Early In-ovo Sexing of Mojosari Duck Eggs on Day 5 Using Hybrid Feature Fusion and Machine Learning
Year: 2026
Type: article
Abstract: Early in-ovo sexing can support more efficient hatchery management by enabling sex determination before hatching.This study investigates a non-invasive framework for sexing Mojosari duck eggs at day 5 by integrating external morphological, embryonic vascular, and Gray-Level Co-occurrence Matrix (GLCM)-based textural features.External morphological measurements, including egg length and width, were obtained directly using a digital caliper, while egg weight was measured using a digital weighing scale; embryonic vascular and textural features were extracted from RGB candling images acquired with the eggshell intact.A total of 503 labeled eggs were analyzed.Statistical differences between female and male eggs were assessed using Welch's t-test with Holm adjustment and Hedges' g effect sizes, showing significant differences in Length, Width, Weight, and Sinus Terminalis Area.Four classifiers Logistic Regression, Random Forest, SVM-RBF, and XGBoost, were evaluated using Bayesian optimization and stratified cross-validation, with model configuration determined exclusively from training and validation data.On the independent test set of 100 eggs, Random Forest achieved 85.00% accuracy, 82.16% balanced accuracy, 83.41% Macro-F1, and 0.9285 ROC-AUC.Batch-held-out evaluation yielded 80.44% accuracy, 75.92% balanced accuracy, 77.22% Macro-F1, and 0.8795 ROC-AUC.The study provides application-level evidence for day-5 in-ovo sexing of Mojosari duck eggs using low-cost RGB candling, while characterizing sex-related features and contribution of feature groups (hybrid feature fusion).Future studies should validate the approach across broader duck populations, incubation stages, and imaging conditions.
Authors: M. Ali Hanafiah(Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing), Edi Noersasongko(Conceptualization, Methodology, Supervision, Validation, Writing – review & editing), Heru Agus Santoso(Formal analysis, Methodology, Resources, Supervision, Validation, Writing – review & editing), Ahmad Zainul Fanani(Formal analysis, Investigation, Methodology, Supervision, Validation, Writing – review & editing)
Institution: Universitas Dian Nuswantoro
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: hybrid
OpenAlex ID: W7235181176
