Μελέτη και κατασκευή μη επανδρωμένου αεροσκάφους (Drone) με δυνατότητα υψηλής ακρίβειας προσγείωσης με χρήση Computer Vision

Abstract

The development of Unmanned Aerial Vehicle (UAV) technology has gained considerable momentum in recent years, with a particular focus on enhancing capabilities for precise and controlled landings. This master's thesis deals with the research and construction of an unmanned aircraft with a specialized emphasis on achieving high-precision landings using computer vision technologies. The study investigates OpenCV, ArUco markers and the Raspberry Pi 5 as a companion computer for the flight controller. The motivation behind this research stems from the growing demand for UAVs in various applications, such as domain surveillance, identification, and delivery services. Achieving high-precision landings is crucial to optimizing these applications, as it allows for safe and accurate landing in restricted or dynamic environments. The thesis aims to contribute to the field by designing and implementing a robust system that leverages computer vision techniques to enhance accurate landing of UAVs. The fundamental technology explored in this thesis is OpenCV, a widely used open source computer vision library. OpenCV provides a flexible set of algorithms for image processing, object detection and feature tracking – key elements for developing a high-precision landing system. The ArUco Marker system is utilized by computer vision and is a central element of the proposed methodology. The hardware architecture chosen for this research includes a Raspberry Pi 5, serving as the flight controller's companion computer. The Raspberry Pi 5 provides the computing power needed for real-time image processing and decision-making during the landing phase. The integration of an companion computer enhances the autonomous functions of the UAV, allowing it to run complex algorithms without overloading the flight controller. An important aspect of the thesis is the development of Python code that implements the precise landing of the drone on a basis consisting of ArUco markers. The code incorporates image processing algorithms from OpenCV to detect and analyze ArUco markers in real time. By utilizing ArUco indicators, the algorithm allows the drone to calculate its position and orientation relative to the indicators, facilitating a controlled descent and landing. The research methodology involves a combination of theoretical analysis, algorithm development and practical application. Theoretical aspects include a comprehensive review of existing literature on UAV landing systems and OpenCV operations. Algorithm development involves the design and coding of image processing and control algorithms adapted for high-precision landings and commands to manipulate the UAV. Practical implementation includes building the drone, developing the Python code, and extensive testing in real-world scenarios. In conclusion, this thesis explores the intersection of UAV technology, computer vision, and high-precision landings. By integrating OpenCV, ArUco indicators and Raspberry Pi 5, the research aims to develop a reliable and effective system to enhance drone landing capabilities. The results of this study have the potential to influence developments in UAV applications, particularly in areas that require precise and controlled landings.