SBIR-STTR Award

Compressive Sensing (CS) is an emerging field based on the realization that a small collection of nonadaptive linear projections of a compressible signal contain enough information for reconstruction and processing. Expanding on this emerging technology, this project will make significant contributions in the following ways to the problem of anamoly detection in hyperspectral imagery: (i) increase our capacity for hyperspectral image acquisition developing new imaging and spectroscopic systems; (ii) expand the frontiers of CS from signal recovery to new applications including knowledge learning including anamoly detection; and (iii) strengthen our ability for extracting knowledge from huge data sets such as hyperspectral images beyong current limits imposed by existing computational resources and methods. BENEFIT

Keywords:
Hyperspectral Image, Compressive Sensing, Spectroscopic Systems, Kernel Methods

We intend to use the emerging technology of compressed sensing and related new algorithms of information science to (1)increase our capacity for hyperspectral image acquisition beyond current capacity by developing new imaging and spectroscopic systems,(2) expand the frontiers from signal recovery to new applications, emphasizing knowledge learning including target and anamoly detection, and (3)strengthen our ability for extracting knowledge from huge data sets beyond current limits imposed by existing computation resources and methods.

Benefit:
Hyperspectral imaging is applied to environmental monitoring, camouflage detection, land mine detection, couterfeit currency and cannabis detection, to name a few. Potential customers include various DoD, police and other government agencies.

Keywords:
Hyperspectral, Compressed Sensing, Target Detection,Spectroscopic System, L1 Optimization, Knowledge