: of
sin-
it is
lly.
de-
mn
the
the
the
up-
ting
| its
ted
on.
sing
f its
DIO-
axi-
rch-
nate
GE
9 ac-
«Jj
ofa
y be
'hite
elet-
ed from image R (i, j). It is possible that a few
polygons link to one another by single lines. These
linked polygons are traced as a complex polygon
and it is necessary to seperate them into indepen-
dant polygons for the automatic recognition of
hatched polygons. The approach to seperating the
linked polygons can be related as following two
steps. ;
a)Removing the pixels on single line linking poly-
gons. The pixels on single line linking polygons
are traced twice in polygon tracing process. If
there are two pixels, say (xi,yi) and (x;,y;), a-
mong the border pixel set of a polygon with the
same position, then these pairs of pixels are re-
moved from the border pixel set. This condition
can be expressed by following formula.
az; A;
VV
where i,j€ [1,n],izZj.
b)Regrouping the remained border pixels into inde-
pendant polygons.
7. RECOGNITION OF HATCHED POLYGONS
Among the ploygons in image R (i,j), there are
some other loops such as symbols of control points
and parts of some Chinese characters besides the
hatched polygons. A decisive step to recognize the
hatched polygons is to exclude the polygons that
do not represent residential sections. By analysing
the charaterics of polygons in image R (i,j), we
can classify them according to three criteria.
a) The circumstance of a polygon. Generally speak-
ing, a hatched polygon is with a longer circum-
stance, which is measured by the number of border
pixels, than other polygons. Suppose n stands for
the number of border pixels of a polygon, then the
polygons can be classifed into two groups, large
polygons and small polygons. If n is larger than a
threshhold N, e. g. N35, then the polygon is
classifed as a large polygon. If n is less than N,
then the polygon is classifed as a small one. Only
the large polygons have the chance to be recognized
as hatched polygons. :
b)The number of black pixels inside a polygon in
image R (i,j). A hatched polygon is with some
black pixels inside it in image R(i,j). On the con-
trary, the paygons standing for symbols of control
points and some Chinese characters do not include
any black pixels inside them. Suppose n; refer to
the number of black pixels inside a polygon in im-
341
age R(i,j), then the polygons can be classified into
two groups, ploygons with black pixels and ploy-
gons without black pixels. If n,>0, then the poly-
gon is one with black pixels. If nj — 0, then the
polygon is one without black pixels. The polygons
without black pixels inside them in image R (i, j)
are definetely not hatched plygons.
c)The ratio of black pixels inside a polygon in orig-
inal image I(i,j). There are still some graphics of
Chinese characters that can not be distinguished by
the above two approaches from hatched polygons |
in image R (i,j). It is discovered that most graph-
ics of Chinese characters are with a higher ratio of
black pixels inside the polygons in original image I
(i,j) than that of hatched polygons. Suppose r
stand for this ratio, then r can be obtained by fol-
lowing formula.
r = nı/n
where n; and n respectively refer to the numbers of
black pixels and of total pixels inside a polygon in
image ICi,j). :
The ratio of black pixels inside a hatched polgyon
mainly ranges from 0. 3 to 0. 6 and that of other
polygon ranges from 0. 6 to 0. 9. The distribution
of r is shown in Figure 12.
n A
— for hatched polygon
- for other polygon
Fig. 12 The distribution of the ratio of
black fixels inside polygons
As shown in Figure 12, the ratio of black pixels
for hatched polygons and other polygons have ob-
viously diffrent ranges. By choosing a threshold R
=0. 6,most of the hatched polygons can be seper-
ated from others. The result of polygon recogni-
tion is shown in Figure 13. Although all of the
hatched polygons can be correctly recognized with
the method described above, a few polygons stand-
ing for Chinese character are misrecognized as
hatched polygons at the same time. These misrec-
ognized polygons should be removed by interctive
editing. Forturately, the operation of removing
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
