Optimizing LiDAR Point Clouds for Mobile Digital Twin Restoration of Cultural Heritage with Trustworthy Validation

Abstract

The preservation and restoration of cultural heritage have long been significant endeavors throughout human history. With the rapid advancement of LiDAR technology, digital methods have become essential tools for heritage restoration. However, efficiently and accurately processing LiDAR point cloud data—particularly in mobile digital twin restoration—poses numerous challenges. Cross-scale point cloud registration and trustworthiness assessment are among the key technical hurdles. Current research methods often face trade-offs between accuracy and efficiency when handling large-scale point cloud data. Moreover, traditional centralized approaches to trustworthiness evaluation raise concerns regarding privacy protection and security. To address these challenges, this study presents two primary contributions. First, an improved PointNet-based method for cross-scale registration of LiDAR point clouds is proposed, enhancing computational efficiency while maintaining registration accuracy. Second, a novel trustworthiness scoring mechanism is introduced, leveraging federated learning to enhance the reliability of restoration results while safeguarding data privacy. These advancements not only drive forward digital twin restoration technology but also offer safer and more reliable solutions for cultural heritage preservation.