Extraction of Characters from 1/25, 000 Scaled Maps by Isolating Lump Objects and Searching Key Characters

Abstract

This paper proposes a new method to extract characters automatically from 1/25, 000-scale maps, the scale widely used in Japan. Characters cannot be independently extracted because they are often connected with other figures. Therefore, large objects are divided into several small character candidates utilizing a boundary tracing method. Character candidates are compared to key characters which appear in maps in high number, and key character positions in the given map are obtained. Characters are searched from positions of obtained key characters. Algorithm of character extraction and experimental results are described.