Speaker
Description
Digital functional models can provide rich 3-D geometric information of structures, such as building information model (BIM) and finite element model (FEM). They are extensively applied for as-built completion acceptance, robotic path planning, structural health monitoring, finite element computation, and other fields. However, for existing digital model reconstruction, how to high-resolution update and reconstruct complex-shape models in com-ponent element level has not been sufficiently investigated. Therefore, this study proposes an approach, which fuses point cloud-to-model mapping, deep learning and model surface reconstruction to update global 3D geometric changes of each component on complex digital functional models at high resolution. The pipeline starts from assigning a corresponding point cloud for each component element. Afterwards, convolutional neural network (CNN) is applied to extract shape features of the corresponding point cloud and update the original model. Finally, model refinement is achieved through triangulation and remeshing. This approach provides an more effective solution for acquiring digital functional model in related fields.
