To Catch a Poacher: How Intel Engineers Brought AI Tech to the Fight Against the Illegal Wildlife Trade

In the wildlife reserves of East Africa, elephants, rhinos, gorillas, and other large mammals are hunted by poachers. All that stands between these animals and harm’s way are small teams of park rangers and conservationists. The danger is very real for these species on the brink: A staggering 35,000 African elephants are killed each year, putting them just a decade away from extinction, according to the non-profit RESOLVE.

Technology is an increasingly critical tool for protecting elephants and other large animals, given their necessarily expansive habitats: A group of just 50 rangers in Kenya, for example, covers a reserve of 3,000 square miles. Park rangers and conservationists have used motion-activated camera traps to catch poachers in action, but the animals are tragically already lost by the time rangers can respond.

But what if you could stop a poacher before he kills?

This was the fundamental question explored by Intel engineers working in partnership with conservationist Steve Gulick and RESOLVE. Radha Mathachetty, an engineer with Intel, was brought on to the project to help the conservationists design a smarter tech solution. It was a project which took his career in a direction he hadn’t expected. “I am grateful for the chance to apply my engineering skills to design a product that will allow for the effective preservation and safety of endangered wildlife all over the world,” Mathachetty says. 

What Mathachetty, who handled hardware, and Intel’s Lucian Vancea, who developed the software, had to work on was this issue: The existing TrailGuard motion-activated cameras provided rangers with thousands of images, mainly of animals triggering the sensors, that must be sifted through after the fact, in the hopes of finding a few that capture a poacher’s likeness or activity. The rangers needed a real-time solution to catch poachers red-handed.

Working off Gulick’s idea, Mathatchetty and Vancea created a new anti-poaching device, TrailGuard AI. TrailGuard AI is a small AI-powered camera that can be hidden along paths used by animals. Once the motion-activated camera captures an image, the Intel Movidius Vision Processing Unit (VPU) runs deep neural network algorithms for object detection and image classification inside the camera. It sorts through the images to quickly determine if any humans or vehicles are present. Any images detecting poachers’ activity—and only those images—are sent to the park rangers’ base of operations by satellite or radio network. It all happens in under two minutes, giving park rangers a chance to stop the killing.

Mathachetty’s first challenge was to get TrailGuard AI in a compact but highly effective form. “In the prototypes [the conservationists] were using, they had all the different pieces of the camera separate,” Mathachetty says. “They came to us with the idea to put everything together— to have the motherboard, camera board, sensor modules—integrated all in one. Putting everything together creates a smaller, portable, easily installable, and power-saving device.” And instead of holding a separate computing unit that requires field maintenance, TrailGuard AI simply sends relevant images to the park rangers’ base by long-range radio, satellite network, or mobile networks. 

While building the custom-designed device, the team faced numerous constraints and requirements. In addition to the AI chip and external sensors, the camera had to be able to connect to Wi-Fi, satellite, and long-range radio to send the images. It needed to consume little power, to reduce the need for frequent battery changes that both take up rangers’ valuable time and also could alert poachers to TrailGuard AI’s presence in the reserve. (The device runs in standby mode until motion is detected.) 

The sensors have to function in all light conditions, and the VPU must recognize humans from every angle and in every light condition since poaching frequently happens at night. “The algorithm is fed thousands of images of humans and animals, and it is able to analyze body shape, facial geometry movement, and more until it can distinguish from any angle and in any light,” Mathachetty says. 

Critically, the “hidden ranger,” as the device is nicknamed, “had to be very compact and easily deployed,” Mathachetty adds. And the design had to be able to be scaled and affordably manufactured, in order to be deployed in wildlife reserves in desperate need of such technology.

The engineers continually adapted the camera’s design over several versions. Mathachetty and the team initially field-tested TrailGuard AI in Big Sur, California, with the forest as a stand-in for the conditions in East Africa. When they encountered a problem with the satellite modem, Mathachetty proposed the idea of using a Wi-Fi hotspot as a means of connecting TrailGuard AI and the host PC to pull off a successful test of the technology. After field testing at the Grumeti Reserve in Tanzania, Mathachetty further evolved TrailGuard AI’s design to be waterproof and moistureproof.

Conservationists are hoping to expand TrailGuard AI’s use to South America and Asia. Perhaps even more exciting, after a year of TrailGuard AI being used in the field, Mathachetty and his team are already improving the device, with the new version to include the next generation of a VPU that utilizes powerful video and audio analytics to better understand poacher activity. 

“In the next generation, we expect the device to run more neural networks to run simultaneous analytical processes,” says Mathachetty. “For example, currently the device is capable of video analytics for object detection. In the future, we hope to also incorporate audio analytics with video action detection as well.” 

The project profoundly affected Mathachetty. “I was able to meet passionate professionals who introduced me to new global issues; they inspired me to take my own steps to save the environment,” he says. “Professionally, I realized the potential and possible applications of Intel products in diverse fields.” Such applications, might, for example, include using tech to predict forest fires. “We can provide excellent solutions to complex problems,” Mathachetty says.