Despite constant innovations, technology never ceases to amaze us, as well as its countless uses to help the most vulnerable. On this occasion, the multidisciplinary research team of the foundation for the conservation of the species ‘WildCRU’, belonging to the Oxford University, with the collaboration of the POT, they managed to develop a technique for surveying elephants from space.
A feat comparable to finding a needle in a haystack, but the dramatic dpopulation census of African elephants due to poaching, killing in retaliation for looting crops and even the decline of their habitat due to development, forced them to find an alternative way to conserve and safeguard the species, through a accurate tracking, without having to intrude on their territory and migration habits, accustoming them to human presence.
Although the vast majority of observation methods use aerial surveys, environmentalists and scientists require a more precise procedure, which allows them to know how many copies are there and where are they, accurately without having to invade their habitat, eliminate the margin of error due to poor visibility and succumb to bias.
So the department of Zoology, the Machine Learning Research Group in collaboration with Dr. Olga Isupova, from the University of Bath and Dr. Tiejun Wang, professor at the University of Twente, in the Netherlands, gave themselves to the task of developing a software of iartificial intelligence, which is capable of detect and monitor the herds of African elephants, through satellite images with automated learning.
Thanks to an approach with NASA, the researchers have managed to detect elephants from space with an accuracy comparable to that of the human eye, using satellite images ‘Worldview 3’, owned by ‘Maxar Technologies’, which can ‘comb’ and collect more than 5,000 km. in a single image, in a matter of minutes.
According to researchers at the University of Oxford, one of the challenges of using satellite monitoring is processing the huge amount of images generated. However, automate detection it means that a process that would formally have taken months can be completed in a matter of hours. Additionally, machines are less error prone, false negatives and false positives in deep learning algorithms are consistent, and can be rectified by systematically improving models; The same cannot be said for humans.
There is hope
To develop this new method, the team created a custom training dataset of more than 1000 elephants in South Africa, which was entered into aneural network and the results were compared with the rhuman performance. It turns out that elephants can be detected in satellite images with as high a precision as human detection ability.
In such a way that, according to the study published in the scientific journal “Remote Sensing in Ecology and Conservation”, the model could even detect elephants from space, even when the animals are in place.s away from the training data site showing the generality of the model. And having trained the machine only on adult elephants, the software was even able to identify the smallest calves.
Without a doubt, this innovation will not only allow elephants to be surveyed from space very soon, but it will also do so with other endangered species, like rhinos; polar bears; mountain gorillas; and even whales and other cetaceans. The important thing now will be to be able to save them from the clutches of their greatest predator: man. Although there is still hope, like the characters who saved an elephant that fell into a well.