ABSTRACT
Using remote sensing technology is an effective way to observe and verify the wise and
sustainable use of natural resources. With its weather and daylight independency Synthetic
Aperture Radar (SAR) data emphasize as particularly suitable for large-scaled monitoring. But
the classification of SAR images is more difficult as of multi-spectral images. Because of the
microwave’s interaction with the object parameters on the earth surface, like roughness or the
complex dielectric constant (CDC), different types of land cover respond very similar. So
separability is not possible without additional information. For the investigated multi-channel
SAR scenes a typical example for this problem is the distinction between rice crops sown on …show more content…
Region of Interest (ROI) of the six different classes are soil,
urban, water bodies, soil with higher moisture, rice date 1 and rice date 2 were created on both
the image. And then classify them to calculate overall accuracy and kappa coefficient.
Confusion matrix is created based on these two classifications.
Where: i = class
x = n-dimensional data (where n is the number of bands).
= probability that class occurs in the image and is assumed the same for all classes.
= determinant of the covariance matrix of the data in class.
= its inverse matrix.
mi = mean vector.
RESULTS AND DISCUSSION
Primary Interpretation
Data of 3 different dates is available in the form of 3 channels of 8 bit information. Thus, the
pixel values vary from 0 to 255. Small training areas of around 300 pixels were selected from
different land cover types from these 3 channels and also from the texture channels. Statistical
data like Mean, Median and Standard Deviation is calculated and based on that, following
conclusions were made.
Numbe
r of
pixels
310 56 56.63 13.61 14.00 89.00 0 15 mean value
Channe
l 1: 6
July
Channe
310 44 45.41 12.45 9.00 74.00 0 15
l 2 :
Using remote sensing technology is an effective way to observe and verify the wise and
sustainable use of natural resources. With its weather and daylight independency Synthetic
Aperture Radar (SAR) data emphasize as particularly suitable for large-scaled monitoring. But
the classification of SAR images is more difficult as of multi-spectral images. Because of the
microwave’s interaction with the object parameters on the earth surface, like roughness or the
complex dielectric constant (CDC), different types of land cover respond very similar. So
separability is not possible without additional information. For the investigated multi-channel
SAR scenes a typical example for this problem is the distinction between rice crops sown on …show more content…
Region of Interest (ROI) of the six different classes are soil,
urban, water bodies, soil with higher moisture, rice date 1 and rice date 2 were created on both
the image. And then classify them to calculate overall accuracy and kappa coefficient.
Confusion matrix is created based on these two classifications.
Where: i = class
x = n-dimensional data (where n is the number of bands).
= probability that class occurs in the image and is assumed the same for all classes.
= determinant of the covariance matrix of the data in class.
= its inverse matrix.
mi = mean vector.
RESULTS AND DISCUSSION
Primary Interpretation
Data of 3 different dates is available in the form of 3 channels of 8 bit information. Thus, the
pixel values vary from 0 to 255. Small training areas of around 300 pixels were selected from
different land cover types from these 3 channels and also from the texture channels. Statistical
data like Mean, Median and Standard Deviation is calculated and based on that, following
conclusions were made.
Numbe
r of
pixels
310 56 56.63 13.61 14.00 89.00 0 15 mean value
Channe
l 1: 6
July
Channe
310 44 45.41 12.45 9.00 74.00 0 15
l 2 :