Keeping animals from tropical regions of the world in a cold climate is an expensive business, they need a warm environment in their pens and sleeping areas. Marwell Zoo was spending a small fortune keeping its herd of nyalas (an antelope, not as the title suggests a deer, native to Southern Africa) warm with electric heating, so they went looking for a technology that could reduce their costs by only heating while an animal was in its pen.
One might expect that a passive IR sensor would solve the problem, but a sleeping nyala too soon becomes part of the background heat for these devices, and as a result, the heaters would not operate for long enough to keep the animals warm. The solution came from an unlikely source, a coffee queue monitoring project at the IBM Watson headquarters in Munich, that used an array of infra-red sensors to monitor the changing heat patterns and thus gauge the likelihood of a lengthy wait for a beverage.
In the zoo application, an array of thermal sensors hooked up to ESP8266 boards talk back to a Raspberry Pi that aggregates the readings and sends them to the IBM Watson cloud where they are analyzed by a neural net. The decision is then made whether or not a nyala is in the field of view, and the animal is toasted accordingly.
This project has some similarities with a Hackaday Prize entry, automated wildlife recognition, in its use of Watson.
Nyala image: Charlesjsharp [CC BY-SA 4.0 ].