Beruflich Dokumente
Kultur Dokumente
M.N.Reddy
PROJECT1 :
Land use/Land Cover Classification from
Remote Sensing Data - IDRISI an Image
processing and GIS Software
PROJECT2 : Characterization:
Characterization of Agro-climatic Zones
in Mahabubnagar District in AP based on
Rainfall and Temperature
IDRISI –not an Acronym
Al-Idrisi–Muslim Scholar
(1100-1166)
Digitization
Decision
support Database
Geographic
Statistical analysis
analysis
Data Products
• Photographic Products
• Digital Products
• Raw data
• Geo-coded Products
• Precision Products
Information extraction
• Visual Interpretation
• Classification
Image Processing:
2. Image Rectification/Geo-
referencing
7. Unsupervised/Supervised
Classification
•Spatial resolution
•Spectral resolution
•Radiometric resolution
•Temporal resolution
Remote Sensing is the science and art of
obtaining information about an object,
area or phenomenon through an analysis
of the data acquired by a device which is
not in contact with the object, area or
phenomenon under investigation.
Spatial resolution
• Pre-processing
• Information Extraction
Pre-processing
• Geometric correction
Image Enhancement
• Contrast Stretching
• Density Slicing
• Colour composites
• Ratio Images
• Principal Components
• Convolution Filtering
• Edge Enhancement and Linear Filtering
Band 2
Band 2
Band 2
2. Second
1. First iteration. 3. N-th iteration.
iteration. The
The cluster centres The centres have
centres move to
are set at random. stabilised.
the mean-centre of
Pixels will be
all pixels in this
assigned to the
cluster.
nearest centre.
ISODATA (unsupervised)
Band 2
Band 2
Band 2
Band 1 Band 1
Band 1
fields
Representative spectral values –
Spectral
signatures
Supervised Classification
Procedures
• Maximum-likelihood Method
• Discriminant Function
• Bayesian Method
• Parallelepiped
• Minimum distance
• Neural network
Parallelepiped (supervised)
• For each training region determine the
range of values observed in each band.
• These ranges form a spectral box (or
parallelepiped) which is used to classify
this class type.
• Assign new image pixels to the
parallelepiped which it fits into best.
• Pixels outside all boxes can be
unclassified or assigned to the closest
one.
• Problems with classes that exhibit high
correlation between bands. This creates
long ‘diagonal’ data-sets that don’t fit well
into a box.
Parallelepiped example
Training classes plotted in spectral
space. In this example using 2 bands.
Maximum likelihood
(supervised)
• For each training class the spectral variance and
covariance is calculated.
• The class can then be statistically modelled with a
mean vector and covariance matrix.
• This assumes the class is normally distributed.
Which is generally okay for natural surfaces.
• Unidentified pixels can then be given a
probability of being in any one class.
• Assign the new pixel to the class with the highest
probability – or unclassified if all probabilities
low.
Concept of Maximum Likelihood Classification
Maximum likelihood example
• Normal probability
distributions are fitted to
each training class.
• The lines in the diagram
show regions of equal
probability.
1
• Point 1 would be assigned to
class ‘pond culture’ as this
is most probable. Equiprobability
contours
• Point 2 would generally be
unclassified as the
2
probabilities of fitting into
one for the classes would be
below threshold.
Problem:
Some people recommend rectifying an image after
it has been classified. The argument is
3) Rectification is quicker since each pixel
contains only one class value instead of many
spectral values.
4) Some spectral integrity is lost during the pixel
resembling process – an un-rectified image is is
spectrally more correct than a rectified image.