AI-Augmented Integration of Smartphone LiDAR and Terrestrial Laser Scanning for Enhanced 3D Building Modelling

Abstract

Artificial Intelligence (AI) is rapidly transforming spatial data acquisition and modelling workflows. The aim of this research focuses on AI-augmented integration of Terrestrial Laser Scanning (TLS) with smartphone-based LiDAR to improve the quality and completeness of 3D building models. Although TLS provides detailed and accurate point clouds, its effectiveness is limited by issues such as occlusions, high costs, and accessibility constraints. Smartphone LiDAR offers a convenient and flexible way to capture data in areas that are difficult for TLS to reach, although it is less precise. In this study, AI techniques were applied to fuse the datasets from both TLS and smartphone LiDAR, filtering out noise and ensuring geometric consistency. The experiment was conducted at Bangunan Wawasan, UiTM Shah Alam, with dimensional accuracy assessed through comparison to Total Station measurements. The results show that TLS alone achieved a Root Mean Square Error (RMSE) of 0.020 meters, while the AI-integrated fusion of TLS and smartphone LiDAR data resulted in an RMSE of 0.025 meters. This small difference confirms that the fusion approach maintains accuracy while expanding spatial coverage. Overall, this work demonstrates that AI-driven integration of diverse LiDAR data sources can effectively address the limitations of individual systems, leading to more complete and accessible 3D models without sacrificing precision.