Feature selection for spectral sensors with overlapping noisy spectral bands

Biliana Paskaleva, Majeed M. Hayat, J Scott Tyo, Zhipeng Wang, Monica Martinez

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

Quantum-dot infrared photodetectors (QDIPs) are emerging as a promising technology for midwave- and longwave-infrared remote sensing and spectral imaging. One of the key advantages that QDIPs offer is their bias-dependent spectral response, which is brought about by the asymmetric bandstructure of the dot-in-a-well (DWELL) configuration. Photocurrents of a single QDIP, driven by different operational biases can, therefore, be viewed as outputs of different bands. It has been shown that this property, combined with post-processing strategies (applied to the outputs of a single sensor operated at different biases), can be used to perform adaptive spectral tuning and matched filtering. However, unlike traditional sensors, bands of a QDIP exhibit significant spectral overlap, an attribute that calls for the development of novel methods for feature selection. Additionally, the presence of detector noise further complicates such feature selection. In this paper, the theoretical foundations for discriminant analysis, based on spectrally adaptive feature selection, are developed and applied to data obtained from QDIP sensors in the presence of noise. The approach is based on a generalized canonical-correlation-analysis framework that is used in conjunction with an optimization criterion for the selection of feature subspaces. The criterion ranks the best linear combinations of the overlapping bands, providing minimal energy norm (a generalized Euclidean norm) between the centers of classes and their respective reconstructions in the space spanned by sensor bands. Experiments using ASTER-based synthetic QDIP data are used to illustrate the performance of rock-type Bayesian classification according to the proposed feature-selection method.

Original languageEnglish (US)
Title of host publicationProceedings of SPIE - The International Society for Optical Engineering
Volume6233 II
DOIs
StatePublished - 2006
Externally publishedYes
EventAlgorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XII - Kissimmee, FL, United States
Duration: Apr 17 2006Apr 20 2006

Other

OtherAlgorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XII
CountryUnited States
CityKissimmee, FL
Period4/17/064/20/06

Fingerprint

spectral bands
photometers
Photodetectors
Feature extraction
Semiconductor quantum dots
quantum dots
Infrared radiation
sensors
Sensors
norms
output
spectral sensitivity
Discriminant analysis
photocurrents
remote sensing
Photocurrents
emerging
tuning
Remote sensing
rocks

Keywords

  • Adaptive feature selection
  • Canonical-correlation analysis
  • Dot-in-a-well
  • Noise
  • Overlapping spectral bands
  • Quantum-dot infrared photodetectors
  • Rock classification

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Condensed Matter Physics

Cite this

Paskaleva, B., Hayat, M. M., Tyo, J. S., Wang, Z., & Martinez, M. (2006). Feature selection for spectral sensors with overlapping noisy spectral bands. In Proceedings of SPIE - The International Society for Optical Engineering (Vol. 6233 II). [623329] https://doi.org/10.1117/12.666773

Feature selection for spectral sensors with overlapping noisy spectral bands. / Paskaleva, Biliana; Hayat, Majeed M.; Tyo, J Scott; Wang, Zhipeng; Martinez, Monica.

Proceedings of SPIE - The International Society for Optical Engineering. Vol. 6233 II 2006. 623329.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Paskaleva, B, Hayat, MM, Tyo, JS, Wang, Z & Martinez, M 2006, Feature selection for spectral sensors with overlapping noisy spectral bands. in Proceedings of SPIE - The International Society for Optical Engineering. vol. 6233 II, 623329, Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XII, Kissimmee, FL, United States, 4/17/06. https://doi.org/10.1117/12.666773
Paskaleva B, Hayat MM, Tyo JS, Wang Z, Martinez M. Feature selection for spectral sensors with overlapping noisy spectral bands. In Proceedings of SPIE - The International Society for Optical Engineering. Vol. 6233 II. 2006. 623329 https://doi.org/10.1117/12.666773
Paskaleva, Biliana ; Hayat, Majeed M. ; Tyo, J Scott ; Wang, Zhipeng ; Martinez, Monica. / Feature selection for spectral sensors with overlapping noisy spectral bands. Proceedings of SPIE - The International Society for Optical Engineering. Vol. 6233 II 2006.
@inproceedings{2fd987fe83784cabaa8e80a03efdcd92,
title = "Feature selection for spectral sensors with overlapping noisy spectral bands",
abstract = "Quantum-dot infrared photodetectors (QDIPs) are emerging as a promising technology for midwave- and longwave-infrared remote sensing and spectral imaging. One of the key advantages that QDIPs offer is their bias-dependent spectral response, which is brought about by the asymmetric bandstructure of the dot-in-a-well (DWELL) configuration. Photocurrents of a single QDIP, driven by different operational biases can, therefore, be viewed as outputs of different bands. It has been shown that this property, combined with post-processing strategies (applied to the outputs of a single sensor operated at different biases), can be used to perform adaptive spectral tuning and matched filtering. However, unlike traditional sensors, bands of a QDIP exhibit significant spectral overlap, an attribute that calls for the development of novel methods for feature selection. Additionally, the presence of detector noise further complicates such feature selection. In this paper, the theoretical foundations for discriminant analysis, based on spectrally adaptive feature selection, are developed and applied to data obtained from QDIP sensors in the presence of noise. The approach is based on a generalized canonical-correlation-analysis framework that is used in conjunction with an optimization criterion for the selection of feature subspaces. The criterion ranks the best linear combinations of the overlapping bands, providing minimal energy norm (a generalized Euclidean norm) between the centers of classes and their respective reconstructions in the space spanned by sensor bands. Experiments using ASTER-based synthetic QDIP data are used to illustrate the performance of rock-type Bayesian classification according to the proposed feature-selection method.",
keywords = "Adaptive feature selection, Canonical-correlation analysis, Dot-in-a-well, Noise, Overlapping spectral bands, Quantum-dot infrared photodetectors, Rock classification",
author = "Biliana Paskaleva and Hayat, {Majeed M.} and Tyo, {J Scott} and Zhipeng Wang and Monica Martinez",
year = "2006",
doi = "10.1117/12.666773",
language = "English (US)",
isbn = "0819462896",
volume = "6233 II",
booktitle = "Proceedings of SPIE - The International Society for Optical Engineering",

