Menu Close

REU Student Project Abstract

Christian Peterson Headshot
Christian Peterson
Faculty Mentor(s): Dr. Angela Green-Miller and Dr. Vikram Adve
Ethogram Development for Computer-Based Detection of Estrus in Individually Housed Goat Does
Detecting estrus in livestock can be a challenging aspect of reproductive management. Installation of a computer-vision tracking system could improve reproduction and labor efficiency. The objectives of this project are to advance the development of tracking technology at the Tuskegee University Caprine Research and Education Unit, including development of a camera monitoring systems and development of an ethogram that can be used to differentiate between goat behaviors. The work proposed here is based on previous computer vision-based behavior analysis experiments done in collaboration at The University of Illinois at Urbana-Champaign. The planned computer vision tracking system will address labor shortage challenges by allowing farmers to remotely monitor goats without having to physically be at the farm, and alert farmers whenever estrus behaviors are being displayed. The installed camera system will be utilized for 24-hour monitoring of individually housed Kiko does. There will be a total of 10 cameras available to record 12 does. The cameras will be connected to an on-site server that will collect and store the data. Establishing a data pipeline to efficiently move all acquired images at the barn to a safe storage area is also an important part of the data management aspect. For analysis, the ethogram will be developed for individually housed Kiko does by watching several videos from different times of the day and writing down every behavior observed. The initial ethogram will then be revised to include desired behaviors relevant to the study. Recordings will be analyzed to assign estrus and non-estrus behaviors to yield an initial ethogram. The ethogram will be used along with a behavior labeling tool to annotate behaviors of does. For model development, another priority is creation of a well-labeled ground truth data set, which can also be used for future projects. Overall, the REU experience was helpful in explaining how behavior studies utilizing cameras and learning models could be designed. I enjoyed the professional seminars presented, and I also enjoyed visiting the research farms and facilities in person because it allowed us to get a sense of the all the video footage we worked on over the summer.