Invariant Features Extraction of Map Symbols
YE Jian-kao, LIU Yue (Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China)
Feature extraction is the key of automated recognition of map symbols. Four two-dimensional invariant features of map symbols are summarized. The methods of the extraction of invariant features and a part of the experiment results are put forward.The four invariant features are the degree of complication of the periphery of the symbols , the breadth length ratio of the symbols, the proportion between the black pixels andall the pixels of a symbol, and the distance between the barycenter and the geometry center of a symbol.These four invariant feature parameters are characterized by invariability of rotation and size. A part of invariant characters has been proved in the paper. They are easy to be picked-up and the calculation is very simple.From the results of the experiment, we can see that the four feature parameters com-mendably describe the characters of the point symbols on maps, and they are not influenced by changing the size of scanning maps or rotating symbols and not disturbed by any noise of the image.The correctness ratio of automated recognition of point symbols on a map could be greatly increased through extracting and analyzing these four feature parameters.