3. Nowadays everywhere remote sensing images are used for wide
variety of applications such as
creation of mapping products for military and civil
applications,
evaluation of environmental damage,
monitoring of land use,
radiation monitoring,
urban planning
growth regulation,
soil assessment
and crop yield appraisal.
A few number of image classification algorithms have proved
good precision in classifying remote sensing data.
An efficient classifier is needed to classify the remote sensing
imageries to extract information. we have used texture based
supervised classification.
Introduction
4. Remote sensing image classification can be viewed as a
joint venture of both image processing and
classification techniques. Generally, image
classification, in the field of remote sensing is the
process of assigning pixels or the basic units of an
image to classes.
Texture classification is a fundamental issue in
computer vision and image processing, which plays a
significant role in a wide range of applications that
includes medical image analysis, remote sensing,
object recognition, content-based image retrieval, and
many more.
5. classification system essentially involves two major steps:
(1) Feature extraction
(2) Classification
supervised classifications usually have a sequence of
operations that must be followed.
(1) Extraction of Signatures.
(2) Defining of Training Sites.
(3) Classification of the Image.
Usually two or three training sites are selected. The more
training site is selected, the better results can be gained.
This procedure assures both the accuracy of classification
and the true interpretation of the results.
Supervised classification is done using neural network
(NN), SVM and K Nearest Neighbor (kNN) classifiers.
6. TECHNICS USED
Gabor Filter
A Gabor filter is obtained by modulating a sinusoid
with a Gaussian. Gabor filter is a linear filter whose impulse
response is defined by a harmonic function multiplied by a
Gaussian function.
Gaussian Filter
The Gaussian smoothing operator is a 2-D
convolution operator that is used to `blur' images and
remove detail and noise.
9. seg1 seg2 seg3 seg4
seg 5
Image Difference
Image in 2000 Image in 2012 Changes(Area=261377.45
sq.mtr)
o Extracted Images
10. 2. RAPID MINER
Input from Matlab
DATA SET
o Training Data
o Test Data
VALIDATION
o Split Validation(Split Ratio- 0.7)
o X Validation
11. The five selected prominent response variables are
1.Water Bodies – identifies all water bodies in the image like lake , garbage
etc.
2.Empty land – identifies all empty land in the image like play ground etc.
3. Form land – identifies all form land used for agriculture in the image.
4. Buildings – identifies all buildings in the image.
5. Trees – identifies all plants , grass , tress in the image.
Confusion Matrix of SVM
13. Application
Automatic Classification Model :
Soil Assessment
Growth regulation
Urban development
Percentage of water bodies
Rate of Deforestation
Estimation of Damage caused by natural disasters
Medical Diagnosis
Satellite Imaging
Shape Analysis
Industrial Inspection
Estimation of object range and orientation
Signature Analysis
14. Conclusions
Remote sensing techniques and geographic information
systems offer a good means of collecting and manipulation
of the data required to assess conservation practices.
We have developed the automatic classification model for
land cover type and land use using texture feature and
classifier.
KNN, SVM and Neural network classifiers are used and
compared. Neural Network gives very good classification
accuracy but time complexity is very high.
Land use mapping has been done by taking images of two
different time period and also area of land use is calculated.