Classification and Recognition of Soybean Quality Based on Hyperspectral Imaging and Random Forest Methods

Area of Science:

• Agricultural Engineering
• Spectroscopy
• Machine Learning

Background:

• Accurate classification of soybean components is crucial for quality assessment during harvesting and storage.
• Hyperspectral imaging offers a non-destructive method for analyzing agricultural products.

Purpose of the Study:

• To develop a rapid and accurate method for classifying soybean components using hyperspectral imaging.
• To assess breakage and impurity levels in machine-harvested soybeans.

Main Methods:

• Hyperspectral image acquisition using Pika L spectrometer.
• Application of eight spectral preprocessing techniques (BC, MA, SGD, etc.).
• Feature wavelength selection via SPA and CARS algorithms.
• Development and optimization of a Random Forest (RF) model using PSO and DE algorithms.

Main Results:

• Optimal feature wavelengths identified using SPA (8) and CARS (10).
• RF model achieved 1.0000 prediction accuracy with SPA-BC preprocessing and DE parameter optimization.
• Demonstrated feasibility of hyperspectral imaging for soybean component detection.

Conclusions:

• Hyperspectral imaging provides a viable solution for rapid and accurate soybean component classification.
• The developed method offers technical support for quality control in soybean harvesting and storage.
• This research serves as a reference for future automated detection systems for machine-harvested soybeans.