Integrating LiDAR Point Cloud Classification and Building Footprints for Enhanced 3D LOD Building Modeling: A Deep Learning Approach
Keywords: LiDAR Point cloud classification, Building Footprint Extraction, 3D Building Modeling, Open3D Library
Abstract. 3D Building modeling is crucial for urban planning, helping stakeholders make informed decisions on critical issues such as flood risk assessment, urban heat island effect, and sustainable infrastructure development. In this research, we use RandLA-Net, a cutting-edge deep learning algorithm to classify LiDAR point cloud data to distinguish building structures. This identification of building points is essential for the subsequent creation of reliable 3D Building LOD models. To enhance the classification accuracy, this research utilizes building footprint vector data as a reference layer, which aids in refining the detection of building points and ensures validation of the results. Once the building points are classified and improvised with Building Footprint vector layer, they are utilized to reconstruct detailed 3D geometric models. This study employs the Open3D library to generate Levels of Detail (LOD) models for buildings. By combining advanced LiDAR point cloud classification with Building Footprint Extraction and 3D Building modeling techniques, this approach maximizes the utility of publicly accessible LiDAR point cloud data, delivering detailed 3D models that support a wide range of spatial decision-making processes.