Editions:  Austria · Denmark · France · Germany · Italy · Norway · Spain · Sweden · Switzerland
Advertisement

How Swiss software is helping drones survey wildlife in Namibia

Share this article

How Swiss software is helping drones survey wildlife in Namibia
Coloured frames help researchers decide if unclear objects in photos taken by drones are animals. Photo: SNSF
11:35 CEST+02:00
A new technique combining drones and automated image analysis is being used to help researchers count animals in Namibia's huge nature reserves.

The work being funded by the Swiss National Science Foundation (SNSF) offers a more accurate and cheaper way of counting gnu, oryx and other large mammals in areas that can be half the size of Switzerland. 

Counting animals in wildlife reserves with the help of drones. Photo: SNSF

Instead of carrying out dangerous manual counts, researchers can employ customized drones designed by Swiss firm SenseFly. These drones take more than 150 images per square kilometre which are then used as the basis for the counting of animals. 

But the process of actually identifying animals in the images is not straightforward because objects such as shrubs and rocks have to be discounted. That's where the artificial intelligence comes in. 

An algorithm developed by PhD candidate Benjamin Kellenberger means now researchers can quickly discount most objects that are not animals during a final manual analysis of drone images, according to a statement from the SNSF

However, getting the artificial intelligence system trained for that image analysis role was a complicated process.  

The training involved an international crowdsourcing campaign launched by Lausanne's EPFL technology institute, which is behind the SAVMAP project designed to promote sustainable development of semi-arid savannahs.

The crowdfunding campaign resulted in 200 volunteers tracking animals in thousands of aerial photographs of the savannah taken by researchers at the Kuzikus wildlife reserve in Namibia. 

The software was then tested for accuracy when it came to identifying animals. 

“For the AI system to do this effectively, it can’t miss a single animal. So it has to have a fairly large tolerance, even if that means generating more false positives, such as bushes wrongly identified as animals, which then have to be manually eliminated,” said Devis Tuia, an SNSF Professor now at the University of Wageningen in the Netherlands. 

Now, however, a single researcher can analyse the images for a 100-square kilometre are in a week. 

This final manual sorting is made easier by coloured frames automatically placed around questionable features in the images. 

 

 

 

 

 

Get notified about breaking news on The Local

Share this article

Advertisement
Advertisement
Advertisement
Advertisement
Advertisement
Advertisement
Advertisement
Advertisement