The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
Figure 1 Position of test site 
3. METHODS 
3.1 Data processing 
The ASTER data used in this study are cloud free level 1B data 
acquired on November 24, 2001. The image has been 
pre-georeferenced to UTM Zone 50 North projection with 
WGS-84 datum. The LIB data are in HDF format which 
contain 15 bands image data, radiance conversion coefficient 
and ancillary data(Fujisada, 1998), when the data were imported 
into the ENVI processing system, the conversion coefficient 
was automatically applied and simultaneously the digital 
numbers of ASTER were calibrated to radiance. Only the first 9 
bands in VNIR and SWIR of ASTER were selected for 
subsequent analysis because the 5 TIR bands were not relevant 
to the reflectance of land surface objects. A Fast Line-of-sight 
Atmospheric Analysis of Spectral Hypercubes- FLASSH based 
radiometric correction was applied for the first 9 
bands .Furthermore, The bands processed above were then 
stacked into one file and resampled to 15*15 m pixel size 
using nearest neighbor algorithm .Finally a 1280*750-pixel 
image was clipped from ASTER data as test site. 
spectral radiances due to the topographic undulation under the 
same sun position. During the selection procedure of 
endmembers in LSMM, terrain undulation could pose 
ambiguity, and even lead to false selection or omission, 
accordingly further sway the unmixing result. 
Terrain illumination could be corrected by several 
methods, but previous literature shows that C-correction is the 
most effective illumination correction algorithm. (Teillet et al., 
1982;Meyer et al., 1993) .C-correction is therefore chosen for 
terrain illumination correction of ASTER data in our study. The 
basic C-correction formula used as follows: 
oos(z) =008(5) • cos(z) +sm(5) • sm(z) • cctfß-ß) 
(l) 
Lt = a+b-cos(/), c-a!b 
(2) 
T COS(z) + C 
Lh — Lt • 
(3) 
cos(z) + c 
Where , COS(z) : cosine of sun’s incidence angle for pixel 
To eliminate the terrain undulation ,A 1:10000 digital elevation 
model (DEM) covering study area was collected and then the 
aspect image and slope image were calculated from the DEM 
with ArcGIS platform. 
3.2 C Terrain undulation correction 
Optical imagery is usually affected by variations in brightness 
due to terrain. The terrain illumination are very common in the 
satellite imagery captured on undulating earth surfaces , and it 
could lead to that the same objects display totally different 
spectral radiance or contrarily different objects exhibit the 
similar spectral value. Specially an object lying in shadow 
receives less reflectance than the same object on the sunny side 
(Klaus I. Itten & Peter Meyer, 1993) . In a word, the same 
targets therefore despite their equal reflectance display varying 
i ; 
Lt : radiance observed for sloped terrain 
(before C-correction); 
Lh : radiance observed for horizontal surface 
(after C-correction); 
a> b : intercept and slope of the linear 
regression line; 
s: slope for pixel i; 
z: zenith of sun on image collecting time; 
(3 : azimuth of sun on image collecting time; 
/3'; aspect for pixel i ;