Intelligent Transportation Systems (ITS) require precise and effective means of vehicle detection and tracking. Real-time traffic monitoring using unmanned aerial vehicles (UAVs) is certainly difficult under certain conditions, but provides valuable aerial perspective. This paper proposes a new approach that incorporates You Only Look Once version 8 Nano (YOLOv8n) with the Convolutional Block Attention Module (CBAM), an object detection model developed with dense prediction networks, and a modified DeepSORT tracker based on Vision Transformers for appearance features and adaptive Kalman filters for trajectory prediction. The framework resolves concerns such as occlusions, traffic density, and illumination. Experiments were conducted on the Aerial Car data set training and the Aerial University Air data set (AU-AIR data set), and the results reflected a marked increase in the level of detection precision, recall, and robustness of tracking. These results show that this framework can be used to transform UAV-aided ITS techniques to enhance accuracy and reliability, especially where environments are dynamic.
CBAM-YOLOv8n and ViT-DeepSORT for UAV Vehicle Tracking
Lect. Notes in Networks, Syst.
International Conference on Information Systems and Management Science ; 2025 ; Malta, Malta February 22, 2025 - February 23, 2025
AI Technologies for Information Systems and Management Science ; Chapter : 16 ; 167-176
2025-08-07
10 pages
Article/Chapter (Book)
Electronic Resource
English