ed one
whole
'SSary.
re map
ou can
bvious-
e com-
be ex-
istance
senting
yurces.
s given
1e two
fore di
erally ,
cal ap-
| expo-
of
rich no
ition €
ructure
map was input into GIS by digitizer and covert to binary
grid map (see Fig. 6.1). Because the structures are all
regional deep-cut faults, and all mineral resources occur
within 5Km of faults, DO was assigned 5Km by geolo-
gists. k was obtained by regression analysis (using DOS
compatible software SYSTAT) of D and fault development
degree at typical mining filed. At Liangcheng, (2) be-
comes
1.0 * EXP(— 0.67 xD), DX 5.0 Km
0.0, D> 5.0Km
IS =
and image of IS was generated after 14 times dilation
(Fig. 6. 2). The values(0. 0<IS(x, y)<]1. 0) of this im-
age represents the qualified influence of structures in this
arca. Fig. 6.2 has been used to predict gold mines at
Liangcheng.
3. CONCLUSIONS AND DISCUSSIONS
The main achievement of this paper is that it provides two
" point" model toextract the structure information on re-
mote sensing image and linear geological map respectively.
This makes it possible to integrate structure information
with other high resolution data in GIS. However, struc-
ture information is very complicated. In (2) , DO isa
threshold given by experienced geologist. It will be differ-
ent if given by different expert, and make the result image
indefinite. So, how to find the objective DO or give a
proper distance threshold is subject to further research.
4. ACKNOWLEDGEMENT
The author would like to thank Professor Xianglin Qian,
Department of Geology, Peking University, Shiwei Shu,
Department of Information Science and Remote Sensing,
IEAS, Academia Sinica, for support.
REFERENCES
[1] Bonham —Cartter, G. F. , Agterberg, F. P. , etc. ,
1988. Integration of geological ^ datasets for gold explo-
ration in Nova Scotia. Photogrammetric engineering and
remote sensing. Vol. 54, No. 11, pp. 1585— 1592.
[2] Singer, D. A, 1981. A review of regional mineral re-
source assessment methods, ^ Econ. Geolo., Vol. 76,
No. 5, pp. 1006— 1015.
[3] Haluk Derin, 1986. Segmentation of textured images
using Gibbs random fields. ^ Computer Vision, Graphics,
and Image processing. 35, pp. 72— 98.
[4] Serra, J. , 1982. Image Analysis and Mathematical
Morphology, Academic Press, London. pp.110—175.
467
Structure Density Profile
e
0 p 30 40
Pixel Positions
Fig. 1 Comparision of original pixel values and
the corresponding structure information
density values on one image profile
Structure Intensity Profile
Original
htensit
20 30 40
0 10
Pixel Positions
Fig. 2 Structure information intensity values
of each pixel on the same profile as Fig. 1
