Algorithms for Multilook Effluent Detection and Identification

78844S Advanced long wavelength infrared (LWIR) hyperspectral-imaging systems are used for the remote detection and identification of gaseous chemical effluents, as a critical technology for monitoring suspected nuclear, biological, or chemical weapons manufacturing sites. Modern sensors have the capability to make repeated observations over a range of view angles (from down looking to long slant-path) and over a range of time intervals (from seconds to months). However, in order to improve plume target detection, identification and false alarm rejection, advanced algorithms are needed to efficiently exploit multiple target observations. This project will develop novel algorithms for exploiting multilook and multi-sensor LWIR hyperspectral imaging data for improved detection, identification, and quantification of chemical plumes. Phase I will apply geo-location and scene registration methods, combined with change normalization, to improve plume detection. Multiple target observations will be used to develop better plume background models for improved identification and quantification capabilities. Performance gains will be quantified and compared to traditional single-look analysis methods. Commercial Applications and Other Benefits as described by the awardee: The new algorithms should permit effluent plume observations from multiple hyperspectral images. Potential users include analysts who interpret LWIR hyperspectral imagery, especially those without advanced scientific training who need to rely on robust and highly automated analysis tools. Such non-expert users may include military image analysts in the field performing time-critical tasks such as WMD threat location or bomb damage assessment, or civilian agencies that require early warning of the presence and concentrations of dangerous chemicals.