AI Case Study

Researchers from the University of Technology Sydney aim at ensure livestock are in peak condition leading up to sale by analysing cattle through AI

Researchers at the University of Technology Sydney have developed technology to determine when livestock are ready for market. Using cameras to analyse cattle as they move, sensors capture contours that reflect fat and muscle depth and size, which are then converted to 3D images. These are processed with AI to suggest a condition score for each animal. As a result farmers may be able to take better-informed decisions to ensure livestock are in peak condition leading up to sale.

Industry

Basic Materials

Agriculture

Project Overview

"Researchers at the University of Technology Sydney (UTS) have developed technology using off-the-shelf cameras to analyse cattle as they move through a crush to determine a condition score for each animal. In 2015, robotics expert Dr Alen Alempijevic from the UTS Centre for Autonomous Systems developed the technology using purebred Angus cattle on farms and in saleyards around Armidale and Grafton in New South Wales. As he explained, operating at 30 frames per second, sensors capture contours that reflect fat and muscle depth and size. This information is converted to 3D images that are processed through artificial intelligence algorithms to provide an accurate condition score for each animal.
By calibrating the fat measurement via ultrasound and a muscle score determined by an expert assessor, the software uses the mathematical description to estimate an animal’s condition based on the 3D shape the machine can “see”. The research was funded by MLA and the NSW Department of Primary Industries.
The analysis may assist other decisions on farm, such as allowing a farmer to select animals with superior measured traits for breeding the next generation."

Reported Results

Research; results not yet available
"It is believed that the technology could result in a transformative shift in livestock management with prediction accuracy of yield potential rising to as high as 80–90%, according to Dr Sangster, program manager for Grassfed Productivity at Meat & Livestock Australia (MLA)."

Technology

The information captured by sensors is converted to 3D images that are processed through artificial intelligence algorithms to provide an accurate condition score for each animal.

Function

R And D

Product Management

Background

According to The Rural Industries Research and Development Corporation (RIRDC), "Australian livestock farmers are facing the challenge of growing individual animals to a condition that matches market specifications and timing. To do this, they proactively manage nutritional inputs, including pasture availability, to achieve growth rate, composition and product quality. However, even farmers with many years’ experience can struggle to predict an individual animal’s yield potential and fat content prior to sale, with some estimates putting prediction accuracy as low as 20–30%.

The benefits of an animal meeting the industry specifications for yield at slaughter can range from $10 per head in sheep up to $80 per head in cattle. For producers marketing under Meat Standards Australia specifications for eating quality, the losses for non-compliance can reach $300 per head, explained Dr Nick Sangster, Program Manager for Grassfed Productivity at Meat & Livestock Australia (MLA)."

Benefits

Data

Contours reflecting fat and muscle depth and size captured by sensors