Non-destructive quantification of egg yolk ratio using visible-near-infrared hyperspectral imaging, machine learning and explainable AI.

Summary

This study investigates the application of visible-near-infrared hyperspectral imaging as a non-destructive analytical technique for quantifying egg yolk ratio, using machine learning algorithms to build predictive models. Explainable AI (XAI) methods are employed to identify the spectral features most influential in model predictions, enhancing interpretability and practical utility. The work contributes to rapid, non-invasive quality assessment methodologies applicable to egg grading and processing contexts.

UK applicability

Whilst the study is not UK-specific, the methodology is directly applicable to UK egg production and food quality assurance sectors, where non-destructive inline inspection technologies are of growing commercial and regulatory interest.

Key measures

Egg yolk ratio (%; predicted vs actual); model prediction accuracy (e.g. R², RMSE, RPD); spectral wavelength importance (explainable AI outputs)

Outcomes reported

The study assessed the ability of visible-near-infrared (Vis-NIR) hyperspectral imaging combined with machine learning models to quantify egg yolk ratio non-destructively. It likely reports prediction accuracy metrics and identifies key spectral wavelengths contributing to model performance via explainable AI methods.

Topic tags