The koala is one of the emblems of Australia. The species is not doing the best, however. Classified as vulnerable by the International Union for the Conservation of Nature (IUCN), it is particularly threatened by the loss and degradation of its habitat linked to logging and urbanization.
But road accidents are another important pressure. Each year between 1997 and 2018, an average of 356 koalas were collected from health centers after being struck by vehicles. It is to better understand this threat that researchers have started a new project.
The initiative aims to study how koalas use wildlife crossings in the south-eastern region of Queensland. In order to reduce collisions with animals, several structures have in fact been built in various places where automobile traffic is important.
Some are underground routes, others are bridges over the busiest roads. If these constructions have proven their worth in other regions of the world, the Griffith University team decided to analyze their usefulness for koalas using innovative technology.
Devices trained to recognize koalas
To observe and track animals, specialists usually use identification tags or GPS devices. The Australian team has chosen to turn to “facial recognition” by training, via artificial intelligence (AI), cameras to recognize koalas.
“Previously, cameras were installed to monitor koala crossings but each of the captured videos must be manually checked to see if the animals filmed were koalas or other species.“said Jun Zhou, assistant professor in charge of the two-year pilot study.
“Now, with artificial intelligence that has been developing very rapidly over the past decade, the technology is powerful enough to help recognize not only koalas in general, but also which individuals are using the passages thanks to the cameras trained by our AI.“, he added in a statement.
By making tracking easier and more automatic, the researchers hope to analyze how koalas use the passages and determine how many frequent them.
About twenty cameras soon to be installed
According to the details unveiled, around 20 cameras will be installed by the end of July at strategic locations in the Redland City region, near Brisbane. At the same time, the team will work with specialized NGOs to train the devices to recognize koalas by their appearance and their movements.
“The movement of the animal will trigger the capture of images and the images will be transferred to a server at Griffith University. Computer vision and machine learning systems will then be used to analyze the images.“and recognize each individual,” said Jun Zhou.
The project has already received support from the Queensland government, which is funding it to the tune of 90,000 Australian dollars (approximately 57,000 euros). Ultimately, the results could help optimize the construction and location of passages, and set up other conservation efforts for the species.
Concerns about the situation of koalas have grown in recent years in Australia, especially since the fires that ravaged the country in 2019 and 2020. According to a report from WWF-Australia, nearly 143 million animal mammals have been found on the path of these fires including more than 61,000 koalas.
Faced with the cry of alarm emitted by NGOs, the Australian government announced at the end of November the implementation of a vast plan to help protect the iconic species. A project which should start with the launch of a national audit intended to identify the global population.
Read also :
⋙ Why do koalas hug trees?
⋙ This koala, born without a foot, can now run and climb with the help of a dentist
⋙ Facial recognition, a new weapon against chimpanzee trafficking?