Issue link: https://iconnect007.uberflip.com/i/1172746
34 SMT007 MAGAZINE I OCTOBER 2019 side—like my colleague does who studies snow leopards—to check camera traps and pull out the SD cards, people could sit in a hut on their laptop to view images of wildlife passing by the cameras. In some places, sav- ing wildlife is the critical objective, while the conservation objective is preventing "human- wildlife conflict." There are about 16 species that this heavily applies to from polar bears entering villages in the Arctic to elephants entering croplands and lions raiding night cor- rals where livestock are kept. There are only 20,000 lions left in the wild, which is only six big populations. Everybody watches "The Lion King," but most don't realize how endangered lions are becoming. The big threat is retalia- tion killing, which is the same for snow leop- ards; that's how most die. When these large cats kill livestock, the herders go after them. If you can prevent that, that's vital. Johnson: How do you accomplish that? Dinerstein: What we're doing for lions right now is testing LionShield, a new system cre- ated by another colleague of ours and inven- tor, Henrik Rasmussen of Savannah Tracking. LionShield works by first identifying "problem lions"; then, you put a radio collar on them with an RFID chip—although going forward, it will be a LoRa chip—which communicates with the base station placed in front of a night corral. When the lion is within about one kilometer of the base station, it starts picking up its sig - nal, which pings out frequently. When the lion is within 200 meters, it triggers flashing lights and an ear-piercing siren that scares the lion off; it also alerts the community to be vigilant because it's very loud. The main problem is you have to capture the lion, tiger, leopard, or chee - tah first. However, this can be extremely diffi- cult in terms of the cost, permissions, logistics, and time and effort required. What if you could replace having to capture the animals with computer vision? You could have a variation of TrailGuard AI called Vil- lageGuard AI in front of livestock corrals. If you have enough of these out there, when it detects a predator, it could trigger the base sta- tion. That's another way that we can use it. We're also developing one called RiverGuard AI for the Amazon to protect indigenous com- munities; most of the traffic—and intruders— comes by boat because there are few roads. If you could monitor boat traffic, you could see who's going in and out of these areas. Indigenous reserves are also being invaded by poachers, gold miners, loggers, and oil and gas extraction explorers. Someone once said, "If you could only put a guard post at the con- Standard camera trap (L) vs. the TrailGuard AI camera (R).