三維房屋模型可利用多種遙測技術所產生之資料進行建置，如衛星、無人飛行載具、航空測量及光達點雲等，而各方法皆有對於三維房屋模型建置的優缺點。近年來，衛星影像的空間解析度逐漸提高，已達數十公分等級，致使利用衛星影像建置三維房屋模型也逐漸受到討論。雖然衛星影像解析度與航測影像或光達點雲資料仍有差距，但衛星影像的涵蓋範圍廣、時間解析度高，因此對於建置三維房屋模型，仍有一定的優勢。
本研究主要針對以高解析光學衛星影像進行影像分析，並建立符合OGC CityGML LOD1等級之三維房屋模型。研究中應用深度學習，從衛星影像自動萃取出房屋平面圖(Building Footprints)，並去除過小或不屬於房屋的多餘區域，之後利用最小包絡矩形(Minimum Bounding Rectangle, MBR)技術、正規化(Regularization)及約化(Generalization)處理後，塑形出較規律及方正的房屋多邊形。最後，匯入數值表面模型(Digital Surface Model, DSM)，在每棟房屋上層萃取附屬結構物，並利用其高程資料與RANSAC (RANdom SAmple Consensus)演算法擬合各多邊形高度，以建立積木式三維房屋模型。
經過各項校正處理與去除異常值後，三維房屋模型平面(X、Y)誤差可達4.044公尺；高程(Z)誤差可達1.517公尺。由於衛星影像解析度較不足以建立複雜的屋頂模型，故建置模型之屋頂屬於平頂，精度符合OGC CityGML LOD1規範。後續三維房屋模型可應用於多種用途，如都市計畫、國土監測、災害重建、三維房屋模擬等，提供更接近真實世界的多元三維應用。

Three-dimensional building models can be reconstructed using data generated by a variety of remote sensing techniques, such as satellite, UAV, aerial survey, LiDAR point clouds, etc. Each method has advantages and disadvantages for the reconstruction of 3D building models. In recent years, the spatial resolution of satellite imagery has gradually improved, reaching tens of centimeters (cm) level, resulting in the use of satellite imagery to reconstruct 3D building models has gradually been discussed. Although there is still a gap between the spatial resolution of satellite imagery and aerial imagery or LiDAR point cloud data, satellite imagery has a wide coverage and high temporal resolution, so there are still certain advantages for reconstructing 3D building models.
This research focuses on image analysis based on high resolution optical satellite imagery and 3D building models reconstruction with an accuracy of the OGC CityGML LOD1 level. Deep learning technique is applied in this study to automatically extract building footprints from satellite images and remove excess areas that are too small or not buildings. Next, Minimum Bounding Rectangle (MBR), regularization and generalization processing are utilized to shape the more regular and square building polygons. Finally, the Digital Surface Model (DSM) is incorporated to extract sub-structures on the upper floor of each building, and the elevation data is used to fit the height of each polygon with the RANSAC (RANdom SAmple Consensus) algorithm to reconstruct block-based 3D building models.
After various corrections and the removal of outliers, the plane (X, Y) error and the elevation (Z) error of 3D building models can reach 4.044 m and 1.517 m respectively. Since the spatial resolution of satellite imagery is not enough to generate complex roof models, the roofs of reconstructed 3D building models is flat roofs, and the accuracy conforms to the OGC CityGML LOD1 specification. Subsequent 3D building models can be applied to a great deal of purposes, such as urban planning, land monitoring, disaster reconstruction, 3D building simulation, etc., providing a variety of 3D applications closer to the real world.

Albawi, S., T. A. Mohammed, & S. Al-Zawi. (2017). Understanding of a convolutional neural network. Paper presented at the 2017 International Conference on Engineering and Technology (ICET), 1-6.

Alidoost, F., H. Arefi, & F. Tombari. (2019). 2D image-to-3D model: knowledge-based 3D building reconstruction (3DBR) using single aerial images and convolutional neural networks (CNNs). Remote Sensing, 11(19), 2219.

Arefi, H., & P. Reinartz. (2013). Building Reconstruction Using DSM and Orthorectified Images. Remote Sensing, 5(4), 1681-1703.

Bittner, K., S. Cui, & P. Reinartz. (2017). Building extraction from remote sensing data using fully convolutional networks. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences-ISPRS Archives, 42(W1), 481-486.

Brédif, M., O. Tournaire, B. Vallet, & N. Champion. (2013). Extracting polygonal building footprints from digital surface models: A fully-automatic global optimization framework. ISPRS journal of photogrammetry and remote sensing, 77, 57-65.

Danilina, N., M. Slepnev, & S. Chebotarev. (2018). Smart city: automatic reconstruction of 3D building models to support urban development and planning. Paper presented at the MATEC Web of Conferences, 251, 03047. EDP Sciences.

Davydova, K., S. Cui, & P. Reinartz. (2016). Building footprint extraction from digital surface models using neural networks. Paper presented at the Image and Signal Processing for Remote Sensing XXII, 10004, 100040J. International Society for Optics and Photonics.

