After evaluating the results for different cases, the selected
impact-factors (case 7) are listed in Table 7. The chosen
combination (i.e., the impact of true-lengths <the impact of
relative angles <the impact of generated-lengths <the impact
of the number of crossing-lines) could be logically true.
• In the matching-phase, different values for the thresh-
olds of,
T
1
,
T
2
, and
T
3
are tested. These thresholds are
selected based on the physical characteristics of high
resolution images. By fixing
T
3
= 6*
NA
, different com-
binations of the values of [1, 10] degrees as well as [0.1,
0.5] are tested for
T
1
and
T
2
, respectively. Part of the
results for these combinations over the 1
st
to 3
rd
datasets
are demonstrated in Table 3.
As shown in Table 3, increasing the thresholds of
T
1
and
T
2
directly increases the computational time without any sig-
nificant effect on the number of matched-lines. Additionally,
all combinations are able to find the matched-lines with high
reliability (validity-factors of 100 percent). By evaluating the
above results, the selected values of
T
1
and
T
2
are represented
in Table 7. However, higher values could also be nominated.
• By fixing the thresholds of
T
1
and
T
2
(based on Table
7), different values of
T
3
are examined over the first to
third datasets (Table 4).
As shown in Tables 3 and 4,
T
3
is the most important
Figure 6. The high resolution images (right side) as well as the related vector maps (left side) as introduced in Tables 1 and 2.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2016
371
