The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
1232
to be inside of the GCP’s/CP’s network, second; the areas
should have different feature types and a significant height
difference, lastly; the areas should be reachable with vehicles
for control operations in field.
1:5.000 scale was selected for compilation processes because of
on-going discussions about usage of high resolution space
imagery in big scale (especially in 1:5.000 or greater scale) map
productions. The selected 3 regions have height differences
about 110 m, 250 m and 220 m respectively.
In mapping researches it is very seldom to see compilation
assessments in which all features have been detected and
investigated using detailed layer information. The most
comprehensive researches in topographic mapping from high
resolution satellite images are applied by the OEEPE (European
Organization for Experimental Photogrammetric Research, now
renamed EuroSDR) using IKONOS imagery and by Ordnance
Survey (UK) using Quickbird imagery (Holland et al., 2002;
Holland and Marshall, 2004; Holland et al., 2006). Especially
in the project applied by the Ordnance Survey, for each of the
features the requirements for capture were broken down into
three different levels - high, medium and low. In high level the
feature must be identifiable and in medium level it would be
desirable to identify the characteristic of the feature while in
low level it would be of some minor interest to identify further
characteristic of the feature. And for each feature type, the
cartographers recorded whether or not the features could be
successfully identified from the image (Holland et al., 2006).
From the other side, in our project, three different experienced
operators compiled all features on these regions (roads,
buildings, water features, forests, hedges, communication and
electricity transmission lines etc.) using mono satellite and
stereo aerial images for 1:5.000 scaled map productions. In
compilation processes, the operators have followed all the
standard procedures applied in General Command of Mapping
for producing big scale map (Figure 3).
numbers of features compiled in aerial photographs were more
than satellite images (Table 5).
Line
%
Polygon
%
Point
%
Total
%
1. Sheet
Aerial
Photo
1894
100
541
100
2242
100
4677
100
Ikonos
776
41.0
338
62.5
1189
53.0
2303
49.2
Quickbird
891
47.0
348
64.3
1068
47.6
2307
49.3
2. Sheet
Aerial
Photo
529
100
53
100
242
100
824
100
Ikonos
295
55.8
26
49.1
186
76.9
507
61.5
Quickbird
285
53.9
18
34.0
111
45.9
414
50.2
3. Sheet
Aerial
Photo
746
100
84
100
384
100
1214
100
Ikonos
397
53.2
77
91.7
471
122.7
945
77.8
Quickbird
424
56.8
61
72.6
475
123.7
960
79.1
Total
Aerial
Photo
3169
100
678
100
2868
100
6715
100
Ikonos
1468
46.3
441
65.0
1846
64.4
3755
55.9
Quickbird
1600
50.5
427
63.0
1654
57.7
3681
54.8
Table 5. The number of details compiled on images
Secondly, the features compiled from different sources have
been located one on the top of the other and detected the
differences between operators.
4.3 Control of Compilations in Field
It is noted that changes of large features such as those
associated with building developments and major road
improvements are often detected using established methods
(either from local authority planning offices, building
developers or bespoke change detection service providers)
(Holland et al., 2006). It is due to the fact that small objects or
boundaries are unclear and can not be identified properly, some
other sources of information like cadastral information or field
survey are used (Alexandrov et al., 2004). Therefore, this study
has been carried out as control and completion applications in
field in October 2005 by two personnel. But because of season
conditions and vehicle capabilities, it was not possible to pass
over some roads.
And lastly, the feature layers have been evaluated in detail. In
compilation processes, 437 features classified in 62 layers have
been used. So, every feature layer in every sheet has been
compared aerial photographs (Table 6).
Figure 3. Compilation of first region by stereo aerial images
4.2 Comparison of Compilations
The map productions from different sources (Ikonos, Quickbird
and aerial images) have been compared within each other and
the compiled features have been controlled in field. In the
comparison studies; firstly, the numbers of features in text, line,
polygon and point layers have been detected, controlled and
compared by using these software. In this stage, aerial
photographs have been selected as reference data because the