Multiple drone tracking and pursuit
In recent years, drones have gained increasing importance in various fields such as agriculture, surveillance, and the military, revolutionizing how these sectors operate. In agriculture, drones are used for crop monitoring, spraying pesticides, and assessing field conditions. In surveillance, they...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2024 |
| País: | España |
| Institución: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglés |
| OAI Identifier: | oai:upcommons.upc.edu:2117/421825 |
| Acceso en línea: | https://hdl.handle.net/2117/421825 |
| Access Level: | acceso abierto |
| Palabra clave: | Drone aircraft Avions no tripulats Àrees temàtiques de la UPC::Aeronàutica i espai::Aeronaus |
| Sumario: | In recent years, drones have gained increasing importance in various fields such as agriculture, surveillance, and the military, revolutionizing how these sectors operate. In agriculture, drones are used for crop monitoring, spraying pesticides, and assessing field conditions. In surveillance, they provide real-time monitoring and data collection for security purposes. In the military, drones play a crucial role in reconnaissance, target tracking, and tactical operations. This is evident in the ongoing Ukraine conflict, where anti-drone technologies are becoming a critical research area. The versatility and efficiency of drones in these diverse applications have spurred significant advancements in drone technology and algorithms. In this context, this thesis proposes a methodology to detect and track multiple drones and then pursue one of them. To achieve this, a combination of hardware and software components is used in a controlled, closed environment. For the detection and tracking of the drones, AI methods are employed, including a Mask R-CNN for object segmentation and a Keypoint R-CNN to extract specific keypoints of the drones captured in the frames, enabling the drawing of a 3D bounding box around the detected drones. These keypoints are also used to determine the pose of the desired drone using the Perspective-n-Point (PnP) method. The flight trajectory control system for the pursuing drone is designed using a PID strategy due to its tuning flexibility. All these tasks, implemented as ROS 2 nodes communicating with each other, enable the successful hardware demonstration of an autonomous drone pursuit system. |
|---|