34.15. RSP: Sensor Response

DIRSIG allows the user to supply a sensor response function that is to be applied to the sensor reaching radiance. The radiance or radiance vector resulting from the application of the response function dictates what type of image is generated for the bandpass defined for each band. The sensor responses and resulting output images can be generalized into three (3) forms:

The type of image that is generated is dictated by the file assigned to the RESPONSE_FILENAME variable for the given band. This filename may have been the IDEAL keyword (resulting in an integrated broad-band Image using an ideal/unity response), the SPECTRAL keyword (resulting in a spectral radiance image) or the name of a RSP format sensor response file. The following table describes each of the variables that may appear in a DIRSIG .rsp file.

Table 34-6. List of the variables that can be set in file.

Variable Name Usage Description
DIRSIG_RSP Required File identification marker (must be first non-comment line)
BAND Required section to define the extents of the response function
TYPE Required Type of response function (INTEGRATED or SPECTROMETER )
SHAPE Spectrometers Only Shape of the channel response (GAUSSIAN or TRIANGULAR )
GAIN Optional A gain to be applied to the radiance spectrum
BIAS Optional A bias to be applied to the radiance spectrum
RESPONSE Required Response data (spectral function or channel widths)
Each file must contain a section to define the extents of the response function. This spectral bandpass is the default for the that this response file is assigned to. This can be overridden by assigning values to either the WAVELENGTH or FREQUENCY variables in the file. The spectral bandpasses (specified by the respective MINIMUM and MAXIMUM variables) and spectral resolution (specified by the respective DELTA variable) only define the extents and resolution that the radiometric calculations will be performed at for this band, and does not place any restrictions on the extents or resolution of the response data being supplied. However, since the response data will be interpolated to the spectral centers defined by the bandpass, the bandpass and resolution should be appropriately chosen so that the response function is not degraded. The GAIN and BIAS variables allow the user to supply a linear amplification to the sensor reaching radiance spectrum. A common use for the GAIN is to perform a unit conversion from the radiance units of [W/(cm^2~sr~ m)] to [W/(m^2~sr~ m)]. If the GAIN or BIAS variables are not set, the default values are 1.0 and 0.0 respectively. The TYPE variable specifies that either an integrated band or series of spectrometer channels will be supplied in the RESPONSE section of the file. More detailed descriptions of these two types are provided in the following sections.

34.15.1. Integrated Broad-band Image

Although ``broad-band'' might imply that there is a width constraint on this output option there is not. The ``broad-band'' image option simply implies that the spectral range specified for the band given is to be output as a single broad-band integrated radiance value for each pixel. The spectral response file in this case contains the spectral response of the sensor for the bandpass region modeled. The sensor reaching radiance spectrum is multiplied by this spectral response before the spectral radiance is integrated. The units for a broad-band radiance image is therefore [W/(cm^2~sr)]. This type of sensor response is specified by setting the TYPE variable to the value INTEGRATED . For this case, the RESPONSE section will contain a series of wavelength/response or frequency/response pairs depending on what spectral basis was used in the section. The following example illustrates the RSP file for the SPOT Pan-Visible band used in our example CFG file:

DIRSIG_RSP
#
# NAME:     SPOT Pan-Visible Channel
# PURPOSE:  Spectral response function for SPOT Pan Channel
#
BAND {
    NAME = SPOT Pan-Visible
    MINIMUM_WAVELENGTH = 0.470
    MAXIMUM_WAVELENGTH = 0.800
    DELTA_WAVELENGTH = 0.010
}

TYPE = INTEGRATED

RESPONSE {
    0.4780 0.0000
    0.4800 0.0050
    0.4900 0.1140
    .
    . [middle wavelengths deleted for documentation purposes]
    .
    0.7800 0.0040
    0.7900 0.0020
    0.8000 0.0000
}
      
There are no assumptions about the spectral spacing or number of response pairs in the RESPONSE section except that the wavelengths or frequencies be in increasing magnitude. If the wavelength or frequencies are found to be out of order, an error will be issued.

34.15.2. Spectrometer Image

To make it easier to model imaging spectrometers, the user can input the center wavelength/frequency and wavelength/frequency full width at half max (FWHM) of each spectral channel in the instrument. The bandpass for this type of response file must cover the extents of the spectral channels. This type of sensor response is specified by setting the TYPE variable to the value SPECTROMETER . For this case, the RESPONSE section will contain a series of channel center wavelength/frequency and wavelength/frequency FWHM pairs. In addition, the SHAPE variable must also be set to define the shape of the channel response function (either GAUSSIAN or TRIANGULAR ). The example below is a condensed version of the HYDICE response file that can be found in the default sensor directory:

DIRSIG_RSP
#
# NAME:     HYDICE
# PURPOSE:  Channel response functions for the HYDICE imaging spectrometer
#
BAND {
NAME = HYDICE Spectrometer
    MINIMUM_WAVELENGTH = 0.470
    MAXIMUM_WAVELENGTH = 0.800
    DELTA_WAVELENGTH = 0.010
}

TYPE = SPECTROMETER
SHAPE = GAUSSIAN

RESPONSE {
    0.394214 0.002225
    0.397510 0.004450
    0.400806 0.004451
    .
    . [middle wavelengths deleted for documentation purposes]
    .
    2.471950 0.011165
    2.480210 0.011124
    2.488430 0.005485
}
      
The channel center/width pairs must be in the same units as the bandpass defined in the section. The extents of the bandpass in this example were chosen to extend well past the extents of the first and last channel in the spectrometer. The spectral resolution (see DELTA_WAVELENGTH ) for this example assures that at least 5 spectral points will be utilized in each channel computation. The computed radiance spectrum for each pixel is then convolved and integrated for each channel specified to produce a pixel spectra with radiometric units [W/(cm^2~sr~ m)]. The pixel spectral are written to the corresponding output image file using a band interleaved-by-pixel (BIP) format. Each pixel spectra consists of N single-precision floating point (float ) numbers (where N is the number of channels in the spectrometer). The dimensions of the output image is therefore X width x Y width x Number of channels. In addition, an compatible header file is created that contains an description of the image format for those attempting to read the imagery into other analysis tools.

34.15.3. Spectral Radiance Image

In some cases the user may wish to avoid any spectral integration effects. Therefore, DIRSIG provides a way to output the high spectral resolution radiance spectrum directly to an output file. This feature is used by setting the response filename to SPECTRAL in the CFG file. The pixel spectrum are written to the corresponding output image file using a band interleaved-by-pixel (BIP) format. Each pixel spectra consists of N single-precision floating point (float ) numbers (where N is the number of points in the spectral region). The bands in the output image are ordered in increasing wavelength, or decreasing frequency. The units for each point in the radiance spectrum is W/(cm^2 sr um) (for both wavelength and frequency based bands). The dimensions of the output image is therefore X width x Y width x Number of spectral points.