DCP-MTL: Vectorization of Agricultural Cultivation Field Parcels via Boundary-Parcel Multi-Task Learning Network in Ultra-High-Resolution Remote Sensing Images

This paper introduces the first UHR UAV dataset specifically for CFP, designed to evaluate the performance of the proposed model in identifying these parcels. The dataset offers ultra-high spatial resolution, various field parcel types, and broad geographic coverage. Figure 8 shows the spatial distribution of the study data. Jilin Province is the primary region for training and evaluating the model, while Hebei, Henan, Anhui, Zhejiang, and Hainan are auxiliary regions for testing the model's transferability.