Reconstructing power cables from LIDAR data using eigenvector streamlines of the point distribution tensor field

Abstract

Starting from the computation of a covariance matrix of neighborhoods in a point cloud, streamlines are utilized to reconstruct lines of linearly distributed points following the major Eigenvector of the matrix. This technique is similar to fiber tracking in diffusion tensor imaging (DTI), but in contrast is done mesh-free. Different weighting functions for the computation of the matrix and for the interpolation of the vector in the point cloud have been implemented and compared on artificial test cases. A dataset stemming from light detect and ranging (LIDAR) surveying served as a testbed for parameter studies where, finally, a power cable was reconstructed.