An annotation-free transient filtering pipeline for organized TLS panoramas, combining self-supervised visual descriptors, text-aligned priors, and geometric cues such as range, gradients, and surface normals to estimate transient-probability masks and back-project them to the original TLS rays.

An end-to-end geometric consistency formulation for robust point cloud registration under challenging overlap and correspondence conditions.

A view-planning approach for communicating cultural heritage content more effectively from 3D documentation.

Jointly improves scan quality and geometric normal estimation for mobile and handheld laser data.

Uses learned contour detection to make heritage-site point clouds more legible and interpretable.

Addresses noise, uneven sampling, and missing structure in portable handheld laser scans.

A framework for associating, aligning, and tracking plant structures over time from point cloud observations.

Studies learning-based upsampling for unevenly sampled natural and scene-scale point clouds.

Focuses on reliable normal estimation for complex heritage geometry acquired with low-resolution noisy scans.

Proposes shape-preserving denoising for portable-laser-scan reconstruction of heritage environments.

An automated method for extracting rail tracks from mobile laser scanning data for railway inspection and mapping workflows.