Out-of-Core Surface Reconstruction from Large Point Sets for Infrastructure Inspection

Abstract

This paper presents a simple, effective, yet fast outof-core method for surface reconstruction based on the vector field surface representation. The algorithm is designed to handle massive amount of real-world point clouds representing large infrastructures (e.g. underwater hydroelectric structure) acquired by LiDAR, sonar or laser scanning system using outof-core techniques. Our method allows performing seamless surface reconstruction from unorganized, unoriented, nonuniform and highly noisy data that include outliers. The applicability of the method has been evaluated in the context of hydroelectric infrastructure inspection, and its performance has been tested using synthetically produced data and field data captured on different Hydro-Quebec's sites by laser line scanning, LiDAR and sonar measurement systems.

A Short Review on Data Modelling for Vector Fields

“…In [46], Nguyen et al introduce vector field representation for 3D point clouds into an automatic coarse-to-fine rigid registration method, which avoids the computation of searching for closest points, compared to the classic Iterative Closest Point(ICP) algorithm. Xu et al [47] designed an out-of-core method for surface reconstruction is another application of the vector field in point clouds. According to [47], in the vector field, each vector serves as the direction and the distance between a point and the closest point on the surface to be reconstructed.…”

“…Xu et al [47] designed an out-of-core method for surface reconstruction is another application of the vector field in point clouds. According to [47], in the vector field, each vector serves as the direction and the distance between a point and the closest point on the surface to be reconstructed.…”