International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
used for different activities within the production area, as
described later in this section.
Area Acquisition Coverage Use
date (km?)
Salisbury 03/03/2003 44 Cl
Manchester 16/03/2003 36 Ch SCA
Cambridgeshire 14/03/2003 25 CA
Walsall 01/10/2003 196 CI
Christchurch 02/06/2001 325 | MU
Table 1: QuickBird data used in the trial
Key:
CA - map Currency Audit
CI = Change Intelligence
MU = Map Update
2.1 Orthorectification
Before any update could be undertaken, the images were
orthorectified. Several different approaches were taken, using
commercial off-the-shelf software. Although in a live
production environment the images would have been rectified
using GPS control points, for this trial the control points were
simply measured from map detail taken from existing large
scale mapping data (OS MasterMap*). Similarly, the digital
terrain model used in the process was taken from the existing
height product, OS Land-Form PROFILE*. Table 2 shows the
resulting orthorectification accuracy figures for two of the study
areas (one urban, one rural). These are slightly better than the
results for the initial study area of Christchurch, which had an
overall RMSE of 2.77m, using 27 control points. Considering
the nature and number of the control points, and the ease with
which the images could be orthorectified using readily-
available software, these results were considered to be very
good.
Manchester (map accuracy 0.4m RMSE)
NO. of RMSE (m)
Point type 4
points
(x) (Y) Overall
control 11 1.18 1.09 1.60
check 15 1.38 1.06 1.74
Salisbury (map accuracy 2.47m RMSE)
RMSE (m)
Point type No. of
points
(x) (Y) Overall
control 9 1.24 431 1.80
check 14 2.65 2.07 3.38
Table 2: Orthorectification accuracy measures, using existing
map detail as control.
748
2.2 Topographic Map Update
The orthorectified imagery was analysed by a small team of
surveyors and cartographers, all of whom were familiar with
the capture of spatial information from imagery in a production
environment. Both positional accuracy and feature attribute
accuracy were analysed and compared with results obtained
from aerial photography. Six sub-areas of the image were
studied, to ensure that the following different types of
topography were investigated:
e Urban — coastal and floodplain
e Urban — inland
® Semi-urban - airport
® Rural — agricultural
* Rural — moorland
In each of these areas, the cartographers attempted to capture
all the features present in the specifications, at the various
mapping scales used in Great Britain. These scales are 1:1250
(urban), 1:2500 (rural) and 1:10 000 (mountain and moorland).
The features collected in this study included roads, railways,
tracks and paths, buildings, vegetation limits, water features and
field boundaries. In addition to the large scale specifications,
the images were assessed against the specifications of the
derived scales of 1:25 000 and 1:50 000. Note that the large
scale data is mainly used by the professional sector (including
national and local government, utility companies and
emergency services) while the smaller scale data is mainly used
to create paper products to serve the consumer sector
(especially the outdoor leisure market). Hence the
requirements of these two sets of products are quite distinct and
the product specifications reflect these differences.
2.2.1 Map Update Results
For each feature type, the cartographers recorded whether or
not the features could be successfully identified from the image,
using the specifications of each of the different mapping scales
as guidelines. Table 3 shows the results of this analysis. It was
found that many of the feature types that are required for
smaller scale mapping ( 1:10 000 — 1:50 000 scale) could be
satisfactorily identified and captured. In some cases, features
required for larger scale mapping (e.g. roads and woodland
boundaries at. 1:2500 scale) could also be identified. As may
be expected, the major exceptions to this are narrow linear
features (such as electricity transmission lines, walls, fences
and hedges), which are generally impossible to distinguish in
imagery of this resolution. A combination of panchromatic and
multispectral imagery can help to differentiate between
vegetation and artificial features (e.g. between hedges and
walls) but in general the imagery is unsuitable for the capture of
these narrow linear features.
When taken together, the results of the feature capture and the
geometric accuracy of the orthorectification indicate that
QuickBird imagery shows potential as a data source for
1:10 000 scale mapping at the current specification, and could
be used to derive topographic data up to scales as large as
1:6 000. The main drawback of the imagery is the inability to
resolve small linear features, which, if required, would have to
be captured in other ways. If QuickBird Imagery were to be
used as the sole data source, some changes to the Ordnance
Survey mapping specifications would be required. In a
commercial climate in which customers demand more and more
information, any weakening of the specification is not likely to
be well received. Hence it is likely that imagery such as this
Internati
should o
informat
Features
] Dual carr!
| Garage bl
; Major sez
: Non-cods
All veget:
! Major lan
: ROnds
, Airports
; Railways
; Non-resic
; Extensior
| Water fea
| Gants
] Heusinae
| Field bou
Minor prc
© Major prc
; Tracks &
Telephon
/ Electricits
: Tide lines
Table 3:
3.
31 Ch:
Althougl
geospati:
“where tc
intelliget
process.
of the m
the fielc
informat
Great [
supplied
consortië
To sup;
change,
