interface to the host and therefore allows fast image transfer. 
The images are stored in such a way that all images grabbed 
at the same time will have a header file that contains the 
necessary information for georeferencing them. This header 
will be filled after processing the INS and GPS data using the 
KINGSPAD software; 
* guiding the operator by a user friendly monitor which 
contains all the important information such as video 
images, position, velocity , attitude, number of grabbed 
images, the computer disk space, the active hard disk, 
number of locked satellites, satellites that have dropped or 
are affected by cycle slips etc. ; 
* switching the cameras off and on in order to limit the 
storage requirements. This feature is very important in 
situations when the vehicle has to stop, for example at 
traffic signals. 
4. GEOREFERENCING OF VIDEO IMAGES 
The problem of georeferencing video images can be defined 
as the problem of transforming the 3-D coordinate vector r9 
of the camera frame (c-frame) to the 3-D coordinate vector r™ 
of the mapping frame (m-frame) in which the results are 
required. The m-frame can be a system of curvilinear geodetic 
coordinates (latitude, longitude, height), a system of UTM or 
3TM coordinates, or any other earth-fixed coordinate 
system. For more details on georeferencing of remotely 
sensed images using INS/GPS data see Schwarz et al (1993b). 
The process of georeferencing the video images includes the 
determination of the camera projective centers (p.c.) in the 
: m 
m-frame, i.e. rpc (D, and the rotation matrix between the c- 
m 
frame and the m-frame R. (0. The problem of determining 
m 
the 3-D coordinate of any point (i) in the m-frame, i.e. Tj » 
can be written as: 
m qm i. ln GE 
r - rye 0* S R, (D*r (1) 
where S! is a scale factor specific to a one point/one camera 
combination, and the time (t) is the time of exposure. 
Equation (1) is the simplest form of the georeferencing 
problem which relates the points in the c-frame and the m- 
frame. It implies that the coordinates of the projective center 
of the camera can be directly determined. This is usually not 
the case because the navigation sensors - GPS antenna/ INS 
gyro's- cannot share the same point in space with the 
imaging sensors. If the vector between the INS body frame 
(b-frame) and the camera is given in the b-frame as ab, ry (O 
can be written as : 
m m m b 
ry 07 TINS (t) + Ry (tea (2) 
m. . 3 
where Ry (t) 1s the rotation matrix between the b-frame and 
: b: 
the m-frame measured at time (t). The vector a is usually 
determined before the mission. 
Similarly, the INS b-frame (gyro frame) cannot be aligned 
with the c-frame. If the small constant misorientation an? 
between the two frames is obtained, R 6 (t) can be written as: 
RC (0 = NO : dR? : 3) 
The computation of 3-D coordinates of any feature includes 
the following major steps : 
* transforming the 2-D image coordinates measured in 
the computer coordinate frame to the 2-D image coordinates 
of the camera CCD chip. This step requires the determination 
of the camera principal point coordinates (Xp. yp). Figure 
(3) shows the relation between the two systems. 
i" 
  
  
  
  
  
  
  
  
  
  
  
Yc 
Computer Coord. 
system 
CCD Coord. systen 
Figure 3 The relation between the computer and CCD chip 
coordinate systems 
* transforming the 2-D CCD chip coordinates to the 3-D 
c-frame coordinates. This step required the determination of 
the focal length (f) of the camera lens. For determining (Xp, 
yp. £f) of step 1 and 2, a bundle adjustment program, 
developed at the UofC, is used (Cosandier and Chapman 
1992), 
* computing the shift component and the 
misorientation between the c-frame of each camera and the 
INS b-frame. For this step, some constraints , are added to 
the bundle adjustment program (El-Sheimy and Schwarz 
1993). The constraint equations make use of the fact that 
both the cameras and the INS are fixed during the mission. 
This is achieved by acquiring a number of stereo pairs at 
different positions and distances from a test field of ground 
control points as shown in figure (4). 
These constraints can be written for two van positions (i) 
and (j) as follows: 
bi=bi and dR{@= dR; (i) (5) 
For the stereo-pair (i), 
Cc. C b m 
dR (i) = dR = R . R. (6) 
  
Mc] ad - x12 + (vi v. -z 7) 
244 
  
Where (X 
camera p 
the bund 
between 
bundle ac 
» int 
cameras : 
This req 
GPS/INS 
e m 
process ' 
containec 
The obje 
through : 
translati 
frame to 
the two 
intersec 
georefer: 
8 
m 
i 
It is cl 
accuracy 
processi 
GPS p 
System « 
As a fir 
factors, 
is perfor 
the seco