Accuracy of automatically extracted geomorphological breaklines from airborne lidar curvature images

摘要

Geomorphological breaklines are 3D polylines extracted from high-resolution digital terrain models (DTMs), which have manifold applications in geomorphology and related fields. Breaklines are used to maintain and refine the characteristic information on discontinuities in the process of DTM filtering and interpolation. Knowledge about the quality of the derived breaklines is important in order to provide adequate input data for DTM generation, geomorphological feature mapping, and change detection analysis. The spatial accuracy and the classification accuracy of derived breaklines depend on data acquisition settings and processing steps. The critical criteria are the precision and density of the LiDAR point cloud, the resolution and quality of the filtered and interpolated DTM and the target scale breaklines are derived for. In this paper a quantitative method for assessing the quality of derived breaklines, which are extracted from regions with high curvature, is presented. Selected components of the breakline detection workflow are analysed. In addition, the reliability of manually digitized reference lines is investigated. Selected geometric properties such as line length, slope and sinuosity are computed and compared to reference data. The properties are discussed regarding their explanatory power with respect to accuracy and precision. Automatic extraction results strongly differ from manually digitized breaklines. Breaklines from manual digitalization are found to vary in coverage and spatial accuracy and thus reflect always the operator's ability, perception and intention. The advantage of automated extracted breaklines is their reproducibility and adaptability to scale.