The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1138
Data
source
( Sensor)
Wav Spatial _ . Image
. . Polan .. Imaging
e resolutio . positio .
. , zation , time
band n n code
Landsat-7
(ETM+
)
Landsat-7
(ETM+
)
Envisat-1
(ASAR
)
Envisat-1
(ASAR
)
1 8 30 200L
1-8 30 - - 08 U
1 8 30 2005 -
30 " “ 09.03
D C ~ nc VV / 2001.6.2
Ban 12.5 IS2 n
j HH y
a
n' n C VV / 2005.7.1
Ban 12.5 IS4 0
. HH 2
d
Table 2. Remote sensing data sources
3. REMOTE SENSING DATA PREPROCESSING
3.1 TM Data Preprocessing
The pre-treatment of TM images including geometric correct
and optimal band combination. In the study area collected
obvious 10 GPS control point of a uniform distribution, for the
ETM+ images geometric correct, adopted Second Order
Polynomial transform, and used three-dimensional convolution
interpolation re-sampling, the precision of geometric correct
was controlled of the RMS <0.5 pixels. The ETM+ optimal
band were combined according to the study area topography,
vegetation distribution, spectral reflectance characteristics.
According to the index-related variance correlation
coefficient matrix, OIF index of the combination bands, in most
cases, TM3, 4,5 and TM1, 4,5-band combination are the best,
most detailed. In the TM345-band, roads, water, the residents,
farmland and other places clear Comparison, better-performing
and useful for the classification between features (Figure 1).
Figure 1. TM345-band combination images
3.2 SAR Data Preprocessing
As ENVISAR satellite imaging to be done in the course of
coherent processing, and thus formed a signal to the noise, it
made great impact with the interpretation of radar images,
reducing the ability to identify targets and structure on the
ground, making thematic feature extraction and classification of
Obstacles, so need for Noise Reduction. After noise reduction,
evaluated the filtering effect with smoothing index F and
smoothing maintain index E of edge. F value is higher means
that the ability smoothing stronger, E value is higher means that
the ability to maintain higher, the formula is as follows:
£|(g„,-g„)|
E =
after filtering
SK G *,- G *2l
before filtering
(1)
The SAR data image pre-processing according to the landscape,
the typical characteristics of surface features and ENVISAT
data in the mining area. As the study area relatively flat terrain,
a typical feature mainly vegetation, soil, water, the residents,
farmland, respectively KUAN filtering, enhanced Lee filtering,
enhanced filtering FROST, GAMMA filtering done a test.
Texture complex information to residents, attention should be
paid in noise reducing, at the same time try to avoid the loss of
texture information leading to the fuzzy of brink; farmland
relatively homogeneous, not too rich texture information,
speckle easy to filter out, but to maximize the ability to
maintain the edge; water for the homogeneous region,
regardless of ability to maintain the edge. Taking all these
factors, according to smoothing effect and maintain ability of
the edge after noise reducing to choose the enhanced Lee
filtering (Figure 2, Figure 3).
Figure 2. Original SAR image
Figure 3. Enhanced Lee filtering