FUZZY-PATTERN RECOGNITION FOR AUTOMATIC DETECTION OF DIFFERENT TEETH SUBSTANCES

A method for characterizing the surface of a test object on the basis of its oscillatory behaviour is presented. The basic principle of the method is the excitation of oscillations in a small surface element on the elastic solid under investigation and the appropriate measurement of the reaction to this excitation. After calculation of the Fourier transform, a total of 400 magnitude values of the oscillation are avai lable as characteristic features for the classification of an unkown s urface. For testing the measuring system experiments were conducted on various test objects and the results of measurements performed on dif ferent tooth surface are presented. The automatic classification of to oth surface is important for the detection of the cementoenamel juncti on in parodontological practice and the soft removal of dental calculu s. Because in this application the spectra are subject to severe fluct uations by variations in the test object a fuzzy classifier with weigh ted distance function is employed for the automatic classification. Fo r solving the essential problem of selecting from a total of 400 spect ral features those which appear to be suited for the respective classi fication problem, a new feature selection algorithm is presented. The highly successful result of these experiments indicates possibilities of applying new diagnostic and therapeutic instruments in the field of periodontology.