Abstract. LiDAR and photogrammetry technologies generate point clouds that serve as a vital source of high-resolution spatial data for accurately reconstructing 3D building models. Along with these advancements, challenges such as occlusions and inconsistencies within individual datasets often lead to incomplete models, resulting in missing structural elements of the building, such as roof sections. The integration methods of multiple point cloud, known as data fusion, enhances data accuracy and completeness by complementing each dataset and addressing these issues. Among these techniques, Iterative Closest Point (ICP) is widely employed for point cloud registration, yet it does not fully eliminate gaps and redundancies. To overcome these limitations, the Laplacian method has been introduced, to refine alignment between point clouds, significantly enhancing overall accuracy and completeness. Despite advancements in point cloud fusion, evaluating the accuracy and completeness of the resulting models remains crucial to ensure their reliability and applicability. Research has shown that higher accuracy and greater point cloud density lead to improved reconstruction quality. This study focuses on evaluating the accuracy and completeness of 3D building models generated through point cloud fusion. Evaluation of the quality of 3D building models involves both qualitative methods, such as visual inspection, and quantitative metrics to measure geometric accuracy and structural integrity. The findings provide valuable insights into the reliability of current modelling techniques, contributing to quality control improvements and advancements in 3D reconstruction methodologies.
Mutiarasari et al. (2026) studied this question.