Reported wildlife trafficking and seizures of animal parts have increased dramatically over the past few years. The illicit wildlife and plant trade is estimated to be worth $70-213 billion a year and infringes on the natural resources of countries and wealth of businesses around the world.

Stopping it can be incredibly difficult given the huge terrain that needs to be monitored.  Researchers at the USC Center for Artificial Intelligence in Society believe artificial intelligence can help matters, and have developed a system that they believe enables poaches to be spotted in real-time.

The work, which was documented in a recently published paper, uses a combination of AI and game theory to better anticipate the spots used by poachers.

Nocturnal activity

Poachers typically operate at night, and whilst infrared cameras can be used, monitoring the footage is incredibly time consuming.  The researchers tagged some 180,000 images of humans and animals that had been captured by infrared cameras.  These images were then used to train a deep learning algorithm to automatically distinguish humans from animals in live footage.

The algorithm was deployed at various base stations out in the wild, with footage streamed from drones that patrol national parks in both Zimbabwe and Malawi.  The initial version of the tool performed well, but took 10 seconds to process each image, which is too long to be practically useful.

The team then moved things to the cloud so that the processing work was largely conducted virtually.  This enabled them to utilize much faster processors, which got identification down to just 3/10s of a second.  They hope to deploy the algorithm, known as “SPOT” or Systematic POacher deTector, across Botswana to begin with, before then hopefully spreading out even further.

“SPOT will ease the burden on those using drones for anti-poaching by automatically detecting people and animals in infrared imagery, and by providing detections in near real time,” they say.