Hyperspectral single-pixel imaging from visible to short-wave infrared

Hyperspectral imaging enables simultaneous acquisition of spatial and spectral information and plays a key role in material identification and compositional analysis. However, existing camera-based hyperspectral imaging systems are constrained by detector materials and spectral acquisition schemes, making it challenging to achieve broadband coverage and high spectral resolution. This limitation restricts their application in scenarios requiring high spectral accuracy and broadband spectral analysis. In this work, we propose a hyperspectral single-pixel imaging approach that realizes high spatial and spectral resolution imaging from 371 to 1725 nm. The proposed method takes advantage of the complementary angular reflection states of a digital micromirror device to partition the modulated visible and short-wave infrared light into two physically separated optical channels, each coupled to an independent spectral detection module. The system achieves spectral resolutions of 1.1 nm in the visible band and 3.3 nm in the short-wave infrared band and reconstructs a 128 × 128 × 770 hyperspectral data cube. Moreover, acceptable reconstruction is achieved at sampling ratios as low as 10% using compressive sensing. Imaging experiments on various samples validate the performance of the system and demonstrate its potential for industrial inspection and biomedical analysis.