Deep Learning based Wall Structure Object Extraction for 3D Building Modeling Automation

To create a digital twin, 3D modeling data that imitated represents the real-world is essential. However, people manually create modeling data by looking at photos or 3D scanning data. To address 3D modeling by hand, it is necessary to automatically extract information required for 3D modeling from 3D scanning data. In this paper, we propose a method based on deep learning-based 3D semantic segmentation and stochastic-based extraction of wall structure object from point clouds. We validate the performance of the proposed method by comparing the extracted wall structure object information from the initial point cloud with the actual 3D modeling.