Digital terrain models from airborne laser scanning for the automatic extraction of natural and anthropogenic linear structures

Digital terrain models (DTMs) acquired from airborne laser scanning (ALS) are available for many mountainous regions. ALS has the capability to penetrate high vegetation, hence high-resolution DTMs can be derived even in forested areas. This terrain information can be used in various geo-related applications such as hydrology, natural hazard process modelling and geomorphological mapping. The high resolution of ALS DTMs shows both natural and anthropogenic terrain features such as erosion scarps, geological lineaments, walking paths and roads. Anthropogenic structure lines often subdivide the terrain into artificial units; therefore, to be able to describe the natural geomorphologic feature as a unit in its original appearance, anthropogenic and natural features have to be separated. Without this modelling, several applications suffer from using ALS data. Standard algorithms previously developed for coarser DTMs are not adapted to the implicit feature representation in the highly detailed terrain description. The classification approach presented here separates road features from geomorphologic line structures based on their slope, curvature and shape properties in three different test sites in the Austrian Alps highlighting different geomorphological situations. The comparison to a road reference map shows the feasibility of the developed workflow. The importance of detecting anthropogenic structures for geomorphological investigations is demonstrated by calculating a line density map, which indicates geomorphological activity on sloped terrain.