A non-destructive testing method for early detection of ginseng root diseases using machine learning technologies...

Area of Science:

• Agricultural Science
• Plant Pathology
• Remote Sensing Technology

Background:

• Ginseng root disease significantly reduces crop yield.
• Manual detection is delayed as leaf symptoms appear late.
• An efficient, non-destructive testing (NDT) method is crucial for early disease identification.

Purpose of the Study:

• To develop and validate a hyperspectral remote sensing method for early detection of ginseng root disease.
• To identify specific spectral bands and machine learning models for accurate disease classification.

Main Methods:

• Collected hyperspectral reflectance data (325-1,075 nm) from ginseng leaves without visual symptoms.
• Utilized machine learning models (Random Forest, Extreme Random Tree, Adaptive Boosting, Gradient Boosting Decision Tree) for classification.
• Analyzed vegetation indices and spectral band importance.

Main Results:

• The Extreme Random Tree model achieved 99% precision for healthy and 98% for diseased ginseng identification.
• The spectral band range of 690-726 nm was identified as crucial for early detection.
• Hyperspectral leaf reflectance effectively indicated underground root disease presence.

Conclusions:

• Leaf hyperspectral reflectance offers an effective NDT method for early ginseng root disease detection.
• This approach aids in timely disease management, preventing yield loss.
• The study proposes a novel strategy for safeguarding ginseng cultivation.