Dumoulin, V., & F. Visin. (2016). A guide to convolution arithmetic for deep learning. arXiv preprint arXiv:1603.07285.

Eriksson, H., T. Johansson, P.-O. Olsson, M. Andersson, J. Engvall, I. Hast, & L. Harrie. (2020). Requirements, development, and evaluation of a national building standard—a Swedish case study. ISPRS International Journal of Geo-Information, 9(2), 78.

Gröger, G., T. H. Kolbe, C. Nagel, & K.-H. Häfele. (2012). OGC city geography markup language (CityGML) encoding standard. v.2.0, OGC-12-019. Open Geospatial Consortium.

Guo, H., X. Su, S. Tang, B. Du, & L. Zhang. (2021). Scale-Robust Deep-Supervision Network for Mapping Building Footprints from High-Resolution Remote Sensing Images. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14, 10091-10100.

Heazel, C. (2021). OGC City Geography Markup Language (CityGML) 3.0 Conceptual Model Users Guide. OGC-20-066. Open Geospatial Consortium.

Ji, S., S. Wei, & M. Lu. (2018). Fully convolutional networks for multisource building extraction from an open aerial and satellite imagery data set. IEEE Transactions on Geoscience and Remote Sensing, 57(1), 574-586.

Kim, J. R., & J. P. Muller. (2002). 3D reconstruction from very high resolution satellite stereo and it’s application to object identification. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 4, 420-426.

Kocaman, S., L. Zhang, A. Gruen, & D. Poli. (2006). 3D city modeling from high-resolution satellite images. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 36(1/W41).

Leotta, M. J., C. Long, B. Jacquet, M. Zins, D. Lipsa, J. Shan, . . . K. Dana. (2019). Urban semantic 3d reconstruction from multiview satellite imagery. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops.

Li, S., Z. Zhu, H. Wang, & F. Xu. (2019). 3D virtual urban scene reconstruction from a single optical remote sensing image. IEEE Access, 7, 68305-68315.

Liu, P., X. Liu, M. Liu, Q. Shi, J. Yang, X. Xu, & Y. Zhang. (2019). Building Footprint Extraction from High-Resolution Images via Spatial Residual Inception Convolutional Neural Network. Remote Sensing, 11(7), 830.

Lokhat, I., & G. Touya. (2016). Enhancing building footprints with squaring operations. Journal of Spatial Information Science, 2016(13), 33-60.

Long, J., E. Shelhamer, & T. Darrell. (2015). Fully convolutional networks for semantic segmentation. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition, 3431-3440.

O'Shea, K., & R. Nash. (2015). An introduction to convolutional neural networks. arXiv preprint arXiv:1511.08458.

Partovi, T., F. Fraundorfer, R. Bahmanyar, H. Huang, & P. Reinartz. (2019). Automatic 3-D Building Model Reconstruction from Very High Resolution Stereo Satellite Imagery. Remote Sensing, 11(14), 1660.

Pokrovac, P. (2021). Comparison of data encodings to deliver 3D terrain models as triangulated irregular networks. (Master Thesis), University of Zagreb, Faculty of Geodesy.

Rau, J.-Y., & L.-C. Chen. (2003). Robust reconstruction of building models from three-dimensional line segments. Photogrammetric Engineering & Remote Sensing, 69(2), 181-188.

Ronneberger, O., P. Fischer, & T. Brox. (2015). U-net: Convolutional networks for biomedical image segmentation. Paper presented at the International Conference on Medical image computing and computer-assisted intervention, 234-241. Springer.

Saito, S., & Y. Aoki. (2015). Building and road detection from large aerial imagery. Paper presented at the Image Processing: Machine Vision Applications VIII, 9405, 94050K. International Society for Optics and Photonics.

Sun, S., & C. Salvaggio. (2013). Aerial 3D building detection and modeling from airborne LiDAR point clouds. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 6(3), 1440-1449.

Walmsley, A., & T. Kersten. (2019). Low-cost development of an interactive, immersive virtual reality experience of the historic city model Stade 1620. The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 42, 405-411.

Xu, Y., L. Wu, Z. Xie, & Z. Chen. (2018). Building extraction in very high resolution remote sensing imagery using deep learning and guided filters. Remote Sensing, 10(1), 144.

Yuan, J. (2017). Learning building extraction in aerial scenes with convolutional networks. IEEE transactions on pattern analysis and machine intelligence, 40(11), 2793-2798.

Zhang, C., & C. Fraser. (2008). Generation of digital surface model from high resolution satellite imagery. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci, 37, 785-790.

林宛蓉（2009），多層次精緻度三維房屋模型之建置，碩士論文，國立中央大學，土木工程學系.

國立成功大學. (2020). 109年度 三維道路模型資料標準制定與鐵路及捷運模型資料標準規劃期末報告, 內政部國土測繪中心.

張智安, & 陳良健（2006），利用光達資料模塑建物之研究, 航測及遙測學刊, 11（2）,頁 175-189.

蔡富安, 張智安, & 黃智遠. (2020). 108年度 三維地形圖資技術發展工作案期末報告, 內政部.