Multi-Perspective RGB and Infrared Dataset for UAV Detection

The increasing presence of Unmanned Aerial Vehicles (UAVs) in both civilian and military domains highlights the need for reliable detection systems for monitoring and surveillance. Although RGB cameras provide high-resolution imagery for UAV detection, their performance degrades significantly under low-light conditions. Conversely, thermal sensors are effective in detecting heat signatures, but lack the spatial resolution required for accurate localization. To overcome these limitations, this study explores a multimodal UAV detection approach based on the fusion of RGB and infrared (IR) data. We introduce a novel dataset comprising synchronized RGB and IR imagery of UAVs, captured from different viewpoints. The dataset includes recordings of several UAV models operating under various lighting and background conditions, ensuring a comprehensive benchmark for detection tasks. To present some metrics, we trained state-of-the-art deep learning-based object detection models on our data. The results demonstrate that by using a specific dataset, the detection accuracy can be enhanced, particularly in challenging conditions such as sunshine, sunset, and darkness, including different and variable backgrounds. This study highlights the potential of multimodal sensing for UAV detection and provides a foundation for future research on robust aerial surveillance systems.