A Method for Intelligent Road Network Selection Based on Graph Neural Network

As an essential role in cartographic generalization, road network selection produces basic geographic information across map scales. However, the previous methods could not simultaneously consider both attribute characteristics and spatial structure. In light of this, intelligent method based on a graph neural (GNN) is proposed this paper. Firstly, case designed to construct sample library. Secondly, some neighbor sampling aggregation rules are developed update features. Then, GNN-based model calculate classification labels, thus completing selection. Finally, few comparative analyses with different conducted, verifying that most accuracy values GNN stable over 90%. The experiments indicate aggregate stroke nodes their neighbors together synchronously preserve semantic, geometric, topological features strokes, result closer reference map. Therefore, paper bridge distance between deep learning facilitating more method.