A hierarchical approach for near real-time 3D surface change analysis of permanent laser scanning point clouds

• Verfasst von: Tabernig, Ronald [VerfasserIn]
• Verfasst von: Albert, William [VerfasserIn]
• Verfasst von: Weiser, Hannah [VerfasserIn]
• Verfasst von: Höfle, Bernhard [VerfasserIn]
• Titel: A hierarchical approach for near real-time 3D surface change analysis of permanent laser scanning point clouds
• Verf.angabe: Ronald Tabernig, William Albert, Hannah Weiser, and Bernhard Höfle
• E-Jahr: 2025
• Jahr: 24.03.2025
• Umfang: 9 S.
• Fussnoten: Gesehen am 07.05.2025
• Titel Quelle: Enthalten in: Joint International Symposium on Deformation Monitoring (6. : 2025 : Karlsruhe)6th Joint International Symposium on Deformation Monitoring
• Ort Quelle: Karlsruhe : KIT, 2025
• Jahr Quelle: 2025
• Band/Heft Quelle: (2025), Seite 1-9
• DOI: doi:10.5445/IR/1000180377
• Datenträger: Online-Ressource
• Sprache: eng
• K10plus-PPN: 1924990032

Abstract

Modern permanently installed laser scanning systems (PLS) allow capturing point clouds in short intervals (e.g., sub-hourly), bringing us closer to the early detection of small surface changes that may precede larger events. Predicting potential hazards necessitates near real-time surface change computation. This requires reliable and efficient methods that can be operated directly on laser scanners in the future. We propose a method that combines low-resolution (meters) change detection with high-resolution (centimeters) change analysis. First, utilizing the Mahalanobis distance, a change detection approach identifies significant intravoxel changes, filtering out temporary changes (e.g., tree movements due to wind) to retain only persistent, relevant changes (e.g., rock movements). Second, sub-point clouds of areas exhibiting significant change are extracted and subjected to point cloud-based surface change analysis. Hierarchical analysis of point Clouds with fine-tuned key parameters results in a data volume reduction of over 95% and a miss rate of less than 6%, both relative to a manually annotated reference point cloud. Furthermore, a computation time decrease of 97% is achieved relative to an M3C2-only run. Our approach is based on the hierarchical detection and Analysis of areas exhibiting surface change. This method is particularly efficient when these areas are considerably smaller than the monitored area, allowing processing within seconds and much faster than data acquisition. A further advantage is that this methodology is implemented using the open-source Python libraries py4dgeo and VAPC, which enables straightforward integration into your own PLS monitoring workflows, allowing processing much faster than data acquisition (e.g. within seconds)