Juergen Oberst
IRDETECTOR
ARRAY RADIAT OR
05-10
micron APS
28 cm
=; —l
3
=
B
SiC
GRATING
UV DETECTOR
m d : oe 30 cm -
Fig. 1: The MICAS camera (left) and the optical light path. Four sensors share one optical system (see text
and Table 1 for details).
Table 1: MICAS Optics First Order Properties
UV VIS(APS) VIS(CCD) IR
Wavelength, nm 80-185 500-1000 500-1000 1200-2400
Focal Length, cm 17.1 67.7 67.7 78.2
Detector Array 35 x 164 256 x 256 1024 x 1024 256 x 256
Pixel Size, 54 12.0 9.0 40
FOV,? 0.63 x 0.03 0.26 x 0.26 0.69 x 0.78 0.7 x 0.003
4 APPROACH
The apparent line/sample positions of the grid holes in MICAS images were measured. It was assumed that the holes
were all equally spaced and that images of the grid differed only by relative lateral shifts and small rotations. Best-fit
shift and rotation parameters for each grid, as well as the grid hole spacing were determined from the measurements
using iterative least-squares methods. This procedure was repeated for the 7 target positions. Fig. 3 (left) shows the
residuals (i.e. the difference between measured position of the grid holes and the position as predicted by the model), all
7 target positions added, after the fit. The residuals scatter by 0.65—1.04 pixels (1 6) for individual images, with peak
magnitudes of the residuals being up to 4 pixels near the image margins. This is well beyond measurement errors
(estimated to be « 1 pixel) and therefore suggests that significant image distortion must be present. From the known
precise grid hole spacings of the target (Table 2), we estimate that the focal length is f = 686.55 +/- 0.05 mm (mean of
all fits).
One may wonder if this pattern seen in the image is perhaps due to the grid's lack of perfection. Fortunately, the object
coordinates of the holes in the target had been explicitly measured in the collimator (Table 2), which allowed us (not
only to determine the absolute focal length of the camera but also) to verify if the target was suited for this type of
calibration approach. The object coordinates were projected to the image plane of the camera adopting the nominal
focal length. This resulted in sets of synthetic line/sample coordinates of the holes, which were subjected to the same
inversion procedure as described above. The residuals show random scatter with o — 0.101 pixels, which demonstrates
the high geometric quality of the grid, more than sufficient for our calibration approach. It also becomes clear that the
distortion pattern seen in the real grid images must be due to an effect caused by the MICAS optical system.
222 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B1. Amsterdam 2000.