International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
2.2 Space resection with adjustment
If we have more than three control points the space resection
should be solved with an adjustment. We adjust only the
projection center coordinates. The rotation angles can be
calculated separately in one step at the end. Let’s list the steps
of the adjustment procedure.
STEP 1. Let the number of control points be 7 . We will
group them by three in every possible combination and for each
group we solve the space resection directly (formulas 2-6). In
general case, at each group we will get more than one solutions
for the projection center.
STEP 2. From each group we choose common solutions, it
means we choose those solutions where the sum of square
differences is minimal.
STEP 3. By the error propagation law we calculate the
M , covariance matrices for each solution by the following
formula:
7
M. =F MF 9)
Yi x14
where
M.:; covariance matrix of control points considered at the
.
geodetic measurements.
The F" dispersion matrix can not be derived directly with
partial derivation from the equations 3-6, so we construct this
matrix from the differences gained from the original solution
and from solutions where each image coordinate is incremented
with a small value one by one. Finally we got the FT matrix
with the dimensions of 3x6, this matrix well approximates the
matrix containing the partial derivatives.
STEP 4. We determine the weight matrices for each solution
by the following equation:
P=0";=c(M,}' (10)
!
i
where Cis a scalar factor, at the space resection we take its
value as 1/1000.
STEP 5. We calculate the adjusted values of unknowns by the
Jacobian Mean Theorem as follows:
Xs By'«N nr) (11)
X = OQ X S (PL)
where
X,
L = Y (12)
Z
and it contains the solutions from each group.
2.3 Gross error detection
During the adjustment we can detect the control points with
gross errors. A gross error can exist in the ground coordinates
or in the image coordinates. By the following procedure we can
detect them affectively and no matter where the gross error is.
Let's group four control points in every combination and solve
the space resection with adjustment (11). Since we made the
space resection in every possible combination, our duty now to
determine which group has a gross error and finally we can
conclude exactly which point or points caused the gross error.
For example if we have 5 control points, we can group them by
four as follows:
23:4] diamo 455 -dp34 $5154,
Let's suppose that the control point No. 1 has a gross error, it
means that the first four solutions will be wrong and only the
{2,3,4,5} group gives a good solution. So, by this logic we
conelude that only the point No. 1 can be the cause for a gross
error. À similar logic can be proved when the number of points
are more than 5 or the number of points with gross-error is more
than one. The only limitation for the detection is that finally at
least four error-free control points should remain (otherwise no
reason to do the space resection with adjustment).
Here is the procedure which helps to decide whether a space
resection made by four control points has a gross error or not:
STEP 1. After getting the adjusted projection center we can
calculate the residuals by the following:
V. -X- L;, where V, E y: (13)
STEP 2. After this we can calculate the 77, weight unit error
(14) and the RMS for each unknown (15) with the help of the
D s covariance matrix:
7
3 V; hr,
My 7 —-
0 (14)
3n—3
where M means the number of control points.
M, = sg
x = M qa
Sy 23
my = My VI. (15)
Intei
STE
adju;
STEI
7 7 7 X
My
n,
Other
RMS
H
0
At p
we ge
as à |
other
equati
It mea
equat
F
SZ
$Zy
F
$27
Otherw
consid: