Geography Reference
In-Depth Information
the data file; it is simple and rapid; and the main drawback is the stair stepped
effect (Liu and Mason
2009
).
Ground control points (GCPs) are pixels with well-defined positions in an
image for which the output map coordinates are previously definite. They must
have the following conditions and characteristics: (1) they have to be recognizable
with a site both on the image and the real world surface; (2) they are accessible in
the field; (3) they are consistently located within the study area of interest; (4)
there are sufficient of them; (5) they have a small feature size; and (6) they have to
be fixed over time. The most frequently used method to select these GCPs is the
visual method (Liu and Mason
2009
). If the point features to be GCPs are difficult
to be exactly located on an image, it is better to select the ground object features to
be a GCP (e.g., intersections of linear features). In an image-to-image registration
model, intersections of highways or main roads are frequently used as GCPs. The
next mathematical statement determines the minimum number of GCPs to be used:
Minimum
number
of
GCPs = ((n1)(n2))/2,
where
(n)
is
the
order
of
polynomial.
To obtain superior classification results, additional GCPs to the minimum
number are commonly used. There is an error measurement technique that can
compute the correctness of selected GCPs. It named the Root Mean Square (RMS)
error, which is the distance between the input (source) position of a GCP in the
input-matrix and the re-transformed position for the same GCP in the output-
matrix. RMS-error is computed using the next mathematical statement:
q
x
r
x
i
Þ
2
þð
y
r
y
i
Þ
2
RMS
Error
¼
ð
where: x
i
;
y
i
: The input source coordinates; x
r
,y
r
: The retransformed coordinates.
The RMS error will be determined for each GCP. Then, the total RMS error
will be computed by calculation the all RMS error of all GCPs. In the third step,
the RMS error will then be tested for accuracy. If the overall RMS error is not
good enough, then those GCPs with high RMS errors must be removed. This
previous step will be repeated until the RMS error is good enough.
For comparison and combination based studies that use diverse sources of data
and information, like remote sensing imagers obtained from diverse sensors (e.g.,
MSS, TM, etc.), field reference points (e.g., GPS-points), topographical data (e.g.,
DEM) and other available data for a study area, it is important to transfer all these
data into a reference cartographic projection system; the result of which is a
generally suitable data basis. The ERB projection parameters are: (Projection:
UTM, Ellipsoid/spheroid: WGS84, Datum: WGS84, Units: Meters, Zone:
37 North). The study area is in one UTM-zone (37 N), which was an advantage for
this work, since no geometric problems occurred due to changes between two
UTM-zones.
All LANDSAT-data with their different sensors have no spatial deformations
among them. However, during the connection and use of mosaic scenes, which
result from gathering individual sensor images together, it was necessary to reg-
ister the mosaic scenes to each other by accrediting one scene as the master-scene
Search WWH ::
Custom Search