Recognition technology – reducing grazing pressure

Managing total grazing pressure on extensive sheep stations is just one of the benefits of soon to be commercialised animal recognition technology which uses machine vision instruments to control animals’ movements.

Rural Pacific Marketing has obtained the licence to commercialise the technology, which has been developed over the past five years by Neal Finch, research officer in wildlife biology from Queensland University’s School of Animal Studies, and Mark Dunn from the University of Southern Queensland. RPM’s CEO, David Lynch said the technology has undergone real time and on site testing at Stuart and Pru Barkla’s 15,000 ha property, Rosco Downs, near Cunnamulla in Southwest Queensland, and is undergoing further trials at a larger property, Tomoo in the Charleville area.

The technology, which was demonstrated on ABC television’s New Inventors last June, uses machine vision technology to identify animals and control their movements through automated gates to allow or deny them access to watering or feeding points. It can also be used to count stock numbers passing through the gate.

“The small, energy efficient hardware running the animal recognition software can also read RFID, which means individual tagged animals can be recognised,” Neal Finch said.

“Combining this technology with the ability to enclose a watering point and operate an automated gate provides an opportunity to manage large areas of grazing land in a revolutionary new way.

“If large enough areas have water access managed, then issues such as total grazing pressure and feral animal control can be addressed without traditional culling methods like shooting or poisoning.”

The equipment is small, light and robust – consisting of solar panels to power the gates, the minicomputer and a web camera.

Of course this does not include all the fencing, races and gates needed to direct the animals towards or away from the valuable water!

How the animal recognition system works

The system has three components. An enclosure, with entry via a narrow laneway, surrounds the watering point. An intelligent camera, based on machine vision (MV) technology, monitors the laneway, and an automated gate controls access to the watering point.
As the animal passes through the laneway, real time video footage is processed by the computer, which traces the outline of the animal. The computer is pre-programmed with a list of animal silhouette ‘templates’ which can be added to as needed. The computer matches the outline of the animal against its templates and identifies the animal.

Only one animal can pass through the laneway at a time. Since animals typically move at a walking pace of around two metres per second and range in length from fifty centimetres to three metres long, Finch said, the full animal will be in the frame anywhere from half a second to two seconds, during which time computer is able to identify the animal’s species before the next one passes through.

Once the computer identifies the animal, it can then activate a gate, which depending on the animal species, allows or denies them access to feed or water.

RPM’s Lynch said designs for the commercial product are being finalised, and it’s planned to launch it on the Australian market some time between April and June. The price of the commercial product is expected to be around $15,000, but the package will include a race to ensure the camera is positioned correctly.

The software can recognise sheep, cattle, horses, goats, pigs, camels and emus, and it can be taught to recognise other animals as needed. “The software is pretty straightforward. It’s just a matter of capturing the image. What do you want it to do -trap animals or allow them to drink?”

He suggested a potential use could be for separating animals at different stages in their life cycle. “For example weaning calves or lambs. It’s just a matter of the software getting the image, and then operating the gates to send the calves or lambs to a different pen.”

Of more importance to potential users than its exposure on the New Inventors, the equipment has been operating on Rosco Downs for almost four years, and has been demonstrated to considerable interest at two field days in the past six months. The first was at Rosco Downs on a cold, grey day in July, when, despite 30 points of very welcome rain, and more threatening, in a black soil area with unsealed roads, more than 120 people turned up to learn about the technology. It was organised in conjunction with UQ Gatton, AgForce Sheep and Wool, RPM and Southwest Natural Resource Management (SouthwestNRM), and Queensland’s Premier, Anna Bligh attended.

Rosco Downs proves technology’s value

Stuart and Pru Barkla have been working with Neal Finch and UQ for four years, trialling the animal recognition technology at their 15,000 ha property near Cunnamulla. They run prime merino meat lambs and wool producing lambs. The property also has feral goats which the Barklas have begun harvesting to send to markets in Asia and the US.

