Deformation Tracking in 3D Point Clouds Via Statistical Sampling of Direct Cloud-to-Cloud Distances

摘要

Dense three-dimensional (3D) point clouds of infrastructure systems, generated from laser scanners or through multi-view photogrammetry, have significant potential as a source of nondestructive evaluation information. The growing maturity of these techniques make them capable of reconstructing photorealistic 3D models with accuracy on the millimeter scale, adequate for inspection and evaluation practices. Manual analysis of these point clouds is often time consuming and labor intensive and does not provide explicit information on structural performance and health conditions, highlighting the need for new techniques to efficiently analyze these models. This paper presents a new 3D point cloud change analysis approach for tracking small movements over time through localized spatial analytics. This technique uses a combination of a direct point-wise distance metric in conjunction with statistical sampling to extract structural deformations. By identifying and tracking these changes, mechanical deformations can be quantified along with the associated strains and stresses. These measurements can then be used to assess both service conditions and remaining system capacity. The results of a series of laboratory experiments designed to test the proposed approach are presented as well. The findings indicate measurement accuracy on the order of +/- 0.2mm (95% confidence interval), making it suitable for accurate and automatic geometrical analyses and change detection in a variety of infrastructure inspection scenarios. Ongoing work seeks to connect this technique to automated finite element model updating, and to field test the measurement technique.