Detecting aircraft on the ground through real-time remote sensing using satellite images and deep learning techniques is crucial for both military and civil aviation. Effective detection enhances airport management, reducing traffic congestion on runways and in hangars. In military applications, accurate aircraft location determination significantly contributes to strategic monitoring and defense planning. This study advances existing methodologies by developing a more comprehensive dataset, incorporating new images and applying extensive data augmentation techniques to improve detection accuracy across diverse scenarios. By leveraging the YOLOv5 algorithm, this approach offers higher accuracy and faster detection compared to traditional methods, thereby enhancing real-time surveillance capabilities. The augmented dataset and optimized training process resulted in improved detection performance, particularly in challenging conditions, such as varying aircraft sizes and occlusions. Performance metrics, including precision, recall, average precision (mAP), and F1 score, yield values of 0.787, 0.751, 0.735, and 0.789, respectively, indicating a strong precision-recall balance and high reliability for the YOLOv5 algorithm. Future work will focus on further expanding the dataset, optimizing hyperparameters, and exploring different model architectures to continue enhancing detection performance. The findings underscore the novelty and robust capabilities of YOLOv5 in real-time aircraft detection, while identifying key areas for future improvement.
Revolutionizing Aircraft Detection with Deep Learning for Satellite Imagery in Aviation
Communic.Comp.Inf.Science
International Conference on Science, Engineering Management and Information Technology ; 2024 ; Ankara, Türkiye September 12, 2024 - September 13, 2024
12.08.2025
19 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch