Radiation Source Localization Using Radio Maps: A Computer Vision Approach

With the advent of 6G, there has been exponential growth in the trend of radio equipment, leading to challenges in electromagnetic spectrum management. Localization of radiation sources plays a pivotal role in effective spectrum management. This paper proposes a computer vision-based approach for addressing the problem of radiation source localization. The method leverages location and received signal strength information from the radio map, employing a YOLO network integrated with the AOT attention mechanism to accurately detect and locate radiation sources within the map. Furthermore, the localization results are depicted directly on the radio map. Additionally, we manually annotate a dataset specifically designed for detecting and localizing radiation sources in radio maps, followed by conducting simulation experiments using this dataset. Finally, simulation experiments are performed on different reconstructed algorithms as well as on a test set to validate their effectiveness while analyzing and illustrating their correlation with fine-grained radio maps.