基于体元分析的三维建筑物模型结构化分割方法

Structural Segmentation Method for 3D Building Models Based on Voxel Analysis

针对现有三维模型分割方法在建筑物基本结构特征识别方面存在的不足,提出一种基于体元分析的三维建筑物模型结构化分割方法。该方法首先通过体元化和内部距离参数计算在模型内部构建分层距离场,然后采用局部极值判别方法从各层距离场中提取出对建筑物结构、形状分布具有代表意义的中心体元,并基于中心体元和内、外部体元聚类找到建筑物在空间组成上相互独立的结构单元,最后依据体元分析结果对原始建筑物模型进行表面分割。试验结果表明,该方法能够对不同风格的建筑物模型进行有效分割,具有极强的稳健性。

To avoid the incapability of existing 3D model segmentation methods in recognition of structural characteristics of buildings,a structural segmentation method for 3D building models based on voxel analysis is proposed. The method firstly constructs distance fields by level in building models through voxelizatien and inner distance parameter calculation, then extracts midvoxels from each level of distance field constructed for representation of buildings＇ structure and shape distribution by means of local maximum search, and finds independent structural units of buildings in composition based on midvexel, inner voxel and outer voxel clustering, finally makes surface segmentation on original building models according to the results ofvoxel analysis, The experimental results show that our method is capable of efficiently segmenting building models of different styles with great robustness.