Abstract: To provide computable models of large-scale cities for numerical simulation, it is necessary to have the ability of fast modeling the computable models for cities with large scale, massive data and frequent updating. However, due to the errors in the data collecting process and the essential simplification operations in model processing, in directly modeled digital city models there exists large amount of CAD model deficiencies such as non-closeness, joint, intersection and isolation, which makes the city models cannot be used in numerical simulation. The cost of repairing these deficiencies using commercial software is extremely high. To solve this problem, this paper proposes a fast modeling method of computable models for large-scale cities. By analyzing the CAD model deficiencies involved in LoD1 city models, we propose the computable conditions that must be satisfied and the corresponding measurements. Our method can directly build the computable city models without CAD model deficiency. To improve the efficiency of computable processing, we also present graphic retrieval acceleration algorithms for large-scale cities. The experiments show that it takes us only about 25 minutes to build a computable city model with 710 km2, 300 thousand buildings and 2.17 million facets, which satisfies the requirements of numerical simulation on large-scale city models.

Key words: computable models, CAD model deficiency, building, terrain