Download - Academy Publisher
Download - Academy Publisher
Download - Academy Publisher
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
xi,<br />
j<br />
y' = εα ( (1 + tanh( )) − y ) (2)<br />
ij , β ij ,<br />
W<br />
NDWI<br />
(, i j; k, l)<br />
=<br />
N (, ij) + Nkl ( ,)<br />
Nij (, ) − Nkl ( ,) + ε<br />
(7)<br />
S = ∑ , ( , ) ( ) ( )<br />
, W ij kl Hx k l +<br />
ij WHp −θ<br />
−<br />
p ij , p WHz − θ (3)<br />
z z<br />
kl , ∈Ni ( , j)<br />
⎡<br />
⎤<br />
P' = (1 − P ) H ⎢ H( x , ) P<br />
i, j i, j<br />
∑ kl −θ<br />
⎥<br />
p −ε<br />
⎢<br />
⎥ i,<br />
j<br />
⎣kl , ∈NP<br />
( i, j)<br />
⎦<br />
⎡<br />
⎤<br />
z ' = H ⎢∑<br />
H ( x kl , ) −1⎥<br />
−<br />
⎢<br />
⎥<br />
z<br />
⎣kl<br />
, ⎦<br />
where I i,<br />
j is the stimulate input of pixel (, i j ). x i , j and<br />
yi,<br />
jare two oscillators on different temporal scale. And<br />
ρ is the variance of Gaussian noise. S ij , was the couple<br />
equation. Wij,<br />
kl is the couple weight of pixel ( ij , ) and<br />
pixel ( kl , ). H is the step function, W p , W z are the weight<br />
coefficient correspond to P ij , , z is the globally<br />
inhibition. P ij , is the later potential of pixel ( i, j ), and<br />
θ<br />
p,<br />
θz<br />
are the threshold of Pi,<br />
jand z.<br />
Equation (1), (2), (3) and (4) are differential equations.<br />
The physical significance of these equations had been<br />
analyzed particularly and comprehensively by Terman<br />
and Wang in [12]. They had proposed a simplified<br />
LEGION method in [18]. The couple weight coefficient<br />
W ij , kl was the important factor, it decided what feature<br />
will be chosen to cut the image.<br />
III. THE LEGION BASE ON NDWI<br />
A. The model of NDWI<br />
The NDWI was defined as follow:<br />
Green−NIR<br />
NDWI =<br />
(6)<br />
Green+<br />
NIR<br />
Based on the NDWI, the distance between water and<br />
other ground objects was enlarged. But because of the<br />
pollution and the magnitude of sandiness, the coherence<br />
on the water surface will be reduced and the NDWI<br />
fluctuated above zero.<br />
Fig.1 shows the spectrum change trend of water area,<br />
vegetation area and building area around the river. Like<br />
paper [5], the spectrum of building on the forth band<br />
(TM4) and the second band (TM2) had the similarity<br />
change trend, but with the different mean. Fifteen sample<br />
points of NDWI were plotted in Fig. 2. The variance<br />
within the water body and the building body were<br />
smaller, and the variance between water and building<br />
was larger.<br />
B. The couple weight based on NDWI<br />
From the above analysis, construct one couple weight<br />
based on NDWI to describe the relationship of pixels:<br />
(4)<br />
(5)<br />
Where N (, ij ) was the NDWI value of pixel Ii (, j. ) The<br />
equation (2) could be rewritten as follow:<br />
S = ∑ W (, ; ,) ( ) ( 0.5) ( 0.5)<br />
, NDWI ijklHN + WHp − −WHz− (8)<br />
ij ij , p ij , z<br />
kl , ∈Ni (, j)<br />
• C. LEGION based on NDWI<br />
The special spectrum property was introduced into the<br />
proposed extraction method which was found by Zhou.<br />
The special spectrum property of water is:<br />
B ( green) > B ( SW ) & B ( red ) > B ( IR)<br />
(9)<br />
3 4 2 1<br />
So the equation (8) could be rewritten as follow:<br />
S = ∑ (, ; ,) ( (, )) (( 3( ,) 2( ,))<br />
ij ,<br />
W NDWI i jkl H Ni j H I kl + I kl<br />
kl , ∈Ni ( , j)<br />
> ( I ( k, l) + I ( k, l))) + W H( p −0.5) −W H( z− 0.5) (10)<br />
4 1 p i,<br />
j z<br />
Where I<br />
1 ( k , l ) represent the spectrum luminance of<br />
pixel I( k, l)<br />
of the first band. H( x)<br />
was the step function,<br />
and H( N( i, j))<br />
compared the NDWI value of pixel Ii (, j )<br />
with zero. Adding H( N( i, j))<br />
onto the second term was to<br />
make the object to be remained which was with the<br />
NDWI larger than zero.<br />
spectrum value<br />
Figure.1<br />
NDWI<br />
0.60<br />
0.40<br />
0.20<br />
0.00<br />
TM1 TM2 TM3 TM4 TM5<br />
water 0.43 0.40 0.47 0.12 0.05<br />
vegetation 0.33 0.28 0.21 0.52 0.48<br />
building 0.48 0.42 0.47 0.31 0.47<br />
0.60<br />
0.50<br />
0.40<br />
0.30<br />
0.20<br />
0.10<br />
0.00<br />
b d<br />
The spectrum change trend of water area, vegetation area<br />
and building area around the river<br />
0 2 4 6 8 10 12 14 16<br />
sample points<br />
NDWI of water and building<br />
Figure.2<br />
building<br />
NDWI of water body and building body<br />
water<br />
259