Basically, imaging spectroscopy, or hyperspectral imaging, is a
happy fusion of spectroscopy and imaging processing. Imaging
spectroscopy can be seen either as an extension of classical image
processing or an enhancement of classical spectroscopy.
The simplest case is using a black and white camera that captures the
grey-scale values of objects - yielding high surface resolution but no
spectral information. A color camera, on the other hand, with three
image sensors or a sensor with Bayer color filter, will deliver a
multi-spectral image with comparably high spatial resolution and three
relatively broad-band color channels of red, green and blue, yet with a
relatively low spectral resolution. Finally, a spectral imaging system -
it operates with just one sensor and a tunable narrow-band filter
placed in the optical path to select a frequency. Alternatively, it
functions as a so called push-broom scanner to perform, usually via
mechanical feed of test object or spectrometer, a line-based scan. For
every pixel in every line, the spectrum is captured and stored.
Both methods use significantly more
color channels. This is why they are called hyperspectral: they deliver
high spatial resolution and, at the same time, high spectral resolution.
The measured data of the X- and Y-coordinates and the radiation
components at certain frequencies are located in a three-dimensional
data space (cube).
Imaging spectroscopy has proved its
worth in the acquisition of geological parameters from the air or by
satellite in order, for example, to answer environment-related questions
such as relating to the water quality of lakes. On a microscopic scale,
the method can be applied to good advantage for multi-channel spectral
analysis of light-emitting semiconductors, or in bio-medical and
chemical sample analysis. Furthermore, hyperspectral imaging can be
perfectly suited for industrial process monitoring for
waste sorting,
fruit and vegetable inspection, moisture measurements, fat analysis, web inspection, … and many more applications.
