Structure Based Multi-resolution Representation Approach For 3D Building Models

Building model is one of the most important parts of urban fundamental geographic information, and its precision degree and complexity can directly influence the quality and efficiency of 3D city model application. With the concepts of digital city and smart city being put forward, the application requirement of 3D urban geographic information system has been growing with each passing day. However, the ever-expanding big data volume of 3D building models has become one of the main factors that greatly restrict the wide application of 3D urban geographic information system. In order to solve various problems in applications brought about by the big data volume, and improve the efficiency of model storage, transmission and visualization, it is very significant to make deep discussion and researches on multi-resolution representation approach of 3D building models.With the aim at the practical requirements of shape recognition consistency and construction recognition consistency in the multi-resolution representation process of 3D building model, this paper makes intensive research on the structures of 3D building models, divides the structures of 3D building models into primary structure and detail structure according to the recognition focuses in different scales, and proposes structure recognition algorithms according to the characteristics of the two kinds of structures. Based on the recognition results of primary structure and detail structure of 3D building model, we propose effective multi-resolution simplification algorithms from the two aspects of primary structure preservation and detail structure deletion. Dividing the common requirements of 3D building model visualization into shape recognition oriented and construction recognition oriented, we make a systematic research on the multi-resolution representation application of 3D building models with the help of the structure based multi-resolution simplification results. The concrete research content of this paper includes following several aspects:1) Research status summary of related areas. As to the structure recognition of 3D models, the current methods for primary structure recognition can be divided into model segmentation based approaches, shape analysis based approaches and structure analogy based approaches, and the existing detailed structure recognition methods can be divided into independent structure recognition approaches and related structure recognition approaches. The existing algorithms for multi-resolution simplification of 3D models can be divided into the simplification for natural object's model, the simplification for man-made object's model and the simplification for building model. And the existing researches for multi-resolution representation of 3D models include multi-resolution representation of complicated object and complicated scenes. Through analysis and comparison, we point out the limitations of the existing works in structure recognition, multi-resolution simplification and multi-resolution representation of 3D building models, which provide the foundation and theoretical references for this paper.2) Structure recognition of 3D building models. Concerning the primary structure recognition, we propose a structural segmentation method based on voxel space analysis, whose main procedures include 3D model voxelization, spatial distance field calculation, voxel clustering, structure extraction and surface segmentation. As to the detailed structure recognition, we propose a classification and extraction algorithm based on the graph search approach, whose main procedures include surface segmentation, adjacency graph construction, and detail structure search.3) Structure based multi-resolution simplification of 3D building models. For multi-resolution simplification of models with primary structure preservation, we adopt multi-resolution reconstruction algorithms based on volumetric technique and propose a multi-resolution simplification framework for 3D building models, whose main procedures include octree structure construction in voxel space, multi-resolution simplification of voxel representation, and model surface reconstruction. As to the multi-resolution simplification of models with detailed structure deletion, we define simplification rules for three kinds of detailed structures, and propose a multi-resolution simplification algorithm for 3D building models based on progressive simplification strategy.4) Multi-resolution representation of 3D building models. Aim at the application demands of 3D navigation and location-based services, we propose a shape recognition oriented multi-resolution representation approach for 3D building models, whose core idea is to organize the multi-resolution simplification results of 3D building models with primary structure preservation in a uniform skeme according to the octree structure in the voxel space, schedule model data strategy in voxel space for multi-resolution representation of 3D building models, and finally realize multi-resolution representation of 3D building models in network environment by means of voxel coding and global surface reconstruction. And as to the application demands of smart city and digital city, we propose a construction recognition oriented multi-resolution representation approach for 3D building models, whose core idea is to take every detailed structure of 3D building models as an independent object, organize the detailed structures together and make incremental storage by objectification management style, make data scheduling with object-oriented method for multi-resolution representation of 3D building model, and finally realize 3D building model representation in network environment by means of structure coding and local reconstruction.