Segmentation and scale detection algorithms for automated analysis of digitized historical maps

Shaw, Tenzing W.

Permalink

https://hdl.handle.net/2142/29683 Copy

Description

Title

Segmentation and scale detection algorithms for automated analysis of digitized historical maps

Author(s)

Shaw, Tenzing W.

Issue Date

2012-02-06T20:10:58Z

Director of Research (if dissertation) or Advisor (if thesis)

Bajcsy, Peter

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Image Analysis
• Segmentation
• Map Scale Estimation
• Historical Maps

Abstract

This thesis addresses the problems of automatic segmentation of objects in historical maps, automatic estimation of map scale and the design of a mathematical framework for understanding the uncertainties associated with map scale estimates. The problems are motivated by the lack of accuracy and consistency in the current analysis of geographical objects found in historical maps, which is conducted by unaided visual inspection. Our approach decomposes the analysis of geographical objects into workflow steps such as object segmentation, spatial scale calibration, extraction of calibrated object descriptors and comparison of descriptors over time and multiple cartography houses. The key computer science contributions are made in the segmentation and map scale calibration workflow steps. The segmentation step is achieved by designing a template-supervised ball-based region growing method employing the Hu moments as shape descriptors. The automation of spatial calibration (map scale estimation) is accomplished by algorithms that detect and classify lines along map borders, searching for dashed neatlines intersected by latitude lines. Thus, descriptors of map objects represented by segmentation results in pixels can be converted to geographical units; for example, the area of a lake can be reported in square miles. Finally, the map scale estimation process is modeled mathematically in order to establish uncertainty of the scale results. The uncertainty framework models contributions from various sources of error in the digitized historical map images, including clutter such as text impinging on the region of interest, low contrast between light and dark dashes of the neatline, as well as other sources. The application of our work has been to compare shape characteristics of the Great Lakes region in a dataset of approximately 40 French and British historical maps created in the seventeenth through the nineteenth centuries. The objective was to determine which colonial power possessed more accurate geographic knowledge of the region, and how this balance changed over time. We report experimental evaluations of automation accuracy based on comparison with manual segmentation results, as well as the knowledge obtained from the area comparisons. We also report the results obtained from experiments designed to allow uncertainty analysis of the scale estimation subsystem.

Graduation Semester

2011-12

Permalink

http://hdl.handle.net/2142/29683

Copyright and License Information

Copyright 2011 Tenzing W. Shaw

Owning Collections