三维建筑模型的规则化重构方法

Regularized Reconstruction Method for 3D Architectural Models

基于无人机照片的倾斜摄影测量技术能够快速重建大量三维建筑模型;但这些模型数据量大、噪声多,极大阻碍了在智慧城市中的应用.针对倾斜摄影测量重建的三维建筑模型屋顶重建精度较高的特点,提出一种基于屋顶轮廓线的三维建筑模型规则化重构方法.首先分割屋顶并提取屋顶的外轮廓线;接着通过改进的平面拟合提取屋顶的内轮廓线;然后由屋顶内外轮廓线进一步拟合出屋顶平面几何基元;最后,针对屋顶中太阳能、烟囱等的非平面部分,进行补充规则化处理,重构出数据小、噪声少的三维建筑模型.本文方法能够将三维建筑模型自动地重构为规则的建筑体,不依赖用户交互和先验信息.大量实验证明了本文方法的有效性、高效性和鲁棒性.

The photogrammetry technology based on drone photos can quickly reconstruct a large number of 3 D architectural models,but these models have much redundant data and noises,which greatly hinders applications in smart cities.According to the characteristics of high-precision roofs for photogrammetry-reconstructed 3 D architectural models,we propose a regularized reconstruction method of 3 D architectural models based on roof contours.Firstly,roofs are segmented out and their outer-contours are extracted.Then innercontours of the roofs are extracted by an improved plane fitting;the roof plane geometric primitives are further fitted by the inner-and outer-contours of the roofs.Finally,the non-planar parts(e.g.solar energy,chimneys,etc.) on the roof are supplemented and simplified to reconstruct regularized models with a small data size and few noises.Our method can automatically reconstruct the 3 D architectural models into regular architectural objects without any prior knowledge and user interactions.A large number of experiments prove the effectiveness,efficiency and robustness of the proposed method.