3D Digital City Modeling from Image-based Dense Point Cloud
| Autor: | Chang, Li-Ying, 張立穎 |
|---|---|
| Rok vydání: | 2017 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 105 3D Digital City Model has gradually developed into 3D Smart City Model, provide high-precision 3D Digital City Model will be the development of 3D Smart City Model of the important issues, building boundary line for the 2D and 3D map of the important spatial information, photogrammetry is still based on 3D mapping system-based production, but its production efficiency is not high. Light Detection And Ranging can directly build the 3D dense cloud of the feature, which is rich in geometric features and 3D spatial information. Theoretically, the boundary of the building can be extracted from the LiDAR point cloud, but the LiDAR point cloud is usually distributed sparsely to the side of the building. It is difficult to extract its characteristics and it is difficult to obtain the boundary of the building. In addition, due to the annex structure of the building makes the distribution of point cloud more chaotic, often make the extraction of the boundary characteristics of incoherent, can not produce high-precision 3D Digital City Model. In this study, Unmanned Aerial Vehicles use multi-view image technology to obtain dense point clouds for building boundary extraction, and improve the accuracy of 3D Digital City Model. Dense point cloud for automated building boundary extraction, this study proposed three methods, the first will be proposed by the use of Concave hull algorithm with Ramer Douglas Peucker, referred to as CRDP. But the extraction of building materials is not complete, the second will be in the building boundary line to the True Orthophoto with 3D dense point cloud, mining artificial way to repair, the third will use the commercial software VRMesh building extract, three extraction results were compared, building boundary line data to FME into Open Geospatial Consortium developed CityGML format to create 3D Digital City Model. In this study, we selected VRMesh and CRDP experiments in three districts with a total of 57 points. The results were analyzed with the results of the existing topographic maps. Under the absence of True Orthophoto images with 3D dense point cloud, the VRMesh its Root Mean Square Error is about 4cm, CRDP RMSE is about 2cm; True Orthophoto images with 3D dense point cloud repair, VRMesh RMSE is about 3 cm, CRDP RMSE is about 1.5cm. From the experimental results, it can be seen that the use of CRDP and the True Orthophoto images with 3D dense point cloud repair method to improve the extraction of building boundary line accuracy is effective, enough to provide high precision 3D Digital City Model production use. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|