Detection of gullies in roughly textured terrain using airborne laser scanning data

Abstract

Airborne laser scanning data contain information about surface features, some of which are of subtle form. These features are usually embedded within the terrain, and rarely form distinct shape-transition to their surroundings. While some efforts have been made in extracting linear elements from laser scanning data, attention was mostly turned to dominant elements that are very clear and distinct. We present in this paper a detection model for gullies of various dimensions using airborne laser scanning data. Gullies are regarded as one of the main landform-reshaping agents, having a pejorative effect on the environment and on regional development. They are commonly observed along receding lakes as a common response to water-level drop. The paper demonstrates how a multi-scale approach enables the extraction of various gully forms, from well developed to subtle. It then proposes an optimization driven model for handling fragmentation in the detection. Results show that using the proposed model, gully networks can be reconstructed and ∼30 cm deep features can be identified and separated from their surroundings using moderate point density data.