Recursive morphological sieve method for searching pictorial point symbols on maps

Point symbols on a map represent interesting and significant positional data, such as airports, ski areas and natural parks. Identification of point symbols and finding their positions are important for automatic map understanding. Point symbols on maps may partially overlap on each other. More often they are touched or connected by line symbols, and their background consists of colored and/or textured regions. In this situation, a straightforward morphological template matching method is not feasible. A new partial ordering method for aligning pictorial point symbols which is based on the relationship of K-tolerance covering and a new general method for searching pictorial point symbols on maps, the recursive morphological sieve method, are described. The main ideas of this new method include (1) finding the coverings of the kernels of a symbol instead of the symbol itself, (2) morphologically preprocessing maps to form multi-level sieves, (3) using the sieves to screen out symbol kernels, and (4) renewing sieves during a recursive screening process. This method may also be used to recognize pictorial point symbols on other types of documental files.