Harry Butler Institute
 

Artificial intelligence a new ecological warrior

A key challenge in preserving Australia’s incredible biodiversity is managing the devastating impacts of invasive species. Murdoch researchers are working with technology giant IBM to develop and trial the use of artificial intelligence (AI) to fight their spread.

 

Areas of research

Biosecurity

Technology utilised

Artificial intelligence

Lead researchers

Andre deSouza

 

What was the need for this project?

The project was initiated by Chevron and involved Harry Butler Institute staff training and testing an AI model to differentiate the invasive Asian House Gecko from native gecko species, which are protected on Barrow Island.

Time is critical when dealing with a highly mobile animal such as a gecko and the remoteness of the island means formal assessments can take up to three days to complete. Using artificial intelligence, the team can test an image and confirm the identity of a suspect species within seconds.

 

How the project was completed

Using a cumulative machine learning process referred to as deep learning, computer software was trained to recognise images of known species and flag the occurrence of potential biosecurity threats.

The technology also helps to provide a rapid triage system to activate and maintain biosecurity responses. If an invasive species entered one of the AI-enabled traps, it would be triggered to close and alert biosecurity staff.

 

Results and achievements for this project

The team is now looking at using AI to recognise weed species in conjunction with an autonomous robot for weed spraying, as well as checking drone footage for animal tracks, such as those left by sea turtles.

Incorporating AI into the research process has broad applications for environmentalists, who may spend several days manually walking paths or beaches in search of animals or animal tracks.

This technology allows them more time to study the footage of the beach and improve the chances of spotting turtle tracks or improve the timeframe for these activities.

 

Other case studies