bul 2004
(3)
old map
it on the
e RMSE
MSE of
: formula
(4)
orithm
old map
essary to
everyone
lon are
merging,
or shape
t curve
between
ver, the
ationship
is small.
m cannot
Id maps.
), We can
ture and
feature
eated by
f the old
r can be
er can be
of back-
n and the
anges. It
mpletely
fers this
hm. The
buffer is
detection
aa
—
Un
—
(6)
(7)
hole line
in back-
in front-
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol XXXV. Part B2. Istanbul 2004
buffer, W is right factor of back-buffer, W
Mcr , 1s the right
beforc
factor of front-buffer. L
Luter after VS the length of old line
feature outside the back-buffer. L.. is the whole leneth
Total -afte: =
of old line feature, Louer —bejore IS the length of new line feature
outside the front-back, T ado is the whole length of new
line feature. Generally the right factor of front-buffer and back-
buffer can be defined as 0.4 and 0.6 separately.
(a) Old and new line feature (b) Front buffer |
a
(c) Back buffer (d) Double buffer
Fig.2 The principle of double-buffer detection algorithm (The
black middle line is an old line feature in 1:5000
map and the blue dotted line is a new line feature in
1:10000 map )
2.2.2 The formula for double-buffer detection distance
Different from buffer detection algorithm, the double-buffer
detection distance is composed of two buffer detection
distances: one is back-buffer detection distance and another is
front-buffer detection distance. Two factors including the
accuracy of new map and the accuracy of old map are important
for the definition of these buffer distances. Another important
factor for
the back-buffer is the generalization effect values and for the
front-buffer distance is the limited distance. So suppose the
RMSE of the detecting feature in old map is Jofd-map » the
T eR
£. 1. it P Ud
: “Fr t
ë E ES
a 14 T
5 > CHEAP
3 Pratt .
5 AB 1-7 L + 14 P
(a) The old chart in May,2001
(b) The new chart in June.2001
the
RMSE of the detecting feature in new map IS Open, map *
generalization effect value is 6,,, , the limited distance is
Climited » the formula for back-buffer and front-buffer detection
distance can be deduced as the following:
? 9
BufjerDis pa EL don map T O pew map + O gen
y ; | 2 à
- A — : em _ 2 s €
Buffer Dis Front = AO old -nmap * Ogsew _ map + Clim red (9)
3. CHANGE DETECTION BETWEEN NEW/OLD MAPS
(8)
The presentation and practice for change detection between new
and old map are mainly based on two reasons: one is that it is
necessary to compare and detect the changes between different
maps for checking compilation errors and controlling quality in
map production procedure; another is that it is a good idea to
reduce work burden using change detection algorithm for
quickly detecting changes before map generalization.
Obviously the former is the change detection between new/old
maps with same scale and the latter is for old and new maps
with different scale.
3.1 Change detection between new/old maps with same
map scale
Suppose new/old maps own same data model and format, then
the above proposed buffer detection algorithm is employed for
solving this problem. So the key problem for change detection
between new/old maps with same scale is depended on the
accuracy of detection features in new and old map. An example
for change detection between new and old charts with same
scale is shown as Fig.3.It can be seen that the positions of the
navigation marks in chart change in different time. In order to
emphasize the changes the old and new navigation marks are all
shown in the chart, shown as Fig.3(c).
= ub = matt s
" i S miu
» aa A 1 :
i = sf pus: " % if
= as E i bm P
; e i f a
3 i.t 5 *
4 ot "iF i 2 So fuu:
PO gh t : v
a f : i . H i
X >
SR
DUE
dy
^ 2
ü 5 QU
; & io BEN.
3 E =; t
(c) The change detection results
Fig.3 An example for change detection example between new/old maps with same scale using buffer detection algorithm
3.2 Change detection between new/old maps with different
map scale
Change detection between maps with different map scale
cannot ignore the effects of cartographic generalization. The
main effects caused by generalization lies on that it is difficult
to quantitatively describe and represent the relationship of the
features before and after generalization. In order to solve this
problem, the change detection expert system is employed
(Sui,2002). Considering the virtues of GIS, it is a good idea to
introduce change detection into GIS environment. On one hand
all kinds of GIS tools can be used for change detection and on
the other hand the change detection knowledge base can be
built based on GIS knowledge base so as to flexibly process the
generalization problems.
Many different change detection algorithms including
knowledge rules algorithm, spatial analysis and detection
algorithm, double-buffer detection algorithm and so on are
employed in detecting changes between new and old maps with
different scale (Sui,2002). One difficult problem for detecting
changes between new and old maps with different scale is the
displacement problem in generalization. When defining the
buffer detection distance, the distance and direction of
displacement must be considered. In general the displacement
objects can be selected according to the attributes or other
conditions before generalization. Sometimes when this
selection is very difficult the manual operation can be used for
this purpose in GIS environment. An example for detecting
