Optical pattern recognition using correlation procedure for rotation- and scale-invariant feature extraction

Abstract

An approach to hybrid optical digital pattern recognition is proposed which uses a correlation procedure for the initial image comparison with the results of its geometrical transformation. It is shown that the obtained description of the object to be recognized provides the invariance with respect to the object scale and orientation and that it contains the information about the object location in the image being analyzed. Results of computer simulation of the correlation procedure are presented along with its experimental implementation in an optical electronic correlator including a liquid-crystal space light modulator in the Fourier plane of the optical system.< >