Satellite image classification and content base image retrieval using type-2 fuzzy logic

By,
Shirish Agale
Exam Seat No: 5076
Under Guidance of,
Prof. Anita Thengade
MIT COE
2-Aug-16 1/37
Dissertation Stage- II

Outline
 Introduction
 Motivation
 Literature Survey
 Problem Statement, Objectives
 Our Contribution
 Architecture of System
 System Modules
 Features of Image
 Mathematical Model
 Algorithms
 Results
 Conclusion
 Future Work
 Publications
 References
2-Aug-16
2/37

Introduction
Image Classification
2-Aug-16 3/37

2-Aug-16 4/37
Global Features
Local Features
Segmentation
Classifiers
Results
Fig. 1. Image Classification

2-Aug-16 5/37
CBIR
 Finding a picture by key words generally is almost implausible.
 CBIR is based on the idea of searching for images by matching their features.
 Can work effectively to a certain degree.
 Relevant techniques
 Feature extraction, indexing, pattern recognition …..
Fig. 2. Content-Base Image Retrieval

2-Aug-16 6/37
Fuzzy Logic
● Fuzzy – “not clear, distinct, or precise; blurred”
Why Fuzzy Logic?
Fig. 3. Structure of Fuzzy Logic
Type 1 Fuzzy Logic Type 2 Fuzzy Logic

Fuzzy Logic
Traditional Representation of
Logic
Slow Fast
Speed = 0 Speed = 1
bool speed;
get the speed
if ( speed == 0) {
// speed is slow
}
else {
// speed is fast
}
Fuzzy Logic Representation
For every problem must represent
in terms of fuzzy sets.
What are fuzzy sets?
Slowest
[ 0.0 – 0.25 ]
Slow
[ 0.25 – 0.50 ]
Fast
[ 0.50 – 0.75 ]
Fastest
[ 0.75 – 1.00 ]
2-Aug-16 7/37

Fuzzy Logic Representation
Slowest FastestSlow Fast
float speed;
get the speed
if ((speed >= 0.0)&&(speed < 0.25)) { // speed is slowest
}
else if ((speed >= 0.25)&&(speed < 0.5)) { // speed is slow
}
else if ((speed >= 0.5)&&(speed < 0.75)) { // speed is fast
}
else // speed >= 0.75 && speed < 1.0 { // speed is fastest
}
2-Aug-16 8/37

Motivation
● During the processing of satellite images, there can be a different types of
imperfection (uncertainties, ambiguities and vagueness).
● To overcome such problems there is need to apply an uncertainty modeling
approach.
2-Aug-16 9/37

Literature Survey
2-Aug-16 10/37

No TITLE METHODS USED ADVANTAGES Disadvantages
1 “Support Vector Machine Training of
HMT Models for Multispectral Image
Classification” IJCSNS International
Journal of Computer Science and
Network Security, VOL.8 No.9,
September 2008
Hidden Markov Tree
(HMT), Support Vector
Machine
Experimental results of the system show
significant improvement over the baseline
HMT system and give a superior
performance in land cover classification.
Does not give more accuracy as
compare with fuzzy logic
2 “Classification of Multispectral Satellite
Images using Clustering with SVM
Classifier” International Journal of
Computer Applications (0975 8887)
Volume 35 No.5, December 2011
Support vector machine The experimentation is carried out using the
multi-spectral satellite images and the
analysis ensures that the performance of the
proposed technique is improved compared
with traditional clustering algorithm.
Result transparency is low. Training
is time consuming. Structure of
algorithm is difficult to understand.
Determination of optimal
parameters is not easy when there is
nonlinearly separable training data.
3 “A review on classification of satellite
image using Artificial Neural Network
(ANN)” 2014 IEEE 5th Control and
System Graduate Research Colloquium,
Aug. 11 - 12, UiTM, Shah Alam,
Malaysia
Artificial Neural
Network (ANN)
Overall classification accuracy and Kappa
Coefficient were calculated to get the
comparison of the performance the image
classification.
It is semantically poor. The training
of ANN is time taking. Problem of
over fitting. Difficult in choosing
the type network architecture.
4 “A Framework for Ocean Satellite
Image Classification Based on
Ontologies” IEEE Journal of Selected
Topics In Applied Earth Observations
And Remote Sensing, VOL. 6, NO. 2,
APRIL 2013
Ontology Engineering Describe how low and high level content of
ocean satellite images can be modeled with
an ontology
Degree of accuracy is less.
2-Aug-16 11/37
Comparative Study

