We propose a hybrid filter, which we call the hybrid optical neural network (HONN) filter. This filter combines the optical implementation and shift invariance of correlator-type filters with the nonlinear superposition capabilities of artificial neural network methods. The filter demonstrates good performance in maintaining high-quality correlation responses and resistance to clutter to nontraining in-class images at orientations intermediate to the training set poses. We present the design and implementation of the HONN filter architecture and assess its object recognition performance in clutter.
The article details a non-linear extension to existing linear filters for explicit optical implementation within our coherent optical pattern recognition hardware. The research demonstrated improved intra-class tolerance of the non-linear filters, as compared to the linear counterparts, whilst maintaining interclass discrimination and tolerance to heavy background clutter. The non-linear filters developed in this research have been incorporated into our hybrid digital/optical correlator hardware architecture and in this have demonstrated improved performance over existing linear filters. The NASA Jet Propulsion Laboratory, Pasadena, California Institute of Technology, is interested in our correlator hardware and filter research (contact: Tien-Hsin.Chao@jpl.nasa.gov,Tel: 001-818-354-8614)