Researchers from AIRI presented a method of hyperspectral shooting through RGB cameras

A group of scientists from AIRI, Samara University and MIPT has developed a technology that allows converting a standard RGB camera into a hyperspectral system. The solution was presented in the scientific paper Diffract and Conquer at the IJCAI 2026 conference.

Instead of expensive equipment, researchers proposed using a special plate with a set of harmonic lenses, which is installed in place of the standard lens. The method allows obtaining hyperspectral images containing **31 spectral channels** per pixel instead of the standard three, in a single exposure.

To restore data from the optically encoded signal, a neural network model is applied. The full text of the study is available in the preprint of the article. The technology allows for spectral analysis of objects in real-time, which previously required specialized cameras costing millions of rubles.

The development is aimed at applied problems, such as pigment identification, determination of fruit ripeness or analysis of material composition. The use of diffractive optics in combination with machine learning algorithms significantly lowers the entry threshold for the implementation of hyperspectral imaging in industry and scientific research.