|M.Sc Student||Karnis Simon|
|Subject||Geospatial Data Conflation Using Agent Based Modeling and|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Joshua S. Greenfeld|
|Professor Sagi Filin|
|Full Thesis text|
Geospatial conflation is the process of integrating two separate data sources containing information with an explicit geographic component and creating a new, better, data source based on both. In this thesis we present a geospatial conflation algorithm for linear vector data using a multi-agent system (MAS) approach. The algorithm is capable of creating full matching of features from the two data sources and enables attribute transfer between matched features including dynamic segmentation of lines where one data source is mapped at a higher level of resolution than the other. The proposed algorithm builds upon existing research in the field and contributes a novel method for spreading feature matching through the linear networks using an agent-based depth-first search style traversal.
The proposed algorithm is tested in a series of experiments designed to identify it's capability to deal with four major challenges to conflation: positional errors in features, missing or unmapped features in both data sources, different levels of resolution and different cartographic representations of the same mapped objects. Tests are conducted with both artificially manipulated data and real-world data. The algorithm is shown to achieve a high percentage of successful feature matches under a variety of conditions which are likely to be encountered while working with different sources of spatial information.