Identification of Rice Sheath Blight through Spectral Responses Using Hyperspectral Images.

Area of Science:

• Agricultural remote sensing
• Plant pathology
• Spectroscopy

Background:

• Sheath blight (ShB), caused by *Rhizoctonia solani* AG1-I, is a major rice disease.
• Hyperspectral remote sensing offers potential for rapid and accurate disease detection.

Purpose of the Study:

• To evaluate spectral responses of ShB-infected rice leaves.
• To construct a spectral feature library for ShB detection using "three-edge" parameters and vegetation indices.
• To identify key spectral indicators for ShB on rice leaves.

Main Methods:

• Analysis of spectral curves from rice leaf lesions at various development stages.
• Calculation of "three-edge" parameters and narrow-band vegetation indices.
• Application of the ReliefF algorithm to determine attribute importance.
• Classification using a decision tree model with selected spectral parameters.

Main Results:

• Spectral curves showed significant changes in blue edge, green peak, yellow edge, red valley, red edge, and near-infrared regions.
• The ratio index of green peak amplitude to red valley amplitude (Rg/Ro) demonstrated high relevance and importance for ShB detection.
• A decision tree classifier achieved a 95.5% estimation rate using Rg/Ro.
• Distinct spectral differences, particularly the ratio index of red edge area to green peak area (SDr/SDg), were found between leaf sheath and leaf blade lesions.

Conclusions:

• Specific spectral features, especially Rg/Ro, are effective indicators for detecting rice sheath blight on leaves.
• Hyperspectral analysis can differentiate between leaf sheath and leaf blade lesions.
• Results provide a foundation for developing targeted sensors for ShB detection in rice.