Parcel-level vector data for scaled land utilization analysis in Xinjiang based on remote sensing image

The scaled utilization of cultivated land has enhanced agricultural development and productivity. Quantifying its spatial distribution is essential for optimizing agricultural decision-making. Xinjiang, a vital grain production region in China, holds immense study significance due to its distinct geographical location and fragile natural environment. However, most studies on cultivated land fragmentation rely on outdated raster datasets. In this study, we introduce a cultivated land dataset of Xinjiang in a vector form with higher boundary accuracy, and more suitable for cultivated land statistics compared to raster datasets. A novel parcel extraction method integrates the Swin Transformer for multi-scale semantic information and DiffusionEdge for capturing fine boundary details is proposed, which improves the precision of land parcel extraction from high-resolution remote sensing images. Finally, we present a practical and up-to-date vector dataset of cultivated land. The scaled utilization analysis demonstrate the effectiveness of the dataset. Through this study, we contribute to developing a replicable methodologies for robust cultivated land extraction and a parcel-wise cultivated land in Xinjiang.