[ad_1]
A new technique developed by Swiss researchers allows a rapid and accurate count of wildebeest, oryx and other large mammals living in wildlife sanctuaries. The drones are used to remotely photograph wild areas, and the images are then analyzed with the help of an object recognition software and verified by humans. The work is reported in an article published in the journal Remote Sensing of Environment .
The challenge is disheartening: some African national parks are expanding to areas that are half the size of Switzerland, says Devis Tuia, SNSF Professor now at the University of Wageningen (The Netherlands) and member of the Savmap project team, launched in 2014 at EPFL. "The automation of part of the animal count facilitates the collection of more accurate and up-to-date information."
No animals left without numbers
UAVs can economically cover large areas. But it's more complicated than simply sending them to photograph the terrain: more than 150 images are taken per square kilometer. At first glance, it is difficult to distinguish animals from other landscape features such as shrubs and rocks.
To decipher this mass of raw visual data, the researchers used a kind of artificial intelligence called "deep learning". "Designed by doctoral candidate Benjamin Kellenberger, the algorithm allows researchers to immediately eliminate most images that contain no fauna." In the other images, the algorithm highlights the models the more likely to be animals.
This first phase of elimination and sorting is the longest and most laborious, says Tuia. "So that the AI system can do it effectively it can not miss a single animal, so it must have a fairly high tolerance, even if it means generating more false positives, such as bushes misidentified as animals, to be eliminated manually. "
The team began by preparing the data needed to train the AI system to recognize the interesting features .In the context of an international crowdsourcing campaign launched by EPFL Some 200 volunteers followed animals in thousands of aerial photographs of the savanna taken by researchers at the Kuzikus Nature Reserve in Namibia.
These images were then analyzed by noting according to different types of animals. errors: the AI system receives a penalty point for confusing a bush with an animal, while the penalty for missing an animal is completely 80. As a result, the software learns to distinguish wildlife from features inanimate, but especially not missing any animal.Once the collection of images has been reduced by the AI system, a human resumes the final sort, made easier by colorful dots automatically placed around dubious features.
100 square kilometers per week
This semi-automated technique was developed in collaboration with biologists from the Kuzikus Wildlife Reserve in Namibia. Since 2014, researchers have been flying over the reserve periodically with the help of drones designed and optimized by SenseFly, a Swiss company, and equipped with standard compact cameras. "At first we were rather skeptical," says Friedrich Reinhard, director of the reserve. "The drones produce so many images that I thought it would be difficult to use."
But, thanks to the sorting done by the AI system, only one person can perform a full count of the Namibian reserve – an area of about 100 square kilometers – in about a week. In contrast, conventional methods involve entire teams taken aboard a helicopter. These methods are both less precise and costly than they are rarely used – once a year at most in Kuzikus.
Swiss researchers continue their work with the Namibian reserve. EPFL students go there regularly. Kenyan authorities have also expressed interest, as has the Veluwe National Park in the Netherlands. Devis Tuia, recently appointed professor at the University of Wageningen (The Netherlands), is still working closely with the University of Zurich (where he was professor FNS) and EPFL, which coordinates the Savmap project.
Source link
