|Ph.D Student||Dalyot Sagi|
|Subject||Hierarchal Modeling and Integration of Topographic|
Databases: Algorithm Development and Geo-Spatial
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Emeritus Yerach Doytsher|
|Full Thesis text|
DTM is a common spatial expression and representation of the topography that is widely used as an infrastructure for a variety of processes and analysis tasks. Accessing and maintaining DTM data is an essential requirement for establishing efficient and computerized management of our environment.
Due to vast data sources available nowadays for DTM production, it is common that different models of a specific area exist. Common geo-related analysis tasks, such as integration and morphologic change-detection, utilize several congruent models simultaneously. These diverse datasets may have different data qualities and representations (datum framework, accuracy, resolution, epochs, level of detailing - to name a few). Instead of coercing a global unique integration framework, which is incorrect due to existing varied-scale discrepancies, this research presents a novel hierarchal integration approach that addresses the different levels of discrepancies quantitatively and qualitatively by implementing a successive triple-sequence procedure, where each handles different data extent.
In the initial stage, the extraction of a qualitative global mutual correspondence is achieved by co-registering individual large congruent frames, which are divided from the entire area. The registration process, which introduces a novel robust morphologic interest point identification algorithm, is based on the forward Hausdorff distance algorithm ensuring correct homologous interest points co-registering.
In the second stage, spatial Iterative Closest Point (ICP) matching algorithm is implemented on local partially-overlapped sub-frames, which are divided from the co-registered frames. The aim is to perform a fine-tuning of the registration values extracted in the initial stage in order to calculate local geometric corrections. The extracted local registration values include translation and rotation parameters relating to the mutual sub-frames spatial correspondence. The complete sets of these local and precise registration values are then utilized on the third stage for a weighted two-way iterative integration process. For each point in the integrated DTM the precise unique transformation values are calculated via the implementation of novel interpolation algorithms that make use of quaternion representation and bi-directional third degree parabolic equation.
This hierarchical framework - implemented almost without any prior considerations or limitations - ensures achieving a spatial-continuous integrated DTM that covers the entire analyzed area. The obtained results preserve the terrain structure and its morphologic details as presented originally by the different DTMs used. The last part of the research employed the proposed approach and its novel data-handling concepts to other geo-related tasks, including: morphologic change detection; morphologic quantification of landslide occurrences; topography animation and simulation.