CAMERA traps could soon get a lot smarter, courtesy of software being built in the minds and computers of a small group of Australian pest animal researchers.
If the work comes to fruition, it will mean that the images snapped by a camera trap - sometimes adding up to tens of thousands in a single download batch - can be rapidly screened to pull out only images of target species, whether they be dog, fox or cat.
Not only that, but the researchers are well along the path of being able to individually identify dogs. If a camera has snapped 20 images of dogs, the algorithms are able to tell how many times the same dog, or dogs, appears in those images.
Camera traps are already a revolutionary piece of hardware for monitoring populations of wild animals. The potential of the software being baked by a small group of scientists associated with the Invasive Animals CRC (IACRC), NSW Department of Primary Industries and University of New England promises to take that revolution to new heights.
For instance, the ability to tell whether one dog has made multiple passes across a camera will mean that camera traps become a much more effective way of monitoring wild dog numbers before and after a control program.
Paul Meek, Invasive Animals Officer with the NSW Department of Primary Industries, said at the launch of the National Wild Dog Action Plan that when the identification software is hooked up to other technologies, the potential for more strategic pest animal control becomes limitless.
The ability to recognise an image that contains a dog, versus all the other things that might trigger a camera trap, opens up the possibility of instantly alerting a landowner to the presence of a dog on or near their property.
The alert could be transmitted across an on-farm wireless network (another technology on the development track) to the landowner’s smartphone - so, in the words of Invasive Animals CRC chair Helen Cathles, the landholder “can go out and deal with the dog before it deals with them”.
Species identification also holds out the promise of very specific baiting technologies.
One idea in development with many of the IACRC partners is technology to incorporate into a “cat-grooming device”. A camera is triggered and its software identifies that the source of the movement is a feral cat. Within milliseconds the software has triggered a device to shoot a squirt of toxic gel onto the cat’s fur. When the cat grooms off the gel, it becomes an ex-cat.
(Such usage would require extreme confidence in the software, Mr Meek said. It would not want to muddle a cat with a small child, threatened species or a prize working dog.)
The team is also working on a project figuring out how to get baits to one species, feral pigs, but in a way that makes the bait inaccessible to other non-target species.
The answer may lie in a closed trap that only unlocks when the camera detects a pig.
Or, Mr Meek said, camera traps aimed at the ground and monitoring data in real-time might also give early warning of mouse plagues in cropping country.
So far, all these ideas are the works-in-progress of a few pest animal researchers, underpinned by the computer science expertise of Greg Falzon at UNE.
Their concepts work, Mr Meek said. Current versions of the camera trap software work to a high degree of accuracy, and individual dog identification works at 85 per cent accuracy on a sample of only 40 dogs, but the research has so far been unfunded. The science is yet to be refined and wrapped into user-friendly software.
That’s the next job, said Mr Meek, who undertook a 2011 Churchill Fellowship to look at camera traps for animal research. Conceiving ideas and the technology is one thing: finding the funds to make them a useable reality is another.