The influence of aerial hyperspectral image processing workflow on nitrogen uptake prediction accuracy in maize

Tyler J Nigon; Gabriel Dias Paiao; David J. Mulla; Fabian G Fernandez; Ce Yang

doi:10.3390/rs14010132

The influence of aerial hyperspectral image processing workflow on nitrogen uptake prediction accuracy in maize

Tyler J Nigon, Gabriel Dias Paiao, David J. Mulla, Fabian G Fernandez, Ce Yang

Research output: Contribution to journal › Article › peer-review

5 Scopus citations

Abstract

A meticulous image processing workflow is oftentimes required to derive quality image data from high-resolution, unmanned aerial systems. There are many subjective decisions to be made during image processing, but the effects of those decisions on prediction model accuracy have never been reported. This study introduced a framework for quantifying the effects of image processing methods on model accuracy. A demonstration of this framework was performed using high-resolution hyperspectral imagery (<10 cm pixel size) for predicting maize nitrogen uptake in the early to mid-vegetative developmental stages (V6–V14). Two supervised regression learning estimators (Lasso and partial least squares) were trained to make predictions from hyperspectral imagery. Data for this use case were collected from three experiments over two years (2018–2019) in southern Minnesota, USA (four site-years). The image processing steps that were evaluated include (i) reflectance conversion, (ii) cropping, (iii) spectral clipping, (iv) spectral smoothing, (v) binning, and (vi) segmentation. In total, 648 image processing workflow scenarios were evaluated, and results were analyzed to understand the influence of each image processing step on the cross-validated root mean squared error (RMSE) of the estimators. A sensitivity analysis revealed that the segmentation step was the most influential image processing step on the final estimator error. Across all workflow scenarios, the RMSE of predicted nitrogen uptake ranged from 14.3 to 19.8 kg ha⁻¹ (relative RMSE ranged from 26.5% to 36.5%), a 38.5% increase in error from the lowest to the highest error workflow scenario. The framework introduced demonstrates the sensitivity and extent to which image processing affects prediction accuracy. It allows remote sensing analysts to improve model performance while providing data-driven justification to improve the reproducibility and objectivity of their work, similar to the benefits of hyperparameter tuning in machine learning applications.

Original language	English (US)
Article number	132
Journal	Remote Sensing
Volume	14
Issue number	1
DOIs	https://doi.org/10.3390/rs14010132
State	Published - Jan 1 2022

Bibliographical note

Funding Information:
Funding: This work was supported by the Minnesota Department of Agriculture (USA) through the Minnesota Clean Water Land and Legacy Act (grant number 153761 PO 3000031069); the Minnesota Soybean Research and Promotion Council (USA) (grant numbers 00079668 and 00071830). Minnesota’s Discovery, Research, and InnoVation Economy (MnDRIVE) and the University of Minnesota also provided financial support in the form of student fellowships and/or research assistantships. The funding sources were not involved with the conduct of the research, preparation of the manuscript, and/or the decision to submit the article for publication.

Funding Information:
This work was supported by the Minnesota Department of Agriculture (USA) through the Minnesota Clean Water Land and Legacy Act (grant number 153761 PO 3000031069); the Minnesota Soybean Research and Promotion Council (USA) (grant numbers 00079668 and 00071830). Minnesota?s Discovery, Research, and InnoVation Economy (MnDRIVE) and the University of Minnesota also provided financial support in the form of student fellowships and/or research assistantships. The funding sources were not involved with the conduct of the research, preparation of the manuscript, and/or the decision to submit the article for publication. The Minnesota Supercomputing Institute (MSI) at the University of Minnesota provided substantial computing and storage resources that made this research possible (http://www. msi.umn.edu). We thank the staff at the University of Minnesota Southern Research and Outreach Center, especially Jeff Vetsch, for help with field activities for the Waseca experiments.

Publisher Copyright:
© 2021 by the authors. Licensee MDPI, Basel, Switzerland.

Keywords

Binning
Cropping
Hyperspec-tral imagery
Segmentation
Spectral smoothing
Supervised regression

