Scientists have actually revealed a brand-new tool for keeping track of threatened wildlife: an AI system that immediately counts elephants from space.
The tech integrates satellite electronic cameras with a convolutional neural network (CNN) to catch African elephants moving through meadows and forests.
In tests, the surveying method discovered elephants as accurately as human observers, while eliminating the threat of disturbing the species.
The research joins a growing variety of AI jobs that are seeking to safeguard endangered animals.
” Accurate tracking is vital if were to save the species,” said Dr Olga Isupova, a computer researcher at the University of Bath who developed the detection algorithm. “We need to know where the animals are and the number of there are.”
[Read: How Netflix shapes traditional culture, described by data] The system also provides a more effective alternative to by hand counting animals from low-flying aircrafts. Each satellite can gather more than 5,000 km ² of imagery every couple of minutes. And if clouds obscure the land, the process can be duplicated the following day, on the satellites next revolution of Earth.
In addition, the method can track animals as they wander throughout countries without having to fret about border controls or disputes.
Credit: 2020 Maxar TechnologiesThe green rectangles reveal elephants identified by the algorithm from area, while the red ones reveal elephants confirmed by humans.The scientists trained and tested their design on threatened African elephants in South Africa, utilizing information from the WorldView‐3 and WorldView‐4 satellites– the greatest resolution satellite imagery thats commercially offered.
They likewise used the design to lower resolution satellite images caught in Kenya– without any additional training information– to evaluate whether it might work outside their research study location.
” Our results show that the CNN performs with high precision, comparable to human detection abilities,” the team wrote in their research study paper.
They picked African elephants for the research study due to the fact that theyre the largest land animal worldwide– that makes them relatively simple to area. But the team believes their system could quickly detect far smaller types.
” Satellite imagery resolution increases every number of years, and with every increase, we will be able to see smaller things in higher detail,” said Dr Isupova.
You can check out the research study paper in the journal Remote Sensing in Ecology and Conservation.
Published January 20, 2021– 17:09 UTC