Detection of Sinkhole Hazards using Airborne Laser Scanning Data

Airborne laser scanning technology is primarily perceived as a means for gathering detailed three-dimensional information about the surface and objects on it. The dense 30 data contain information about surface features and geohazards, some of which are of subtle form. Geohazards are usually embedded within the terrain, and scarcely form distinct shape-transition to their surroundings; therefore their detection is challenging. We address in this paper detection of subtle terrain features and demonstrate it on collapse sinkholes. Collapse sinkholes are surface depressions whose formation has severe effect on the environment and on regional development. We present an autonomous model for their extraction and characterization. Sinkholes within the studied regions appear in various size, forms, from their embryonic to a well developed formation. The level of sinkhole detection is high, and as demonstrated, the model performs well under varying landforms and surface texture, with little influence on the correctness of the extracted sinkholes. As the results show, features of approximately 20 cm deep can be identified and separated from their surroundings in the data.