Hyperspectral-based Estimation of Leaf Nitrogen Content in Corn Using Optimal Selection of Multiple Spectral...

Area of Science:

• Agricultural Science
• Remote Sensing
• Plant Physiology

Background:

• Accurate monitoring of crop nitrogen status is crucial for effective fertilization strategies and optimizing crop yields.
• Spectral analysis offers a non-destructive method for assessing plant physiological conditions, including nutrient levels.

Purpose of the Study:

• To compare the effectiveness of different spectral variable types for estimating corn leaf nitrogen content (LNC).
• To identify the optimal spectral variables and modeling approach for accurate LNC estimation.

Main Methods:

• Spectral data preprocessing using the Savitzky-Golay technique.
• Extraction and screening of spectral variables (sensitive bands, feature positions, vegetation indices) using the successive projections algorithm (SPA).
• Estimation of LNC using partial least squares (PLS) regression and random forest (RF) algorithms.

Main Results:

• The integrated variable set, selected by SPA from multiple spectral types, yielded the best LNC estimation performance.
• Partial least squares (PLS) regression with the integrated variable set achieved R² of 0.77, RMSE of 0.31, and NRMSE of 17.1%.
• Random forest (RF) model showed moderate performance with R² of 0.55, RMSE of 0.43, and NRMSE of 23.9%.

Conclusions:

• Combining optimally selected multitype spectral variables significantly improves the accuracy of corn LNC estimation.
• Partial least squares (PLS) regression provides a robust and effective tool for evaluating corn LNC using integrated spectral variables.
• This approach supports data-driven fertilization decisions for enhanced crop management.