近赤外線分光を使った肥料栄養成分濃度の予測
キーワード:
Near-infrared Spectroscopy (NIR),
Prediction of Fertilizer Concentration,
t-SNE,
K-means++,
Calibration Curve
2023 年 38 巻 p. 1-8
詳細
2023 年 38 巻 p. 1-8
Appropriate fertilization in farmland is quite important. However, it takes a long time to analyze the concentration of fertilizer components. To solve this difficulty, the concentration analysis and fertilization in the farming area at the same time surely provide convenience and efficiency of the farm work. Near-infrared spectroscopy (NIR) provides one of the solutions: NIR is a spectroscopic method that uses the near-infrared region with wavelengths from 800 nm to 2500 nm and utilizes the harmonic and combinatorial vibrations of molecules. As fertilizer components, we focused on phosphorus (P2O5), calcium (CaO), magnesium (MgO), and potassium (K2O). The relationship between NIR spectra and these concentrations in soil has been calibrated using the partial least squares (PLS) method. Furthermore, in principal component analysis on NIR spectra, the principal components were evaluated by the magnitude of eigenvalues and cumulative eigenvalues. The results were visualized using the t-SNE and K-means++ methods, and the samples were classified into clusters with similar NIR spectral characteristics. The results for each cluster were used to obtain a calibration curve to accurately determine the relationship between the NIR spectra and the concentration of fertilizer components in the soil.
