Translucency of chicken eggs: proposal of an automated system for analysis and classification

Eggshell quality is a determining factor in food safety, production efficiency, and the commercial acceptance of eggs. This study proposed the development and validation of an automated system for measuring eggshell translucency using computer vision and machine learning. A total of 326 commercial eggs from different production systems, with white and brown shells, were analyzed. Images were captured in a controlled environment and digitally processed to extract quantitative translucency measurements. The obtained values were compared with traditional visual classification and used in supervised classification models (KNN, SVM, and Random Forest). The SVM model showed the best performance, with accuracy exceeding 90 % in distinguishing translucency levels. Additionally, predictive models (Multiple Linear Regression and SVM) were tested to estimate intrusive variables based on translucency, revealing moderate correlations, particularly with shell thickness and shell weight. It is concluded that translucency can be accurately quantified through automated techniques, with potential application in the screening and quality control of commercial eggs, although it should be used as a complementary indicator alongside other technical parameters.