Algorithms for Multilook Effluent Detection and Identification

Hyperspectral imaging systems are used for the remote detection and identification of gaseous chemical effluents, which is needed for monitoring suspected nuclear, biological, or chemical weapons manufacturing sites. This project will develop automated software, which uses a time series of remote target observations from infrared hyperspectral imaging sensors, to dramatically improve the detection and identification of chemicals in effluent plumes. In Phase I, optimal clutter suppression, based on co-registered multi-look images, yielded gains in plume detection sensitivity on the order of 100% for many industrial chemicals. Multi-look methods of chemical identification realized an initial 10% improvement in false alarm reduction, and the potential for much greater improvement was indicated. Phase II will: (1) refine and optimize the multi-look algorithms, (2) implement the software prototype, and (3) demonstrate prototype performance using airborne and ground-based sensor data from target sites with simulated proliferant activities. Commercial Applications and other Benefits as described by the awardee: The software package for detecting and quantifying effluent chemicals from multi-temporal hyperspectral imagery should be of interest to: intelligence image analysts performing time-critical tasks such as WMD threat location or bomb damage assessment, military units concerned with tactical chemical threat warning, or civilian agencies requiring warning of the presence of dangerous chemicals