Публікація: Analysis of Object Identification Methods for FPV Drones
Завантаження...
Дата
2025
Автори
Назва журналу
ISSN журналу
Назва тома
Видавництво
Анотація
The abstracts of the report consider modern methods of object identification for FPV drones with an emphasis on their application in real time and under conditions of limited computing resources. Classical approaches based on keypoint extraction, deep convolutional neural networks, semantic and instance-segmentation methods, as well as state filters and lightweight optimized models are analyzed. The study shows that each of the methods has its advantages and limitations depending on the accuracy, processing speed and complexity of the
environment. Special attention is paid to hybrid approaches that combine the advantages of several methods to ensure stable and effective object identification on board FPV drones. The results obtained emphasize the need to optimize algorithms and adapt models to the resource constraints of drones to ensure reliability and accuracy of operation in dynamic conditions.
Опис
Ключові слова
FPV drones, object identification, deep neural networks, segmentation, Kalman filter, lightweight models
Бібліографічний опис
Chebanchyk D. Analysis of Object Identification Methods for FPV Drones / D. Chebanchyk, V. Yevsieiv // Manufacturing & Mechatronic Systems 2025 : Thesises of Reports of IX-st International Conference, October 25-26, 2025. - Kharkiv, 2025. - P. 30-33.