The proposed project will explore the potential of neural network algorithms to yield significant performance improvements needed for future automatic target recognition processors. Neural processing opens new possibilities for automatic target recognition by offering the potential for compact. Massively parallel computational hardware and increased flexibility in accorrinodating new targets and environments. The proposed project will result in techniques for template creation, methodologies for pretraining neural networks to incorporate a priori information, and approaches to feature extraction using target signature compression. The effectiveness of the algorithms resulting from the proposed study will be demonstrated using synthetic aperture radar and forward looking infrared sensor data provided by the government. The proposed project will identify template generation, signal decomposition. and data compression algorithms that can potentially improve both processing speed and classification accuracy in ATR applications. The project will provide a demonstration of new and enhanced AIR algorithms using synthetic aperture radar and infrared images provided by the government. There are numerous commercial applications of the algorithms that will emerge.
