Modern sensors have given conservationists a powerful new tool in the fight against poaching, with the ability to track the whereabouts of endangered animals as they wander through the wild. A new research project at the University of Twente could harness this technology in yet another useful way, by mixing motion sensors with machine learning to recognize when wildlife is responding to a nearby threat.
The work was carried out by University of Twente’s Jacob Kamminga who is a member of a research group developing small, autonomous sensors for a range of applications. Kamminga developed the wildlife motion sensor as part of his PhD, hoping to offer a way of detecting distinct movement patterns in response to the presence of humans.
The sensor is an inertial measurement unit that includes an accelerometer, gryoscope and magnetometer, which can be attached to the animal to gather motion data as it carries out activities. This is analyzed by onboard artificial intelligence, which then classifies the movements and relays them to a mobile network or satellite connection when it detects a sudden change.
Kamminga used the sensor to classify eleven different movement patterns of goats, sheep and horses using mostly what he calls “unlabeled data,” meaning that the sensor only had to be trained on a small set of animal movements. He also found that in most cases, the sensor worked just as well when reduced to a single accelerometer, which made it far more efficient.
“I added a gyroscope as well, that measures rotation,” Kamminga says. “This can make it some more accurate, but this comes with a cost. It consumes 100 times more energy than the accelerometer. In most cases, just the accelerometer is accurate enough.”
Using artificial intelligence to analyze animal motion in this way could lead to some interesting possibilities for anti-poaching groups, who could setup the system so they receive alerts when an endangered species is moving in response to a very specific threat. Beyond that, the system could combine with other wildlife tracking tools to assist in overall efforts to preserve biodiversity.
"Linking wild animal movement recorded using sensors with remotely sensed imagery and GIS (geographic information systems) models is promising technology to better understand the ecological requirements of species, as well as inform management and policy decisions with conservation outcomes and biodiversity,” says Professor Andrew Skidmore, who was involved in the work.
Kamminga’s PhD thesis is available online here.
Source: University of Twente