Mitigating human-wildlife conflict and monitoring endangered tigers using a real-time camera-based alert system

Abstract

The recovery of wild tigers in India and Nepal is a remarkable conservation achievement, but it sets the stage for increased human-wildlife conflict where parks are limited in size and where tigers reside outside reserves. We deployed an innovative technology, the TrailGuard AI camera-alert system, which runs on-the-edge artificial intelligence algorithms to detect tigers and poachers and transmit real-time images to designated authorities responsible for managing prominent tiger landscapes in India. We successfully captured and transmitted the first images of tigers using cameras with embedded AI and detected poachers. Notifications of tiger images were received in real time, approximately 30 seconds from camera trigger to appearing in a smart phone app. We review use cases of this AI-based real-time alert system for managers and local communities and suggest how the system could help monitor tigers and other endangered species, detect poaching, and provide early warnings for human-wildlife conflict.

Keywords: embedded-AI camera-alert systems; endangered species; human–wildlife conflict; poaching; tigers.

Figure 1.

(a) The human population surrounding the world's tiger conservation landscapes (TCLs) increased by over 19.5 million people between the years 2000 and 2020. These population increases are especially acute for TCLs in northern and central India, Cambodia, and Sumatra. (b) Our initial deployment of the TrailGuard AI camera-based alert system focused on central India in the partially protected forest corridor between Kanha Tiger Reserve in the northeast and Pench Tiger Reserve to the southwest. An additional deployment occurred in northern India within Dudhwa Tiger Reserve. Source: The population data were generated using global population estimates from the WorldPop Hub (WorldPop 2018). The data were clipped to include the area within a 5-kilometer buffer around all TCLs to capture population growth in close proximity to potential tiger habitat. The inset map of the Russian Far East TCL is not at the same scale as the overall map.

Figure 2.

The TrailGuard AI camera system (left, US quarter for scale) can be well camouflaged near a trail such as this one in Dudhwa Tiger Reserve, as seen from the trail, that sent multiple real-time notifications of poachers without being detected (right).

Figure 3.

Images captured by TrailGuard AI and then transmitted via the cellular network as (a) tiger detection notifications. The tiger edge detector was extremely accurate, with a median probability value of .9883. We also captured and transmitted detections of (b) other conflict-prone species, including leopard and sloth bear, by setting the detection probability for an animal species to be very low, turning, in this case, the detector targeting tigers into a megadetector (see the supplemental material). Images of humans were also successfully transmitted from a camera location where (c) cell connectivity was absent, instead relying on long range radio and a range-extending repeater unit to a gateway within range of a cell tower.

Figure 4.

Detections of poachers and a tiger in Dudhwa Tiger Reserve, India. A TrailGuard AI camera detected the same gang of poachers on two different nights, including an image of one poacher carrying a dead chital (the lowest image). Authorities identified the individuals from the real-time image notifications. Tiger detections were also transmitted from the same location between the poacher detections.

Figure 5.

Conceptual matrix of where real-time camera deployments directed toward abating human–tiger conflict or for stopping poaching may be most appropriate. Strategies for only preventing poaching would be appropriate with increasing tiger density but low human density (the lower quadrants), whereas with increasing human density (the upper-right quadrant), camera deployments may shift to preventing more human–tiger conflict and poaching. With most of the tiger populations of India applied to the matrix, it is notable that some of the largest and densest tiger populations are also in areas with relatively high human density, indicating camera deployment to prevent conflict and poaching would be most appropriate. The circle size is scaled by land area, and the Dudhwa–Rajaji–Corbett and Kaziranga–Orang populations are outliers that could extend further beyond the figure boundaries relative to the other tiger populations.