Raw data that is output by an imaging spectrometer represents a quantity that is ideally related to incident pupil-plane spectral radiance as a function of spatial or angular position and spectral band. Pupil-plane spectral radiance is dependent on the material properties of the scene of interest. However, these data are integrated over an SRF that can be wavelength dependent, perturbed in spatial and spectral scale by optical distortions of the instrument, and also dependent on the effective optical transmission and stray radiance of the optical system as well as the responsivity and dark current of the FPA. Extracting quantitative spectral information about material properties from this raw data requires a method to determine the system response spectrally, spatially, and radiometrically. This method is referred to as spectrometer calibration. One theoretical construct for calibration is to precisely measure the characteristics of every system component and determine system response through an extensive model, such as that outlined in Chapter 7. This is generally complex and impractical. An alternate, more practical approach is to directly measure the full system response using sources with known characteristics. The basic methodology of this approach is described in this chapter.

10.1 Spectral Calibration

Spectral calibration refers to the determination of spectral band centers for all samples in a hyperspectral data cube. The case of a dispersive imaging spectrometer is treated first and involves characterizing the center wavelength for each element of a 2D FPA to correct any potential spatial-spectral distortion. To perform a thorough spectral calibration, the spectrometer can be illuminated by a monochromator that produces monochromatic illumination at controllable wavelengths. Theoretically, a monochromator can be stepped across an entire spectral range at a sufficiently fine spectral sampling such that the center wavelength at each detector pixel can be determined as the illuminating wavelength producing a maximum response. Such a thorough spectral calibration is sometimes impractical and generally unnecessary because the spectral-spatial distortion of an imaging spectrometer is often a somewhat well-behaved function.