Automated generalization of road network data is of great concern to the map generalization community because of the importance of road data and the difficulty involved. Complex junctions are where roads meet and join in a complicated way and identifying them is a key issue in road network generalization. In addition to their structural complexity, complex junctions don’t have regular geometric boundary and their representation in spatial data is scale-dependent. All these together make them hard to identify. Existing methods use geometric and topological statistics to characterize and identify them, and are thus error-prone, scale-dependent and lack generality. More significantly, they cannot ensure the integrity of complex junctions. This study overcomes the obstacles by clarifying the topological boundary of a complex junction, which provides the basis for straightforward identification of them. Test results show the proposed method can find and isolate complex junctions in a road network with their integrity and is able to handle different road representations. The integral identification achieved can help to guarantee connectivity among roads when simplifying complex junctions, and greatly facilitate the geometric and semantic simplification of them.

To analyze the topological properties of hazardous materials road transportation network (HMRTN), this paper proposed two different ways to construct the cyberspace of HMRTN and constructed their complex network models, respectively. One was the physical network model of HMRTN based on the primal approach and the other was the service network model of HMRTN based on neighboring nodes. The two complex network models were built by using the case of Dalian HMRTN. The physical network model contained 154 nodes and 238 edges, and the statistical analysis results showed that (1) the cumulative node degree of physical network was subjected to exponential distribution, showing the network properties of random network and that (2) the HMRTN had small characteristic path length and large network clustering coefficient, which was a typical small-world network. The service network model contained 569 nodes and 1318 edges, and the statistical analysis results showed that (1) the cumulative node degree of service network was subjected to power-law distribution, showing the network properties of scale-free network and that (2) the relationship between nodes strength and their descending order ordinal and the relationship between nodes strength and cumulative nodes strength were both subjected to power-law distribution, also showing the network properties of scale-free network.