Every rock on a planetary surface looks the same to an RGB camera. Hyperspectral imaging sees beyond colour — encoding what a surface is made of, not how it appears.

Your phone camera captures three channels — red (~630–700 nm), green (~520–560 nm), and blue (~450–490 nm). These correspond to the three cone types in the human eye. Conventional cameras are therefore designed to sense only the narrow slice of the electromagnetic spectrum that humans perceive as visible light.

But objects, surfaces, and materials reflect and absorb light across a much wider range of wavelengths. Extend sensor measurements beyond the three RGB channels into the broader electromagnetic spectrum and you arrive at Spectral Imaging.

Hyperspectral imaging (HSI) captures tens to hundreds of narrow, contiguous spectral bands per pixel — producing a complete reflectance spectrum at every spatial location. This spectrum is a material fingerprint: stable across illumination changes, invariant to viewpoint, and encoding chemical composition rather than visual appearance. The data is stored as an image cube (Height × Width × Bands) — a 3D tensor where each pixel is its own spectrum.

Look at the chart above in the amber-shaded region (600–860 nm) — the range of the HyperLoop sensor (Ximea IMEC camera, 15 bands). Every mineral curve converges to a similar reflectance level. The terrain is spectrally flat: minerals that are chemically completely different look almost identical to the sensor.

Now look at the teal-shaded region (900–2500 nm) — the SWIR diagnostic window. The curves diverge dramatically. Orthopyroxene plunges to near-zero at 920 nm (Band I) and 1850 nm (Band II). Phyllosilicate shows sharp notches at 1400, 1900, and 2200 nm. Jarosite has its own distinctive dip at 2200 nm. Each mineral becomes uniquely identifiable.

The HyperLoop framework demonstrated that integrating the Ximea IMEC Red-NIR sensor (600–860 nm) into SLAM loop closure provides limited additional discriminability for planetary analog terrain. The sensor was designed for agricultural vegetation analysis — a domain where Red-NIR reflectance differences encode chlorophyll content and plant health.

For rocks and regolith, the diagnostic mineral absorption features only appear in SWIR (900–2500 nm). The framework is ready. The spectral pipeline works. The field needs the right sensor.