5 “Decision Tree Classification of
Remotely Sensed Satellite Data using
Spectral Separability Matrix”
International Journal of Advanced
Computer Science and Applications,
Vol. 1, No.5, November 2010
Decision Tree Classifier
(DTC), Separability
Matrix, Maximum
Likelihood Classifier
(MLC)
Can handle nonparametric training data.
Does not required an extensive design and
training. Provides hierarchical associations
between input variables to forecast class
membership and provides a set of rules n are
easy to interpret. Simple and computational
efficiency is good.
The usage of hyperplane decision
boundaries parallel to the feature
axes may restrict their use in
which classes are clearly
distinguishable. Becomes complex
calculation when various values
are undecided and/or when various
outcomes are correlated
2-Aug-16 12/37

1 “Type-2 Fuzzy Logic in Image Analysis
and Pattern Recognition” Advances in
Type-2 Fuzzy Sets and Systems, Studies
in Fuzziness and Soft Computing 301,
DOI: 10.1007/978-1-4614-6666-6_12,
Springer Science+Business Media,
New York 2013
Sobel Edge Detector,
Morphological Gradient
Edge Detector
Comparing Sobel+FIS1, Sobel+FIS2,
Morphological Gradient+FIS1,
Morphological Gradient+FIS2 and finally
conclude that Type-2 have better
performance than the Type-1 fuzzy edge
detector and traditional method
Works on ORL, Cropped Yele and
FERET which are face database
2 “A Region-Based Fuzzy Feature
Matching Approach to Content-Based
Image Retrieval”
IEEE Transactions On Pattern Analysis
And Machine Intelligence, VOL. 24,
NO. 9, September 2002
Fuzzy Logic,
Unified Feature
Matching(UFM)
Size, shape and absolute and relative
location of the regions used for retrieval
Location for image retrieval are
insufficient and not fully explained
3 “Fuzzy Content-Based Image Retrieval
for Oceanic Sensing” IEEE
Transactions On Geo-Science And
Remote Sensing, VOL. 52, NO. 9,
September 2014
Neuro-Fuzzy Provides a tool for working with
multispectral images from different
satellite sensors
Uses standard image processing
technique where edges of images can
miss
4 “Feature Selection in AVHRR Ocean
Satellite Images by Means of Filter
Methods Fuzzy", IEEE Transactions
On Geo-Science And Remote Sensing,
2010.
K-Means, Bayesian
Classifier
Detail information of image features.
Bayesian Network is used for identity
relationship of features.
Computationally complex and give
less accuracy
2-Aug-16 13/37
Fuzzy Logic

14/37
5 “OBIA System for Identifying Mesoscale
Oceanic Structures in SeaWiFS and
MODIS-Aqua Images", IEEE Journal Of
Selected Topics In Applied Earth
Observations And Remote Sensing, Vol. 8,
No. 3, March 2015.
Type-1 Fuzzy Logic Detail information of image features
and information about color
clustering
Less accuracy compare with Type-2
Fuzzy Logic
6 “Analytical Structure and Characteristics of
Symmetric Karnik–Mendel Type-Reduced
Interval Type-2 Fuzzy PI and PD Controllers”
IEEE Transactions on Fuzzy Systems, VOL.
20, NO. 3, June 2012
Zadeh AND operator,
Karnik–Mendel(KM)
Type-Reduced Method
KM algorithm uses a new switch
point to compute the bounds of the
type-reduced set
Only used for type reducer module
in Type-2, it is computationally
expensive.
7 “Observations on Using Type-2 Fuzzy Logic
for Reducing Semantic Gap in Content–
Based Image Retrieval System”
International Journal of Computer Theory and
Engineering, Vol. 7, No. 1, February 2015
Type-2 Fuzzy Logic The performance of the system was
evaluated on real world images of
different categories and promising
research results were obtained
Feature extraction was done on only
one parameter.
8 “Satellite Image Classification Using Fuzzy
Logic”
International Journal of Recent Technology
and Engineering (IJRTE)
ISSN: 2277-3878, Volume-2, Issue-2, May
2013
Fuzzy Logic Mixed pixel is classified to a
specific category with the help of
the fuzzy logic
They worked on already existing
tool.
2-Aug-16

Problem Statement
● To design system for identifying ocean condition and CBIR for satellite images
using type 2 fuzzy logic.
● Ocean Conditions
● Upwelling’s
● Wakes
● Cyclones
● Design system for Image Classification and CBIR with improved accuracy.
● Global Change
● Security
● Economic Importance
2-Aug-16 15/37
Problem Statement
Objective
Applications

Our Contribution
● Find out oceanic conditions with improved accuracy
● K-Means Algorithm for color clustering used to label oceanic condition
● Type-2 Fuzzy logic instead of Type-1 Fuzzy Logic
● KM Type Reducer for minimizing rules in Type-2 Fuzzy Logic
● KNN algorithm for Content-Based Image Retrieval.
2-Aug-16 16/37

