. The experiment also shows that when the abundance fraction of target increases, the detection performance improves.

Original languageEnglish
Title of host publicationCAR 2010 - 2010 2nd International Asia Conference on Informatics in Control, Automation and Robotics
Pages105-108
Number of pages4
DOIs
StatePublished - 2010
Event2010 2nd International Asia Conference on Informatics in Control, Automation and Robotics, CAR 2010 - Wuhan, China
Duration: 6 Mar 20107 Mar 2010

Publication series

NameCAR 2010 - 2010 2nd International Asia Conference on Informatics in Control, Automation and Robotics
Volume3

Conference

Conference2010 2nd International Asia Conference on Informatics in Control, Automation and Robotics, CAR 2010
Country/TerritoryChina
CityWuhan
Period6/03/107/03/10

Keywords

  • Modified correlation matrix
  • Multi-target detection
  • Winner-take-all CEM

Fingerprint

Dive into the research topics of 'A modified algorithm for multi-target detection in hyperspectral image'. Together they form a unique fingerprint.

Cite this