Master's Thesis

KURATA, Yohei
Department of Urban Engineering, University of Tokyo

Advisor: Okabe, A., Asami, Y. and Sadahiro, Y.

Abstract

The objective of this study is to develop a method for importing digitalized rough maps into Geographic Information System (GIS). Rough maps are frequently used in our daily life. If they can be imported into GIS, vast amount of information accompanying rough maps can be managed, analyzed and utilized by using various GIS techniques.

In order to achieve this objective, i) rough maps should be properly modeled by computer, and ii) an algorithm for matching rough maps to precise maps should be developed. Thus, based upon the examination of 87 rough maps, a representative model for rough maps is proposed. Then, by modeling a human reasoning process that is used for interpreting rough maps, an algorithm that matches elements in a rough map to those in its corresponding precise map is proposed (Fig. 1).

The proposed algorithm proceeds the matching by the following three steps. First, each landmark drawn on a rough map is identified with one of the elements in a precise map. Second, for each node and link adjacent to the identified landmarks, an appropriate counterpart is detected around the corresponding element of each landmark. Third, other unidentified elements on the road network of a rough map are sequentially identified . In each step, geometrical similarity, topological consistency, commonality of attributes (such as name) and so forth are comprehensively used as clues for appropriate matching.

In our experiment, about 70% of 178 sample maps were sufficiently matched by the proposed algorithm. The result implies that rough maps people draw are so diverse that still more flexible reasoning algorithm should be developed in order to realize practical matching.

Finally, using a metaphor from natural language processing, a concept named enatural map processingf is discussed (Here natural map means the map designed by people not following any strict rules). Thus, the methodology of matching rough maps is reviewed in this framework.