A Hyperspectral Architecture for Fast Anomaly Recognition

Hyperspectral and multispectral imaging sensors have enormous potential for military reconnaissance, surveillance and targeting missions. However, rapid detection of spectral, spatial, and/or temporal anomalies, which represent potential targets, is not possible with current sensor designs and algorithms, due to limitations in optical throughput, time-consuming data reduction requirements, or both. Spectral Sciences, Inc. (SSI) hereby proposes to address these needs through the integration of a novel multiplexed spectral imager, currently under development, with very fast anomaly detection algorithms that take advantage of the instrument’s unique characteristics. The instrument provides strong advantages over other multiplex systems by virtue of its robust mechanical design (no macroscopic moving parts), maintenance of spatial registration across wavelengths, on-the-fly programmable wavelength resolution, very fast computation of spectra, and, due to the mathematical properties of the multiplexed images, the ability to perform anomaly detection with little or no pre-processing of the raw data. In Phase I we will develop and prototype a software architecture for instrument control and data reduction. In Phase 2, the software will be implemented for the developmental sensor, with the goal of anomaly detection on a ~1 second time scale.