| In the dynamic realm of aerial surveillance, detecting drones, especially those equipped
with advanced camouflage techniques, is an escalating challenge. This study delves into the
detection of camouflaged drones, examining various concealment methods, such as drone-birds
and commercial and military drones, that are meticulously designed to merge with their
surrounding environment. Given the lack of existing datasets to address these intricate
scenarios, we have created a new dataset considering evasion strategies. This dataset covers a
wide range of variables, including different angles, technologies, and environmental conditions.
Through this approach, we aim to boost the efficiency of drone detection systems
against the sophisticated evasion tactics employed by various drone models. Our method
not only enhances detection rates under diverse conditions but also lays the groundwork
for future research in aerial surveillance technology. By merging these synthetic datasets
with current detection frameworks, we showed how the detection systems perform in order to
detect the evading drone.
Keywords: Drones, Drone detection, Camouflage, Evasion, Unmanned Aerial Vehicle,
Machine Learning, Artificial Intelligence, You Only Look Once (YOLO), Object Detection,
RetinaNet, ResNet60, Radio Frequency |