Abstract:
Abstract: Evaluated from remote sensing perspective, urban region is a real
district heterogeneous, what gives reflectance from different some land cover
type and material. The limitation of the spatial resolution from middle resolution
sensor such as Landsat requires analysis at level sub-pixel. Mixture pixel in
remote sensing data is one of the source of error in accuracy assessment result
in conventional classification. This research tries to apply Linear Spectral Mixture
Analysis (LSMA) method to detect land cover change (vegetation, impervious
surface, bare soil and water) at level sub-pixel in Banjarbaru City based on
Landsat temporal data. LSMA is approach with analysis sub-pixel which can give
information of the fraction in each pixel, so that is a potential solution to classify
one pixel. Maximum Likelihood Classifier applied as comparable from LSMA.
Accuracy assessment to this method use a higher spatial resolution IKONOS
image. Some processing phases applied in this research to increase the
accuration, are Atmospheric Correction, Minimum Noise Fraction (MNF) and
Pixel Purity Index (PPI). The percentage of each land cover component in each
pixel shown by fraction image from method LSMA with RMS Error average is
0,016 indicated that each endmember land cover has been dissociated well with
small deviation standard. The accuration test result of abundance for each
endmember using IKONOS image equal to 95%, indicates that LSMA have a
high accuration to detect the endmember land cover at level sub-pixel.
