Efficient Method on Integration of Construction Site digital maps Using Voxelized Pointclouds

Abstract

In the construction industry, efforts to improve productivity and safety through digital twin technology have gained significant attention. To build an effective digital twin, it is essential to rapidly and accurately acquire highquality digital maps using equipment such as UAVs and UGVs. This requires aligning PCD Data (PCD) in accordance with the specific characteristics of construction sites. In this study, we propose a voxel-based registration method for generating digital maps that capture the dynamic and large-scale nature of construction sites. Unlike traditional methods that focus solely on registration outcomes, our approach emphasizes both speed improvement and quality retention by leveraging the strengths of PCD during the registration process. By applying voxel registration to PCD from actual construction sites, we confirmed that our method maintains the quality of PCD obtained from different devices while significantly accelerating the registration process. The registration results showed a more than 30 times reduction in processing time compared to the commonly used ICP (Iterative Closest Point) algorithm, while maintaining approximately 86.56% of data quality. Additionally, the site coverage increased by 45.76% for UGVs and 7.95% for UAVs after registration. Furthermore, our method demonstrated superior performance in both processing time and quality retention compared to the NDT (Normal Distributions Transform) algorithm. This research made a significant contribution to digital twin technology, which prioritizes real-time updates and accuracy.