Compressive hyperspectral imaging with free-form plasmonic filter arrays

Snap-shot hyperspectral imaging systems are commercially available from several manufacturers and typically use narrow-band filters exploiting Fabry-Perot resonances or thin-film interference effects to spectrally filter light reaching the detector. In limiting cases, particularly where line widths are narrow and spectral resolution is high, less than 5% of available light reaches the detector. Rejecting such a large portion of the visible spectrum reduces signal to noise ratio and requires longer integration times – limiting the frame rate in video rate applications. In recent years, however, a proliferation of work with compressive sensing (CS) algorithms has enabled the deconvolution of spectral data from a relatively small number of filtered measurements with overlapping bands of sensitivity. Simultaneously, new nanophotonic simulation tools and AI based design have allowed fast optimization of plasmonic filters with large structural degrees of freedom. ‘Free-form’ plasmonic filters are unconstrained by traditional metamaterial design shapes and may have unique broad-band transmissivities from the UV to the LWIR. Our team will combine both of these recent advances to design a free-form plasmonic filter based snap-shot hyperspectral imaging solution where spectral data is reconstructed from a set of sparse measurements using CS algorithms. The proposed approach uses broad-band filters that admit as much as 25x more light when compared with camera-integrated Fabry-Perot based filter arrays. Since spectral data is reconstructed in software using CS, the resolution is no longer constrained by the filter bandwidth and the entire hyperspectral data cube can be approximated. Finally, the free-form plasmonic filters can be fabricated in a single lithography step significantly reducing cost. We anticipate the solution presented in the Phase I will enable a new class of snap-shot hyperspectral imagers, capable of high refresh rate video, produced at a much lower cost than the current state-of-the-art and ideal for integration with a wide range of both manned and unmanned aircraft systems.