lld be
[ON
on of
lucing
uld be
uding
ce the
imited
color
itor to
rithm
hm as
ion is
varies
e ofa
ablish
t here
Roads
n any
1uman
)ne as
2 ona
st site
| to 24
colors
ige of
e by
ion of
5, Was
'arried
mbers
pixels
ixel in
pixels
value
t was
r each
paring
10ises,
t 1084
laying
cen on
it. By
some
re this
but it
cannot be used for automatic revision. Another problem is that
it is not clear whether this method can improve efficiency of
change detection because this method requires conversational
operation of a human operator.
de. vs à “a 3 -— 4 Á4 2 e
51 [ti a it Er oy *
EU Y q^
Tk
s gm
* ^ - A a act x x v Y
xz 1 ^ d os i
M ee à
Tarn
YT. ie Pr,
; x e ^ i
Figure 3 Extracted road components of 1984
—
Figure 4 Extracted road components of 1992
4. ENHANCEMENT OF EFFICIENCY OF MANUAL
CHANGE DETECTION WITH OLD MAP DATA AND
NEW AERIAL PHOTOS
The objective is to make it easy to manually identify changed
areas in updating work of digital cartographic data. In this
situation, unrevised digital map data are available. Hence these
old map data are used to mask out unnecessary part on the new
aerial photos which are used to find out changed parts. Another
idea is to apply an emboss filter onto aerial photo images to
enhance visibility of images for change detection purpose. The
efficiency was tested by measuring the time required for
revision work.
4.1 Method
We set up here the target of detection to be the new
construction of houses (disappearance of houses is ignored
551
here). Then already existing houses and buildings need not be
checked. It is unlikely that new buildings are constructed on
existing roads. Therefore existing houses and road surfaces are
masked using old map data and the remaining part on the photo
was checked by operators on the CRT. Required time for
revision work and its accuracy were tested. Single aerial photo
is used and a mask made of digital map data is overlaid on it
after projection transformation to locally register the position.
Masks of existing houses are expanded with 5 m width in order
to fill out unnecessary space between houses.
Emboss filter modifies an image like a relief picture and
eliminates unnecessary details. This filter was applied together
with the mask and the effect was examined.
Examples of masked aerial photo image and that with emboss
filtering are shown in Figure 5 and Figure 6.
The following data are used for the test. These are the same data
used in section 2.
® Old digital map: Fukuyama city, 1:2,500, 1987
€ New aerial photo: ditto, 1:25,000 B/W, 1994
scanned in 25 xm sampling pitch
Figure 5 Masked aerial photo image
Figure 6 Masked aerial photo image with emboss filtering
4.2 Results
Four test areas are selected (These are different from those of
section 2). Time of digital map revision work was measured for
a non-experienced operator and an experienced operator. The
results are shown in Table 5 and Table 6.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
