Theses and Dissertations

Issuing Body

Mississippi State University

Advisor

Chesser, Gary Daniel Jr.

Committee Member

Zhao, Yang

Committee Member

Purswell, Joseph L.

Committee Member

Du, Qian

Committee Member

To, Filip Suminto D.; Linhoss, John

Date of Degree

4-30-2021

Original embargo terms

Worldwide

Document Type

Dissertation - Open Access

Major

Agricultural Science

Degree Name

Doctor of Philosophy

College

College of Agriculture and Life Sciences

Department

Department of Agricultural and Biological Engineering

Abstract

Appropriate measurement of broiler behaviors is critical to optimize broiler production efficiency and improve precision management strategies. However, performance of different precision tools on measuring broiler behaviors of interest remains unclear. This dissertation systematically developed and evaluated radio frequency identification (RFID) system, image processing, and deep learning for automatically detecting and analyzing broiler behaviors. Then different behaviors (i.e., feeding, drinking, stretching, restricted feeding) of broilers under representative management practices were measured using the developed precision tools. The broilers were Ross 708 in weeks 4-8. The major findings show that the RFID system achieved high performance (over 90% accuracy) for continuously tracking feeding and drinking behaviors of individual broilers, after they were customized and modified, such as tag sensitivity test, power adjustment, radio wave shielding, and assessment of interference by add-ons. The image processing algorithms combined with a machine learning model were customized and adjusted based on the experimental conditions and finally achieved 85% sensitivity, specificity, and accuracy for detecting bird number at feeder and at drinkers. After adjusting labeling method and hyperparameter tuning, the faster region-based convolutional neural network (faster R-CNN) had over 86% precision, recall, specificity, and accuracy for detecting broiler stretching behaviors. In comprehensive algorithms, the faster R-CNN showed over 92% precision, recall, and F1 score for detecting feeder, eating birds, and birds around feeder. The bird trackers had a 3.2% error rate to track individual birds around feeder. The support vector machine behavior classifier achieved over 92% performance for classifying walking birds. Image processing model was also developed to detect birds that were restricted to feeder access. Broilers had different behavior responses to different sessions of a day, bird ages, environments, diets, and allocated resources. Reducing stocking density, increasing feeder space, and applying poultry-specific light spectrum and intensity were beneficial for birds to perform behaviors, such as feeding, drinking, and stretching, while using the antibiotics-free diet reduced bird feeding time. In conclusion, the developed tools are useful tools for automated broiler behavior monitoring and the measured behavior responses provide insights into precision management of welfare-oriented broiler production.

Sponsorship

USDA-ARS cooperative agreement (6064-32630-008-01-S), USDA National Plan Program (Project No. 6064-32630-008-00-D), USDA Non-Assistance Cooperative Agreement (58-6064-9-015), USDA National Institute of Food and Agriculture (Accession number: 1020090),

Download

Share

COinS