The development of the digital economy has highlighted the important value of geospatial data across numerous domains, with data trading being a pivotal link in activating this value. The current user base engaging in data trading is diverse, while trading platforms encounter problems such as disorganized data management and oversimplified retrieval methods. These concerns lead to inefficient retrieval for users with minimal domain knowledge. To address these complexities, this study proposes an intelligent retrieval method for geospatial data oriented toward data trading. This method establishes a geospatial data knowledge graph based on a standardized ontology model and innovatively utilizes large language models to assess user requirements in data trading. It effectively addresses the problems of standardized management for multi-source heterogeneous geospatial data and the poor adaptability of traditional retrieval methods to the ambiguous requirements of users lacking professional domain knowledge. Thus, it improves the efficiency and universality of geospatial data trading while guaranteeing the semantic interpretability of retrieval results. Experimental results confirm that the proposed method considerably outperforms traditional keyword-based retrieval methods. It exhibits particularly notable performance enhancements in scenarios with ambiguous requirements. This research not only effectively extends management approaches for geospatial data but also strengthens the inclusivity of data trading. Thus, it provides technical support for maximizing the value of geospatial data.

Terraced fields are difficult to delineate from high-resolution imagery because narrow and irregular boundaries, fragmented patches, and visual similarity to adjacent cropland can lead to omission errors and boundary confusion. This study adapts DeepLabV3+ for terraced field extraction from 2 m GF-1 RGB imagery through a task-oriented integration of spatial-detail preservation, contextual modeling, and boundary refinement. The resulting framework incorporates a dilated residual encoder, a gated global–local context module, and a boundary-supervised PointRend refinement head to address spatial-detail loss, background confusion, and boundary uncertainty in terrace mapping. On the re-evaluated test set, the framework achieved 86.68% precision, 87.77% recall, 87.22% F1 score, and 77.34% target-class IoU. Relative to the unmodified DeepLabV3+ baseline, the F1 score and IoU increased by 10.01 and 14.46 percentage points, respectively. These results indicate that task-oriented integration of spatial-detail preservation, contextual modeling, and boundary refinement can improve fine-scale terraced-field delineation from high-resolution optical imagery.