Automatic Detection of Poachers and Wildlife with UAVs

Poaching has been on the rise recently, especially poaching of elephants and rhinoceroses in Africa. The recent Great Elephant Census (2016) shows the effect on elephants in particular. With elephant and rhinoceros numbers dropping rapidly, it is imperative that we act swiftly before they are hunted to extinction. Multiple strategies exist to combat poaching, including park ranger patrols, and more recently, the use of unmanned aerial vehicles (UAVs or drones), such as in the work of Ivoševic et al.(2015). In particular, UAVs equipped with long-wave thermal infrared (hereafter referred to as thermal infrared) cameras can be used for nighttime surveillance to notify park rangers of poaching activity, because there is increased poaching activity at night, and because animals and humans are warm and emit thermal infrared light. However, the video stream from these UAVs must be monitored at all times in order to notify park rangers of poachers. Monitoring of streaming footage is an arduous task requiring human supervision throughout the night, and is also prone to systematic lapses in quality as human detection often degrades with fatigue (Porikli et al. 2013). Furthermore, as more drones are added to the system, more resources are required to monitor the additional videos.Whereas previous work in AI has focused on game theory for patrol planning, such as Xu et al.(2017) and Wang et al.(2017), and machine learningbased poaching prediction, such as Gholami et al.(2017) and Critchlow et al.(2015), to assist human patrollers in combating poaching, little effort has been focused on decision aids to assist the UAV crew in detecting poachers and …