Thomas Rommel (2015)

Automatic Detection of Line Structures in Airborne LiDAR Data

Airborne LiDAR-derived digital terrain models can be used to detect skid trails overgrown by dense forest vegetation. This study proposes a model approach for an automatic detection and vectorization of skid trails using a 0.5 meter digital terrain model (DTM) of Rhineland-Palatinate, Germany. Due to the steep terrain and the resulting terrain height differences, a moving window filter method, using the height differences along a defined distance to extract skid trails from the DTM, is created. For the vectorization of the resulting binary image, digitized start lines are used for an automatic connection of related nodes, with a given search angle and a maximum distance. Appropriate results are obtained and evaluated by features derived from hillshades of the test area. In comparison to an active contour model, the derived feature had a length of 2,193 meters, with a completeness value of 67 percent, a correctness value of 68 percent and a root mean squared error value of 2.7 meters.