Drone detection and tracking on embedded device

Abstract

Thanks to rapid advancements in technology and manufacturing, drone usage has significantly expanded across both civilian and defense sectors. These Unmanned Aerial Vehicles (UAV) offer a relatively low- cost solution for a wide range of applications. In the civilian domain, they are increasingly employed in agriculture (for crop monitoring and pesticide spraying), logistics (transporting medical supplies, food and retail products), entertainment (aerial cinematography, live event coverage) and more. In the defense sector, their utility spans surveillance, reconnaissance, and combat operations (kamikaze drones, drones transporting grenades etc.). This thesis specifically focuses on the defense-related applications of drones. Recent conflicts in Ukraine and the Middle East have highlighted both the strategic value of drones and the urgent need for effective anti-drone systems. In this context, the objective of this work is to enhance an existing system known as the Victor, a firearm-mounted accessory designed to aid soldiers in neutralizing drone threats. The Victor is equipped with an optical sensor, an ARM-based CPU, an inertial measurement unit (IMU) and a battery. Its primary function is to emit either a sound or light signal when a particular condition (in our case, when a drone is detected) is met so as to indicate to the solider when to shoot. The goal of this thesis is to refine the Victor system to make it more effective and reliable in real-world defense scenarios. This work will therefore be split in two parts : Firstly, improving the detection accuracy of the model used to detect the drones and perform com- parison studies testing different configurations. Secondly, developing and testing a tracking pipeline in addition to the detection in order to improve the system’s performance.