ISPRS Commission III, Vol.34, Part 3A ,,Photogrammetric Computer Vision“, Graz, 2002
is required in this approach as listed in Table (3). Results from
manual SPR are also listed in the same table. One can see that
the results from the traditional manual orientation and our
approach (even in Experiment 4 where fewer object space roads
were used and digitisation errors were introduced in the image
space features) are comparable. In addition to the estimated
parameters, we had obtained the correspondences between
points in image and object space.
Table 2: Experiments Summary.
image space that had no correspondences in the object space in
experiment 4 will be considered changes, as they belong to the
same boundary of the object space entities that had been
examined. In the next paragraph, we are examining changes that
occurred locally between object and image space entities.
Table 3. Estimated EOP and their initial (approximate) values
together with the results from manual SPR.
Object space data sets
O1 (Fig. 4-a) O2 (Fig. 4-b)
15 1572 5 799
roads | points | roads | points
[Image Il 15 Experiment 1 Experiment 2
space | Fig. 4-c | roads
data 55178
sets points
D 15 Experiment 3 Experiment 4
Fig. 4-d | roads
63397
points
Xo(m) Yo (m) Zo (m)
Manual 600.00 -26.781 1014.894
Appx. 450.0 100.0 900.0
Exp. 1 599.762 -26.937 1014.842
Exp. 2 599.797 -26.663 1014.699
Exp. 3 599.722 -26.974 1014.818
Exp. 4 599.245 -27.081 1014.754
a Qo x
Manual 0.584667 -0.867300 1.191474
Appx. 9.0 -9.0 10.0
Exp. 1 0.590318 -0.872063 1.185914
Exp.2 0.572123 -0.871997 1.182790
Exp. 3 0.589120 -0.870109 1.189792
Exp. 4 0.594399 -0.895585 1.183058
30 4» 80 mU 1X0 120 up Ju
SST NTE Ew RU e
(a) (b)
150 150
100 E à 100 i A t ;
e m NE M
agb A M 13 5 i
| mme 4 77655
| f | A Ï A A ) v \ |
a | e ZA e [sf Ls]
| Nam 121 Sa d
100 m i4 [s] * ts] 100 \ 4 3]
a ur 0 mw we Ba 100 — 50 a 50 100150
(c) (d)
Figure 3: Object space linear features O1 (a), and O2 (b), and
image space linear feature without digitisation errors
Il (c) and with digitisation errors I2 (d). Roads are
labelled for further references.
The correspondence between the image and the object space
linear features as well as the consistency check described in
section 3.2 have been performed. All the roads were reliably
matched (see Figure 3). In this figure, corresponding road
segments were given the same label. One should notice that we
had realised the correct correspondences between higher level
entities (road segments). Even when there is no one to one
correspondence between higher-level image and object space
entities (as in experiment 4, where 10 out of 15 road segments
in the image space are not present in the object space),
correspondences are reliably obtained. Moreover, the quality of
the estimated parameters did not deteriorate. The 10 roads in the
Non-matched points in the object space, after being projected
into the image space, were significantly far from any of the
points in the image space. Therefore all of them were
considered as discrepancies. Among these non-matched points
consecutive points were segmented and the longitudinal and the
lateral distances from the corresponding road line were
computed. Results from experiment 4 are listed in Table 4. We
can see that the changes had been reliably detected.
Table 4. Changes (discrepancies) between image and object
space linear features, experiment 4.
Discrepancies (Changes) in the object space
o —
T = Location 3 8 : 3 8
2 m 5 2 = OQ Z ©
2E| BE xm | Ym [zm | £2 £2
© > "uu EÀ
zZ SS | 33
2 3 -70.81 309.98 | 34.18 | 1.69 16.02
2 17 56.71 200.27 | 3723 | 5649 | 211.53
2 19 403.81 | 263.37 | 43.59 | 5746 | 141.1
2 25 855.49 248 S3.03 5 67.6 | 20122
2 19 1227.19 | 222.05 | 60.53 | 40.61 | 148.97
3 13 1238.2 | -9128 | 67.05 | 44.96 | 114.61
3 A 910 -104.28 | 61.59 | 28.59 | 30.53
3 7 327.42 | -64.31 | 55.89 | 30.97 | 7242
3 13 199.75 32.54 | 4781 | 4162 | 81.31
3 12 -16.3 74.74 | 42.04 | 65.7 | 149.53
5 13 1167.07 | -39973 | 76.26 | 23.1 49.28
s 4 1169.49 | -444.77 | 76.8 1.04 13.29
5 8 1199.87 | -564.32 | 8422 | 10.8 12.52
3 10 1259 -592.19 | 94.94 | 6.84 16.94
8 17 302.33 87.72 | 46.44 | 27.6 60.32
8 31 274.23 | -141.86 | 51.63 | 56.78 | 105.73
8 27 240.81 | -451.65 | 58.09 | 31.65 | 115.76
15 8 590.49 | 351.71 | 46.08 | 14.74 | 28.01
Examples of changes, which were detected, are shown in Figure
4. It has to be noted that all the changes were reliably detected
and their existence does not contribute to the estimated
parameters. Therefore, we realised a robust estimator for the
A- 154