Doctoral Dissertations
Date of Award
12-2025
Degree Type
Dissertation
Degree Name
Doctor of Philosophy
Major
Animal Science
Major Professor
Yang Zhao
Committee Members
Yang Zhao, Hao Gan, Tom Tabler, Maria Prado, Hairong Qi
Abstract
Poultry is the leading source of animal-based protein worldwide. Its growing demand has significantly increased pressure on the stakeholders to handle major industry-wide challenges. Therefore, poultry research has been highly invested in applying technological advancements. Specifically with the advent of deep learning methods (DL) feasible livestock solutions through precision technologies have gained traction. This dissertation explores the application of DL technology to address key issues in broiler welfare management and broiler breeder production. Specifically, it focuses on automatic quantification of mobility and exploring dynamics of mating behavior in broiler breeders and its effects on egg fertility through DL model applications. Chapters one and two addressed the gap in detecting and tracking individual broilers in pen settings. The feasibility of DL models and existing algorithms was explored in this context. While DL models could effectively detect broilers but tracking individual broilers was not as successful. In chapter three, the continuous identification issue of broilers was tackled by color-coding them uniquely. A DL model was applied to detect and classify individual color-coded broilers, hence providing their continuous mobility in the pen, leading to an alternative automatic gait scoring approach. Chapter four provides a study about understanding the tibia bone strength in broilers as affected by their mobility and physiological parameters such as weight. The mobility of individual broilers, provided by a DL model application, and the tibia bone strengths did not show significant correlations, partly due to the multifactorial nature of bone health. Chapter five is about mating behavior in broiler breeders detected through vision-based DL models. The mating sub-actions, mounting and tail movement, were accurately detected and classified by the model. The results showed a decreasing trend in rooster mating behavior as affected by age, weight, gait and footpad dermatitis. Chapter six, logistic regression statistical modeling was utilized to analyze mating behavior sub-actions’ effect on egg fertility levels in four pens. The results indicated that roosters were impacting fertility levels by the number of mating per day, selective behavior in mating, the hens were consequential with mating receptivity, gait and welfare conditions.
Recommended Citation
Jaihuni, Mustafa, "Deep Learning Model Applications in Broiler Welfare and Breeder Production Management. " PhD diss., University of Tennessee, 2025.
https://trace.tennessee.edu/utk_graddiss/13606