Road extraction from airborne LiDAR data using SBF and CD-TIN

Abstract

This paper proposes a method for automated road extraction from airborne Light Detection and Ranging (LiDAR) data. The method combines Segmentation Based Filtering (SBF) with Triangular Irregular Network-based segmentation to extract the road points. The method contains two major steps. Firstly, Segmentation Based Filtering (SBF) is applied to LiDAR data for initial segmentation of road regions. Here, region growing algorithm is applied after detecting outliers and subsequently, ground reference points are also identified using impainting with interpolation method. Secondly, road extracted by the SBF method is refined through the help of Constrained Delaunay Triangular Irregular Network based segmentation and then, road contour is extracted from road point image. For the experimental validation, the proposed and existing method is tested against ISPRS reference dataset. The experimental results showed that proposed method achieved that the completeness of the approach is 83.40 % and the correctness value is 83.02 % and the accuracy is 83.16%.

IQPC 2015 Track: EVALUATION OF AUTOMATICALLY GENERATED 2D FOOTPRINTS FROM URBAN LIDAR DATA

“…Completeness, correctness and RMS differences were used to evaluate performance of 6 submitted results, where the completeness and correctness were, respectively, around 0.7 and 0.85, while the RMS difference was no more 3.74 pixels. The criteria for evaluating complete extracted road were to test automatic road detection for individual roads (Cheng et al;Kumar et al, 2013;Miraliakbari A. et al, 2015;Narwade and Musande, 2014;Yang et al, 2013). Amongst these, Kumar et al (Kumar et al, 2013), evaluated the extracted roads by using a point cloud of road edges, while Miraliakbari et al (2015) evaluated their work by using detected road areas.…”