Non-destructive detection of internal egg defects: Transmission imaging for blood spots and Vis-NIR spectroscopy for runny yolk.

Ensuring the internal quality of eggs is essential for food safety and industrial-scale grading. While current systems can detect blood spots, non-destructive identification of runny yolk remains a major challenge. This study evaluates a parallel, non-destructive approach utilizing transmission imaging specifically for blood spot identification, alongside visible-near-infrared (Vis-NIR) spectroscopy for the detection of both blood spots and runny yolk. In the imaging system, blood spots were detected based on pixel area thresholds, achieving up to 92.00% accuracy in white eggs and 88.00% in brown eggs. Using supervised classification, Linear Discriminant Analysis (LDA) yielded the best performance among image-based methods. However, Vis-NIR spectroscopy outperformed imaging, with Quadratic Discriminant Analysis (QDA) achieving 97.50% and 98.37% accuracy for blood spot detection in white and brown eggs, respectively. Notably, this study reports the first successful non-invasive detection of runny yolk, reaching 99.01% accuracy in white eggs and 100.00% in brown eggs using LDA. This was enabled by identifying distinct lipid absorption features in the 500- 600 nm range. Although induced defects were used for validation, the results demonstrate the feasibility of a robust, scalable framework for intelligent egg quality assessment, advancing automation in the poultry industry.