Hyperspectral System Coupled With a Global Spectral Feature Classification Network to Identify the Origin of Mung...

Area of Science:

• Agricultural Science
• Data Science
• Food Science

Background:

• Geographical origin fraud in mung beans undermines quality, brand integrity, and supply chain trust.
• Accurate traceability is crucial for maintaining the specific attributes and market value of mung beans.

Purpose of the Study:

• To develop and validate a novel method for identifying the geographical origin of mung beans.
• To address the challenge of origin fraud and enhance supply chain integrity in the mung bean market.

Main Methods:

• A hyperspectral system was used to collect spectral data from mung beans originating from six different locations.
• A Global Spectral Feature Calculation Module (GSFCM) was developed, integrating convolution, self-attention, and residual connections for deep feature extraction.
• A Global Spectral Feature Classification Network (GSFC-Net) was designed to map spectral information to origin labels, validated through structural optimization and ablation experiments.

Main Results:

• The GSFC-Net achieved high classification performance, with an accuracy of 98.10%, precision of 98.09%, recall of 98.55%, and an F1-score of 98.32%.
• The proposed method demonstrated superior classification performance and stability compared to existing spectral information classification techniques.
• Ablation experiments confirmed the effectiveness of the GSFC-Net architecture and its components.

Conclusions:

• The study presents an effective analytical method, GSFC-Net, for authenticating mung bean origin using hyperspectral data.
• This technology provides a robust solution to ensure product authenticity and combat geographical origin fraud.
• The findings support fair trade practices and enhance consumer trust in the mung bean market.