Publisher link

10.3390/rs14010132

Find It

Find It at U of M Libraries

Cite this

@article{e4889a5ce1b24bf5b54f0d61ef5c96ef,

title = "The influence of aerial hyperspectral image processing workflow on nitrogen uptake prediction accuracy in maize",

abstract = "A meticulous image processing workflow is oftentimes required to derive quality image data from high-resolution, unmanned aerial systems. There are many subjective decisions to be made during image processing, but the effects of those decisions on prediction model accuracy have never been reported. This study introduced a framework for quantifying the effects of image processing methods on model accuracy. A demonstration of this framework was performed using high-resolution hyperspectral imagery (<10 cm pixel size) for predicting maize nitrogen uptake in the early to mid-vegetative developmental stages (V6–V14). Two supervised regression learning estimators (Lasso and partial least squares) were trained to make predictions from hyperspectral imagery. Data for this use case were collected from three experiments over two years (2018–2019) in southern Minnesota, USA (four site-years). The image processing steps that were evaluated include (i) reflectance conversion, (ii) cropping, (iii) spectral clipping, (iv) spectral smoothing, (v) binning, and (vi) segmentation. In total, 648 image processing workflow scenarios were evaluated, and results were analyzed to understand the influence of each image processing step on the cross-validated root mean squared error (RMSE) of the estimators. A sensitivity analysis revealed that the segmentation step was the most influential image processing step on the final estimator error. Across all workflow scenarios, the RMSE of predicted nitrogen uptake ranged from 14.3 to 19.8 kg ha−1 (relative RMSE ranged from 26.5% to 36.5%), a 38.5% increase in error from the lowest to the highest error workflow scenario. The framework introduced demonstrates the sensitivity and extent to which image processing affects prediction accuracy. It allows remote sensing analysts to improve model performance while providing data-driven justification to improve the reproducibility and objectivity of their work, similar to the benefits of hyperparameter tuning in machine learning applications.",

keywords = "Binning, Cropping, Hyperspec-tral imagery, Segmentation, Spectral smoothing, Supervised regression",

author = "Nigon, {Tyler J} and Paiao, {Gabriel Dias} and Mulla, {David J.} and Fernandez, {Fabian G} and Ce Yang",

note = "Funding Information: Funding: This work was supported by the Minnesota Department of Agriculture (USA) through the Minnesota Clean Water Land and Legacy Act (grant number 153761 PO 3000031069); the Minnesota Soybean Research and Promotion Council (USA) (grant numbers 00079668 and 00071830). Minnesota{\textquoteright}s Discovery, Research, and InnoVation Economy (MnDRIVE) and the University of Minnesota also provided financial support in the form of student fellowships and/or research assistantships. The funding sources were not involved with the conduct of the research, preparation of the manuscript, and/or the decision to submit the article for publication. Funding Information: This work was supported by the Minnesota Department of Agriculture (USA) through the Minnesota Clean Water Land and Legacy Act (grant number 153761 PO 3000031069); the Minnesota Soybean Research and Promotion Council (USA) (grant numbers 00079668 and 00071830). Minnesota?s Discovery, Research, and InnoVation Economy (MnDRIVE) and the University of Minnesota also provided financial support in the form of student fellowships and/or research assistantships. The funding sources were not involved with the conduct of the research, preparation of the manuscript, and/or the decision to submit the article for publication. The Minnesota Supercomputing Institute (MSI) at the University of Minnesota provided substantial computing and storage resources that made this research possible (http://www. msi.umn.edu). We thank the staff at the University of Minnesota Southern Research and Outreach Center, especially Jeff Vetsch, for help with field activities for the Waseca experiments. Publisher Copyright: {\textcopyright} 2021 by the authors. Licensee MDPI, Basel, Switzerland.",

year = "2022",

month = jan,

day = "1",

doi = "10.3390/rs14010132",

language = "English (US)",

volume = "14",

journal = "Remote Sensing",

issn = "2072-4292",

publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",

number = "1",

}

TY - JOUR

T1 - The influence of aerial hyperspectral image processing workflow on nitrogen uptake prediction accuracy in maize

AU - Nigon, Tyler J

AU - Paiao, Gabriel Dias

AU - Mulla, David J.

AU - Fernandez, Fabian G

AU - Yang, Ce

N1 - Funding Information: Funding: This work was supported by the Minnesota Department of Agriculture (USA) through the Minnesota Clean Water Land and Legacy Act (grant number 153761 PO 3000031069); the Minnesota Soybean Research and Promotion Council (USA) (grant numbers 00079668 and 00071830). Minnesota’s Discovery, Research, and InnoVation Economy (MnDRIVE) and the University of Minnesota also provided financial support in the form of student fellowships and/or research assistantships. The funding sources were not involved with the conduct of the research, preparation of the manuscript, and/or the decision to submit the article for publication. Funding Information: This work was supported by the Minnesota Department of Agriculture (USA) through the Minnesota Clean Water Land and Legacy Act (grant number 153761 PO 3000031069); the Minnesota Soybean Research and Promotion Council (USA) (grant numbers 00079668 and 00071830). Minnesota?s Discovery, Research, and InnoVation Economy (MnDRIVE) and the University of Minnesota also provided financial support in the form of student fellowships and/or research assistantships. The funding sources were not involved with the conduct of the research, preparation of the manuscript, and/or the decision to submit the article for publication. The Minnesota Supercomputing Institute (MSI) at the University of Minnesota provided substantial computing and storage resources that made this research possible (http://www. msi.umn.edu). We thank the staff at the University of Minnesota Southern Research and Outreach Center, especially Jeff Vetsch, for help with field activities for the Waseca experiments. Publisher Copyright: © 2021 by the authors. Licensee MDPI, Basel, Switzerland.

