misrecognized polygons is easy to carry out. After 
the process of extracting the vertexes of the recog- 
nized polygons, the hatched polygons in vector 
form can be used in geographic information sys- 
tems. 
At last, a process of deleting the recognized 
hatched polygons from the original image I (i,j) 
should be performed. It is very helpful to recognize 
other kinds of map information. ; 
Ya 50 
055 © 8% 
228$ 
qb 
00s 
Fig. 13 The result of recognized polygons 
8. EXPERIMENTS AND CONCLUSIONS 
In order to verify the feasibility and the efficiency 
of the method proposed in this paper, we arbitrari- 
ly select a 1 : 50000 topographical map as experi- 
mental data. The scanning resolutions in both hor- 
izontal and vertical directions are 250dpi. This 
piece of map includes 834 hatched polygons as well 
as many other kinds of graphics. Among the 
hatched polygons on this map, 829 hatched poly- 
gons are correctly recognized and extracted with 
the proposed method. The ramained 5 hatched 
polygons are rejected because they are wrongly 
drown with unclosed borders on original map. Be- 
sides the recognized hatched polygons, 26 polygons 
standing for parts of Chinese characters are misrec- 
ognized as hatched polygons. The final ratio of 
correct recognition of hatched polygons is 96%. 
Experiments show that the automatic recognition 
method for hatched polygons put forward in this 
paper is effective and feasible. It can reach a high 
ratio of correct recognition as about 96%. The 
misrecognized ones are exclusively parts of chinese 
characters. This means that a even higher recogni- 
tion ratio of hatched polygons may be reached if 
342 
this method is performed to maps of western coun- 
tries since the western texts on maps are more dif- 
firent from hatched polygons. The method has a 
strong capability of anti — interference. The 
hatched polygons linking to other graphics in origi- 
nal scanning image can be correctly recognized and 
extracted. According to the recoginition strategy 
of layer by layer, it is important to remove the 
recoginized map symbols from the original map im- 
ages because it will be helpful to extract other lay- 
ers of map information afterwards. 
References 
Capson, D. W. 1984. An improved algorithm for 
sequential extraction of boundary from a reaster 
scan. Computer Vision, Graphics, and Image Pro- 
cessing, 28(109). 
Fulford, M. C. 1981. The FASTRAK automatic 
digitizing system. Pattern Recognition, 14(1). 
Wahl, F. M. etal. 1982. Block Segmentation and 
text extraction in mixed text/image documents, 
Computer Graphics and Image Processing, 2 (6) 
pp375— 390. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996 
  
KEY 
ABS 
Kobi 
and 
Instit 
docu 
simil 
prim: 
Worl 
Com 
area 
disas 
groui 
geot 
Com 
mate 
Digit 
coop