Hyperspectral imaging for non-destructive quality evaluation and grading of chicken eggs under diverse breeds and storage conditions

Proper storage practices are critical to maintaining the freshness, safety, and nutritional quality of chicken eggs throughout the supply chain. Traditional evaluation methods such as candling and Haugh Unit test are destructive, labor- intensive, and unsuitable for large-scale grading. To address these challenges, this study introduces a non-destructive framework that integrates hyperspectral imaging with machine learning to monitor and classify egg quality across breeds and storage conditions. Freshly laid eggs from three breeds Hy-Line, Badcock, and Rajasree were stored at 3 °C, 20 °C, and ∼ 33 °C for 42 days, with hyperspectral imaging performed at three day intervals over the 400–1000 nm range. Maximum and average spectral reflectance features were extracted and correlated with Haugh Unit, serving as the reference standard for internal quality. The results demonstrated clear spectral differentiation of freshness grades (AA, A, B, and C), achieving grading accuracies exceeding 92% at 3 °C, 88% at 20 °C, and 85% at ∼ 33 °C. Regression analysis further confirmed the strong predictive capability of hyperspectral data, with the best-performing models varying across breed and storage temperature, and Random Forest frequently yielding superior performance, reaching coefficients of determination up to R 2 ≈ 0 . 95 with RMSE values as low as ≈ 3 . 0 . The adoption of leakage-free preprocessing and nested cross-validation ensured robust and generalizable predictions. Overall, this work establishes hyperspectral imaging combined with machine learning as a practical, scalable, and non-invasive solution for reliable egg quality monitoring and automated grading, with strong potential to improve efficiency, reduce food loss, and enhance consumer confidence in the poultry industry.