Modified Rapid Transform Features in Information Symbols Recognition system

摘要

Various transformations have been suggested as a solution of the problem of high dimensionality of the feature vector and long computation time. Transforms which do not change with cyclic shifts in the sequence are called translation invariant. Fast translation invariant transforms are valuable tool for pure shape-specific feature extraction in pattern recognition problems (1). In the field of pattern recognition and also scene analysis is well knwon the class of fast translation invariant transforms-certain transforms (CT) (2) based on the original rapid transform (RT) (#). More recently was introduced the modified rapid transform (MRT) (5) which can distinguish many more patterns from one another that the original RT can. The MRT was presented to break undesired invariances of the RT which leads to a loss of information about the original pattern. In the paper application of fast translation invariant modified rapid transform (MRT) in feature extraction stage of Information Synbols recognition system are described. Experimental results are given of applying the proposed recognition system to recognition Airport Passenger Orientation Symbols and Meteorological Symbols, including the dependence of recognition efficiency on the number of selected features and noise.