Using neural networks to enhance the operation of optical correlators in image recognition applications

Much research has been devoted to the development of compact optical correlators, optimal correlation filters, and high frame rate optical modulators for applications in automatic pattern recognition. However, little has been done to automate the analysis of correlation plane imagery. Artificial neural networks process information in a distributed, highly parallel fashion, and they perform as well as humans in many pattern recognition tasks. It is theorized that a neural network might enhance the operation of an optical correlator by providing an effective means of automatically identifying auto-correlation peaks. This thesis investigates the use of a multi-layer feed-forward neural network in a hybrid automatic target recognition device.