Abstract: Analysis and comprehensive identification of radar characteristics of low-altitude targets， such as unmanned aerial vehicles （UAVs）， play a vital role in various fields， including low-altitude economy applications， air traffic management， urban counter-drone operations， and airport safety surveillance. Technologies such as deep learning and large-scale models have been widely applied in radar target recognition research. However， existing datasets are primarily generated in laboratory simulation environments. The few available real-world measurement datasets provide limited radar frequency coverage and environmental diversity， thereby failing to adequately meet practical research needs. To address this limitation， this study releases a publicly available dataset of radar target characteristics in the S-， X-， and Ku-bands， focusing on small and lightweight UAVs， birds， and meteorological clutter. This multi-frequency， multi-scenario， and multi-type dataset includes wideband and narrowband radar data for six categories of small and lightweight UAV targets， two types of bird targets， and four typical meteorological conditions. It serves as a benchmark for tasks such as radar echo-level genuine/fake classification， fine-grained classification， multimodal fusion recognition of wideband and narrowband signals， algorithm design and validation， intelligent model transferability assessment， and intrinsic feature extraction and comparison of low-altitude targets. The dataset provides a standardized foundation for radar-based low-altitude target recognition. Comparative analyses of typical scenarios， including “same target， different radar bands” and “different targets， same radar band” were conducted to investigate the wideband and narrowband echo characteristics of targets. Using the proposed dataset， a convolutional neural network-based genuine/fake classification algorithm for UAV targets was validated. Representative examples of classifying and identifying small and lightweight UAVs and other targets， including birds and meteorological clutter， are also presented. The dataset provides high-quality support for the development of intelligent radar recognition algorithms.