A New Learning Method for Continuous Hidden Markov Models for Subsurface Landmine Detection in Ground Penetrating Radar

Author

Xuping Zhang ; Bolton, Jeremy ; Gader, Paul

Author_Institution

Dept. of Comput. & Inf. Sci. & Eng. (CISE), Univ. of Florida, Gainesville, FL, USA

Volume

7

Issue

3

fYear

2014

fDate

Mar-14

Firstpage

813

Lastpage

819

Abstract

A new learning algorithm based on Gibbs sampling to learn the parameters of continuous Hidden Markov Models (HMMs) with multivariate Gaussian mixtures is presented. The proposed sampling algorithm outperformed the standard expectation-maximization (EM) algorithm and a minimum classification error algorithm when applied to a synthetic data set. The proposed algorithm outperforms the state of the art when applied to landmine detection using ground penetrating radar (GPR) data.

Keywords

Gaussian processes; expectation-maximisation algorithm; ground penetrating radar; hidden Markov models; landmine detection; mixture models; pattern classification; radar detection; EM algorithm; GPR data; Gibbs sampling algorithm; HMM; continuous hidden Markov model; ground penetrating radar; learning method; minimum classification error algorithm; multivariate Gaussian mixture; standard expectation-maximization algorithm; subsurface landmine detection; synthetic data set application; Ground penetrating radar; Hidden Markov models; Image color analysis; Landmine detection; Learning systems; Remote sensing; Standards; Gibbs sampling; Hidden Markov Model (HMM); Markov Chain Monte Carlo (MCMC) sampling; ground penetrating radar (GPR) imagery; multivariate Gaussian mixture;

fLanguage

English

Journal_Title

Selected Topics in Applied Earth Observations and Remote Sensing, IEEE Journal of

Publisher

ieee

ISSN

1939-1404

Type

jour

DOI

10.1109/JSTARS.2014.2305981

Filename

6767137