| dc.description.abstract | The Local Hawk drone project, currently using Single Board Computers (SBCs), faces difficulties in simultaneous task execution such as motor control and object detection due to the limitations of the SBCs. The necessity to improve frame rates and efficiency led to an exploration of various hardware and software configurations. The primary aim of the research was to evaluate and compare four different configurations to optimize object detection capabilities in lightweight Unmanned Aerial Vehicles (UAVs).
The key areas of focus were processing power, accuracy, and energy efficiency. The study conducted an exhaustive benchmarking of four hardware and software configurations, evaluating them on various parameters such as detection accuracy (precision, recall, and F1-score), frame rate, power consumption, weight efficiency, and complexity of setup and operation. The analysis demonstrated the importance of carefully selecting hardware and software configurations to achieve optimal object detection performance within the constraints of lightweight UAVs. It was found that configurations varied greatly in their precision, recall, and F1-scores, with different trade-offs between frame rate and power consumption.
Additionally, weight efficiency and complexity of setup and operation played crucial roles in determining the overall suitability of each configuration. This research significantly contributes to the understanding of edge image processing for lightweight UAVs, serving as a foundation for future investigations in this area. The results hold practical relevance for the Local Hawk project and similar endeavors aiming to enhance the capabilities of lightweight UAVs in applications ranging from surveillance to search-and-rescue operations. | en_US |
