errors
ler and
vation,
well as
ointing
minate
stment
ort on
'search
in the
t. The
ans of
en the
le RC7
d time
ss met,
| trian-
y have
stereo
ilation,
1ded in
I. The
he dif-
ch. give
) trian-
y block
would
hod of
ow the
^re ob-
port of
ent of
ions of
tal co-
cedure
ch the
points,
ns, are
| cards.
degree
a pro-
n using
twenty
ed and
istment
cause
degree
results.
hat the
| strips
can be
of the
LA COMPENSATION DES BLOCS DE BANDES, DISCUSSION 65
third degree, or maybe even the fourth degree.
Therefore, it will take some experience with
block adjustment before we can decide whether
conformal transformations of the second degree
are sufficient in each case. In making this
decision, we have to consider the accuracy that
we require of the mapping and the plotting scale.
The first indications are that the transformation
of the second degree is quite satisfactory for
strips of about 25 photographs, as we have
used in this test.
For our participation in the international
test, we requested the coordinates of eighteen
ground control points: two in each of the four
corners, in the middle of each of the four sides,
and in the centre of the block.
For the preliminary adjustment we used only
the co-ordinates of the eight points in the four
corners of the block. The eight strips were
transformed to one system and the resulting
block was transformed by means of the eight
points. The result of this adjustment has been
published in the General Report of Commis-
sion III. The largest errors remain in the
TABLE 1.
adjustments, in meters.
Key to distribution of
corners, and amount to 20 meters.
After this we performed three iterative ad-
justments. In the first of these, the eight corner
points were used. This adjustment reduced the
errors in the corner points to a few meters, while
the largest errors in the other points became
only slightly smaller. In the second iterative
adjustment, the east and west perimeter control
points were used and in the third, all perimeter
control points. The residual errors after these
three adjustments are listed in Table I.
The errors in the control points that are
used only as check points cannot be computed
yet. However, from the distribution of the errors
it can be concluded that much greater errors
than those in the known points will not occur.
Therefore, for the required plotting at scale
I : 100,000 the iterative adjustment on corner
points is sufficiently accurate. For plotting at
scale 1 : 50,000 the iterative adjustment on all
perimeter control is sufficiently accurate. An
adjustment using also points in the centre will
not result in any appreciable improvement and
hardly seems worthwhile.
Residual errors in known ground control points after three iterative
Adjustment on eight corner points
ground control points 1X AY [X AY AX AY
19 4] - 12 0.0| — 1.5 + 13.0] — 2.9 — 2.7
s 25 40 — 2.8 + 0.31. 88 + 93 28 + 1.4
153] 133] 118 + 10.6 F 143] 1-72 +. 2,.5| + 7.1 + 0.1
159] 136] 192 E 2.7 - 113]. 11.1 T2231 1.9.4 + 8.2
2071 286] 275 + 1,8 — 2.0} + 9.0 + 6.4} — 32 + 19
299 | 287 | 271 — 0 +241 +798 1 10.0] 4- 3.5 — 1.9
Adjustment on west and east
perimeter control
[X AY 1X [Y [X (Y
Adjustment on all perimeter control
(X IY. 4X AY AX AY
+04 --0.4]—3.5 4
8.41— 3.9 —2.7
—4.1 — 0.8 69 +4
+29 —0.1
+14 + 03]—48 + 3.6l— 5.5 —2.5
— 38 —2.21+ 56 +01|/+26 —12
0.0 —3.3
$32 +26/1+50 —21
+89 —27|+24 + 5.0
—01 +17
+17 +2.01+ 3.5 —4.11— 0.5 — 33
—08 +07 +74 —47- 16 +41
+12 —30}+65 +2.6/—44 +08
00 +15|+73 +63|+38 — 1.6
—0.6 —2.6|4 1.7. —0.7|— 5.8 —0.7
—01 +1.6J+ 23 +30|+ 49 —2.1
We define the relative error in the position of
à strip tie point as the deviation from the
average of the two values obtained in two over-
lapping strips. After the preliminary adjustment
the relative error has a mean square value of
2.5 meters. The iterative adjustments improve
the fit of the individual strips on ground con-
trol points at the cost of the agreement among
EEE
EMS
A
TT EET
SENSE Se rer
sm
sesso
