Laboratory demonstrator of a super-resolved, compressive sensing-based imager working in the VIS-NIR-MIR

In the framework of the H2020 SURPRISE project (SUper-Resolved comPRessive InStrument in the visible and medium infrared for Earth observation applications), a demonstrator of a super-resolved, Compressive-Sensing (CS)-based instrument operating in whiskbroom mode has been designed and assembled in the laboratory. The demonstrator relies on the concept of the single pixel camera, in which a sequence of acquisitions of the observed target is spatially coded and integrated on a single element detector. While the acquisition of an entire sequence of coded measurements allows the reconstruction of the target image at the same spatial sampling of the coding mask, the adoption of the CS paradigm optimises the target image reconstruction with a limited number of measurements. A scanning system is used to implement the whiskbroom operation mode of the demonstrator for the acquisition of an entire scene, by acquiring measurements on subsequent target areas. The demonstrator has 10 spectral channels in the VIS-NIR and two channels in the MIR. The super-resolution factor, that is the ratio between the number of pixels of the reconstructed image and the number of pixels of the detector, can be adjusted from 4 x 4 to 32 x 32. The instrument uses one fore-optics based on a custom Newtonian telescope focusing the observed image on a Digital Mirror Device which implements the coding mask. The coded image reflected by the DMD is spectrally split by a system of dichroic mirrors towards, on one hand, a spectrometer in the VIS-NIR wavelength range and, on the other, the two MIR channels for the simultaneous acquisition of the observed target in the VIS-NIR-MIR bands. Another optical path in the demonstrator was exploited to integrate a high-resolution camera that was used for optical alignment and system setting. The DMD used in the demonstrator was a DLP®7000, manufactured by Texas Instruments Inc.. The DMD was reworked in order to replace the glass window to gain transparency in both VIS-NIR and MIR bands. The acquisition of an entire scene, composed of multiple target areas, is performed by using a target scanning system. The target image reconstruction is obtained by using ad-hoc CS reconstruction algorithms. Here we present and discuss the results of the measurements, acquired in the laboratory. To the authors' best knowledge, the results are the first examples of super-resolved CS acquisitions in both VIS-NIR and MIR spectral range and have potential for future development of novel remote sensing instruments with enhanced capabilities in terms of target spatial sampling and native compression.