AUTOMAnc SEGMENT SHIFTER STUDY
GEORGE MAY AND NED JONES
.'.J. r ..
One of the tasks in preparing Landsat data for analysis is the proper location
of the ground-truth data (JES segment) in the Landsat scene. This task,
known as registration,
consists of two parts,
Methodology used by USDA, SRS is detailed
in a paper by Paul W.
In global registration, Landsat row-and-column coordinates are matched
with map latitude-and-Iongitude.
A regression equation is estimated
to [row, column] and vice versa. Global registration
gives approximate locations of segments in the scene (.:!:5 rows and.:!:5 columns).
is an exact alignment of the segment and the Landsat
This step utilizes a grey scale and plot corresponding to each segment.
is a computer generated representation
of one channel of Landsat data.
levels of reflectance
by levels of grey ranging from
black to white. The grey scale locations are calculated from the global-registration
function, and the grey scale edges correspond to a border of 20 pixels about the
segment •. The segment plot is drawn at the same scale as the grey scale.
segment plot is moved over the grey scale until
the best match of segment
patterns and grey levels is found. The difference between this best match location
and the global registration
location is referred
to as the segment shift.
The shift is
recorded as [.:!:rows, .:!:columns].
The sign indicates
the direction of the shift.
This allows the analyst
the exact window--that
is, box of pixels,--
containing the segment's Landsat data.
To assist with segment shifting, the Automatic Segment Matching Algorithm
(ASMA) was developed by Graham of NASA/NSTL (2). The algorithm compares
segment boundaries with pixel reflectance
to find the maximum difference
reflectance which coincides with the segment boundaries.