Discrimination of rice panicles by hyperspectral reflectance data based on principal component analysis and support...

Area of Science:

• Agricultural Science
• Remote Sensing
• Spectroscopy

Background:

• Accurate crop health monitoring is crucial for disease and pest management.
• Late-stage detection of rice panicle conditions impacts yield and quality.

Purpose of the Study:

• To evaluate the feasibility of using visible and near-infrared spectroscopy for discriminating rice panicle health.
• To develop a classification model for identifying healthy, empty, and diseased panicles.

Main Methods:

• Hyperspectral reflectance measurements of rice panicles in the visible and near-infrared regions.
• Application of first and second-order derivative spectra and Principal Component Analysis (PCA) for spectral dimension reduction.
• Support Vector Classification (SVC) using principal component spectra (PCS) for health condition discrimination.

Main Results:

• Support Vector Classification achieved high accuracy in discriminating panicle health.
• The highest classification accuracy (99.14%) and kappa coefficient (98.71%) were obtained using PCS from the first-order derivative spectra.
• The study demonstrated the effectiveness of spectroscopy for differentiating healthy, Nilaparvata lugens-induced empty, and Ustilaginoidea virens-infected panicles.

Conclusions:

• Visible and near-infrared spectroscopy is a viable technique for assessing rice panicle health.
• Derivative spectra combined with PCA and SVC offer a robust approach for crop disease and damage detection.
• This method supports timely interventions for improved rice production.