}

TY - GEN

T1 - Feature selection for spectral sensors with overlapping noisy spectral bands

AU - Paskaleva, Biliana

AU - Hayat, Majeed M.

AU - Tyo, J Scott

AU - Wang, Zhipeng

AU - Martinez, Monica

PY - 2006

Y1 - 2006

N2 - Quantum-dot infrared photodetectors (QDIPs) are emerging as a promising technology for midwave- and longwave-infrared remote sensing and spectral imaging. One of the key advantages that QDIPs offer is their bias-dependent spectral response, which is brought about by the asymmetric bandstructure of the dot-in-a-well (DWELL) configuration. Photocurrents of a single QDIP, driven by different operational biases can, therefore, be viewed as outputs of different bands. It has been shown that this property, combined with post-processing strategies (applied to the outputs of a single sensor operated at different biases), can be used to perform adaptive spectral tuning and matched filtering. However, unlike traditional sensors, bands of a QDIP exhibit significant spectral overlap, an attribute that calls for the development of novel methods for feature selection. Additionally, the presence of detector noise further complicates such feature selection. In this paper, the theoretical foundations for discriminant analysis, based on spectrally adaptive feature selection, are developed and applied to data obtained from QDIP sensors in the presence of noise. The approach is based on a generalized canonical-correlation-analysis framework that is used in conjunction with an optimization criterion for the selection of feature subspaces. The criterion ranks the best linear combinations of the overlapping bands, providing minimal energy norm (a generalized Euclidean norm) between the centers of classes and their respective reconstructions in the space spanned by sensor bands. Experiments using ASTER-based synthetic QDIP data are used to illustrate the performance of rock-type Bayesian classification according to the proposed feature-selection method.

AB - Quantum-dot infrared photodetectors (QDIPs) are emerging as a promising technology for midwave- and longwave-infrared remote sensing and spectral imaging. One of the key advantages that QDIPs offer is their bias-dependent spectral response, which is brought about by the asymmetric bandstructure of the dot-in-a-well (DWELL) configuration. Photocurrents of a single QDIP, driven by different operational biases can, therefore, be viewed as outputs of different bands. It has been shown that this property, combined with post-processing strategies (applied to the outputs of a single sensor operated at different biases), can be used to perform adaptive spectral tuning and matched filtering. However, unlike traditional sensors, bands of a QDIP exhibit significant spectral overlap, an attribute that calls for the development of novel methods for feature selection. Additionally, the presence of detector noise further complicates such feature selection. In this paper, the theoretical foundations for discriminant analysis, based on spectrally adaptive feature selection, are developed and applied to data obtained from QDIP sensors in the presence of noise. The approach is based on a generalized canonical-correlation-analysis framework that is used in conjunction with an optimization criterion for the selection of feature subspaces. The criterion ranks the best linear combinations of the overlapping bands, providing minimal energy norm (a generalized Euclidean norm) between the centers of classes and their respective reconstructions in the space spanned by sensor bands. Experiments using ASTER-based synthetic QDIP data are used to illustrate the performance of rock-type Bayesian classification according to the proposed feature-selection method.

KW - Adaptive feature selection

KW - Canonical-correlation analysis

KW - Dot-in-a-well

KW - Noise

KW - Overlapping spectral bands

KW - Quantum-dot infrared photodetectors

KW - Rock classification

UR - http://www.scopus.com/inward/record.url?scp=33748675094&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33748675094&partnerID=8YFLogxK

U2 - 10.1117/12.666773

DO - 10.1117/12.666773

M3 - Conference contribution

AN - SCOPUS:33748675094

SN - 0819462896

SN - 9780819462893

VL - 6233 II

BT - Proceedings of SPIE - The International Society for Optical Engineering

ER -