Architecture
17/37
Fig. 4. Satellite Image (08/10/1993) showing (1) western
Africa, (2) the sea and the Canary Islands (3) La Palma, (4)
El Hierro, (5) La Gomera, (6) Tenerife, (7) Gran Canaria,
(8) Fuerteventura, and (9) Lanzarote]. The oceanic
structures are enhanced: (green color) upwellings, (red
color) warmcore eddies, (blue color) cold-core eddies, and
(yellow color) wakes
Fig. 5. Architecture of system for Image Classification and CBIR
2-Aug-16
DataSet

System Modules
Fig. 6. System Modules; (a) Type-2 Fuzzy Classification; (b) Content-Based Image Retrieval
18/372-Aug-16
(a) (b)

Features of Image
● CBIR Features
1. RGB values (Red, Green, Blue)
2. Autocorrelation
3. ColorMoment
4. Cooccurence
5. EdgeDirectionHistogram
6. GeometricMoment
7. InvariantMoment
19/372-Aug-16
● Image Classification Features
1. HuMovement (Hu1,…Hu7)
2. Circularity
3. AreaConvex
4. AreaToPerimeterRatio
5. EnclosedArea
6. Elongation
7. GetSolidity
8. OrientationMajorAxis
9. OrientationMinorAxis
10. LongSideMBR
11. ShortSideMBR
12. PerimeterConvexHull
13. Perimeter
14. NumberofHoles
15. ZernikeMoments (Z1, .. Z15)
16. FractalBoxDimension

Mathematical Model
Let S be a system which is defined in the following manner
S= {I, IP, CBIR, T2C, Er, Th, Dk, R, OIPI, GSV, IC, OC, FE, F| f1, f2, f3, f4, f5, f6, f7, f8, f9, f10, f11, f12}
2-Aug-16 20/37
S.No. Function Description
1 F1 I1 Detects cloud pixel in original satellite image and remove such pixel that hide data.
2 F2I1 Histogram equalization improve contrast of image, it will useful to identify required
area
3 F3I1 Identify Island continent i.e. it find land and water area
4 F4I1 Delete regions larger and smaller than given threshold value.
5 F5I1 Edge Detection using Morphological Gradient
6 F6IK Color Clustering using K-Means algorithm for labeling images
7 F7(I1)FE On Preprocess image we perform feature extraction
8 F8(FE)FE The fuzzifier transforms the crisp set values to fuzzy sets by applying fuzzification
function
9 F9(FE)R Automatic rule generation in the form of IF–THEN
10 F10(R)R’ Sugeno fuzzy Inference system will decide output rule which satisfy all conditions.
11 F11(R’)R-K Karnik-Mendel Type Reducer for reducing number of rules
12 F12Crisp output transform the aggregated fuzzy set into a crisp value using the defined defuzzification
algorithm

Success Conditions
2-Aug-16 21/37
Fig. 7. State Diagram
S.No. Function Description
1 F1I1
I1 ¢ I
(if cloud pixels remove)
2 F2I1
I1 ¢ I
(if contrast level get improve)
3 F3I1
I1 ¢ I
(if identity water and land surface)
4 F4I1
I1 ¢ I
(if surfaces which are less or more
than specified threshold values)
5 F5I1
I1 ¢ I
(if morphological edge detection
perform )
6 F6IK
IK ¢ I
(Color clustering perform)
7 F7FE (all features get calculated)
8 F8(FE)FE’
FE’ ⊂ FE
(Fuzzification successful)
9 F9(FE’)R (all rules get generated)
10 F10(R)R’
R’ ⊂ R
(Sugeno fuzzy inference)
11 F11(R)R-K
R-K ⊂ R
(Rule get reduced)
12 F12Crisp output (Defuzzification)

Algorithms
Algorithm Name Used in Module Function
Morphological Gradient Edge
Detection
Image Processing Draw an edge around selected
region
Sugeno Fuzzy Inference Type-2 Fuzzy
Classification
Apply weight and select best rule
Karnik-Mendel (KM) Type
Reducer
Type-2 Fuzzy
Classification
Conversion of Type-2 rule into
Type-1
K-NN Algorithm Content-Based Image
Retrieval
Based on Cosine Similarity
measure retrieve similar images
K-Means Color Clustering Image Labeling To label images like upwellng’s,
cyclones, eddies
2-Aug-16 22/37

2-Aug-16 23/37
K-Means Color Clustering Algorithm:
K-Means (image, no of clusters)
{
1. Create clusters
2. Create clusters lookup table by multiplying width and height of image
3. Set pixel changed cluster true, as first loop all pixel will move their clusters
4. Loop until all clusters are stable
{
update clusters
for( i < cluster length )
clear i.e. set r=0; g=0; b=0;
for( y < height of image)
for( x < width of image)
Add pixels to cluster
}
5. Once we get all stable clusters assign unique color to each cluster
6. Create result image
}

