|Ph.D Student||Ben-Haim Gev|
|Subject||Local Accuracy Analysis of 3D Topographic Databases|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Emeritus Yerach Doytsher|
|Professor Joshua S. Greenfeld|
|Dr. Sagi Dalyot|
|Full Thesis text|
A Digital Terrain Model (DTM) represents a continuous terrain surface via a set of discrete points. This digital infrastructure is widely used for many applications and tasks. It constitutes an important requirement for establishing an efficient management of our environment. As DTMs become more available, different types of data describing the same relief are increasingly common. These diverse models may be fundamentally discrepant, having different data qualities and representations, such as reference system, accuracy, resolution, coverage area - to name a few. A problem arises when a single DTM contains different accuracy levels, thus harming its homogeneity and affecting its reliability and characteristics at different magnitudes. Moreover, different areas present different topographic attributes, thus altering the qualities of the model.
The majority of data-processing operations used to generate DTMs lack the capability to monitor these accuracy measures correctly, thus, spatial inconsistencies and ambiguity already existing in DTMs are preserved. Accuracy analysis approaches commonly assess the vertical accuracy of the final DTM as a consequence of all errors resulting from production processes or by comparing the analyzed model with sampled heights. This usually results in a single global measure or a range of values that rarely express the actual accuracy, not to mention inner-local accuracies. Relying solely on such approaches might result in an erroneous topographic analysis and might not suffice for a comprehensive analysis of DTMs.
The research presented here suggests a qualitative and quantitative modeling approach for analyzing accuracies of DTMs. The methodology computes the absolute vertical accuracy values of DTMs globally and locally. It utilizes the data stored in several different DTM sources (covering the same relief) simultaneously, while thoroughly analyzing, characterizing, and modeling the inner spatial characterization of the terrain and varying localized qualities.
The spatial alignment of the participating DTMs is implemented, to minimize any ambiguities and discrepancies that may result. A spatial topographic framework of all participating DTMs is established via a precise identification of unique topographic features. Having achieved this, an algorithm for the automatic matching and registration of these features is developed based on the correspondences of triangulation structures, aimed at evaluating the geometric similarity of the data. Assuming that a qualitative spatial alignment has already been achieved, multiple different DTMs with a similar coverage area are simultaneously analyzed to compute the absolute vertical accuracies. Since the accuracies are not uniform and homogeneous, the localized errors existing in the models are also addressed.
This research is an important step towards a thorough quality analysis of DTMs. It relies on the data stored in diverse types of DTM sources, i.e., post-processing analysis. Global and local absolute accuracies can be successfully, reliably and precisely determined. It is achieved regardless of any preliminary information and almost without prior considerations and limitations. The outcome is local accuracy measures expressing the varying existing accuracy trends. Experimental results proved to be sound, supporting the overall assumptions and considerations made during the course of this research, demonstrating the feasibility of using the proposed methodology and developed algorithms.