The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLIII-B3-2020
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B3-2020, 533–539, 2020
https://doi.org/10.5194/isprs-archives-XLIII-B3-2020-533-2020
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B3-2020, 533–539, 2020
https://doi.org/10.5194/isprs-archives-XLIII-B3-2020-533-2020

  21 Aug 2020

21 Aug 2020

AUTOMATED ROAD INFORMATION EXTRACTION FROM HIGH RESOLUTION AERIAL LIDAR DATA FOR SMART ROAD APPLICATIONS

M. Previtali, L. Barazzetti, and M. Scaioni M. Previtali et al.
  • Department of Architecture, Built environment and Construction engineering (ABC), Politecnico di Milano, via Ponzio 31, Milan, Italy

Keywords: ALS, LiDAR, Road extraction, Random forests, Point cloud classification

Abstract. Automatic extraction of road features from LiDAR data is a fundamental task for different applications, including asset management. The availability of updated and reliable models is even more important in the context of smart roads. One of the main advantages of LiDAR data compared with other sensing instruments is the possibility to directly get 3D information. However, the task of deriving road networks form LiDAR data acquired with Airborne Laser Scanning (ALS) may be quite complex due to occlusions, low feature separability and shadowing from contextual objects. Indeed, even if roads elements can be identified in the ALS point cloud, the automated identification of the network starting form them can be involved due to large variability in the size of roads, shapes and presence of connected off-road features such as parking lots. This paper presents a workflow aimed at partially solving the automatic creation of a road network from high-resolution ALS data. The presented method consists of three main steps: (i) labelling of road points; (ii) a multi-level voting scheme; and (iii) the regularization of the extracted road segments. The developed method has been tested using the “Vaihingen”, “Toronto” and “Tobermory” data set provided by the ISPRS.