Next generation phenotyping for production and efficiency traits in beef cattle

Hanson, Erin Marie

Permalink

https://hdl.handle.net/2142/130220 Copy

Description

Title

Next generation phenotyping for production and efficiency traits in beef cattle

Author(s)

Hanson, Erin Marie

Issue Date

2025-07-23

Director of Research (if dissertation) or Advisor (if thesis)

Bresolin, Tiago

Department of Study

Animal Sciences

Discipline

Animal Sciences

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Biometric Traits
• Digital Phenotypes
• Depth Images
• Feed Efficiency
• Novel Phenotypes

Language

eng

Abstract

There has been a growing need to improve feed efficiency traits in beef cattle to counteract the high costs of feed, which are part of the reason for the decline in beef cattle operations within the US. Advancements in sensor technologies have enabled the development of automated phenotyping tools to support management and breeding selection decisions. This study evaluated the potential of biometric body traits derived from depth images to predict production and efficiency traits in beef cattle. Specifically, we investigated the predictive performance of image-derived biometric features for body weight (BW) and feed efficiency traits including dry matter intake (DMI), residual feed intake (RFI), and residual average daily gain (RADG), the comparative performance of different predictive models, and the influence of image quantity and selection strategies on model accuracy. A total of 196 commercial Angus steers were video recorded using an Intel RealSense D455 camera. Biometric body traits, including projected volume, surface areas, length, width, and height, were extracted and summarized per animal using the median of multiple frames. Prediction models were developed using linear regression, partial least squares, elastic net, random forest, support vector machine, gradient boosting machine, and neural networks. Linear regression emerged as the most effective and practical model to predict all traits, achieving high accuracy with fewer computational requirements. For BW prediction, the combination of projected volume, flat surface area, and body length produced the most accurate results (R2 = 0.96; MAE = 10.12 kg). For DMI, the model using projected volume and flat surface area was the most effective (R2 = 0.59). However, biometric traits were not effective in predicting RFI, RADG, or residual intake and body weight gain (RIG) (R² < 0.01). Prediction accuracy was highest when biometric traits were summarized using all available frames, though models using 15 or more randomly or centrally selected frames also performed well. These findings demonstrate that biometric traits derived from depth imaging can accurately predict key production traits and may serve as valuable inputs for automated, non-invasive monitoring systems.

Graduation Semester

2025-08

Type of Resource

Text

Handle URL

https://hdl.handle.net/2142/130220

Copyright and License Information

Copyright 2025 Erin Hanson

Owning Collections