Study On Urban Environment Procedural Modeling Techniques

Large-scale urban environment has become significant basis for various applications, such as urban planning, driving simulation, virtual reality and so on. Highly efficient modeling of the urban environment is the essential technique to digitalize city and improve urban management level. In this thesis, it is unveiled that modeling technique for urban environments is to find out such algorithms that emphasizes the static scenes of high degree of realistics, as well enhances the dynamic behaviours to the utmost with the characteristics of life. Meanwhile, it is pointed out that there are three proved solutions to accelerate the city modeling procedure as procedural modeling based on grammar, mathematics and artificial intelligence, model reconstruction based on image and video, traffic flow modeling based on multi-agent.The main work stated in this thesis focuses in the following aspects:Firstly, rule-driven modeling for urban roads is proposed. The modeling procedure is expressed as the intuitive operation rules to automatically generate large scale road network. The modeling results are controllable under the user's pre-processing design and post-processing interactive modification. Furthermore, road structure is the key to the spatial configuration of the road network. The typical road structures are well studied and formulated. In order to extend the typical ones' limitation, a statistical learning model is adopted to address that the road structures are extracted from samples and reused to generate the new roads with richer spatial configuration.Secondly, the constrained optimization approach is applied to model the urban zones. It is very efficient to construct and distribute the urban elements on the land. The elements layout is defined as a cost function subjected to the layout templates and planning constraints. The optimization algorithm includes particle swarm optimization and path planning. It searches and interprets the optimal solution through minimizing the cost function to obtain the optimal layout. The optimization efficiency is improved greatly by means of the pattern tree of the spatial configurations, because the pattern tree builds up the tight correlations among the optimized parameters, so decreases the searching dimensional complexity, as well enhances the layout fidelity. Finally, the large urban scene is achieved in the process of 3D assembling and synthesis according to the layouts in the 2D blueprints. A model library is created beforehand, in which all elements models are stored, from which those required are chosen and instantiated. The approach shows three advantages: it enables the whole layouts to be controllable for introducing the urban planning constraints and layout templates; It improves the realistic by adopting the virtual and real mixed methods to construct the urban elements models efficiently; It tradeoffs the scene complexity and the memory requirement for rendering by simplifying both geometries and textures of the models.Thirdly, traffic flow modeling based on driving behaviour simulation is put forward. It reproduces the temporal and spatial viariation of traffic flow, vitalizes the urban environment adding the vehicles' motion and the drivers' behaviours into the urban scene. The driver's perception and decision model based on multi-agent is used to simulate various driving behaviours in the traffic flow. Compared with the previous works on traffic flow modeling, this model integrates the vehicle with its driver and controls the driver's observation, analysis and decision process with quantification and individuation so as to enhance the capability of coordinating the complicated driving behaviours. Generated with the believable driving behaviour simulation, the effective simulation data about the traffic flow exert great potential in practice to evaluate the traffic flow and service level at the macro-level, as well to compare different traffic planning schemes.In brief, the approaches discussed in this thesis are efficient enough to procedurally model large-scale realistic urban environment and dynamic traffic flow. The work is a helpful exploration for solving many problems based on urban environment, exhibits powerful practical significance and application worth.