ABSTRACT. As a key component of land use data, land type parcels are spatially characterized by seamless coverage, non-overlapping boundaries and full-area enumeration. Variations in operational methods and proficiency among different data producers often lead to data quality issues. In particular, inconsistent boundary problems such as gaps and overlaps tend to occur more frequently in areas with complex topological relationships. Therefore, inspecting and processing boundary consistency of land type parcel data constitutes an essential step in data processing, database loading and quality inspection..

It is difficult for traditional inconsistency detection and processing methods to take into account multiple constraints such as topological correctness, semantic consistency, and area balance, and they usually require manual intervention, which takes a long time. In view of this, an inconsistency detection and processing method for patch boundary based on topological decomposition model is proposed. First, the topological decomposition model was built, topological polygons were converted from patches, and the polygons were dimensionally reduced to agent points. Second, 4 boundary inconsistency patterns were defined, and they were recognized through the topological relationship between agent points and patches and the semantic information of the arcs which were composed of topological polygons. Finally, the merger method and the dissecting method were proposed to modify boundary inconsistency patterns, taking into account semantic similarity and space adjacent indicators.

Experimental verification is carried out on the land-use data in Guizhou Province, and the result shows that the algorithm for detecting and correcting inconsistencies in parcel boundaries proposed in this paper can automatically detect and handle inconsistencies in the boundaries of land-use parcels. While ensuring the accuracy and rationality of the results, it overcomes the shortcoming of mainstream software that requires extensive manual intervention, and improves the efficiency of consistency inspection and processing for shared parcel boundaries.

ABSTRACT. To address the issue of excessive matching granularity in geographic entity matching, which hinders the capture of subtle inter-entity differences, this paper proposes a generalized geographic entity matching method that accounts for both spatial and semantic features. The method introduces a novel matching primitive—the feature line segment—and establishes initial matches for linear and polygonal entities by combining these segments in various patterns and calculating spatial similarity based on projection distance and relative offset rate. Subsequently, semantic similarity is computed to detect and refine the matching results. Finally, experimental validation was conducted using two real-world datasets from a region in Southeastern China: one comprising polygonal residential areas at different scales and the other consisting of linear river data from different periods. The experimental results demonstrate that the proposed method is applicable to both polygonal and linear entities. Specifically, nine matching patterns were identified for polygonal entities and six for linear entities. Furthermore, both precision and recall for the matching of polygonal and linear entities exceeded 90%.

ABSTRACT. Aiming at the problems of insufficient effect in maintaining road network spatial structure and poor rationality of selection scheme recommendation in existing studies, this study proposes a reasoning and recognition method for typical road network spatial structures coupling semantic and spatial features, as well as an intelligent recommendation method for road network selection schemes based on meta-paths and user preference diffusion. Taking the spatiotemporal knowledge graph for road network selection as the core framework, this study, based on the high correlation between semantic and spatial information in the graph, clearly selects semantic features such as road hierarchy and level affiliation, as well as spatial features such as mesh density and node density as identification factors. It adopts a calculation method combining rule reasoning based on knowledge graph and feature weighted calculation, designs specific reasoning elements and rules, and realizes the knowledge reasoning and recognition of five typical road network spatial structures: dendritic, regular grid, cellular, annular, and trumpet-shaped. To ensure the effectiveness and generalization of the method, this study selects the spatiotemporal knowledge graph data of the road network in Beijing, China as the training set and test set for the development and verification of the method. At the same time, it constructs meta-paths for scheme recommendation, optimizes the recommendation algorithm by combining user preference data, and can automatically push and chain-organize selection methods, models and algorithms suitable for various road network spatial structures. The study finally achieved the following results: first, it realized the accurate and automatic reasoning and recognition of five typical road network spatial structures. Experimental results show that the recognition accuracy rates of dendritic, grid, cellular, annular and trumpet-shaped road networks reach 89.36%, 86.87%, 95.86%, 98.26% and 82.59% respectively. Its recognition accuracy, recall rate and F1-score are all significantly improved compared with existing methods, effectively solving the problems of low efficiency and strong subjectivity of traditional recognition methods; second, the optimized selection scheme recommendation algorithm has greatly improved the rationality and adaptability of recommendation. Verified by F1-score, the F1-score of scheme adaptability is increased to 0.88, which can accurately match the selection needs of different road network spatial structures; third, the proposed method can effectively maintain the consistency of the spatial structure of typical road networks before and after selection, significantly improve the defect of easy distortion of road network structure in existing studies, and clearly define the expected application scenarios of the method as urban road network planning, map making, intelligent traffic management and control, etc. It can provide more accurate technical support for road network expression in map making, improve the scientificity and practicality of map making, and ultimately provide scientific, efficient and accurate technical support and method reference for road network selection work.