Esri China (Hong Kong) Limited
With the rapid growth of digital twin buildings (DTBs) worldwide, there is an increasing need for scalable, low-cost, and labour-saving DTBs solutions. In contrast to conventional BIM methodologies that rely on high-resolution Lidar and significant manpower for Lidar scanning and 3D modeling, our proposed methodology is an efficient and scalable methodology for constructing DTBs using low-cost portable sensors and generative models. Three algorithms will be deployed and verified in this project. First, a novel algorithm for generating 3D floor plans is introduced, which utilizes cross-room-NeRF and fuses iPhone Lidar point clouds, thus allows for the use of affordable and publicly accessible portable sensors. Second, we introduce an algorithm for local 3D view alignment, which allows for partial scanning of a building. This enables the scanning phase to be carried out by regular employees and facilitates dynamic data updating. Third, we propose a 3D scene editor that refines scan-to-modeling errors by optimizing unidentifiable objects with compact shapes using generative models. Finally, we visualize the impact of the DTB using a game engine and create a plugin for efficient 3D mesh rendering. We believe the proposed methodology presents an opportunity to save costs in texture scanning and significantly boosts 3D modeling efficiency.
