FOURIER-PLANE HOLOGRAPHIC ASSOCIATIVE FILTER BASED COMPACT OPTICAL CORRELATOR FOR TARGET RECOGNITION

Physical Optics Corporation (POC) proposes to apply its Fourier-plan Holographic Associative Filter (FPHAF) based optical correlator to high speed target identification. The proposed correlator eliminates the spatial light modulator (SLM) from the Fourier plan of an optical correlator by using neural network models to train a set of filters and multiplex them onto a volume holographic plate, to simultaneously correlate the input object with multiple filters. The system can be compactly packaged in a rugged and light box suitable for airborne operations. In Phase I, such pattern recognition algorithms as synthetic discriminant functions (SDF's), circular harmonic filters (CHFs) and moving target indication (MTI) techniques will be examined. A proper combination of the filters for shift-scale-rotation and projection invariant target recognition will be selected for neural network training and feature extraction. Neural network models such as the Inter-Pattern Association (IPA) and backpropagation models will be studied for use in the construction of the associative filters. Experimental results of the proposed correlator will be demonstrated. Software and hardware engineering design of a prototype FPHAF based compact optical filter will be provided for Phase II implementation of the system.