Shape recognition using spectral features

The classification of planar shapes using spectral features is presented in this paper. The contour of a planar shape is represented by the magnitude and phase of radial vectors drawn from a centroid, and they are modeled by an autoregressive process. The spectral features are extracted from the least squares estimators of the model parameters, and planar shapes are classified by an artificial neural network. In the experiment with contours of aircraft, machine parts and handwritten numerals, more than 96 percent of the shapes are correctly classified. (C) 1998 Elsevier Science B.V. All rights reserved.