The requirements to the methods of embryo sex determination in an egg have been formulated and substantiated in accordance with the tightening of the previously accepted norms of cockerel culling during incubation. New methods under development for identifying and culling of egg embryos within 7 days of incubation have been analyzed, and their advantages and disadvantages have been described. Two non-invasive techniques have been identified that have some potential for commercialization in the poultry industry (infrared spectroscopy and computer vision). The purpose of the study is to determine the possibilities of a non-invasive method for determining the sex of an embryo in an egg prior to incubation based on intelligent analysis of the proposed morphometric features of poultry eggs. The scientific novelty of the research lies in the fact that for the first time a method of determining sexual dimorphism based on the analysis of egg asymmetry parameters by three spatial coordinates determined by computer vision methods with the use of machine learning has been developed. An experimental unit for viability assessment and establishment of the necessary conditions for incubation and hatching of chicks has been developed to validate the implementation of the proposed method. It includes a smart incubator "Smart Nest", a brooder, a thermal imaging micro-camera TE-Q1, an oil-filled radiator POLARIS model PRE T 0915, an air humidifier Ergopower ER 604, a bactericidal air irradiator-recirculator DEFENDER 2-15C, a thermohygrometer RGK TH-30 and a laptop. For image acquisition, the setup utilized a Canon EOS 2000D EF-S 18-55 III Kit digital camera with a state-of-the-art CMOS sensor (22.3 × 14.9 mm) and a powerful processor. The geometric spatial digital model of each egg was artificially divided into a set of elements by software, by which the asymmetry of the egg shape was determined. In doing so, their shape indices, area, volume and perimeter were determined from the measured linear dimensions of each element. Incubation of 72 fertilized eggs of Dekalb White cross hen was carried out. Following the incubation, it was possible to reliably determine the sex of 38 chicks. Applying machine learning methods in solving binary classification problems for a small sample (38) with high dimensionality of the initial feature set yielded three final models with accuracy metrics AUC = 73–72% and F1 = 69–72%: Random Forest classifier with 4 evaluators and maximum depth of 3; Random Forest classifier with 10 evaluators and maximum depth of 5 and AdaBoost classifier with 4 decision tree evaluators and maximum depth of 3. Experimental confirmation of the relationship between the egg shape asymmetry and its sexual dimorphism will make it possible to approach the solution of the world scientific problem of reliable determination of the egg sex before incubation.
