|Ph.D Student||Baruch Amit|
|Subject||Characterization of Subtle Topographic Features within|
Airborne laser Scans
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Sagi Filin|
|Full Thesis text|
Characterization of topographic features using digital surface models has focused on dominant elements while overlooking surface texture patterns and features of subtle appearance. The latter are however valuable for detailing geographic databases and for documenting and understanding undergoing process that contribute to landscapes reshaping. In the past, difficulties in their characterization could have been explained by resolution of the prevailing surface models. However, with the introduction of laser technology as a means to acquire detailed 3D data, their detection becomes feasible. This thesis develops methodologies for detection and the characterization of geomorphic features from laser scanning data. Review of related studies shows that current tools are inapplicable for such tasks and require research into new methodologies. The thesis introduces feature detection and extraction methodologies which are based on surface curvature analysis. Contrasting previous studies, the proposed methodologies consider measurement noise, surface texture and clutter, and the variety of shapes and forms that the features wear. To account for the feature diversity, the research introduces adaptive level-of-detection tests, which incorporate roughness and data accuracy parameters; thereby, it enables detecting subtle features while attenuating noise effect. For the characterization phase, the research focuses on two forms, closed entities and linear ones. A boundary delineation algorithm, which is based on incorporating surface and internal shape considerations, is proposed for development of the detected seed-region into the edge of the closed-form object. For the linear entities, a novel gap bridging model is introduced, which proposes an optimization driven model for handling fragmentation. Finally, the research proposes a method for grouping linear entities that share shape characteristics. Such grouping is important for analyzing sequences and patterns. Similarity is measured while considering shape variation as a function of the underlying surface topography.
The proposed models were applied on regions featuring different surface characteristics and geomorphic entities of various size and forms. The extraction accuracy was on order of 97%, and of features as shallow as 20 cm in depth were detected. Landform and surface texture had little influence on the detection. The high level of detection relieves the need to manual digitization, which is incomplete and error prone. The ability to detect autonomously natural features of 20 cm depth suggests great prospects in utilizing laser scanning data for characterizing geomorphic entities, and extraction of geometric data about them. It may thus provide a valuable tool for morphological studies.