Ensemble Feature Selection for Plant Phenotyping

A Journey from Hyperspectral to Multispectral Imaging

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Hyperspectral imaging is becoming an increasingly popular tool for high-throughput plant phenotyping, because it provides remarkable insights about the health status of plants. Feature selection is a key component in a hyperspectral image analysis, largely because a significant portion of spectral features are redundant and/or irrelevant, depending on the desired application. This paper presents an ensemble feature selection method to identify the most informative spectral features for practical applications in plant phenotyping. The hyperspectral data set contained the images of four wheat lines, each with a control and a salt (NaCl) treatment. To rank spectral features, six feature selection methods were used as the base for the ensemble: correlation-based feature selection, ReliefF, sequential feature selection, support vector machine-recursive feature elimination (SVM-RFE), LASSO logistic regression, and random forest. The best results were achieved by the ensemble of ReliefF, SVM-RFE, and random forest, which drastically reduced the dimension of the hyperspectral data set from 215 to 15 features, while improving the accuracy in classifying the salt-treated vegetation pixels from the control pixels by 8.5%. To transform the hyperspectral data set into a multispectral data set, six wavelengths as the center of broad multispectral bands around the most prominent features were determined by a clustering algorithm. The result of salt tolerance assessment of the four wheat lines using the derived multispectral data set was similar to that of the hyperspectral data set. This demonstrates that the proposed feature selection pipeline can be utilized for determining the most informative features and can be a valuable tool in the development of tailored multispectral cameras.

Original languageEnglish (US)
Article number8479334
Pages (from-to)56870-56884
Number of pages15
JournalIEEE Access
Volume6
DOIs
StatePublished - Jan 1 2018

Fingerprint

Feature extraction
Imaging techniques
Salts
Support vector machines
Pixels
Clustering algorithms
Image analysis
Logistics
Pipelines
Cameras
Throughput
Health
Wavelength

Keywords

  • Band selection
  • classification
  • ensemble feature selection
  • hyperspectral imaging
  • machine learning
  • multispectral imging
  • plant phenotyping
  • salt stress
  • wheat

Cite this

Ensemble Feature Selection for Plant Phenotyping : A Journey from Hyperspectral to Multispectral Imaging. / Moghimi, Ali; Yang, Ce; Marchetto, Peter M.

In: IEEE Access, Vol. 6, 8479334, 01.01.2018, p. 56870-56884.

Research output: Contribution to journalArticle

@article{b8b5a6b47c904a5299144cda644e6dca,
title = "Ensemble Feature Selection for Plant Phenotyping: A Journey from Hyperspectral to Multispectral Imaging",
abstract = "Hyperspectral imaging is becoming an increasingly popular tool for high-throughput plant phenotyping, because it provides remarkable insights about the health status of plants. Feature selection is a key component in a hyperspectral image analysis, largely because a significant portion of spectral features are redundant and/or irrelevant, depending on the desired application. This paper presents an ensemble feature selection method to identify the most informative spectral features for practical applications in plant phenotyping. The hyperspectral data set contained the images of four wheat lines, each with a control and a salt (NaCl) treatment. To rank spectral features, six feature selection methods were used as the base for the ensemble: correlation-based feature selection, ReliefF, sequential feature selection, support vector machine-recursive feature elimination (SVM-RFE), LASSO logistic regression, and random forest. The best results were achieved by the ensemble of ReliefF, SVM-RFE, and random forest, which drastically reduced the dimension of the hyperspectral data set from 215 to 15 features, while improving the accuracy in classifying the salt-treated vegetation pixels from the control pixels by 8.5{\%}. To transform the hyperspectral data set into a multispectral data set, six wavelengths as the center of broad multispectral bands around the most prominent features were determined by a clustering algorithm. The result of salt tolerance assessment of the four wheat lines using the derived multispectral data set was similar to that of the hyperspectral data set. This demonstrates that the proposed feature selection pipeline can be utilized for determining the most informative features and can be a valuable tool in the development of tailored multispectral cameras.",
keywords = "Band selection, classification, ensemble feature selection, hyperspectral imaging, machine learning, multispectral imging, plant phenotyping, salt stress, wheat",
author = "Ali Moghimi and Ce Yang and Marchetto, {Peter M}",
year = "2018",
month = "1",
day = "1",
doi = "10.1109/ACCESS.2018.2872801",
language = "English (US)",
volume = "6",
pages = "56870--56884",
journal = "IEEE Access",
issn = "2169-3536",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - JOUR

