Principle of Snapshot Multispectral Imaging: Single‑Frame Simultaneous Capture of Spectra & Images to Reconstruct Material “Optical Fingerprints”

Traditional spectral imaging mostly adopts push‑broom or point‑scanning schemes that collect spectra line‑by‑line via mechanical structures, featuring low speed and large size, unsuitable for dynamic devices such as high‑speed assembly lines, smartphones and automotive terminals.Snapshot Multispectral Imaging, the mainstream global commercial solution, integrates micro optical filters directly into the CMOS pixel layer, enabling “one pixel for one band”. High‑definition spatial images and complete spectral datasets are output simultaneously in one single‑frame exposure.

Technical implementation consists of three key steps. First,Pixel‑Level Filter Array: narrow‑band filters are deposited during wafer fabrication to divide 4–40+ visible‑to‑near‑infrared bands with a typical 10nm interval, matching characteristic absorption peaks of molecular chemical bonds of different materials. Second,Back‑Illuminated (BSI) Pixel Architectureimproves photon utilization under low‑light conditions, reduces noise and ensures high‑speed imaging sensitivity. Third,AI‑powered Spectral Unmixing Algorithmdecomposes mixed spectra into pure material spectral features, eliminates ambient light interference, and accurately identifies materials, compositions and defects. Compared with RGB imaging with only 3 bands, multispectral imaging with 40+ bands captures micro material differences invisible to human eyes to realize “same‑color‑different‑material” identification. Compared with hyperspectral imaging, snapshot technology features simple structure, controllable cost and higher frame rate, making it the optimal route for large‑scale commercialization. It is widely applied in industrial quality inspection, mobile health detection, automotive harsh‑weather perception and food pesticide residue screening.