Imaging technology could help producers track broiler activity more easily, giving the entire supply chain a common assessment tool for assessing and monitoring welfare.
“Society is more and more concerned about broiler welfare nowadays so. For the broiler industry, gait score is considered one of the most important parameters,” Dr. Yang Zhao, assistant professor, University of Tennessee, explained at the 2023 Georgia Precision Poultry Farming Conference.
Currently, gait score assessments are conducted manually. Broilers are given a score ranging from one to five, where one means normal and a five indicates a severely impaired gait.
However, this approach is highly subjective and lack consistency.
Automating broiler gait scores
Automated gait scoring uses top view cameras to quantify bird activity and floor distribution at the flock level.
“With this information, combined with broiler age and body weight, we were able to understand the correlation between gait score and bird activity to predict a gait score,” said Zhao.
The system to calculate gait scores for individual broilers is slightly different.
“This system involves a side-view camera. This camera identifies a moving bird in the video and separates it from the background,” he said.
They’ve developed a deep learning method that uses neural networking to pinpoint parts of the broiler, such as the head, center and feet, to determine a gait score at the individual level. Deep learning and neural networking are two types of machine learning, which uses data to improve computer performance on a task over time.
It can be difficult to calculate gait scores if several broilers are clustered in the area where the side-view camera is recording. To circumvent that challenge, Zhang’s lab is testing modifications to an overhead camera as a potential solution.
The project was initially selected as a finalist for Phase I of the SMART broiler initiative from the Foundation for Food & Agriculture Research’s (FFAR).