T1 - Ensemble Feature Selection for Plant Phenotyping

T2 - A Journey from Hyperspectral to Multispectral Imaging

AU - Moghimi, Ali

AU - Yang, Ce

AU - Marchetto, Peter M

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Hyperspectral imaging is becoming an increasingly popular tool for high-throughput plant phenotyping, because it provides remarkable insights about the health status of plants. Feature selection is a key component in a hyperspectral image analysis, largely because a significant portion of spectral features are redundant and/or irrelevant, depending on the desired application. This paper presents an ensemble feature selection method to identify the most informative spectral features for practical applications in plant phenotyping. The hyperspectral data set contained the images of four wheat lines, each with a control and a salt (NaCl) treatment. To rank spectral features, six feature selection methods were used as the base for the ensemble: correlation-based feature selection, ReliefF, sequential feature selection, support vector machine-recursive feature elimination (SVM-RFE), LASSO logistic regression, and random forest. The best results were achieved by the ensemble of ReliefF, SVM-RFE, and random forest, which drastically reduced the dimension of the hyperspectral data set from 215 to 15 features, while improving the accuracy in classifying the salt-treated vegetation pixels from the control pixels by 8.5%. To transform the hyperspectral data set into a multispectral data set, six wavelengths as the center of broad multispectral bands around the most prominent features were determined by a clustering algorithm. The result of salt tolerance assessment of the four wheat lines using the derived multispectral data set was similar to that of the hyperspectral data set. This demonstrates that the proposed feature selection pipeline can be utilized for determining the most informative features and can be a valuable tool in the development of tailored multispectral cameras.

AB - Hyperspectral imaging is becoming an increasingly popular tool for high-throughput plant phenotyping, because it provides remarkable insights about the health status of plants. Feature selection is a key component in a hyperspectral image analysis, largely because a significant portion of spectral features are redundant and/or irrelevant, depending on the desired application. This paper presents an ensemble feature selection method to identify the most informative spectral features for practical applications in plant phenotyping. The hyperspectral data set contained the images of four wheat lines, each with a control and a salt (NaCl) treatment. To rank spectral features, six feature selection methods were used as the base for the ensemble: correlation-based feature selection, ReliefF, sequential feature selection, support vector machine-recursive feature elimination (SVM-RFE), LASSO logistic regression, and random forest. The best results were achieved by the ensemble of ReliefF, SVM-RFE, and random forest, which drastically reduced the dimension of the hyperspectral data set from 215 to 15 features, while improving the accuracy in classifying the salt-treated vegetation pixels from the control pixels by 8.5%. To transform the hyperspectral data set into a multispectral data set, six wavelengths as the center of broad multispectral bands around the most prominent features were determined by a clustering algorithm. The result of salt tolerance assessment of the four wheat lines using the derived multispectral data set was similar to that of the hyperspectral data set. This demonstrates that the proposed feature selection pipeline can be utilized for determining the most informative features and can be a valuable tool in the development of tailored multispectral cameras.

KW - Band selection

KW - classification

KW - ensemble feature selection

KW - hyperspectral imaging

KW - machine learning

KW - multispectral imging

KW - plant phenotyping

KW - salt stress

KW - wheat

UR - http://www.scopus.com/inward/record.url?scp=85054471259&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85054471259&partnerID=8YFLogxK

U2 - 10.1109/ACCESS.2018.2872801

DO - 10.1109/ACCESS.2018.2872801

M3 - Article

VL - 6

SP - 56870

EP - 56884

JO - IEEE Access

JF - IEEE Access

SN - 2169-3536

M1 - 8479334

ER -