4 Airborne Measurements and Preprocessing of the Data
4.3 Geometric Rectification
For the radiometric examination of the data, geometric rectification is not mandatory but helpful for the orientation within the image as well as for the presentation of the data. Sources of geometric distortions of particular concern to airborne sensor data are related to an unstable platform, low acquisition altitude and large FOV causing changes in pixel size as well as pixel displacements.
Low altitude acquisition combined with the large FOV causes pixels in the nadir line to be smaller than pixels on the edges of the swath due to the panoramic effect (equation (4.3), Richards, 1998):
θ
θ
β⋅
⋅
⋅
=
⋅
⋅
=
2 2 0h
sec
p
sec
p
(4.3) wherep0 across-track pixel size [m],
θ
scan angle [°],β
instantaneous field of view (IFOV) [rad],h altitude above ground [m],
p pixel size at nadir [m].
The largest AVIS view angle is 34° (FOV/2), which results in p0=1.45. This indicates that the geometric distortion at the edge of the swath due to the panoramic effect is 45%. The altitude is given with 600 and 1200m above ground (which corresponds to 4000 and 10000ft asl). This caused the nadir pixel size of 1.86 and 3.72m to increase to 2.49 and 5.39m respectively.
The displacement of pixels is further influenced by variations in platform velocity, altitude and attitude due to factors such as atmospheric turbulence. An increase in altitude will change the geometry of the pixel as illustrated in Figure 4-19a. Variation in velocity changes the along-track pixel size (Figure 4-19b), attitude variation in the form of roll, pitch and yaw results in along track displacement, across-track displacement and scan line rotation (Figure 4-19c, d, e).
Because of their small frames, fast frame acquisition times, low altitude and narrow swath width, airborne sensors do not cause problems in connection with earth rotation and earth curvature as do satellite based sensors.
The aircraft motions can be fully retrieved by using a dGPS and an inertial navigation system. The former provides the geographical position, velocity and altitude and the latter the roll, pitch and yaw.
At the time of this thesis only dGPS data were available for the geometric correction of the AVIS data. Therefore the geographical position and major variations of the flight path could be corrected. Figure 4-20 shows an example of the kind of geometric correction that is possible with the available dGPS data. The selection of bands corresponds to a false colour composite. The image capture during flight is adapted to a specific velocity of the aircraft, namely 180km/h, but the real velocity lies between 180 and 195km/h. Therefore the actual along track pixel spacing is different from the nominal across-track pixel size and varies in flight direction. Thus the surface looks compressed in along-track direction. The angular motion of the aircraft is clearly visible at roads and borders between fields. The dGPS data provide information about geographical position, altitude and velocity of the aircraft (Figure 4-20 centre). The former enables the alignment to north while the latter are used to calculate square pixels. The dGPS data are stored in the header of each image line and enable the correction of each pixel within that line.
The geometrically corrected image stripe is aligned to geographic north and has a square pixel spacing, which is consistent over the whole image. In this case the resulting pixel spacing is 3x3m.The angular aircraft motion could not be corrected and is still noticeable. The remaining distortions in the flight stripes can only be corrected with the aircraft angular motion data and a topographic map. This distortion correction by ground points is very time-consuming and was carried out for just a few flight stripes. The results can be combined to form a mosaic, which is shown in Figure 4-21 for the test site Gilching.
Figure 4-19: Image distortions due to altitude (a), velocity (b), pitch (c), roll (d) and yaw (e) (Richards, 1998)
The rectified scenes can be used for the examination of some system parameters such as the FOV, which can be derived using equation (4.4).
⋅ = − h D FOVreal 2 sin 2 1 (4.4) where
FOVreal Field of view derived in the field [°],
D across-track distance covered on the ground [m],
h flight altitude above ground [m].
The altitude above ground can be derived from flight height above sea level, which is measured with the dGPS and stored in the image header. The ground level as well as the across track distance on ground can be derived from a topographic map or from the DEM.
The FOVreal was calculated for several AVIS scenes and the results were in the range of
66° to 68°. Therefore the specifications of the contractor could be confirmed.
Figure 4-20: Geometric correction of preprocessed AVIS data (left) with dGPS data (geographical position, flight altitude as well as velocity of the aircraft) and resulting geometrically corrected image (right) (R/G/B = 582/674/729nm), acquired on 19th July 1999
Figure 4-21: Mosaic of the test site Gilching assembled using flight strips acquired on 3rd and 10th May