|M.Sc Student||Barak Kohavit|
|Subject||Fuzzy Logic Based Boundaries Determination of Vectors Based|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Joshua S. Greenfeld|
|Professor Sagi Filin|
|Full Thesis text|
In this study we introduce a new method for improve existing polygon boundaries in
GIS databases, using Fuzzy Logic theory on vector based data sets. The method works on points from different vector based background layers. Fuzzy logic algorithms are not new in GIS but in most cases they are performed on raster data sets, not on vector data sets. It is rare to find fuzzy logic methods applied to vector based data sets.
The most fundamental spatial feature is a point. In this work the polygon is first decomposed to its vertices. Then, each point (vertex) is passed through the fuzzy logic engine to improve its location. Finally, a fine tuning process, if necessary, is used to add additional vertices to improve the final fuzzy logic result. It is done with a convex hull algorithm on segments of the boundaries when needed.
The fuzzy logic approach is performed in two levels. The first one is in the determination of the density membership score in the fuzzification state. The first fuzzy state of the data is when each point in the buffer gets the density membership values for membership in each other point in the buffer. The second level is Defuzzification when fuzzy state of the data has to be re-focused into a unique value for the feature. By using a selected ratio between the density membership grade and the source membership grade, we produce one crisp value.
We conducted several tests involving several layers of the national Geo database. The first test was conducted on a synthetic polygon with four vertices to establish the method to create the density membership grade. The next test was conducted on a real world village named "Ofer" and was extended with a more complex algorithm that was developed. "Ofer" polygon is a more complex case. The next test was conducted on another village polygon named "Kerem Maharal. The basic fuzzy logic process was tested with several buffer radii and several ratios between the density membership grade and the source membership grade.
At the end of the basic fuzzy logic analysis there was a need to conduct a fine tuning process.
At the end of the fine tuning process there were no odd points compared to the reference polygon. Our research showed promising results and can be applied to improve the national Geodatabase.