P. Boccardo : Environmental Impact Analysis Using Remotely Sensed Data
155
If the degree of understanding is sufficient or almost complete, it is useful to
proceed with a supervised classification that can link the known information (the
training samples and the derived spectral signatures) with the raw data stored in the
digital image.
Sometimes the information can be extracted by using different procedures (such
as radiometric processing) by means of low complex techniques and algorithms
(contrast stretching, digital filtering, etc.).
The methodology presented here is a combination of the different classification
algorithms (supervised and unsupervised) and of the simple radiometric
processing.
2.6 Thematic map generation and editing
The last step of this digital image analysis methodology is the generation of
thematic maps, that is the spatial representation of the analyzed phenomena. A
thematic map is the classification result corrected from the interpretation errors,
and plotted at the right scale.
In this map the information should not be redundant (otherwise it will be
necessary to reprocess the data redefining the spectral signatures), and the selected
classes (the organization of the information) should be consistent, both statistically
(classes should represent the analyzed phenomena) and geometrically (spatial
positioning should be precise).
3
CASE STUDIES
3.1 An experimental test to obtain a Bormida valley land
use map using Landsat 5 Thematic Mapper data
The first results of a land use study of a valley in the Piedmont region are shown
in this experimental test, using remotely sensed data. In particular, attention was
paid to the data collected by the American satellite Landsat 5 in seven different
spectral bands of the multi spectral Thematic Mapper scanner (T.M.).
The digital classification was conducted using different classification algorithms
(both supervised and unsupervised) and then compared to previous classification
data, carried out by using the classical photo interpretation technique on
aerophotogrammetric color pictures and field data.
Due to the particular conditions of the analyzed area (a hilly valley with frequent
small area agricultural crops), different digital image processing procedures were
used; the results obtained using these procedures were then merged in order to
extract as much information as possible from the different band images.
155
If the degree of understanding is sufficient or almost complete, it is useful to
proceed with a supervised classification that can link the known information (the
training samples and the derived spectral signatures) with the raw data stored in the
digital image.
Sometimes the information can be extracted by using different procedures (such
as radiometric processing) by means of low complex techniques and algorithms
(contrast stretching, digital filtering, etc.).
The methodology presented here is a combination of the different classification
algorithms (supervised and unsupervised) and of the simple radiometric
processing.
2.6 Thematic map generation and editing
The last step of this digital image analysis methodology is the generation of
thematic maps, that is the spatial representation of the analyzed phenomena. A
thematic map is the classification result corrected from the interpretation errors,
and plotted at the right scale.
In this map the information should not be redundant (otherwise it will be
necessary to reprocess the data redefining the spectral signatures), and the selected
classes (the organization of the information) should be consistent, both statistically
(classes should represent the analyzed phenomena) and geometrically (spatial
positioning should be precise).
3
CASE STUDIES
3.1 An experimental test to obtain a Bormida valley land
use map using Landsat 5 Thematic Mapper data
The first results of a land use study of a valley in the Piedmont region are shown
in this experimental test, using remotely sensed data. In particular, attention was
paid to the data collected by the American satellite Landsat 5 in seven different
spectral bands of the multi spectral Thematic Mapper scanner (T.M.).
The digital classification was conducted using different classification algorithms
(both supervised and unsupervised) and then compared to previous classification
data, carried out by using the classical photo interpretation technique on
aerophotogrammetric color pictures and field data.
Due to the particular conditions of the analyzed area (a hilly valley with frequent
small area agricultural crops), different digital image processing procedures were
used; the results obtained using these procedures were then merged in order to
extract as much information as possible from the different band images.