on 
'se 
ng 
nd 
ral 
or 
a 
pe 
nd 
he 
ce 
re 
nt 
rs. 
on 
er 
ce 
of 
al 
de Jong, Steven 
  
3 DAIS IMAGERY AND PROCESSING 
3.1 The DAIS 7915 instrument 
The Digital Airborne Imaging Spectrometer (DAIS7915) is a 79 channel, high-resolution spectrometer. The DAIS is an 
experimental scanner, financed by the Furopean Union and DLR, and built by the Geophysical Environmental Research 
Corporation (GER). The 79 spectral bands cover the range from the visible- to the thermal wavelengths. The scanner can be 
mounted on the Dornier DO 228 aircraft. Table 1 gives the specifications for the different spectrometers. Six spectral bands 
between 8 en 12 jum are used for the measurement of the radiant temperature of land surfaces. Besides the high spectral 
resolution, the DAIS 7915 also has a very high spatial resolution. At a flight altitude of approximately 3000 m. the spatial 
resolution of the recorded image is 6 by 6 meter. 
Before processing, the 79 bands were used for a visual inspection of image quality. Image quality is generally good in 
the visible and near infrared bands. Striping occurred in most of the SWIR bands. The striping in some SWIR bands is 
probably caused by vibration of the Dornier DO228 prop-engines (Strobl, 1996b). 
3.2 Georeferencing of the DAIS imagery 
' The Peyne flight strip was flown from the north-west to the south-east. The first step in the georeferencing of the image 
is a rotation of the flight strip to the geographical north. The flight strip is rotated over 137 degrees with the cubic 
convolution resampling method. A spatial subset of the flight strip is taken to reduce the size of the image. This spatial 
subset corresponds with the size of the study area. Fifty DGPS measurements were taken in order to correct for the 
distortions in the image. The RMS-error of the image, caused by the distortions, comes to 7.856 pixels. Transformation 
with a second-degree polynomial with the cubic convolution resampling technique gave the best results for the Peyne 
flight strip. Most of the large-scale distortions are corrected, but some of the small-scale distortions cannot be corrected 
because of the rough transformation algorithm. 
3.3 Processing of the optical DAIS bands 
In order to perform noise reduction, the Minimum Noise Fraction transformation (MNF) can be used. This 
transformation algorithm can be described as a cascaded Principal Component transformation (PCA). The MNF 
transformation decorrelates and rescales the noise in the data. The output of the MNF transformation is a given 
eigenvalue for each MNF band. When the eigenvalue of an MNF band is small (less than one), the image is noise 
dominated (Green, 1998). The MNF images with the largest eigenvalues can be transformed back to the original data 
space. The original data space is restored, however without the determined noise. 
To correct for atmospheric effects, the empirical line method is applied. This method uses field measurements to correct 
the atmospheric effects. Reflectance spectra of homogeneous targets measured in the field are compared to the cor- 
responding areas in the image. A linear regression function is computed for each spectral channel in the image. The 
regression has been calculated by fitting the regression line through the spectra. The advantage of this calibration 
technique is the removal of the solar irradiance curve and the atmospheric path radiance, because of the difference in 
path length through the atmosphere. The distance between the DAIS sensor and the ground is approximately 3000m, 
while the path length of the field spectrometer is only less than one meter (de Jong, 1998b). In order to obtain a good 
empirical line conversion function, reflectance spectra of dark and bright targets are used, like water, asphalt and soil. 
To classify the DAIS image into land cover classes, the Spectral Angle Mapper (SAM) Classification algorithm is used. 
This classification method is a supervised classification technique, which uses field derived spectral endmembers 
(Kruse et al., 1993). The SAM classification algorithm results in a continuous land cover map. 
3.4 Processing the thermal DAIS bands 
The TIR spectrometer on the DAIS scanner measures the radiant temperature, radiated from objects on the earth 
surface. The unit of the thermal DAIS bands is given in radiance in [mW/(cm*ster*um)’]. In order to study the thermal 
properties and behaviour of surfaces in the study area, the radiance was converted to absolute temperature in degrees 
Celsius of the various land cover types in two steps: the radiant temperature is converted to blackbody temperatures and 
next, the blackbody temperature is converted to absolute temperatures. 
In order to obtain a blackbody image from the six thermal DAIS bands of the 1998 DAIS flight, an empirical approach 
is applied. The following steps are taken: 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 349