Page 13 - 中国仿真学会通讯2020第1期
P. 13
buildings, which provides a new technical way to process. Therefore, our research is also based on
acquire the 3D scene data quickly and efficiently. this.
At present, a series of 3D reconstruction methods
for oblique photogrammetry have been proposed In this paper, we propose an optimized
[10] [ 11 ] . Sun [ 12 ] proposes an automatic modeling method based on hierarchical
modeling method for urban buildings. This method reconstruction for 3D building point cloud, which
generates dense point cloud by SFM technology can make full use of the geometric information of
and projects them into regular meshes of the XY the original building facade point cloud and the
plane. By minimizing energy function,it identifies roof point cloud. Firstly, the point cloud of the
points belonging to buildings and reconstructs 3D building roof and the building facade are
prismatic building models. In view of the low segmented and extracted. Further, the 3D model
accuracy of the 3D reconstruction of oblique of building facade is reconstructed by hierarchical
photography, Wu [ 1 ] proposes an optimized reconstruction and the 3D model of building roof
method for 3D urban modeling, which combines is reconstructed by block reconstruction. Finally,
aerial oblique images with ground images. the two models are merged and clip unnecessary
However, this method adds a lot of work in the polygons,leading to the final model.
surveying and the mapping and requires the fusion
of different data. Existing commercial solutions, The major contributions of our method are:
such as ContextCapture ( Bentley, 2018 [ 13] ) , 1. A reconstruction method that can utilize
PixelFactory ( Airbus Intelligence,2018 [ 14] ) , geometric information of the building point cloud
and PhotoScan ( Agisoft, 2018 [ 15 ] ) , can and the roof point cloud, and can automatically
efficiently generate 3D meshes with textures from perform segmentation, clustering, and modeling
point cloud and oblique images. However, processes on the point cloud of a building.
commercial solutions focus on the process of 2. A hierarchical reconstruction method that
reconstructing 3D point cloud based on images, establishes a realistic building facade model.
but not the process of modeling based on point 3. This method can generate an authentic
cloud. texture with a certain resolution instead of a fake
texture using the original point cloud.
Generally speaking, there are still many The rest of the paper is organized as follows.
problems to be studied and solved in the Section 2 describes the method proposed in this
reconstruction of 3D building models. Due to the paper as a whole. In Section 3, the key steps of
aerial photography occlusion and the precision of the proposed method are presented in detail.
LIDAR equipment, the point cloud of 3D Section 4 presents the modeling results of this
buildings collected or generated may have serious method and compares the modeling results with
geometric defects and deformation, especially on the business modeling software. Conclusions are
the facade of buildings. Through the investigation drawn and discussed in Section 5.
of the relevant work, it can be found that there is
a lack of a modeling method to ensure the 2 The Workflow of The Building
accuracy and integrity of the roof and facade of Reconstruction
the 3D building. So, how to ensure the integrity
and construction details of the model is the focus The original point cloud used in the modeling
and difficulty of the research in the reconstruction process of this paper comes from Tanzhe temple in
Beijing. We collect data through aerial oblique
10 photogrammetry and perform 3D reconstruction