Advanced search
1 file | 3.72 MB Add to list

Detection of leek rust disease under field conditions using hyperspectral proximal sensing and machine learning

(2021) REMOTE SENSING. 13(7).
Author
Organization
Abstract
Rust disease is an important problem for leek cultivation worldwide. It reduces market value and in extreme cases destroys the entire harvest. Farmers have to resort to periodical full-field fungicide applications to prevent the spread of disease, once every 1 to 5 weeks, depending on the cultivar and weather conditions. This implies an economic cost for the farmer and an environmental cost for society. Hyperspectral sensors have been extensively used to address this issue in research, but their application in the field has been limited to a relatively low number of crops, excluding leek, due to the high investment costs and complex data gathering and analysis associated with these sensors. To fill this gap, a methodology was developed for detecting leek rust disease using hyperspectral proximal sensing data combined with supervised machine learning. First, a hyperspectral library was constructed containing 43,416 spectra with a waveband range of 400-1000 nm, measured under field conditions. Then, an extensive evaluation of 11 common classifiers was performed using the scikit-learn machine learning library in Python, combined with a variety of wavelength selection techniques and preprocessing strategies. The best performing model was a (linear) logistic regression model that was able to correctly classify rust disease with an accuracy of 98.14%, using reflectance values at 556 and 661 nm, combined with the value of the first derivative at 511 nm. This model was used to classify unlabelled hyperspectral images, confirming that the model was able to accurately classify leek rust disease symptoms. It can be concluded that the results in this work are an important step towards the mapping of leek rust disease, and that future research is needed to overcome certain challenges before variable rate fungicide applications can be adopted against leek rust disease.
Keywords
General Earth and Planetary Sciences, hyperspectral, proximal sensing, disease detection, leek, rust, machine learning

Downloads

  • published.pdf
    • full text (Published version)
    • |
    • open access
    • |
    • PDF
    • |
    • 3.72 MB

Citation

Please use this url to cite or link to this publication:

MLA
Appeltans, Simon, et al. “Detection of Leek Rust Disease under Field Conditions Using Hyperspectral Proximal Sensing and Machine Learning.” REMOTE SENSING, vol. 13, no. 7, 2021, doi:10.3390/rs13071341.
APA
Appeltans, S., Pieters, J., & Mouazen, A. (2021). Detection of leek rust disease under field conditions using hyperspectral proximal sensing and machine learning. REMOTE SENSING, 13(7). https://doi.org/10.3390/rs13071341
Chicago author-date
Appeltans, Simon, Jan Pieters, and Abdul Mouazen. 2021. “Detection of Leek Rust Disease under Field Conditions Using Hyperspectral Proximal Sensing and Machine Learning.” REMOTE SENSING 13 (7). https://doi.org/10.3390/rs13071341.
Chicago author-date (all authors)
Appeltans, Simon, Jan Pieters, and Abdul Mouazen. 2021. “Detection of Leek Rust Disease under Field Conditions Using Hyperspectral Proximal Sensing and Machine Learning.” REMOTE SENSING 13 (7). doi:10.3390/rs13071341.
Vancouver
1.
Appeltans S, Pieters J, Mouazen A. Detection of leek rust disease under field conditions using hyperspectral proximal sensing and machine learning. REMOTE SENSING. 2021;13(7).
IEEE
[1]
S. Appeltans, J. Pieters, and A. Mouazen, “Detection of leek rust disease under field conditions using hyperspectral proximal sensing and machine learning,” REMOTE SENSING, vol. 13, no. 7, 2021.
@article{8709528,
  abstract     = {{Rust disease is an important problem for leek cultivation worldwide. It reduces market value and in extreme cases destroys the entire harvest. Farmers have to resort to periodical full-field fungicide applications to prevent the spread of disease, once every 1 to 5 weeks, depending on the cultivar and weather conditions. This implies an economic cost for the farmer and an environmental cost for society. Hyperspectral sensors have been extensively used to address this issue in research, but their application in the field has been limited to a relatively low number of crops, excluding leek, due to the high investment costs and complex data gathering and analysis associated with these sensors. To fill this gap, a methodology was developed for detecting leek rust disease using hyperspectral proximal sensing data combined with supervised machine learning. First, a hyperspectral library was constructed containing 43,416 spectra with a waveband range of 400-1000 nm, measured under field conditions. Then, an extensive evaluation of 11 common classifiers was performed using the scikit-learn machine learning library in Python, combined with a variety of wavelength selection techniques and preprocessing strategies. The best performing model was a (linear) logistic regression model that was able to correctly classify rust disease with an accuracy of 98.14%, using reflectance values at 556 and 661 nm, combined with the value of the first derivative at 511 nm. This model was used to classify unlabelled hyperspectral images, confirming that the model was able to accurately classify leek rust disease symptoms. It can be concluded that the results in this work are an important step towards the mapping of leek rust disease, and that future research is needed to overcome certain challenges before variable rate fungicide applications can be adopted against leek rust disease.}},
  articleno    = {{1341}},
  author       = {{Appeltans, Simon and Pieters, Jan and Mouazen, Abdul}},
  issn         = {{2072-4292}},
  journal      = {{REMOTE SENSING}},
  keywords     = {{General Earth and Planetary Sciences,hyperspectral,proximal sensing,disease detection,leek,rust,machine learning}},
  language     = {{eng}},
  number       = {{7}},
  pages        = {{25}},
  title        = {{Detection of leek rust disease under field conditions using hyperspectral proximal sensing and machine learning}},
  url          = {{http://doi.org/10.3390/rs13071341}},
  volume       = {{13}},
  year         = {{2021}},
}

Altmetric
View in Altmetric
Web of Science
Times cited: