Design and development of a broiler mortality removal robot

Authors: LI, GUOMING - Mississippi State University; CHESSER, JR, GARY - Mississippi State University; Purswell, Joseph - Jody; Magee, Christopher; GATES, RICHARD - Iowa State University; XIONG, YIJIE - University Of Nebraska

Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/22/2022
Publication Date: 12/14/2022
Citation: Li, G., Chesser, Jr, G.D., Purswell, J.L., Magee, C.L., Gates, R.S., Xiong, Y. 2022. Design and development of a broiler mortality removal robot. Applied Engineering in Agriculture. 38(6):853-863. https://doi.org/10.13031/aea.15013.
DOI: https://doi.org/10.13031/aea.15013

Interpretive Summary: Manual collection of broiler mortality is time-consuming and laborious. The objectives of this research were to design and evaluate a robotic arm to locate and remove broiler mortality under production conditions. The broiler shank was the target anatomical region for detection and picking. Deep learning models and image processing algorithms were employed to provide location and orientation of the shank to guide the gripper mechanism under varying light intensities. Results showed that increased light intensities improved the performance of the deep learning model detection, image processing orientation identification, and final picking performance. The final success rate for picking dead birds was 90.0% at the 1000-lux light intensity. The research has shown that mortality removal can be implemented with robotics, but requires further research to optimize detection across a broader range of bird age and lighting environments.

Technical Abstract: Manual collection of broiler mortality is time-consuming and laborious. The objectives of this research were to 1) design and construct a broiler mortality removal robot from commercially available components to automatically collect dead birds; 2) compared and evaluate deep learning models and image processing algorithms for detecting and locating dead birds; and 3) examine detection and picking performance of the robot under different light intensities. The robot consisted of a hand-mounted camera, a two-finger gripper, a robot arm, and a computer connected with an ethernet cable. The robot arm was mounted on a table, and 64 Ross 708 broilers between 7 and 14 days of age in weeks 1 and 2 were used for the robot development and evaluation. The broiler shank was the target anatomical part for detection and picking. Deep learning models and image processing algorithms were embedded into the vision system and provided shank information of location and orientation of the shank of interest, so that the gripper can arrive at the shank of interest and be positioned rotate to be perpendicular with the shank for precise picking. Light intensities of 10, 20, 30, 40, 50, 60, 70, and 1000 lux were set for the testing. Results indicated the You Only Look Once (YOLO) V4 deep learning model was able to detect and locate shanks more accurately and efficiently than the YOLO V3. Higher light intensities improved the performance of the deep learning model detection, image processing orientation identification, and final picking performance. The final success rate for picking dead birds was 90.0% at the 1000-lux light intensity. In conclusion, the developed system is helpful to robotize the broiler mortality removal process, which contributes to further development of integrated autonomous solutions for improvement of production and resource use efficiency in commercial broiler production and improves well-being for workers.