Noninvasive multiclass milk contaminant detection using hyperspectral imaging and hybrid ensemble learning

Area of Science:

• Food Science
• Analytical Chemistry
• Spectroscopy

Background:

• Milk contamination is a significant global health risk, particularly in developing nations.
• Accurate detection methods are vital for consumer safety and food industry integrity.
• Existing detection methods can be invasive or lack comprehensive analysis.

Purpose of the Study:

• To develop and evaluate a novel noninvasive method for detecting milk contamination.
• To compare the efficacy of invasive and noninvasive techniques for milk quality assessment.
• To introduce a hybrid ensemble learning framework for enhanced contamination detection.

Main Methods:

• Invasive analysis using the Lactoscan system for key milk parameters.
• Noninvasive analysis via hyperspectral imaging (400-1,000 nm) with Specim FX-10.
• Data preprocessing including radiometric correction and spectral smoothing.
• Classification using a hybrid ensemble learning (HEL) framework combining multiple models.

Main Results:

• The HEL framework demonstrated superior performance compared to existing methods.
• Achieved 100% accuracy in training and 96% accuracy in validation.
• Hyperspectral imaging effectively captured spectral and spatial features for contamination detection.

Conclusions:

• The developed HEL approach offers a highly accurate, noninvasive solution for milk contamination detection.
• This method holds significant potential for real-time food quality assurance and safety monitoring.
• Noninvasive hyperspectral imaging combined with advanced machine learning is a promising direction for food analysis.