Researchers have used drones equipped with koala-seeking heat sensors to map the marsupials' population in Brisbane from 60 metres above the tree canopy.
The joint Queensland University of Technology-Brisbane City Council pilot study was in addition to a project in which koala-sniffing dogs and human field workers walked through bushland to identify where Brisbane's koalas could be found.
Sniffing dogs and field walkers identified “high koala activity” in bushland at Belmont, Burbank, Kuraby, Wishart, Bardon, Mt Coot-tha, McDowall, Anstead, Moggill, Alderley, Nathan and Mt Gravatt.
Researchers hoped drones could be a more cost-effective method of identification and mapping of koalas and other endangered species than walk-throughs of bushland.
QUT koala ecologist Grant Hamilton and his team have developed a way of using computer algorithms to read data from a heat-sensing camera flown by a drone.
“We are flying high-tech sensors using drones and we are taking the data that comes from the thermal imagery, which means it captures the heat from animals,” Dr Hamilton said.
New drone-supported technology is emerging which could make it faster and cheaper to map koalas.CREDIT:BRISBANE CITY COUNCIL
“We feed that thermal imagery into an artificial intelligence algorithm and we use those algorithms to automatically determine the kind of organisms are using that system, so it might be koalas or it might be kangaroos, for example.
“What we do when we fly over the area, we can count the number of koalas in the area.
“We are getting up to 100 per cent accuracy at the moment using this system.”
The algorithm identified everything that was “not a koala”, Dr Hamilton said, so only koalas showed up in the imagery.
Trials have been flown in bushland near Petrie, where the University of the Sunshine Coast Moreton Bay campus was being built.
“We have also flown in northern New South Wales and we hope to do some extra sites in Brisbane soon,” Dr Hamilton said.
The thermal imagery could cover about 20 hectares in about two hours.
“The aim here is to find a methodology to identify a system where we can help many species, not just one,” Dr Hamilton said.
He said better data was a conservationist's best friend.
“Realistically, the conservation budget is always limited, so if you can do the same job more cheaply that is great thing to do.
"If you can do it better – more cheaply – that is an even better outcome.”
Dr Hamilton said in certain environments the team - which included data scientist Evangeline Corcoran and artificial intelligence algorithm specialist Simon Denman – was getting 100 per cent accuracy.
“We are doing it much more cheaply and we are getting better results than humans are getting," he said.
He said it would take time – and extra research grants – before the research would be used by planning bureaucrats.
“It would be a couple of years away but our aim is to be able to go out with a drone to a particular site, get the data and either have it sent back directly or take the SIM card out of the drone once it comes back, insert it into the computer and get the answer within a half an hour, or an hour.”
Tags: Drone Photography, Conservation, Environment