Automatic recognition of point symbols on topographic map

摘要

This paper presents a new method of extracting and identifying point-shaped symbols of scanned topographic maps based on Distance Transformation. First, we remove of the interferences of transition color pixels on symbol edge, and filter the lines which do not meet the features of point symbols according to curve density and line density, and extract point symbols from the map based on adding hull transformation and removing skin transformation. Then we judge topology consistency by skeleton of symbol and shape consistency by weighted distance function, and use the both consistency to identify symbols. The algorithm can extract and identify rotated or distortional point symbols in the condition of character & line conglutination or complex background, and the identification accuracy is 98%.