Abstract

This paper presents a new deep-learning-based method for 3D Point Cloud Semantic Segmentation specifically designed for processing real-world LIDAR railway scenes. The new approach relies on the use of spatial local point cloud transformations for convolutional learning. These transformations allow an increased robustness to varying point cloud densities while preserving metric information and a sufficient descriptive ability. The resulting performances are illustrated with results on railway data from two distinct LIDAR point cloud datasets acquired in industrial settings. The quality of the extraction of useful information for maintenance operations and topological analysis is pointed together with a noticeable robustness to point cloud variations in distribution and point redundancy.

Details

Title
RAILWAY LIDAR SEMANTIC SEGMENTATION WITH AXIALLY SYMMETRICAL CONVOLUTIONAL LEARNING
Author
Manier, A 1 ; Moras, J 2 ; J-C Michelin 3 ; Piet-Lahanier, H 2 

 SNCF Réseau – DGII TTD, 9 Avenue François Mitterand, 93210 Saint-Denis, France; SNCF Réseau – DGII TTD, 9 Avenue François Mitterand, 93210 Saint-Denis, France; ONERA – DTIS, 6 chemin de la Vauve aux Granges, 91120 Palaiseau, France 
 ONERA – DTIS, 6 chemin de la Vauve aux Granges, 91120 Palaiseau, France; ONERA – DTIS, 6 chemin de la Vauve aux Granges, 91120 Palaiseau, France 
 SNCF Réseau – DGII TTD, 9 Avenue François Mitterand, 93210 Saint-Denis, France; SNCF Réseau – DGII TTD, 9 Avenue François Mitterand, 93210 Saint-Denis, France 
Pages
135-142
Publication year
2022
Publication date
2022
Publisher
Copernicus GmbH
ISSN
21949042
e-ISSN
21949050
Source type
Scholarly Journal
Language of publication
English
ProQuest document ID
2665343459
Copyright
© 2022. This work is published under https://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.