Application of Hierarchical Memory Models to Automatic Target Recognition Modeling and Simulation

This SBIR Phase I project proposes a visual processing system for automated target recognition. Inspired by biological vision systems and hierarchical memory models, the proposed system is capable of learning hierarchical invariant features from unlabeled data that are independent of object labels. The model exploits these learned features to create hierarchical representations of target memories with the ability to learn from a small number of examples. The proposed biologically inspired solutions provide robust target recognition capabilities with missing data, view point changes and illumination variations. This Phase I effort will leverage Novateur Teams expertise in the areas of deep learning and convolutional neural network, top-down and bottom-up visual attention, development of computational memory models, exploitation of remote sensing imagery, and image and scene understanding.