On average they have 6-8,000 sheep, and 4,000 breeding ewes. The goats are migratory, but Stuart Barkla said there are around 1500 feral goats. He sells 1000 goats a yeas. The goats and the large numbers of feral pigs and kangaroos have been a big problem, causing overgrazing and environmental damage. The area is hot, dry, and as a result of the recent long-lasting drought, generally arid.

“We’re trying to improve our land management by reducing the total grazing pressure,” he said. “If we can trap or keep out the feral animals we can reduce the grazing pressure without lowering our sheep numbers.”

The technology is set up in nine yards. There are three sets of equipment, which are moved around as needed. Barkla built the yards and fences himself. “It’s not cheap; it takes about 80 man hours and cost about $9,000 in labour and materials to build each yard.”

The computer-controlled gates proved a boon during the drought, as feral animals competed for stock with the dwindling water supplies. Fencing off the main water holes and only allowing sheep to drink, saved the stock. The 1500 mm high weldmesh gates are too high for kangaroos and pigs to jump or climb, so they are forced back to other small waterholes.

Barkla said the technology does more than just protect water resources. “It allows us to manage our stock more efficiently. For instance, we can take off different weights of sheep if they need a supplement when we weigh them in the yard.”

Managing feral animals has also been made easier and more efficient. Pigs and foxes were attacking the lambs, reducing the lambing average to 50% or even 40%. Barkla uses 1080 to control the ferals, and while it has wiped out the foxes, the pigs are harder to eradicate. “The technology allows us to monitor if the pigs are turning up at the watering holes, and we can trap them or shoot them. Our lambing average has gone from 40-50% to 120% over the last five years.”

The set-up

The animal recognition technology set-up (Courtesy Neal Finch)

Following the Rosco Downs demonstration, SouthwestNRM purchased three trial units for graziers in its region, which covers four river catchments, the Bulloo, Warrego, Paroo and Nebine-Mungallala-Wallam over 187,000 square kilometres of southwest Queensland. The Warrego, Paroo and Nebine-Mungallala-Wallam catchments are part of the Murray Darling Basin, while the Bulloo River is an internally draining river system.

“The success of the technology at Rosco Downs sparked our curiosity about its robustness and commercial application on a bigger scale,” SouthwestNRM CEO, Neil Judd said. “We’re exploring the merits of this technology in assisting graziers and others to better manage the environment.

The units were purchased so the organisation could assist not just graziers but Land Care and community groups to better manage the environment. He stressed that his organisation and the network of cooperative farmers linked to it would be exploring more applications than simply controlling stock access to water.

“We’re looking at working with graziers, and others working on revegetation and reclamation of degraded areas – salt areas, claypans, river banks.

“People and grazing are an important part of the whole ecosystem. We need to develop more efficient grazing, better subdivision fencing, new water systems.”

The three lots of animal recognition technology are installed in a 10,000 acre paddock at Tomoo, midway between Bollon, Charleville and Morven. Adam and Renee Bradshaw run sheep, cattle, a small group of domesticated camels, and feral goats, which are now being captured for the market. The property also has feral camels, pigs, kangaroos and wallabies, and has weed control and grazing issues.

Like the Barklas, SouthwestNRM and the Bradshaws will be working with UQ and Neal Finch to finetune the technology before it is released on the market.

“We will be exploring its robustness in a commercial setting and seeing what are benefits for the greater good of people in the community.” Judd said.

“Times have been really tough with the extended drought. People are scarce here, the distances are vast, and if they are going to invest $10,000 in this technology, they need to see both the ROI and better land management.”

The equipment is to be trialled in two stages, he said. Firstly, its robustness and reliability in identifying all the species – sheep, cattle, camels, goats, pigs, kangaroos and wallabies.

“Step two is to identify how we can use it. We’re interested in exploring how we can use it not just to assist with the exclusion of animals from feed or water, but, for instance, in trapping the feral goats being opportunistically farmed; to see if there are enough goats to be worth trapping in a different set of yards ready for market.”

In October, SouthwestNRM held a field day at Tomoo which 200 people attended.
“We’re asking them what they wanted from the equipment. What would they need to make it worth spending $10,000 on and be able to get a quick return?”

© Sue Cartledge

« Back to writing examples