What is compressive hyperspectral imaging?
A plain-language explanation of hyperspectral imaging, compressive sensing and single pixel imaging, and how Photonics Open Projects combines them.
Hyperspectral imaging, in plain terms
Hyperspectral imaging captures many narrow, contiguous spectral bands at every point in a scene, instead of the three broad red, green and blue bands recorded by a standard camera. The result is a full reflectance spectrum for each pixel, a spectral signature that can reveal composition, ripeness, moisture content, disease or material identity long before any of it becomes visible to the naked eye.
Compressive sensing
A conventional hyperspectral camera has to record hundreds of spectral bands for every pixel in the scene, which multiplies the volume of raw data and the cost of the optics and detectors involved. Compressive sensing takes a different route: it acquires far fewer measurements than a conventional sensor would need, then reconstructs the full data cube algorithmically. This works because most real scenes are redundant, or sparse, in an appropriate mathematical basis, so a comparatively small number of well-chosen measurements is enough to reconstruct the complete signal. The payoff is lower hardware cost, faster acquisition and less raw data to move and store.
Single pixel imaging
Single pixel imaging is one practical way of implementing compressive sensing. Instead of an array of thousands or millions of detectors, it uses a single physical sensor, in our case a spectrometer, paired with a spatial light modulator, typically a standard video projector, that projects a sequence of structured light patterns onto the scene. For each pattern, the spectrometer measures the spectrum of the light reflected by the whole scene. After enough projections, a reconstruction algorithm combines all of these single-point spectral measurements to recover a full hyperspectral image, spatial map and spectral signature together.
How ONE-PIX and PRO-PIX apply these principles
ONE-PIX turns any Python-controllable spectrometer into a hyperspectral single pixel camera by pairing it with a standard video projector, an onboard Raspberry Pi for synchronisation, and open source reconstruction software, making the technique accessible to classrooms, fablabs and research labs. PRO-PIX packages the same compressive single pixel principle into an industrial, IP65-rated sensor with hardware and software reconfigurability, built for continuous use in precision agriculture, food processing, geology and health care.