Meat & Livestock Australia: Instant animal counting

This project improves the welfare and mortality rates of livestock during transport by developing AI-based video analytics technologies for detection, counting and monitoring.

Challenge

Australia exports red meat and livestock to more than 100 countries, representing over 60% of the industry’s production. Livestock must be counted before overseas transport but the current method requires intensive physical resources that are time-consuming and often inaccurate, leading to over or under crowding in holding pens.

MLA needs a more reliable method of monitoring sheep loading and pen capacity in order to improve livestock welfare and mortality rates.

Solution

The UTS Tech Lab team developed an automated sheep counting technology that provides irrefutable numbers at loading and unloading.

The system uses two cameras mounted on a central pole and suspended four metres above the ship ramp. The cameras detect sheep movements from point-to-point, and only animals successfully entering and exiting the counting zone are recognised.

Using advanced computer vision and pattern recognition algorithms, the video-based system can achieve sheep counting in real time and variable weather conditions, for multi-loading ramps.

This revolutionary application of AI determines ship capacity, pen density and whether sheep are standing or lying down. This detection can trigger staff to check on an animal’s condition and can flag ‘bullying’ amongst pen occupants.

Duration
12 months

Academic team
Associate Professor Jian Zhang

Lab
Multimedia Data Analytics

Engagement model
Joint-funded project

Funding
Livestock Export Program (LEP)
Meat & Livestock Australia

Future applications
Export terminals
Sale yards
Feed lots
Water pools

Area of expertise
Artificial intelligence
Data science