Arash Farahmand
An Automated Geometric-Topology Pipeline for Railway Centerline Extraction from Dense MLS Point Clouds.
Rel. Marilena Cardu. Politecnico di Torino, Corso di laurea magistrale in Georesources And Geoenergy Engineering, 2026
|
Preview |
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (5MB) | Preview |
|
|
Archive (ZIP) (Documenti_allegati)
- Altro
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (49kB) |
Abstract
In railway tunnel inspection, Mobile Mapping Systems (MMS) play a key role in creating a dense 3d model of the tunnel. Extracting accurate railway centerlines from raw point clouds remains challenging due to point density, noise, and environmental obstacles. This thesis presents an automated pipeline for railway centerline extraction from MMS point clouds using a combination of Digital Terrain Modeling (DTM), Principal Component Analysis (PCA), geometric topology constraints, density-based filtering, and a Spherical Snap smoothing algorithm. Unlike conventional approaches based on RANSAC line fitting, the proposed methodology exploits local geometric dimensionality and parallelism in rail topology to improve robustness while reducing computational complexity.
The extracted centerline is further enhanced through an embedded Data Quality Indicator (DQI) that provides point-level confidence estimates based on local point density
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
URI
![]() |
Modifica (riservato agli operatori) |
