Identifying Invasive Predators In New Zealand
The Cacophony Project, a New Zealand based organisation, is using the latest technologies to lure, identify and eliminate invasive predators. This involved heat cameras, AI identification, sound lures and novel elimination tools.
Grant Ryan, The Cacophony Project
The Conservation Challenge
New Zealand is one of the worst places in the world for the number of species endangered or still in decline. This is because our native fauna did not evolve with predators and the introduced species continue to decimate local birds and insects.
The first part of the solution is a heat camera that can identify all introduced predators that are out there. The most common tools used at the moment are tracking tunnels and chew cards. Even off the shelf tracking cameras are designed for pigs and deer rather than smaller faster mammals.
Description of Technology Used
To make the new camera much less labour intensive and useful, we have applied AI to identify rats, stoats, possums, cats, birds etc. This is the most sensitive tool around to see the predators and we know exactly when they turn up each night. Already we have over 95% accuracy so only 5% of footage need to be looked at manually.
The AI part is relatively easy compared to the data management part to ensure that we have high quality properly tagged data for the AI to learn from.
This tool already looks to be much more sensitive and cost effective to monitoring predator numbers.
Opportunities/Call To Action
The opportunity it to use the AI to be able to rapidly test digital lures (social sounds) as we can apply big data tools to test lots of potential lures.
The ultimate next step is to use the AI to trigger more effective trapping. This is likely to start with an AI triggered open live capture trap and ultimately a paintball-like poison firing system to enable all predators to be eliminated with one too.
The Cacophony Project
The Cacophony Project is developing a set of technologies that will be deployed throughout New Zealand. These will:
- Lure invasive predators with sound and light
- Observe predators using a thermal camera
- Identify predators automatically using machine learning algorithms
- Eliminate positively identified predators
- Monitor the bird song over time to measure the impact