Presents a method for the recognition of symbols on binary images,especially tailored for geographic maps. The prototyping stage isperformed by means of a geometric approach: using a suitable distancedefinition, the prototype of a class is obtained as the patternminimizing the sum of distances among itself and all the samples of theconsidered class. In this phase some statistical parameterscharacterizing the distortions occurring on the symbols are alsoobtained, so allowing the estimate of the recognition reliability. Onthe basis of the misclassification and reject costs, some tuningparameters, affecting the classification criteria, are also evaluated inorder to maximize the classification performances. Experimental resultsfor several test maps are finally presented
展开▼
