Vector Feature Extraction Using Object-Oriented Image Analysis Techniques from Scanned Maps

Author

Wook Rak Jung, Western Michigan University

Date of Award

6-2009

Degree Name

Master of Arts

Department

Geography

First Advisor

Dr. Charles Emerson

Second Advisor

Dr. David Dickason

Third Advisor

Dr. Gregory Veeck

Access Setting

Masters Thesis-Campus Only

Abstract

Deriving GIS data from historical maps will provide the ability to compare historical geographic patterns to more recent GIS-layers. In this example, USGS topographic maps, scanned at Western Michigan University's W.E.Upjohn Center for Global Change provide experimental sources for the feature extraction techniques. USGS topographic maps contain standardized symbols and uniform colors, and the Digital Line Graphs that correspond to the scanned quadrangle provide a baseline for checking the accuracy of extractions. Vector line features and polygon features were extracted from scanned maps using an object-oriented image analysis technique. The Definiens eCognition^TM software provides the benefit of isolating the target objects in the map according to the difference in spectral, shape, and context. Polygon features could extract with spectral difference, but vector lines' extraction is required to employ all three factors. Extracted polygon features are acceptable to export as a shapefile directly, but extracted vector line features usually contain a lot of "noise data". When extracted vector line features are exported as shapefiles, the exported vector lines are constructed with many discrete lines. These problems were overcome to some extent with the ArcScan extension of ArcGIS™. The quality of extracted vector line features also improved using morphological processes in ArcScan.

Recommended Citation

Jung, Wook Rak, "Vector Feature Extraction Using Object-Oriented Image Analysis Techniques from Scanned Maps" (2009). Masters Theses. 234.
https://scholarworks.wmich.edu/masters_theses/234