Human Vision Model for Advanced Autonomous Seekers

WaveBand Corporation proposes to develop a novel, biological vision model that improves autonomous seeker performance by performing multi-resolution feature extraction at the sensor level, and integrating these extracted features into a consistent representation of the combat scene using a parallel attention model of the Human Visual Systems (HVS). The method, which incorporates our present understanding of the early stages of the HVS, was specifically designed to encode the optimal sets of target features that best characterize any particular class while creating a well separable recognition space that is robust to large variations in viewing/imaging conditions and noise. Specifically, it employs a novel feature extraction process (including light adaptation mechanism) integrated into a multi-resolution attention scheme. The subsequent recognition systems uses parallel visual streams to implement evidence accumulation, pattern completion and recognition, and temporal processing. Based on the proposed biological vision model Waveband will build a robust ATR engine using simulated target data or data provided by the sponsor. The Phase I & II efforts will focus on developing a test bed prototype which will be used to evaluate the system's performance in the context of autonomous seeker technology and real-world target recognition scenarios.