Invariant pattern recognition: smart algorithms for segmented two-dimensional patterns with an axis of symmetry

We propose an approach to realize distortion-invariant pattern recognition for segmented 2-D patterns with an axis of symmetry. Parameters of distorte d input pattern are first estimated and normalized. Invariant pattern recog nition is then achieved by correlating the normalized input pattern with th e undistorted reference, using a classical matched filter or its modified v ersion. This approach realizes pattern recognition invariant to rotation, s cale, and projection without the need to synthesize a distortion-invariant filter. Simulation results demonstrate the effectiveness of this approach. (C) 1999 Society of Photo-Optical instrumentation Engineers. [S0091-3286(99 )01012-0].