is in
patial
cuses
'S are
as a
maps.
e the
'e, he
. All
his is
n the
'es in
0, the
"945
these
user-
The
> easy
This
fined
those
cial,”
t, the
m to
nitial
uced,
en a
0 the
lation
s are
ty, or
‘able.
flora
dary-
trees,
coded
restry
base
of the
| base
s the
Xf the
h the
ed to
work
other
amely
i] and
> map
dated
Data base. This solution, although simple to operate, is
costly in the long run, as the cartographers work will
duplicate for each map revision [Tait, 1991]. Here again, the
SOURCE-CODE attached to each feature in the GIS Data
base, plays a major role. The date of mapping is part of the
information within the SOURCE-CODE table. Thus, all
features acquired after the date of providing the cartographers
with data for last revision, are selected for the current map
revision process.
These features are collected in separate layers, according to
their TYPE-CODE, and are superimposed over the
cartographic copy files, using somewhat different symbols.
The cartographers may now detect the superimposed features,
and treat them according to their protocols. These treatments
include some changes that are made with the data depicted
already in the old file. Once all treatments are done, the
“new” features are imbedded in the “old” file in original
layers and hierarchy levels, in order to produce the revised
map.
Old files are archived in a separate workspace. This is done
both for the direct extraction from the GIS library, and for
the final cartographic copies. This is performed in the same
mode as keeping old separations, in order to keep track of
changes and to enable any repetition of the GIS Data base
updating, or map revision and reproduction. This protocol
has not yet been tested in its full length, but the process was
proven when adding additional features surveyed in-situ to
the cartographic copy processed before the field completion
step was implemented.
DATA BASE UPDATING
The National GIS Data base updating is planned as a 5-6
year circle. Within this time frame, all the data base should
be rechecked and updated. In addition, two more instances of
updating may take place: (a) When the amount of changes
detected will pass a certain criteria (not yet devised); and
(b) Under special order of digital data in a particular area.
This strategy aims to update the populated areas more
frequently, while keeping up with customer demands and
with the minimum updating time.
Change Detection
As reported by Peled [1994], the key to a successful updating
would be the development of advanced change detection
capabilities at the Survey. This year, only two quads are to
be updated; traditional processes are still in use. Comparing
two sets of air photographs, it took a trained photo interpreter
four days to mark the changes, using a work map, produced
directly from the data base, as reference. This map directed
the operators to map and acquire the new features detected by
the photointerpreter. Just recently, the Survey commissioned
à photogrammetric company to update a full quad. The
company is giiven an extraction of the GIS Data base that
will be superimposed on the stereoscopic model, and the
operators may map only the changes they detect. It was
clearly observed that, although possible, this procedure is
error-prone and time-consuming. The pre-process of change
detection is important, as it removes the burden from the
647
operators, who are then able to carry-out their work, free of
the tension involved in the original protocol. In addition , the
Survey has just made a grant to research and develop
customatic change detection capabilities. This research is
based on modular steps, starting with epoch-based change
detection to continue with GIS-driven change detection and
Data base updating [Peled, 1994].
Epoch-Based Change Detection
Epoch-Based change detection appears to be the simple and
rapid solution to aid both the updating priority and traditional
map revision, and as a basis for the GIS Data base updating.
The Survey of Israel is covering the country with 1:12,500-
scale photographs at two-year intervals. Experiments were
carried out to test whether these data may be used to detect
changes, and also to quantitize them. These photographs
were found very useful to achieve clear change detection.
Nevertheless, this process demands the scanning of thousands
of photographs, and the work involved with georeferencing
them, solely for the purpose of change detection, does not
seem worthwhile. Thus, the experiments are now focused on
two parallel efforts. Spot-HRV panchromatic images are
tested as a tool to quantitize the “amount of changes,” in
order to solve the updating priority problem. The question
here is to devise some area-wide criteria, as the geometric
resolution does not permit to continue with the freature-based
criteria devised for the 1:12,500-scale photographs. In a
second effort, epoch-based change detection is tested on
1:40,000-scale photographs. These are the same photographs
that are used in the remapping effort. As orthophotos are
becoming a by-product of the remapping project, this appears
an affordable solution, as most of the georeferencing work
has already been performed. The idea is to use satellite
images in order to define whether a certain area should be
updated. Then, by using the 1:40,000-scale photographs, a
second epoch-based change detection process will serve the
actual updating.
GIS-Driven Change Detection
The basic idea in GIS-Driven change detection is to use the
inherent information in the GIS data base, in order to find
the changes, toward updating the very same data base. This
research effort commenced only at the end of 1995, after
some progress was made in detecting changes in roads
[Peled, 1993], and buildings. Here again, modular process is
envisaged, first, to replace the photointerpreter’s work. The
idea is to use more lenient restrictions in the position
accuracy, and to serve the stereoplotter operators with a
change model that will also show the type of change detected.
In the long run, the idea is to vectorize the detected changes,
and to import them directly to the GIS data base, bypassing
the photogrammetric mapping process.
SUMMARY
Automatic change detection methods seem to be the only
solution to large-scale operation of map revision and GIS
data base updating. The transition from traditional mapping
to the new era of digital mapping is not straightforward. As
both processes are still operable side-by-side, a modular
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
