Plasmonic Superpixel Sensor for Compressive Spectral Sensing

In multispectral and hyperspectral sensing, there is a growing need for a versatile sensor capable of adapting response and improving detection of hard-to-find dynamic targets of interest in contested environments. Such on-the-fly adaptivity in current systems requires significant data resources and computation time for data analysis. In order to implement practical systems with this capability, sensors that reduce data loads and computational requirements while maintaining performance are required. To this end, we report a novel hybrid algorithm sensor method using plasmon-based tunable superpixels and a compressive spectral sensing (CSS) algorithm for the next generation of hyperspectral sensors. The benefit of our hybrid approach is that it enables us to effectively sense a minimal data set and only performs simple arithmetic such as linear superposition to extract spectral features of a target without requiring actual spectral filters. In this paper, we focus on the selection of a minimum basis of plasmonic spectral bands, the configuration of superpixels using selected plasmonic structures, and finally the generalization of a CSS algorithm to process superpixel data for feature extractions. The performance of algorithm-driven superpixels has been successfully demonstrated with the context of reconstructing infrared spectral signatures.