TER 
nd con- 
empt to 
ollowing 
'oduced 
cessing 
) photo- 
/ (input) 
duction 
cessing 
by the 
viewing 
hich will 
acecraft 
ribe the 
teristics 
'esent a 
mulated 
NT 
S AM-1 
by the 
ation of 
cession 
! Space- 
) as the 
m about 
1ave an 
ameras. 
J toward 
s point- 
nd Df in 
of four 
e same 
ges are 
surface 
à +70.5° 
aly. The 
Jirection 
> Figure 
> (CCD) 
line arrays in a single focal plane. The line arrays consist of 
1504 photoactive pixels plus 16 light-shielded pixels per 
array, each 21 um x 18 um. Each line array is filtered to pro- 
vide one of four MISR spectral bands. The spectral band 
shapes will nominally be Gaussian, and centered at 443, 
555, 670, and 865 nm. Because of the physical displace- 
ment of the four line arrays within the focal plane of each 
camera, there is an along track displacement in the Earth 
views at the four spectral bands. This must be corrected for 
within ground data processing. The cross-track instanta- 
neous field of view (IFOV) and sample spacing of each pixel 
is 275 m for all of the off-nadir cameras, and 250 m for the 
nadir camera. Along-track IFOV's depend on view angle, 
ranging from 250 m in the nadir to 707 m at the most oblique 
"Physical" MISR instrument 
    
"Flight 
direction 
   
Figure 1: MISR imaging event 
angle. Sample spacing in the along-track direction is 275 m 
in all cameras. 
In order to find the geolocation corresponding to the pixel's 
field of view, the pixel pointing direction is expressed in the 
Geocentric coordinates system, as follows: 
B TT (1) 
where ? is the pixel pointing direction relative to the instru- 
ment coordinate system defined by the observable image 
coordinates and the set of constants representing the instru- 
ment interior orientation parameters. T, represents the 
transformation between the instrument and spacecraft coor- 
dinate axes. T, , defined by the ephemeris and attitude data 
at the time of imaging, represents the transformation 
between the spacecraft and Geocentric coordinate system. 
Equation (1) is an often used photogrammetric model suit- 
able for various image-ground point determinations required 
for satellite based imagery. 
3. TERRAIN-PROJECTED RADIANCE PRODUCT 
As described in the previous section the MISR instrument 
acquires, at any given time, push-broom imagery from nine 
widely separated locations along the sub-spacecraft track. It 
177 
takes about seven minutes for any single location along this 
track to be observed by the nine MISR cameras. The sci- 
ence objectives require this set of multi-angle images to be 
geolocated and coregistered, in order for the higher-level 
geophysical retrievals to occur. Furthermore, within any sin- 
gle camera, the images in the four MISR spectral bands are 
slightly displaced in the along-track direction, as a conse- 
quence of the instrument design. Therefore, ground data 
processing must also compensate for this, by effecting a 
superposition of the multi-spectral data. 
In an abstract world the terrain-projected radiance product 
may be looked upon as the data collected by a "virtual" 
MISR (see Figure 2.) 
“Virtual” MISR instrument 
     
grid 
Figure 2: Terrain-projected radiance product: output 
from a “virtual” MISR. 
Effectively, this product represents collection of about 140 x 
9 digital image maps of the entire globe, over land and 
inland water, obtained during a period of 6 years. Each digi- 
tal map contains radiances from four spectral bands. We 
have selected Space Oblique Mercator (Snyder, 1987) as 
the reference map projection grid, because it is design to be 
suitable for continuous mapping of satellite imagery. The 
ground resolution of the map grid is 275 m. 
The spatial horizontal accuracy goal associated with these 
maps and requested by the science algorithms, is an uncer- 
tainty better then +275 m at a confidence level of 9596. Obvi- 
ously this kind of accuracy requires knowledge of a DEM 
and removal of the displacement due to relief. In addition, 
the accuracy specifications for the supplied spacecraft navi- 
gation and attitude data suggest the possibility of horizontal 
errors of about 2 km in the most oblique cameras. The next 
section discusses the algorithm which accounts for the dis- 
placement due to the topography and errors in the space- 
craft data prior to the resampling of the acquired MISR 
imagery to the map grid. 
4. GEORECTIFICATION ALGORITHM 
4.1 Overview 
In addition to the spatial accuracy requirements, the georec- 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996