In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
Image H5 ATM vs. L7 ATM 
Q 
o 
c 
0 
Urn 
CD 
a= 
Q 
0 
> 
JS 
0 
01 
Image H5 ATM vs. L7 ATM 
0 
> 
25 
0 
01 
Figure 5. Relative differences of averaged image reflectances 
for different targets viewed from two flight lines from different 
altitudes, but with the same flight heading (N-S). 
This is due to the fact that the applied BRDF correction uses a 
correction function derived from an average statistics of the 
whole image. So some rare targets with a BRDF different from 
the average will not be corrected adequately. 
Image L4B ATM vs. L7 ATM 
Image L4B ATM+BRDF vs. L7 ATM+BRDF 
Q 
0 
> 
22 
0 
01 
For a comparison of the atmospheric correction results of two 
different flight levels, again lines L7 and H5 were used (Figure 
5). The relative difference of both was below 10% forAsphalt l 
and below 20% for Grass l, and the colour tarps (Figure 5). 
For the additional vegetation targets in Figure 5 (e.g. Stub- 
ble_#), it can be seen that by averaging a large number of pixels 
(Table 2) the difference reduced to below 10% for R, G, and 
NIR. Only for the blue channel a larger discrepancy was found. 
This is attributed to atmospheric effects. 
3.2 Validation of BRDF correction quality 
To see the effect of the BRDF correction, the relative differ 
ences of the lines L4 (Backward) and L7 (Nadir) before and 
after the BRDF correction were evaluated using several large 
vegetation targets. As expected from the discussion in sec. 3.1, 
Figure 6 shows differences of up to 70% for the ATM corrected 
reflectances between the W-E and N-S flight lines L4 and L7. 
After the BRDF correction, the differences for vegetation tar 
gets were generally below 10% for R, G, and B, and 20% for 
the NIR. Only the asphalt target showed a larger discrepancy. 
Figure 6. Relative differences of averaged image reflectances 
for different targets from a WE and a NS flight line of the same 
flight level before (top) and after (bottom) BRDF correction. 
4. CONCLUSIONS 
The project showed that ground reference targets used for vali 
dation (and also for in-flight calibration) have to be prepared 
carefully of non-specularly reflecting material and measured at 
the exact time of overflight. This is a tedious procedure and may 
be prohibitively expensive in most cases for routine remote 
sensing campaigns. 
The ADS40/80 sensor together with the XPro “atmospheric” 
correction option enables the production of images with relative 
reflectance differences of less than 10% even for dark objects. 
The relative pixel-to-pixel accuracy is much higher, since the 
calibration and correction functions are smooth functions of the 
view angle. 
To achieve this absolute accuracy, the following precautions 
have to be made. First, in the flight planning the flight line