00 
ving Smile 
ation 
  
00 
Showing 
libration 
  
00 
ion 
  
000 
ction 
before the 
and 8 show 
all of the 
alibration. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004 
  
The standard error of the control point movement in the 
adjustment was (0.4, 0.4, 0.1) meters (1 sigma). The. a 
posteriori standard error of the focal length estimate was 48 
micrometers (1 sigma). This equates to less than 0.1 pixels 
at the end of the linear array. The a posteriori standard error 
estimate for the roll and pitch camera alignments were 0.5 arc 
seconds. 
The completion of the camera calibration is the last step in 
the initial geometric calibration of OV-3. Various test site 
data used to validate the calibration is shown in the next 
section. 
5. GEOLOCATION ACCURACY RESULTS 
The validation of the calibration and the end-to-end system 
performance evaluation is done through a series of 
geolocation accuracy assessments. These assessments are 
evaluations of the system performance compared to exterior 
control data. 
This section shows the results of two types of geolocation 
accuracy assessment of OV-3 imagery: monodrop, and stereo 
comparisons. 
The check points used in the geolocation accuracy 
assessments consist of photo-identifiable ground survey 
points in sites around the world. The imagery was collected 
over a period of 5 months and demonstrates the stability of 
the calibration parameters. 
Generally, a geolocation accuracy assessment tends to show 
a small diameter cluster of differences for the points from a 
mono image or stereo pair. However, the mean of the cluster 
is not usually zero and represents a bias for the image or 
stereo pair. These biases change from image to image and are 
said to be random. Random bias scatter plots are shown in 
the figures below. 
5.1 Monodrop Comparison 
The monodrop comparison used imagery had not been 
adjusted through the inclusion of tie points or control 
points and represents the direct positioning capability of 
the system. To perform a monodrop comparison, the 
measured image coordinates were projected down to the 
elevation of the check point to determine the latitude and 
longitude observed by OV-3. The difference between the 
latitude and longitude observed by OV-3 and the check 
point are computed and aggregated to form test statistics 
such as CE. The imagery was collected in a variety of scan 
directions that included: north-south, south-north, east-west 
and west-east scan directions. In all, 72 images were used in 
this assessment. 
The monodrop CE histogram has a median value of 9 meters 
and a mean value of 10 meters. The 90-th percentile of the 
CE histogram is 18 meters. There is a significant skew to 
this CE sample, which is expected when considering CE as a 
random variable. There seems to be some variation as to 
which metric to use in discussing the geolocation accuracy 
of the system. The median CE and the mean CE give 
somewhat optimistic views to the system performance and 
do not address the variability in performance. 
The 90-th percentile of the monodrop random biases sample 
ranked according to radial distance from the origin is 14 
meters and is shown as the blue dashed line in Figure 10. 
  
0.1 
0.05 
  
  
  
CE 
(meters) 
Figure 9: Monodrop Circular Error Histogram 
  
latitude (meters) 
  
  
  
725 
725720715 710-5 9 35 10 15 20 25 
longitude (meters) 
Figure 10: Monodrop Random Bias 
5.2 Stereo Pairs Comparison 
The stereo pairs comparison also used the measured image 
coordinates with the post-processed ephemeris and attitude 
history data. A relative triangulation was performed using 
only tie points. The 3D coordinates of the check points were 
determined by space intersection within the triangulation. 
The differences between the latitude and longitude observed 
by OV-3 and the check points were determined and 
aggregated to form the CE test statistic. The difference in 
elevation observed by OV-3 and the check points was used 
to form the linear error (LE) test statistic. 
The stereo pairs comparison test data consists of 15 stereo 
pairs located world wide. Due to the smaller sample size, the 
histogram bins were enlarged to 3 meters. 
  
0.15 
0.05 
  
  
  
0 10 20 30 
CE 
(meters) 
Figure 11: Stereo Pair Circular Error Histogram