: eoe Mo pow LANI P ^ t ET - EET TN - STE P " " " S - - " »
be niii LH AA M ENS are Sy uy mo el a SLR EI CYS ST BN RE ER NR SR Le alm Una Ue oninia ami aeu toi andi AA NEA Nas GERI SRE Si AY
15.
persion of the subject edge location of the 4 observers was very large
in relation to the standard error of a single observation.
Decutance is derived from the density spread function as the ratio
of the amplitude and the width of the spread function. Thompson found
that his measure of image quality correlated well with Decutance.
Image quality and hence reliability of edge pointing improved, as the
spread function decreased. That is the measure of image quality is
inversely proportioned to the spread function width.
A firm estimate of the magnitude of the systematic errors is not
available as they vary so greatly, but Thompson believes that it may
be possible to determine the personal bias and then compensate for this
error during the observations. Since the errors vary with spread
function and contrast it appears this proportion would present some
difficulties in practical situations where contrasts and spread
functions would be variable over an image.
Welch and Halliday performed an extensive series of pointing
observations to the edges of bar targets, and isolated square objects
photographed at a scale of 1:24,000. Measurementsof the sizes of the
objects were made on a stereocomparator at an optical magnification of
35X to 40X. Errors in the measurements of the widths of the bars and
the actual edge pointing locations of the squares were presented in
graphical form. From the plot of pointing positions to the edges of .
the squares, average errors were derived and an attempt made to relate
these errors to imaging quality of the system, for this paper. A
summary of these mean errors is given in Table 3.
The systematic errors increased with target contrast. Indeed the
error of pointing to high contrast targets was approximately twice
(7.6um or 250 secs of arc) that for medium contrast targets (3.5um or
100 secs of arc). There is no significant overall change in the
systematic errors for measurements in the "with flight direction
compared with measurements in the 'Against flight'direction. Welch and
Halliday also found that for bar target width measurements, higher
width measurement errors occurred for the high contrast bar targets
than for the medium contrast (2.5 to 1 to 6:1), for which errors of 3-7
um regularly occurred. (The precision of pointing was 2-3um). For
both low and high contrasts systematic errors deteriorated rapidly.
Martucci (1972) found a similar relationship and claimed that medium
contrasts resulted in the highest quality of bar width measurement.
In order to relate the results of Table 3 to a measure of imaging
quality of each photogrammetric system, MIF curves were derived from
each of the square wave functions given by Welch and Halliday. An
estimate was then made of the spread function width, assuming each MIF
was approximately Gaussian. The approximations in general were
acceptable, but in some cases they were poor. The spread function
widths however were used only as a guide. A plot was then made of the
reciprocal of the spread function width against the systematic errors
mi
Hi
EU
d cce
P^