Surveillance of African elephants under threat using AI

According to the article published in Wild Life Conservation Research Unit Oxford University, high-resolution satellite images of elephants in the forests are taken from space to count and conserve them. The photos are sent from a satellite orbiting 600km above the surface of Earth.

The researchers have chosen the Addo Elephant National Park, South Africa to test this new method involving satellite imagery.

The population of elephants in Africa had deteriorated over the past century due to poaching, retaliatory killing from crop-raiding and habitat fragmentation.

Accurate monitoring of these mammals, including the details of how many are there and where they are is an essential information for the envisioned conservation plans.

The new technique is trained to use a computer algorithm to recognize the adult elephants in a variety of backdrops. The images from the satellite could survey up to 5000 square kilometers of area in one single pass in a matter of minutes. It helps avoid double-counting, makes the previously inaccessible areas and cross border areas which are important in conservation planning easily accessible.

Getting an accurate number of elephants is crucial to track and conserve them. This unobtrusive method is comparatively safe as it does not disturb the species or risk human lives in the process of manual counting. It is also an efficient method to count the elephants in times of pandemics as it does not involve people on grounds.

Usually, the most common method used to survey elephant population is aerial counts from aircraft but this is an expensive and exhausting technique which at times is hindered by poor visibility resulting in a counting error. Inaccurate counting can result in misallocation of conservation resources and misunderstanding their population trends.

The research highlights the importance of technology in conservation processes and how they open up new possibilities to prevent the mass extinction of wildlife on the planet.