PY - 2022/1/1

Y1 - 2022/1/1

N2 - A meticulous image processing workflow is oftentimes required to derive quality image data from high-resolution, unmanned aerial systems. There are many subjective decisions to be made during image processing, but the effects of those decisions on prediction model accuracy have never been reported. This study introduced a framework for quantifying the effects of image processing methods on model accuracy. A demonstration of this framework was performed using high-resolution hyperspectral imagery (<10 cm pixel size) for predicting maize nitrogen uptake in the early to mid-vegetative developmental stages (V6–V14). Two supervised regression learning estimators (Lasso and partial least squares) were trained to make predictions from hyperspectral imagery. Data for this use case were collected from three experiments over two years (2018–2019) in southern Minnesota, USA (four site-years). The image processing steps that were evaluated include (i) reflectance conversion, (ii) cropping, (iii) spectral clipping, (iv) spectral smoothing, (v) binning, and (vi) segmentation. In total, 648 image processing workflow scenarios were evaluated, and results were analyzed to understand the influence of each image processing step on the cross-validated root mean squared error (RMSE) of the estimators. A sensitivity analysis revealed that the segmentation step was the most influential image processing step on the final estimator error. Across all workflow scenarios, the RMSE of predicted nitrogen uptake ranged from 14.3 to 19.8 kg ha−1 (relative RMSE ranged from 26.5% to 36.5%), a 38.5% increase in error from the lowest to the highest error workflow scenario. The framework introduced demonstrates the sensitivity and extent to which image processing affects prediction accuracy. It allows remote sensing analysts to improve model performance while providing data-driven justification to improve the reproducibility and objectivity of their work, similar to the benefits of hyperparameter tuning in machine learning applications.

AB - A meticulous image processing workflow is oftentimes required to derive quality image data from high-resolution, unmanned aerial systems. There are many subjective decisions to be made during image processing, but the effects of those decisions on prediction model accuracy have never been reported. This study introduced a framework for quantifying the effects of image processing methods on model accuracy. A demonstration of this framework was performed using high-resolution hyperspectral imagery (<10 cm pixel size) for predicting maize nitrogen uptake in the early to mid-vegetative developmental stages (V6–V14). Two supervised regression learning estimators (Lasso and partial least squares) were trained to make predictions from hyperspectral imagery. Data for this use case were collected from three experiments over two years (2018–2019) in southern Minnesota, USA (four site-years). The image processing steps that were evaluated include (i) reflectance conversion, (ii) cropping, (iii) spectral clipping, (iv) spectral smoothing, (v) binning, and (vi) segmentation. In total, 648 image processing workflow scenarios were evaluated, and results were analyzed to understand the influence of each image processing step on the cross-validated root mean squared error (RMSE) of the estimators. A sensitivity analysis revealed that the segmentation step was the most influential image processing step on the final estimator error. Across all workflow scenarios, the RMSE of predicted nitrogen uptake ranged from 14.3 to 19.8 kg ha−1 (relative RMSE ranged from 26.5% to 36.5%), a 38.5% increase in error from the lowest to the highest error workflow scenario. The framework introduced demonstrates the sensitivity and extent to which image processing affects prediction accuracy. It allows remote sensing analysts to improve model performance while providing data-driven justification to improve the reproducibility and objectivity of their work, similar to the benefits of hyperparameter tuning in machine learning applications.

KW - Binning

KW - Cropping

KW - Hyperspec-tral imagery

KW - Segmentation

KW - Spectral smoothing

KW - Supervised regression

UR - http://www.scopus.com/inward/record.url?scp=85122042093&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85122042093&partnerID=8YFLogxK

U2 - 10.3390/rs14010132

DO - 10.3390/rs14010132

M3 - Article

AN - SCOPUS:85122042093

SN - 2072-4292

VL - 14

JO - Remote Sensing

JF - Remote Sensing

IS - 1

M1 - 132

ER -

The influence of aerial hyperspectral image processing workflow on nitrogen uptake prediction accuracy in maize

Abstract

Bibliographical note

Keywords

Publisher link

Find It

Other files and links

Fingerprint

Cite this