This dataset was derived from field experiments conducted in 2024 within the Ningxia Yellow River Irrigation District, establishing the first UAV-based RGB image collection of alfalfa inflorescences that encompasses multiple flight scenarios, harvests, cultivars, and temporal stages. Through rigorous sample selection, hybrid filtering, and image enhancement procedures, 5,000 high-quality samples were extracted from low-altitude aerial imagery, covering two flight altitudes (5 m and 7.5 m), three harvests, and four alfalfa cultivars (Junene 401, Zhongmu No. 3, Gannong No. 4, and Algonquin). The dataset contains imagery capturing various developmental stages of alfalfa flowering, with a total size of 2.72 GB, providing a systematic characterization of the dynamic processes of inflorescence development. This resource offers a reliable data foundation for intelligent phenological monitoring of alfalfa and supports phenomics research as well as cultivation decision-making, holding substantial practical value for the precision management of large-scale grasslands in the agro-pastoral ecotone of Northwest China.