Poult Sci. 2025 Dec 9;105(2):106228. doi: 10.1016/j.psj.2025.106228. Online ahead of print.
ABSTRACT
Identification and removal of cracked perforated embryo eggs are of great importance in the vaccine production industry. To address the issues of low efficiency, high labor intensity, and high misjudgment rates associated with manual crack identification after perforation, this study employed the box-counting fractal method. This method calculates the the fractal dimension of preprocessed embryo egg images, enabling differentiation between qualified eggs and those with linear or star-shaped cracks. However, misjudgment occurred due to overlap between the fractal dimensions of eggs with net-shaped cracks and those of qualified ones. To overcome the aforementioned limitations, a fractal matrix was constructed in the perforation region, from which 4 damage feature values were extracted. These values were combined with the fractal dimension to form the crack feature vector ζ. A logistic regression model was then used to establish a discriminant function, achieving an accuracy of 98.9 %. This method successfully digitizes crack information on the eggshell surface, establishes precise discriminant criteria, and provides a reliable basis for the automated removal of defective cracked embryo eggs.
PMID:41442919 | DOI:10.1016/j.psj.2025.106228