Rhinoceros horn is more expensive than gold on the black market, and demand is huge. It is used in traditional oriental medicine to “cure” everything from the common cold and an upset tummy to cancer – completely without effect, since the main component in the horn is keratin, the material of fingernails. Poaching of rhinoceros is the basis of a multibillion dollar industry. During the past century, the number of rhinos in Africa has fallen dramatically, and both species found on the continent – the black rhino and the white rhino – are under serious threat of extinction. And many other African animals are also experiencing the same sorry fate, as a result of poaching.
Fredrik Gustafsson is professor of sensor informatics at Linköping University and is leading a project with the goal of being able to monitor animals in African nature reserves with the aid of sensors, cameras and machine learning.
“The park rangers must be able to monitor the animals’ well-being on the savannah. At the moment, a great deal of manual work is needed from far too few rangers to watch over the animals. And the job is full of risk. Fredrik Gustafsson, professor. Photo credit Fredrik GustafssonIf we can automate parts of the surveillance, the rangers can work more effectively and more safely”, says Fredrik Gustafsson.
Machine learning
Game cameras will be deployed across the savannah to monitor the animals in places where they are often found, such as at waterholes. These cameras are activated by motion and connected wirelessly to a server in which software identifies the animals. Sara Olsson and Amanda Tydén are authors of the software, having developed a recognition algorithm for their degree project for the Media Technology and Engineering master’s programme at Linköping University.
“We used machine learning to enable the algorithm to identify the different animals and distinguish them from humans. We collected many images of animals such as rhinos, elephants and large cats to use as training material. The current version of our model can distinguish between seven animal species and humans, and can follow the motion of an individual in a sequence of images to determine observations of a unique animal”, says Sara Olsson.
The software has also been tested on rhinos in the Kolmården Wildlife Park outside of Norrköping, Sweden, which gave many valuable insights. Amanda Tydén has found it rewarding to work with a problem from concept to completion.
“We have tested several variants of the system using, for example, different hardware and different cameras. We found out very quickly that something that works well in the lab may run into problems in a real-life situation. For example – how can we deal with loss of network connectivity, and how can we keep the software as efficient as possible so that the system can work for long periods on power from solar cells”, she says.
Reduced poaching
The next step of the project is to install the technology at more parks in Kenya to prevent the extinction of many large animals. Further, a new student is working on a degree project to extend the algorithm to identify poachers on the savannah. It is hoped that the game cameras will become a natural part of the park ranger’s work, and poaching can be reduced.
“This research contributes directly to Number 15 of the UN’s 17 Sustainable Development Goals: Life on Land. It will be key in preserving biological diversity. If we are successful, we will contribute to the conservation of animal life and the natural world for coming generations”, says Fredrik Gustafsson.
Principal partners in the project have been the Kolmården Wildlife Park and the HiQ consultancy firm.
Read more about the project at .
Translated by George Farrants
