77ze International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
A test field suitable for such procedures is seen in Figure 1. A 
closer look at the extracted point and line features is given in 
Figures 2a and 2b, respectively. In Figure 2b it is clearly seen 
that the line features are composed of individual points. 
Figure 1: Suggested calibration test field with 
automatically extracted point and linear features 
Figure 2a: Point feature 
Figure 2b: Line feature 
3. CAMERA STABILITY 
It is well known that professional analogue cameras, which 
have been designed specifically for photogrammetric purposes, 
posses strong structural relationships between the focal plane 
and the elements of the lens system. Amateur digital cameras, 
however, are not manufactured specifically for the purpose of 
photogrammetric mapping, and thus have not been built to be as 
stable as traditional mapping cameras. Their stability thus 
requires thorough analysis. If a camera is stable, then the 
derived IOP should not vary over time. In the work done by 
Habib and Pullivelli (2006b), three different approaches to 
assessing camera stability are outlined, where two sets of IOP 
of the same camera that have been derived from different 
calibration sessions are compared, and their equivalence 
assessed. In their research, different constraints were imposed 
on the position and orientation of reconstructed bundles of light 
rays, depending on the georeferencing technique being used. 
The hypothesis is that the object space that is reconstructed by 
two sets of IOP is equivalent if the two sets of IOP are similar. 
The three different approaches to stability analysis are briefly 
outlined in the following sections. In these methods, two sets of 
IOP are used to construct two bundles of light rays. A synthetic 
regular grid is then defined in the image plane. The distortions 
are then removed at the defined grid vertices, using the two sets 
of IOP in order to create distortion-free grid points. The 
distortion-free grid points of each IOP are then compared to 
assess their similarity. 
3.1 Zero Rotation Method (ZROT) 
In the ZROT method, a constraint is applied on the bundles 
such that they must share the same perspective centre and have 
parallel image coordinate systems. If the two IOP sets are 
equivalent, then the coordinates of the distortion-free vertices in 
the two synthetic grids should be the same. Therefore, the 
differences in the x and y coordinates between the two 
distortion-free grids are used to estimate the offset between the 
two sets of IOP. When the principal distances of the two sets of 
IOP are different, the distortion-free grid points from one IOP 
are projected onto the image plane of the other, before the x and 
y coordinate offsets are measured (Figure 3). The similarity 
between the two bundles is then determined by computing the 
Root Mean Square Error (RMSE) of the offsets. If the RMSE is 
within the range defined by the expected standard deviation of 
the image coordinate measurements, then the camera is 
considered stable. This similarity imposes restrictions on the 
bundle position and orientation in space, and thus has similar 
constraints to those imposed by direct georeferencing with 
GPS/INS. Therefore, if the IOP sets are similar according to the 
ZROT method, the relative quality of the object space that is 
reconstructed based on the direct georeferencing technique 
using either IOP set will also be similar. 
Ray from Bundle I 
Ray from Bundle II 
Original Image Grid Points 
Distortion-free Grid Point using IOPi 
Distortion-free Grid Point using IOP n 
Projected Grid Point of IOP tI 
Figure 3: The offset between distortion-free coordinates of conjugate 
points in the ZROT method 
3.2 Rotation Method (ROT) 
In comparison with the ZROT method, which restricted the 
bundles orientation, this method allows the comparison of 
bundles that share the same perspective centre but which have 
different orientation in space (Figure 4). The purpose of the 
stability analysis is to determine if conjugate light rays coincide 
with each other, and this should be independent of the bundle 
orientation. This method checks if there is a set of rotation 
angles (cd, (p, k) that can be applied to one bundle to produce the 
other. A least-squares adjustment is performed to determine the 
rotation angles, and the variance component of the adjustment, 
which represents the spatial offset between the rotated bundles 
in the image plane, is used to determine the similarity of the 
two bundles. The bundles are deemed similar if the variance 
component from the least squares adjustment is in the range of 
the variance of the image coordinate measurements. This 
similarity imposes restrictions on the bundle positions in space, 
and thus has similar constraints to those imposed by GPS 
controlled photogrammetric georeferencing. Therefore, if the 
IOP sets are similar according to the ROT method, the relative 
quality of the object space that is reconstructed based on the 
GPS controlled georeferencing technique, using either IOP set, 
will also be similar. 
P.C. (0, 0,0) 
Pi fri, yi,-Ci) 
R (co.iil), k) 
Pafrn,Vn,-cn) 
Figure 4: The two bundles in the ROT method are rotated to reduce the 
angular offset between conjugate light rays 
1061 
y'M-S 
SÈSSI