An effective method for detecting the wheat freshness by integrating biophotonics and machine learning algorithm.

Area of Science:

• Agricultural Science
• Biophysics
• Data Science

Background:

• Accurate wheat freshness assessment is vital for grain storage safety.
• Traditional methods can be time-consuming and lack precision.
• Novel approaches are needed for efficient and reliable quality evaluation.

Purpose of the Study:

• To develop an innovative method for quantitative wheat freshness evaluation.
• To integrate machine learning algorithms with Biophoton Analytical Technology (BPAT).
• To optimize machine learning models for enhanced accuracy in freshness detection.

Main Methods:

• Measuring spontaneous ultraweak photon emissions from wheat.
• Constructing feature vectors using statistical descriptors.
• Employing Particle Swarm Optimization (PSO) to tune Support Vector Machine (SVM) parameters.
• Validating the approach with K-Nearest Neighbors (KNN), Multilayer Perceptron (MLP), and decision trees.

Main Results:

• Both machine learning algorithms and input features significantly impact model performance.
• Central tendency factor features alone provide commendable recognition results.
• Variability factor features are not essential for achieving high accuracy.
• The proposed BPAT-SVM model demonstrates significant potential for wheat freshness assessment.

Conclusions:

• Machine learning combined with BPAT offers a novel and effective approach for quantitative wheat freshness evaluation.
• Feature selection, particularly focusing on central tendency factors, is crucial for model efficiency.
• This research contributes to improved grain storage safety through advanced analytical techniques.