Trap maker unleashes AI in the war against possums

NZ AutoTrap's possum and rat trap.

A trap maker is using artificial intelligence (AI) in the fight against possums.

The firm, NZ AutoTraps, says AI, wireless connectivity and robust, efficient trapping equipment could be potential game-changers in protecting New Zealand’s native birds and wildlife from pests.

AutoTraps says its latest trap incorporates an automatic resetting trap, with an AI-powered on-board camera and management system that engages the trap only when it recognises a target species – such as a possum, rat or stoat – and reports back when it kills a pest via a remote network of “yarn mesh” nodes.

The system, designed to create a network of connectivity across New Zealand’s remote, rugged and mountainous environments, is the result of a collaboration between NZ AutoTraps and FTP Solutions, a global technology provider focused on helping businesses capture and manage data.

NZ AutoTraps is 50% owned by New Zealand Carbon Farming, which uses the traps in its forests around the country.

Operations manager Haydn Steel said AutoTraps had looked to improve its initial trap design by avoiding the risk of by-kill.

“It’s also important, particularly for groups which are results-funded, to have a reliable record of when the trap has worked,” he said.

FTP Agriculture manager Jonathan Clark says the new AT520-AI uses a machine-learning model to train the AI on thousands of images so it can accurately distinguish between a target and non-target species.

“As an animal approaches the trap, it detects that movement through the proximity sensors on the AT520, and then the AI camera starts firing off a sequence of images – classifying each image in about one 20th of a second – and arming the trap only when we get a positive ID on a target species,” Clark said.

“The AT520-AI can also alert other nearby traps on the network to the presence of a targeted pest, which can then also deploy their lures, increasing the probability of removing the animal.”

In the next iteration of the AT520-AI, in development for 2025, the AI will also be able to instruct the traps to deploy one of three species-specific lures.

FTP’s network means the system is ideal for areas with limited internet connection, especially where direct line of sight is not possible, with traps able to be connected up to 25km apart.

The traps record and send valuable surveillance data for predator management.

Steel said the company has also been attracting interest from organisations looking to use the yarn mesh network to provide health and safety support for employees and volunteers working on predator control programmes in remote areas.

Jamie Gray is an Auckland-based journalist, covering the financial markets and the primary sector. He joined the Herald in 2011.