(Semi-) Automated extraction from airborne laser scanning (ALS) data for road & path detection in forested areas


Until recently, the application of prospection techniques in heavily vegetated or forested areas has been very problematic. This has had a negative impact on research of roads and paths, especially as they tend to be better preserved in these kind of areas. Over the last decade, Airborne Laser Scanning (ALS) has changed the situation. With this remote sensing technique it is possible to visualise on a large scale the microtopography of landscapes. Moreover, it can also be applied in heavily vegetated or forested areas. Therefore, it is now possible to conduct research on road networks on a large scale in these areas. Nevertheless, compared to ALS data filtering, only few articles have been dedicated to the use of ALS data for road and pathway detection (Mallet et al., 2010; Štular, 2011; Štular, 2011) and work still has to be done for automatic extraction of linear features (Mallet et al. 2010, Doneus & Briese 2004). However, good suggestions (Hough transform, geostatistical kriging, automatic breakline extraction, openness) have been made to further the development of automated road and path network extraction for archaeological purposes (Mallet et al., 2010; Humme et al., 2010; Briese, 2004; Doneus, 2013). In this paper we will present different workflows for the extraction from ALS data for road and path detection in forested areas based on the concepts of break lines and openness. The visualisation concept openness was first introduced by Ryuzo Yokoyama, Mlchlo Shlrasawa, and Richard Pike in 2002. Both the positive and negative openness are an angular measure denoting the aperture angle of a cone centred at a grid point and constrained by the neighbouring elevations within a specified radial distance (Pregesbauer, 2013). This technique is considered as ideal tool for mapping and outlining of archaeological features. Firstly, because it has some advantages over several other visualisation techniques, as it does not produce directional bias and horizontal displacement. Secondly, openness provides a clear distinction between relief features and the surrounding topography and it highlights both the highest and lowest parts of features (Doneus, 2013). In the break line approach a grouping of the point cloud into two groups takes place. Each group includes the points on either side of the break line, and subsequently reconstructs the surface on each side independently. This leads to two surface descriptions, which are valid for one side only. So, in a more technical sense, the definition of break line is: the intersection of two smooth surfaces, each surface interpolating the points on either side. The high degree of accuracy and completeness makes this approach to be preferred over the raster, grid and tin based approaches (Briese, 2004). The products of both openness and the break line technique are followed up by further elaboration steps in a GIS environment to finalize to extraction workflow. The obtained results based on openness and break lines demonstrate that this line of research is worthwhile for road and path detection in forested areas and invites to further investigation.