The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1139
3.3 TM and SAR Image Registration
Image matching error is impacted with results on land-use
change monitoring, to get reliable analysis results of the images
need high match accuracy. As a matching error of the element
space will result in 50 percent of the false changes. To control
the error rate of 10 percent, matching error on the image must
be less than 0.2 pixel, or more like small, the prerequisite and
basis of remote sensing image fusion is the registration for
multi-source remote sensing images, the registration error
requirements less than one pixel. TM and SAR image resolution
were 30 meters and 12.5 meters, a greater difference, and the
SAR images have noise dot, in order to improve the accuracy of
registration, uniform selection for homonymy points, as clearly
building, rivers changes smaller, texture is clear such as roads
for the registration, a total of 26, using a polynomial correction,
the error in control as 0.3 pixel. It was satisfied for the
registration accuracy requirements in the test.
4. TM AND SAR IMAGE FUSION
TM images received surface features information by the
electromagnetic spectrum stretching from 0 to 255 between the
Gray values; SAR images recorded form 0 to 65535 between
the integer values in the microwave band features the return
value, different Gray Represents the relative characteristics of
different features. The homology sensors, can direct use of
video Gray value integration and subsequent processing, and
sensors for different sources, because of its imaging, spectral
range and the characterization of surface features and other
characteristics are different, then different images Gray value
was not comparable, pixel Gray value for fusion will bring
about the corresponding error. Sensor characteristics need to,
through the use of physical corresponding model and the
reflectivity of the true features and value after the scattering,
accordingly for images fusion and information extracted.
The SAR and TM images using the fusion algorithm integration
of wavelet and HIS, its integration with the basic approach: the
largest amount of information on the composition of the TM5,
TM4, TM3 for IHS transform, I get the strength component
wavelet transform, using the wavelet function for Sym4 wavelet.
SAR image and the image / used Sym4 wavelet for a wavelet
decomposition, with the approximate SAR images AS replace
the approximate / image AI, to impose anti-wavelet transform
to generate 1' image and then IHS inverse transform, and you
will get a new fusion images of R\ G\ B' (Figure 4).
Figure 4. Fusion of TM and SAR images
5. MINE TYPICAL FEATURE EXTRACTION
According to the fusion of SAR and TM images, classified the
mining area features. In order to enhance the classification
results credibility of features, using remote sensing feature of
the land cover information from the tiered approach that,
according to the different and various targets Characteristics,
take the appropriate information extraction methods,
respectively, the establishment of thematic information layer,
and then merger the thematic information layer to the overall
classification, hierarchical extraction methods to take fully into
account the different characteristics of typical features, the
results of the classification can be making area measurement
accuracy estimates, such as change detection analysis, and
generate thematic maps throughout the category.
According to the land surface monitoring purposes, the study
area is divided into hydro-geological environment (include
natural water and geological conditions), agricultural land
(include grassland, paddy fields, dry land, others) and
vegetation growth, land use and land cover (include traffic,
residents, Mining land, forest land) and soil conditions as
several major categories. The special features which not
recognized by the fusion images, to further explore the solution
method of microwave remote sensing and interference SAR
images.
(1) Water extraction (canals, reservoirs, ponds and stagnant
water of the collapse)
Waters of the study area include canals, reservoirs, ponds and
stagnant water of the collapse, can exclude the water impact
from the mountain shadows and the shadow of buildings. On
the basis of the results from water extraction, conducted
extraction of canals, reservoirs, ponds, according to the
differences characteristics in space, such as canals were bending
the long strips; reservoir area are relatively large, border is
relatively smooth; ponds shape are sleek, oval or rules similar
to the quadrilateral. Because microwave remote sensing
particularly sensitive to water, on the radar vegetation canopy
soil or snow cover has some penetration, we can detect these
objects on the surface and sub-surface information, so it can use
the SAR images for the water body Monitoring.
(2) Building land extraction (urban land, rural settlements, an
independent mining site, transportation sites)
Towns, settlements, land transport and mining sites are easy to
distinguish from the fusion images, the merits to retain the
spectral characteristics of TM images, but also retained texture
information of the high-resolution SAR images.
(3) Green space and farmland extraction (paddy field, dry land)
The study areas include green woodland and grassland,
farmland including rice fields and mainly dry. In the TM
images similar spectrum information, information also is not
very clear after the fusion, distinction between the regional
crops can use SAR images to distinguish between them by
scattering differences, related to integrated data, will distinguish
types of crops.
(4) Mining subsidence land extraction
As Mining long time and scope, ground subsidence is quite
serious, especially in the rainy season there are stagnant water