2026-04-17T17:44:58.797Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:10124 2024-12-03T20:10:47Z 3

Signals Intelligence Based Drone Detection Using YOLOv8 Models 2024 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/10124 https://www.atlantis-press.com/proceedings/iciitb-24/126002409 M. Protic L. Jovanovic M. Dobrojevic M. Cajic M. Zivkovic H. Shaker N. Bacanin The reduced costs associated with deploying and utilizing Unmanned Aerial Vehicles (UAVs) have spurred their widespread adoption across various industries, including aerial photography, information gathering, and search and rescue operations. However, this rapid uptake has also raised concerns regarding safety and privacy, particularly due to instances of misuse and potential hazards posed by convertible drone technology. Addressing these concerns, this study investigates the application of emerging Artificial Intelligence (AI) techniques in computer vision for the detection and classification of ISM band transmissions, distinguishing between conventional Bluetooth signals and those used for drone control. Several YOLOv8 architectures, optimized for lighter hardware, are evaluated using a publicly available ISM band visual dataset. Results demonstrate that even lighter models, such as nano and small architectures, can achieve significant precision rates, with the best-performing models reaching a peak precision of 90%. However, medium-sized architectures are recommended for optimal performance. bookPart Atlantis Press 74 86 10.2991/978-94-6463-482-2_6 Proceedings of the 2nd International Conference on Innovation in Information Technology and Business (ICIITB 2024), Chapter in Advances in Computer Science Research