N
1.2 One of the most promising possibilities for improving the precision
of point transfer is the application of digital image correlation. It
has been shown that digital image correlation applied to aerial photo-
graphs is capable of very high precision, /2/. Standard errors for
x-and y-parallaxes in the order of 1 - 2 um can be obtained with well
textured image areas. It seems feasible, therefore, to apply digital
image correlation for high precision point transfer which consists es-
sentially of x- and y-parallax assessment. It could exceed and replace
the stereoscopic performance of the human operator.
There is a second aspect of digital point transfer which is independent
of the precision aspect. The repeated manipulation of photographs is a
most tedious necessity of conventional point transfer operations. This
could, in principle,be avoided by digital point transfer, if organized
accordingly. Instead of referring back to the original photograph when
transferring a point into several other photographs, the digital point
transfer could refer to the digitized array of the local image area
which would temporarily be stored in the computer. Thus a photograph
would only once be operated upon in the measuring instrument or system,
and the measurement of plate coordinates would be obtained simultaneous
with point transfer. The real problem here is to obtain suitable
transfer points and approximate values automatically, with as little
operator interference as possible. It will have to be investigated
whether and how far computer assistance will be helpful. Anyhow, such
operational advantages alone would justify the method to be pursued.
1.3 In this paper results of 2 practical aerial triangulation tests for
point transfer with digital image correlation are presented. The tests
are only concerned with the precision of digital transfer of arbitrary
terrain points. It is demonstrated that the test results indeed repre-
sent high precision aerial triangulation and that they compare very
well with the results obtained conventionally with signalized points.
The above mentioned operational aspect of digital point transfer is not
part of the tests, as the automatic system is not yet developed. It
means that the initial approximate values for the digitization of local
image areas were derived from previous measurements or controlled by
the operator.
The hardware system used is an analytical plotter Zeiss Planicomp C 100
which is equipped with Hamamatsu CCD video cameras for digitization of
small image aereas/3/. The digitized image data, after being transfer-
red on-line to the HP 1000 of the instrument, are normally directly
processed and the resulting parallaxes and plate coordinates fed back
into the Planicomp system. In these tests, however, the digitized data
were stored externally and processed separately. Nevertheless, the re-
sulting image coordinates refer to the image coordinate system as the
digital arrays are calibrated to the Planicomp system and the instru-
ment was used in the monocomparator mode. The digitization has 20 um
pixel size and a range of 256 steps of grey level values. The image
correlation was calculated with digital arrays of 32 x 32 pixel for the
search matrix and 16 x 16 pixels for the reference matrix. The digital
correlation is in fact a least squares area match of the 2 arrays. If,
however, a weighted central point is defined in one array the cor-
responding point in the second array is identified through the trans-
formation by which the match is obtained. Thus, digital area correla-
- 19 -
