Integrated ATR with Fuzzy IR Image Processing

The integrated automatic target recognition (ATR) design provides a promising approach to achieving high performance based on multisensor data fusion, processing methodology integration, and target/enviromental/contextual information incorporation. In this proposal, American GNC Corp. (AGNC) will take advantage of the tolerance of imprecision offered by fuzzy logic and apply it to IR image scene modeling and pattern matching. It will also investigate the integration of fuzzy logic and neural network to blend the approximate reasoning capability of the fuzzy logic and the adaptive learning feature of neural network in target recognition. Phase I will demonstrate the benefits of such a fuzzy system in the integrated ATR system for tactical applications such as cruise missile looking for relocatable targets. The main innovations of the proposed approach are: (1) Integration of methodologies in every level of automatic target recognition processing: (2) Organization of scene models and input image structures based on fuzzy membership functions and fuzzy restrictions to increase tolerance to imprecision: (3) Augmentation of robustness of scene models and matching process and reduction of sensitivity to image quality and preprocessing: (4) Scene model generation which takes into account individual characteristics, deviations in image acquisition, and noise effects present in the training images. In Phase II, the algorithms of scene modeling and image processing for expert system and integrated ATR system will be fully developed, tested, and documented. The validated algorithms will be reduced to integrated circuit scale chip sets.