Year: 2026
Type: preprint
Abstract: Egg size and weight are very important for the poultry industry. Automated egg morphometry provides an efficient alternative to manual measurements for poultry breeding, hatchery management, and grading. This study was undertaken to automatically detect and measure eggs using computer vision and predict their weight. We also used species-specific calibration and evaluated weight prediction in chicken and goose eggs. A total of 661 eggs (504 chicken and 157 goose) were photographed using a smartphone together with a calibrated reference object. Computer vision-derived length and width measurements were compared with manual calliper measurements. We used Bland–Altman analysis and species-specific cross-validated linear calibration. Egg weight was predicted using linear, power-law, and Ridge regression models based on morphometric features. The YOLO11x-OBB detector had precision of 99.9%, recall of 100%, and [email protected] of 99.5%. Species-specific calibration substantially improved measurement agreement, reducing chicken length MAE to 0.088 cm and goose length MAE to 0.283 cm. Ridge regression achieved the best weight prediction performance, with out-of-fold R² values of 0.962 for chicken eggs. These findings demonstrate that 2D imaging combined with machine learning is practical for non-contact egg morphometry and weight prediction.
Source: Research Square
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: green
OpenAlex ID: W7213451539