Applications of machine learning in the field of nature conservation

Machine learning has become one of the three leading technologies in the field of nature protection. The article provides an overview of the tasks solved with the help of machine learning – from assessing the population of chimpanzees to determining the location of whales.

Artificial intelligence can learn to identify which photos contain rare species of animals or accurately identify an animal by its voice signals from hours of recordings. In addition to monitoring tasks, AI algorithms are used in the fight against poaching and the collection of scientific data.

Fighting poachers

Kafue National Park in Zambia has installed a 19 km long “fence” of infrared cameras that detects suspicious activity in real time. The algorithms are trained to ignore the passage of birds through the cameras and notify the guards when poachers pass through the fence.

Tracking water losses

The MapBiomas water project has published the results of the analysis of more than 150,000 satellite images and revealed the loss of more than 15% of surface water over 30 years. Without artificial intelligence, researchers would not be able to analyze changes in water resources across the country on the required scale and with the necessary degree of detail.

Whale search

Knowing where the whales are is the first step to taking measures such as creating protected areas to protect them. Visually locating whales in the oceans is difficult, but their characteristic signals can travel hundreds of miles underwater. The National Association for Oceanic and Atmospheric Research uses acoustic recorders to monitor marine mammal populations on remote and hard-to-reach islands. Google AI participates in the development of algorithms.

Koala protection

The population of Australian koalas is seriously declining due to habitat destruction, attacks by domestic dogs, road accidents and forest fires. Using unmanned aerial vehicles and infrared cameras, the algorithm analyzes the infrared footage and determines whether the heat signature is a koala or another animal. The system was used after wildfires in Australia in 2019 and 2020 to identify surviving koala populations.

Population assessment

The rescue of endangered species was implemented using an image classification algorithm for large-scale monitoring of biodiversity in Congo national parks. The algorithm classifies up to 3000 images per hour with an accuracy of 96%.

Notify of

Inline Feedbacks
View all comments