TWO DIMENSIONAL CODING OF STRUCTURE INFORMA- 
TION IN GIS FOR MINERAL RESOURCE PREDICTION 
Feng Ji-lu, Zhou Qin 
Institute of Remote Sensing, Peking University, P. R. China. IEAS, Academia Sinica 
ABSTRACT 
In the integration of geological datasets in GIS, while making mineral resource predication, structure information is 
always hard to be coded , because it is difficult to be quantified. This paper introduces two new approaches to code the 
structure information, [1] information intensity techniques for structureor texture operation of remote sensing images 
and [2] mathematical morphology technique for structure line operation or linear geological map. Results of practical 
application are effective. 
KEY WORDS: Structure, Coding, Intensity, Mathematical Morphology, Resource 
1. INTRODUCTION 
It has been proved effective to make mineral resource pre- 
[1] 
diction by multi-information synthesis in GIS*-. Among the 
fundamental geological materials, structure information 
(mostly faults on geological maps or linear structure on re- 
mote sensing images) is very important due to their close 
relation with resources, but it is hard to integrate with other 
data and put to further use. That is, it is still a problem to 
relate the geographical positions of a given structure to their 
corresponding attribute values. In the previous work, the 
working area was divided into many units of same size 
(Concept of " unit" was once recommended by IGCP in 
1975). Value of each unit is characterized by the structure 
length or structure number in it. But, this approach needs 
large unit area (not less than 1km X 1km), and the preci- 
sion of it does not match with the other newly employed 
geophysical / geochemicalor remote sensing data of high 
resolution (less than 100m X 100m). The new method in 
this paper does not use the old concept of "unit" , and create 
direct point model of relation between structure position and 
the importance (attribute). The coded information map has 
high spatial resolution, and may be put to integration with- 
out losing the resolution of other materials. The concept of 
direct point model is from the following basic geological 
ideas, 
[1] Any structure is not a theoretical line or curve, though 
represented by line and curve, but a zone, which is possibly 
fracture zone, alteration zone, plastic deformation zone or a 
strip constructed by many low order fractures. The anomaly 
caused by this belt is two dimensional. 
[2] The distance to a given structure central line is an im- 
portant factorre presenting the function range of the struc- 
ture. Generally, the closer to thecenter, the larger its im- 
portance is. 
[3] Occurrence of mineral resource is also related to struc- 
ture patterns. Many Gold mines occur near the cross-section 
465 
of multi-structure or many linear tones on image. But, 
these patterns still cannot be recognized effectively by 
computer pattern recongnition techniques. 
2. 2D CODING OF IMAGE STRUCTURE 
INRMATION 
Haluk Derin? (1986) once extracted strucutre and tex- 
ture information of graphics by Gibbs random approach. 
But, the following two aigorithm will work more effi- 
ciently for operation on remote sensing images and geo- 
logical datasets. 
2. 1 Techniques of Structure Intensity of Image 
For a digital image, any structure (including texture) 
may be ultimately expressed by definite number of line 
segments of different directions (ideas from differential 
calculus). If the directional information of each pixelon 
the whole image is figured out, structure information will 
be easily coded and put to further use. 
Let G(x,y) be a digital image. For any pixel (x,y), the 
parallel- Y directional structure information may be rep- 
resented by D(x,y), 
(x.y) . 1l 
ax G 
  
D( x,y) == 
wheressa Cyl - i560 JG i 053] 
So, one image of D(x, y) can be made out likewise for 
any given direction nwithin 0?— 360? by 
2G Go, y) % b. 
an G(x,y,n) 
Define I(x, y, n) as the structure information density of 
D( x:y»n) = 
direction perpendicular to n. Fig.1 is the comparison of 
the original pixel values and the structure information 
density at each pixel on one image profile. It is obvious 
that some densityvalues are negative. 
The structure information intensity perpendicular to n is 
given by IG,y n) 9