Stock up on our livestock innovations and technology

By  Pamela Tyers 15 February 2024 5 min read

In support of our work with the livestock industry, we’ve got research stations near Armidale in New South Wales, near Townsville in Queensland and Floreat in Western Australia? One of our favourite things to do is take visitors behind the scenes to share our innovations in the livestock sector and the cool technology we work on.

We recently opened the doors of our site near Armidale, on Anaiwan Country, on the picturesque Northern Tablelands of New South Wales (NSW). Our Armidale Connected open day celebrated country, community and agriculture. Here is some of the current research we had on show.

Keeping an eye on cattle

Some of our work in digital technologies for monitoring livestock, like eGrazor, involves accelerometers. These are like the pedometers we use to measure our steps. They measure things like when an animal is walking, grazing or lying down chewing its cud. Beefed up with GPS trackers, they can tell us a lot about what animals are doing and where they are at any point in time. 

We’ve now also started researching computer vision coupled with machine learning as a way of monitoring and understanding the behaviour of livestock in different environments. Computer vision and machine learning are artificial intelligence (AI) techniques. 

We’re using cameras in the field to detect individual animals. Machine learning can then remove off-target objects like trees and identify individual features and behaviours such as grazing.

But it's still challenging for the technology to track multiple objects, or animals in our case, over time. Its orientation to the camera might make it hard to recognise as an animal. Bushes and trees can be mistaken as animals. It can be hard to tell animals in a herd or flock apart. And a big problem is obstructions like other animals getting in the way.

In the image above, how many cattle do you see? How many can the computer see? In the large green box, you can just see two cows but currently the computer only sees one. We're developing ways to teach the computer to recognise these differences to increase the accuracy of animal identification.

The research is at early stages but we’re hoping it will be another way we can monitor livestock and their well-being. 

He says, sheep says 

Like humans, sheep, cattle and other livestock vocalise to communicate what they’re feeling – desires, contentment, stress or pain. We hope if we can better understand sheep and cattle, it will offer another way we can help with their productivity and welfare. 

So we're using microphones and AI to recognise when livestock animals are talking. We then want to know what they’re saying. In other words, we want to go from ‘is that a baaa’? to ‘is that a happy baaa?’

Natural environments like livestock production form a ‘soundscape’. There are lots of sounds – not just animals, but birds, humans, weather and more. Soundscapes also change throughout the day. It’s a complex task to sift through the ‘noise’ to find important sounds, but AI helps us tackle the problem.

Farmers are better attuned to the sounds their animals make than technology could ever be, but they can’t be around them 24/7. With tech like this, we don’t aim to replace farmers. Rather, it will support their deep existing knowledge by providing continuous real-time data on important livestock production and welfare traits.

Combining audio with other data sources such as accelerometers, video and GPS can give us a more detailed picture of what’s going on, particularly relating to animal behaviour and health.

It’s early work but there are many potential applications. In addition to welfare monitoring, tech like this could one day also help with detecting disease symptoms or when animals go into labour.

Taking stock during droughts

In times of drought, farmers sell off some of their stock to make it easier to feed those they have left. We modelled strategies farmers can take when ‘destocking’ to use it as an opportunity to improve the quality and productivity of their herd or flock in the long run. 

Farmers aim to own animals of high genetic value and expect superior performance from them. Drought destocking can be an opportunity to sell off older animals or those with lesser genetic value.  However, in many commercial operations there is little information on individual animals, often making it difficult to identify which animals to select. 

In addition, farmers may be sceptical whether superior genetics perform well in challenging seasons. 

In this work, we asked the question whether targeted destocking could have an advantage over just selling off older animals. We then modelled the consequences of different destocking strategies over time. We were surprised to find the benefits of value-based targeted selection led to higher productivity in not only in good conditions but even more so in projected dry seasons.  

These are early results so far, but they mean that even when the climate outlook is poor there is something farmers can do to improve the long-term drought resilience of their operation.

This destocking research is part of our Drought Resilience Mission and delivered with the University of New England, CQ University and industry with funding from the Australian Government’s Future Drought Fund.

Virtual technology for our woolly friends

We’ve talked before about our virtual fencing technology for cattle. It’s a welfare-friendly system that uses a neckband and a pulse to guide the animals to stay in a grazing area or out of a sensitive area such as a creek. There doesn’t need to be an actual fence – the farmer outlines one on a device. The system is being commercialised by Gallagher.