Table 3: Testblock Appenweier, precision and accuracy results after
bundle block adjustment with 12 additional parameters
tie-points from check points
Version | measurement point transfer 99 Wy uy Hy y u
[um]| [um] — [um] (um ] — [ym]
a operator no 3.3 7.6 5.4 6.6 14.5
C100 (sign. points)
b dig.correlation yes 4.0 73 773 73 9.8
C 100
Assessing the results we look first at the g, values of 3.3 ym and 4.0
um of version a) and b), respectively. The difference is almost entire-
ly caused by the digital point transfer. It can be concluded, there-
fore, that the transfer of natural tie points by digital image correla-
tion contributes to the total error budget of the image coordinates
with a r.m.s. value of 2.3 um. (If the setting errors of the signalized
points still included in version b) are considered the r.m.s. value in-
creases to 2.4 ym.)
The absolute accuracy of the adjusted blocks a) and b) (signalized tie
points without point transfer against natural tie points with digital
point transfer), as assessed from the 23 planimetric check points, is
represented by the r.m.s. values of Urine 6.6 um and Ry yf 7.3 um,
respectively (or 6.3 pm and 7.0 ym, cóifdidering the inacéuracy of the
check points). Thus, the additional contribution of digital point
transfer to the absolute planimetric accuracy of the block amounts to
3.1 um. The comparison with regard to the vertical accuracy (14.5 um
against 9.8 um) is contradictory, as the block with digital point
transfer has, in this test, considerably better vertical accuracy than
the block with signalized transfer points. The apparent incongruency
may be caused by the small number of only 10 vertical check points. It
cannot be considered significant. Remark: The results of version a)
(signalized points, operator measurements) differ from previously pub-
lished results of the total test block Appenweier, which were more pre-
cise /6/. The differences may have several reasons (13 years old photo-
graphy, different geometry, different control points, tie points and
measuring instruments). They are of no particular concern here, as this
test is primarily looking for the performance of digital point transfer
against a comparable set of manual measurements of signalized tie
points.
4. CONCLUSION
The experimental investigation attempted to demonstrate and confirm the
high accuracy capability of digital transfer of natural tie paints by
digital image correlation. The experiments were carried out with 2
small blocks of black and white, wide-angle photographs (Schnürpflin-
gen, 26 photographs, 1:4 000 and Appenweier, 28 photographs, 1:7 800).
In both cases the results of manual measurement of signalized tie
points were compared with digital transfer of natural tie points. The
- 25 -
