Non-invasive classification of single and double-yolk eggs using Vis-NIR spectroscopy and multivariate analysis.

1. This study was conducted to develop an efficient technique for separating double-yolked (DY) from single-yolked (SY) light brown broiler eggs with comparable shape and size, that were hard to distinguish merely by their external characteristics, using Vis-NIR transmission spectroscopy combined with multivariate analysis.2. Spectroscopic transmission (200-900 nm) was measured after collecting the eggs, and the yolk number was verified by breaking the eggs after boiling. The absorbance of important spectral wavelengths sensitive to yolk amount were identified using feature selection techniques (Principal Component Analysis and Genetic Algorithm).3. Discriminant analysis (DA) and support vector machine (SVM) classifiers were used to develop classification models for DY and SY eggs using the selected important spectral wavelengths.4. When compared to alternative nonlinear techniques, the developed model applying linear discriminant analysis produced greater accuracies in the first (96%) and second (100%) experiments, implying lower inter-egg variability from spectral data and a linear relationship between classes. However, the position and orientation of yolks in DY eggs may limit the classification accuracy of the eggs.