International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
Landsat-7 ETM images includes the topographic maps, aerial
photographs, orthophotos and personal knowledge about the
area. However, before the image classification processes,
geometric correction of Landsat images was completed. For this
purpose, 21 uniformly distributed GCPs digitized from the
1:25,000 scales topographic maps of the interest arca were used.
Planimetric accuracy of these GCPs can be expected in the
range of 7.5m. On Landsat images, linear features appeared
sharp enough, so GCPs are mainly selected from road crossings
and bridges. Digital image coordinates for GCPs were measured
manually using the GCPWorks module of PCI system with the
sub-pixel point determination. Then, affine transformation was
applied between the GCPs's image and ground coordinates.
Root means square errors for X and Y directions were found to
be 0.69 pixels (20.7m) and 0.67 pixels (20.1m) respectively.
After producing transformation function, for image registration,
bilinear resembling method was used to determine the pixel
values to fill into the output matrix from the original image
matrix.
Table 1. Phenomena revealed by different bands of Landsat 7
ETM+ data
Band Phenomena revealed
Shorelines and water depths (these
n 45.0 5^ EY 3
0.45-0.52 jam (visible blue) wavelengths penetrate water)
Plant types and vigor (peak vegetation
0.52-0.60 um (visible green
Spm e green) reflects these wavelengths strongly)
Photosynthetic activity (plants absorb
0.63-0.69 pm (visible red) these wavelengths during
photosynthesis)
Plant vigor (healthy plant tissue reflects
-() © ar
0:76-0:90 um (near IR) these wavelengths strongly)
Plant water stress, soil moisture, rock
1 75 TD
1.55-1,75 pum nid IR) types, cloud cover vs. Snow
10.40-12.50 pum (thermal IR) Relative amounts of heat. soil moisture
Plant water stress, mineral and rock
types
2.08-2.35 um (mid IR)
4. CLASSIFICATION AND RESULTS
4.1 Pixel-based classification
Pixel-based classification of Landsat ETM image of interest
area was realized in two steps. In the first phase, ISODATA
(Iterative Self Organizing Data Analysis Technique) has been
applied and thus, spectral clusters have been determined, which
gives pre-knowledge about the site. Amongst the obtained
clusters, some have been eliminated or combined based on the
ground truth materials. Finally, 7 main classes have been used
as training areas for the classification procedure (see Figure 3).
Value Name Color
| |sea |
2 |damlake
3 |settlement areas|
_ 4 |dense forest |W
jopenareas |
coal waste —
wood land —
Figure 3. Result of ISODATA unsupervised classification.
In the second stage, supervised classification algorithms
(parallelepiped, minimum-distance and maximum-likelihood)
have been applied respectively to the Landsat image based on
the determined training patterns and reference materials. For
comparative analysis of each method, same training sites have
been utilized with the same colour information. Classifications
have been undertaken by the related module of PCI Geomatica
V9.1.2 software package and respective results are given in
Figure 4.
Name
a
sea
dam lake
settlement areas;
dense forrest
areas
coal waste
1
2
3
4
5
6
7
wood land
c. Result of Maximum-Likelihood classification.
Figure 4. Results of classical supervised classification
techniques (a. Paralelpiped, b. Minimum Distance and €.
Maximum Likelihood).
1120
Intel
4.2 (
Obje
para
scale
the
segm
Level
Level 1
Level 2
Level 3
Fig
From t
selecte
propert
With di
promin
Correpe
comple
and cla: