hul 2004
with the
y. Both
lem lies
half of
er-class
(lower-
eak and
SS roads
ance on
t on the
the 60-
ds, not
may be
| size.
n, ar
xel size
ovided)
ch with
maging
use of
rol on
0 road
pleted.
or the
er and
inel. of
ension
orders
future
led for
Office
vided
ieneva
'omial
ng Li,
Zurich
F road
"Data
hede,
IMI —
from
ation.
nsing
1998.
axes.
1-94.
se by
and
79.
Es for
n of
y and
UPDATING MAPS IN A WELL-MAPPED COUNTRY USING HIGH RESOLUTION SATELLITE
IMAGERY
David Holland (corresponding author), Paul Marshall
Ordnance Survey, Romsey Road, Southampton, SO16 4GU UK
david.holland@ordnancesurvey.co.uk paul marshall@Gordnancesurvey.co.uk
Inter Commission WG II/VI
KEY WORDS: High Resolution, Satellite, Change Detection, QuickBird, Mapping, Cartography
ABSTRACT:
Since the successful launch of IKONOS in 1999, a new source of imagery has been available to the civilian spatial-data user. In many
parts of the world high resolution satellite imagery from KONOS, QuickBird and, more recently, ORBVIEW 3, has proved to be a
useful data source for the creation of orthorectified images and associated mapping products. One of the great advantages of satellite
imagery is the ease of access to areas which have previously been too remote or too dangerous to reach using conventional aerial
photography. However, for areas of the Earth which are not difficult to reach, and which have a tradition of high resolution mapping
from aerial photography, this advantage is of rather limited importance. Ordnance Survey, Great Britain's national mapping agency,
currently makes extensive use of aerial imagery in the collection of large scale geospatial data. In 2003, a project was initiated
which would determine whether satellite imagery could replace, or complement, aerial photography in this data collection process;
or could be used in other ways within a production environment to make the process more efficient. Interim results of this research
are presented in this paper.
1. INTRODUCTION
High resolution images from satellites such as IKONOS and
QuickBird have proved their usefulness over the past few years,
especially in the mapping and surveillance of otherwise
inaccessible areas, for example in areas of military conflict,
such as Afghanistan and Iraq (see, for example, Kumar and
Castro, 2001, Petrie 2003). Such images have also been used to
update maps, or generate completely new mapping, in many
areas of the world, including Saudi Arabia, Indonesia
(Mandeville, 2001) and Alaska. In most cases up until now
these applications have been in parts of the world that do not
have a tradition of detailed mapping. In these areas, which
often do not possess the resources required to collect and
process aerial photography, high resolution satellite imagery
can provide a rapid, high-quality data source for the production
of image maps, thematic maps and vector-maps, to satisfy a
variety of needs.
In traditionally well-mapped areas of the world, such as
Western Europe, the position is different. In these areas,
detailed databases of geospatial information have been built up
over many years, using both field survey and photogrammetric
techniques. In addition, the infrastructure required to capture
and process aerial photography is already well established. In a
country such as Britain, the maintenance of existing mapping,
rather than the creation of new information, is the main
business of the national mapping organization. High resolution
satellite sensor imagery has only recently been adopted as a
data source in large projects in these well-mapped areas of the
world (e.g. European Space Imaging, 2003). This paper seeks
to discover whether high resolution satellite imagery could be
used to assist in the revision of mapping in these well-mapped
countries; specifically in Great Britain.
At Ordnance Survey, Great Britain's national mapping agency,
a project was initiated in 2003 to investigate the potential of
QuickBird imagery as a source for updating mapping at mid
scale (1:25 000 and 1:50 000) and large scale (1:10 000 and
larger). The initial findings of this research were presented at
the ISPRS *High Resolution Mapping From Space" Workshop
747
in Hannover [Holland and Marshall, 2003]. During this
research, the use of QuickBird imagery for topographic
mapping, change detection, and quality auditing was
investigated. The initial findings indicated that it was feasible
to produce a topographic map at a scale of 1:6000, using
QuickBird imagery. This paper follows on from the above
research, and presents new findings from the second phase of
the project. At the time of writing, a full production trial is due
to start soon, in which a mid-scales mapping product will be
updated using both high resolution satellite imagery and a
traditional revision method. Two further aspects of the project
have already been completed: an assessment of QuickBird
imagery for topographic change detection; and an evaluation of
QuickBird imagery for auditing purposes. The results of these,
and a recap of the results of the original map update project, are
discussed in this paper.
2. UPDATING MAPS FROM SATELLITE IMAGERY
Ever since the first commercial high-resolution satellite sensors
were heralded in the mid 1990s, mapping agencies have shown
a keen interest in the possible use of satellite imagery in their
data collection programmes. Since as long ago as 1996
(reported in Ridley ef al, 1997) Ordnance Survey has been
investigating this potential, initially using images synthesized
from aerial photography, more recently using satellite imagery
from the IKONOS and QuickBird sensors. A recently-published
OEEPE (now EuroSDR) report presented findings of an
investigation into the potential of IKONOS data for mapping.
undertaken by several European mapping agencies and
institutions (Holland e/ al. 2003). The results suggested that
there is potential in this field, especially in rural areas at scales
between 1:10 000 and 1:50 000.
To extend the OEEPE research further, Ordnance Survey
purchased QuickBird images of several areas in Great Britain
and began a series of investigations into the potential of such
imagery in a geospatial data production context. The study
arcas covered several different types of land, including urban.
suburban and rural, and are listed in Table ]. The images were