Research Methodology - Historical Land Use/Land Cover Classification Using Remote Sensing - page 69

Fig. 5.4 The distribution of
the 14-GCPs, used for
registration of the tow data
set (MSS-June-1975
and TM-August-2007),
image-to-image concept
(2) compute the transformation matrix using the GCP editor and the transformation
editor until the RMS error is small enough. A first-order polynomial was sufficient
for the transformation; and (3) resample the image data. The nearest neighbor
resampling technique was applied for rectifying the multispectral imagery.
For example, the geo-registration for the two remotely sensed data coverages
LANDSAT-MSS-June-1975 and LANDSAT-TM-August-2007, was carried out
using 14 GCPs (Fig. 5.4 ) which distributed across the image, especially on the
margins (the number was dependent on the size and image spatial resolution of the
used remote sensing data set). Table 5.1 lists the GCPs coordinates. It was simple
to gather and present good results. The nearest neighbor 1st order polynomial
correction was also used. According to the criteria presented in the literature of
remote sensing, the RMS error per image must be always less than the half of
spatial resolution of the image pixels, namely, \15 m (0.36) (Townshend et al.
1992 ; Mather 2004 ; Jensen 2007 ).
5.2.2 Atmospheric Correction
Electromagnetic energy detected and recorded above the atmosphere by remote
sensing sensors (here, those that work in the optical section of the EM spectrum/
especially in the visible and near-infrared regions) includes two sources of energy:
reflected and/or emitted from the ground surface; and energy scattered within and/
or emitted from the atmosphere. The quantity of this electromagnetic energy is
dependent on the quantity of exhaustive solar energy (irradiance), which is
reduced due to many factors: atmospheric absorption; the reflectance character-
istics of the various ground surface features; the differences in path length; the