Adaptive Hyperspectral Imager – ImHypAd

The ImHypAd project (Adaptive Hyperspectral Imager) will develop a new generation of hyperspectral imagers, with algorithms and strategies for the acquisition and processing of associated hyperspectral data. This new system is very flexible, offering the possibility of numerous acquisition schemes under software control, including adaptive acquisition schemes, that allow the near real-time acquisition of all or part of the hyperspectral information present in a scene or an observed object. This adaptiveness departs from existing hyperspectral data acquisition principles.

A first instrumental proof of concept and a working prototype have been demonstrated by the LAAS. The configuration of these devices can be modified in real time by controlling a matrix of micro-mirrors (Digital Mirror Device - DMD), acting as multiple slits whose width and position are fully controllable. This allows to control the spectral and spatial resolutions of the acquired data. Rather than the traditional scanning of the whole hyperspectral data cube (two spatial dimensions and one spectral dimension), the control of the device makes it possible to focus on spatial and / or spectral region of interest, with variable spectral and spatial resolutions.

The main objective of this project is to explore and develop data processing algorithms several and acquisition schemes to exploit the specificity of this device. These algorithms and acquisition schemes can be classified by the kind of information required on the object exhaustive (reconstruction of the whole hyperspectral cube) or task-driven (only a part of the information is relevant) and by the way the DMD is controlled: a priori (predefined pattern), random (random pattern), adaptive (the pattern is dynamically controlled in a closed-loop acquisition process).
This project relies on a challenging co-design approach where both accurate device modeling, control of the DMD, data acquisition and processing of the acquired data are interleaved. It will open the possibilities for new applications of hyperspectral sensing, with constrained computation or communication resources.