Hyperspectral Imagery

Multispectral imaging measures light in a small number (typically 3 to 15) of spectral bands.  Landsat or Sentinel 2 are good examples.

Hyperspectral imaging is a special case of spectral imaging where often hundreds of contiguous spectral bands are available.  The bands have a much smaller spectral bandwith (many shades each of red, green, and blue, as well as the NIR), allowing very precise determination of reflectance spectra and the possibiliy to surface material determination.

Hyperspectral imagery is particularly suited for:


Hyperspectral image cube.  You a selected single band or and RGB merge on the top, and see the image in all  bands (1 on top, highest in the sensor on the bottom) for front and right side.

You can see:

  • Bands that are high/low reflectance in all bands.  High reflectance is white, and low reflectance black.
  • Bands that are missing or distorted. 
    • If the band is all black, is is likely missing (two prominent water absorption bands occur in the limits for many hyperspectral sensors).
    • If it is speckled white with some black, it is likely very noisy, perhaps at the edge of an absorption band.
    • Typically the sensor actually has two different detectors, since designing one sensor to run the entire range of waverlengths is too hard, and in that case, they sensors may not overlap well, leading to bands that are not collected
  • Individual pixels that are high/low in reflectance in some or all bands, which have the same shade of gray top to bottom.

Hyperion is a good example of a hyperspectral imager, and a reasonable amount of archived data is available.

Last revision 12/22/2017