Hyperspectral S3 imager using liquid crystal polarized grating and its demonstration of material classification with deep learning

We have constructed a near-infrared hyperspectral S₃ imager consisting of a circularly polarized broadband light source, a liquid crystal polarization grating, and a commercial hyperspectral camera. The circular polarization diffraction efficiency of the polarization grating was over 99 percent at 1550 nm. This imager is capable of obtaining both hyperspectral circular polarization distributions and conventional hyperspectral images. Using the S₃ hyperspectral imager, we demonstrated the classification of plastic samples with a deep learning algorithm, which can be applied to waste classification in recycling plants.