ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001
133
)n study
;ting the
to auto-
film, we
>are with
ind band
re. If the
idow the
:eled. So
ot image
bands of
hen build
from all
ige Grey
orrelation
ess of oil
And use
•iminating
to extract
d-oil
a Surface
reflectivity
ndency in
D.033 and
in red light
so low. It’s
1.005. The
range of peak value’s change is correlated with the thickness
of oil film and the category of oil. Heavy diesel oil’s peak value
decreases more distinctly than others. Lube and light diesel
oil's reflectivity will increase as the thickness of oil film adds.
The amplitude of lube is greater, while the one of light diesel oil
is little. The peak value change of coal oil is similar to the one
of heavy diesel oil, but the range of drop of it is not as
intensive as the one of heavy diesel oil. Crude oil is very
special. According to spectrum data line of two kinds of crude
oils that are Middle East crude oil and Daqing crude oil
selected by experiment. Middle East crude oil that compromise
little quantity of waxiness has the obvious wet characteristic
and can form oil film. When this oil film is thin, its reflectivity is
higher than seawater’s. When oil film becomes thicker its
reflectivity is lower than seawater’s. While Daqing crude oil
compromise more quantity of waxiness. It is very mucous,
easy to agglomerate and close to solid. Its average reflectivity
is low.
Considering from the image feature the image hue of heavy
diesel oil is darker than water’s in channell 2, 3 of TM.
Information we can get from channel 4, 5, 7 are very little. Hue
is light in TM6. Lube image has delicacy information in all
channels; the information is strong in TM3, TM4, TM5, and
TM7. While in TM1, TM2, TM3 information is delicate and hue
of image changes from light to dark. The reflectivity of light
diesel oil is between heavy diesel oil and lubes. Its spectrum
data line is similar to seawaters’. Only when thickness of oil
film that is in 0.4 - 0.45 nm is not great there are differences .
When the thickness of oil film increase it is in 0.45 -0.57 m m.
Therefore we can get obvious oil film information in TM1, TM2,
TM3 and CH1, CH2, CH3 from AVHRR. However light diesel
oil's film thickness has very little influence on its reflectivity.
3. The Model of Extracting Spilled-oil Information and
Information Extraction
By analyzing the spectrum feature of spilled-oil and processing
image datum, make sure of the hot-spot area of spilled-oil, and
take synchronous sample from all channels in this area on the
base of it. The processing is as follows:(fig 1)
Fig. 1 Information Extracting process
According to the results of sample analysis form a mode curve
as practice sample, and then decide if pixels in image accord
with discriminating condition of a certain set by computer
program. When original data are transferred to EMS memory,
program check out every pixel, put the one into a virtual set
Rnoaa ~ A + YM
/=1
and endow with a special value according to a certain
discriminant. In the end output the result, which is the
processing of auto-extracting the sea oil information.
The key problem to extract oil film information on the surface of
sea by Remote Sensing is to set up the quantificational
relations between Remote Sensing data and a certain thickness
of oil film. Based on the result of spectrum feature data line all
kinds of oil film’ reflectivity feature data line is different from
each other, but beside data of the reflectivity feature data line
that is based of the different thickness but the same kind of oil
film is different. The shape of data line is same basically. So we
can believe that the thickness of oil film has direct ratio relations
with reflectivity of oil film, select reflectivity of feature data line of
a certain thickness of oil film of every kind of oil to multiply a
certain different coefficient. Then reflectivity of feature data line
of another thickness of oil film can be obtained.
Assume that the function value of f(x) in different Point is
known: f(X1), f(X2) ,f(Xn). We can construct a simple
multinomial by interpolation.