טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentGabbay Meir
SubjectDeveloping Procedures for an Automatic Conflation between
Vectorial Datasets of Geospatial Information
DepartmentDepartment of Civil and Environmental Engineering
Supervisor Professor Emeritus Yerach Doytsher


Abstract

Over decades, large quantities of geographically overlapping geo-spatial digital information were accumulated. This led to the development of a new Information Technology technique known as map conflation which is actually a process for integrating two or more GIS databases. It is progressed in three main stages: (1) Initial matching; (2) Alignment; (3) Final matching.  This research thesis presents a new approach towards implementing these stages.  The method originates in the computer vision pattern recognition domain, a method known as geometric hashing.  This method is divided into two main stages: a pre-processing stage and a recognition stage, and it uses vertices for encoding the object patterns.  In this respect, the developed new method uses Hough transform parameters for encoding lines instead of points as in the standard geometric hashing mechanism. 


The aim of this research is to develop new algorithms for matching linear features.   The geometric hashing technique is used for the detection of mutually corresponding linear features. The technique is applied on two layers of an overlapping geographic area. The aim is to automatically match features in both layers. The mutual features which are being identified are used for computing the transformation parameters between the two coordinate systems. The transformation is performed, bringing the two layers to an approximate alignment setting.  Subsequently a linear feature transformation is performed using the ICP algorithm to improve the accuracy of the mutual positions of the features for a better alignment. Finally the overall process is completed by identifying the edges of the overlapping section of each correspondent feature pair.


At every stage of the algorithm, graphic and numeric results are presented and a discussion of the results is conducted. Finally, conclusions followed by a discussion regarding future research directions are presented.