A dataset of alfalfa lodging images in the Yellow River irrigation area of Ningxia based on unmanned aerial vehicles

Rapid identification of alfalfa lodging and accurate acquisition of the lodging area are key steps for modern grassland animal husbandry to achieve high efficiency and intelligence. This dataset is derived from the alfalfa planting areas in the Yellow River irrigation zone of Ningxia from April to September 2024, covering 1,564 UAV images of alfalfa lodging in three harvests, two growth stages (branching stage and initial flowering stage), and three flight scenarios (5 m, 7.5 m, and 10 m). After preprocessing the images, three categories were set, and detailed pixel-level manual annotations were conducted using the LabelStudio annotation tool, with cross-checking to ensure the reliability, completeness, and consistency of the data. Four semantic segmentation models were selected for training and validation, and multiple evaluation metrics were used to assess the models. The results show that the precision rates of this dataset in the YOLOv5x, YOLOv8x, YOLOv10x, and YOLOv11x segmentation tasks are 76.0%, 79.9%, 74.5%, and 75.3%, respectively. This dataset can be used for rapid monitoring of alfalfa lodging in large-scale fields and for the application of deep learning methods in lodging area identification, and can further provide basic data support for the research and development of visual intelligent systems for alfalfa intelligent management and harvesting equipment.