2-Aug-16 24/37
Fig. 8. T2 FS. UMF: upper membership function; LMF: lower membership
function. The shaded region is called a footprint of uncertainty (FOU)
KM algorithm:

2-Aug-16 25/37
KNN Algorithm:
K-NN(image, no of images to show as output)
{
1. Read text file which contain features for CBIR and store in ArrayList
2. Create Vectors for each of the feature
3. Calculate CoSine Similarity
for( i < ArrayList Size)
where A and B are Vectors
4. Sort all vectors of CoSine Similarity
5. Display output images
}

Conclusion
● We analyzed the different Oceanic Structures and classify the images according
to the features calculated for upwelling’s, cyclone and eddies.
● Using K-Means algorithm we are doing color clustering so that we can easily
label the images.
● The location of oceanic structure is relevant for study of marine circulation,
atmosphere and also global change.
● Type-2 fuzzy system provides greater accuracy also knowledge generated
during execution is reusable.
● The goal of this work will be to find natural disasters such as cyclones and
tropical storms by studying SST, chlorophyll concentration and meteorological
data.
2-Aug-16 32/37

Future Work
 Reduce Computational Time
 Use Hierarchical Type-2 Fuzzy Logic which can reduce even more number
of rules
2-Aug-16 33/37

Publication
1. Shirish Agale, Prof. Anita Thengade “Remote Sensing Image Classication and Content-
Based Image Retrieval using Type-2 Fuzzy Logic", paper id-946, cPGCON-2015, MET's
Bhujbal Knowledge City, Nashik.
2. “Satellite Image Classification and Content-Base Image Retrieval Using Type-2 Fuzzy
Logic” IOSR Journal of Computer Engineering (IOSR-JCE) e-ISSN: 2278-0661,p-ISSN:
2278-8727, Volume 16, Issue 6, Ver. IV (Nov – Dec. 2014), PP 74-78. Impact Factor 1.68
3. “Health Health assessment of Nuclear Reactor using Fuzzy Logic Controller”, Equinox,
International Conference On Engineering Confluence, TPCT’S Terna Engineering College,
Nerul Navi Mumbai, ISSN:2320-0332
Paper Acceptance:
1. Paper acceptance notification for “Satellite Image Classification Using Type-2 Fuzzy
Logic”, from 2014 IEEEICCIC(IEEE Conference) PARK College of Engineering and
Tekhnology, Coimbatore, Tamilnadu, India.
2. Paper acceptance notification for “Health assessment of Nuclear Reactor using Fuzzy Logic
Controller”, from CIPECH 14(IEEE Conference) Krishna Institute of Engineering and
Technology (KIET), Ghaziabad, Uttar Pradesh.
Competition and Exhibition
1. Participation in AVISHKAR-2014 Zonal Level Research Project Competition
2. Participation in MIT TechnoGenesis-2015 Project, Research and Technology Exhibition
2-Aug-16 34/37

References
1. Reda A. El-Khoribi, “Support Vector Machine Training of HMT Models for Multispectral Image Classification", IJCSNS International
Journal of Computer Science and Network Security, Vol.8, No.9, pp.224-228, September 2008.
2. S.V.S Prasad, Dr. T. Satya Savitri, Dr. I. V. Murali Krishna, “Classification of Multispectral Satellite Images using Clustering with SVM
Classifier”International Journal of Computer Application, VOL. 35, No. 5, December 2011
3. Nur Anis Mahmon and Norsuzila Yaacob, “A review on classification of satellite image using Artificial Neural Network (ANN)" 2014
IEEE 5th Control and System Graduate Research Colloquium, Aug. 11 - 12, UiTM, Shah Alam, Malaysia
4. Jesus M. Almendros-Jimenez, Luis Domene, and Jose A. Piedra-Fernandez, “A Framework for Ocean Satellite Image Classification Based
on Ontologies", IEEE Journal of Selected Topics In Applied Earth Observations And Remote Sensing, VOL. 6, NO. 2, APRIL 2013
5. M. K. Ghose , Ratika Pradhan, Sucheta Sushan Ghose, “Decision Tree Classification of Remotely Sensed Satellite Data using Spectral
Separability Matrix", International Journal of Advanced Computer Science and Applications, Vol. 1, No.5, November 2010
6. Yixin Chen, James Z. Wang, “A Region-Based Fuzzy Feature Matching Approach to Content-Based Image Retrieval” IEEE Transactions
on Pattern Analysis and Machine Intelligence, VOL. 24, NO. 9, September 2002
7. Jose A. Piedra-Fernandez, Gloria Ortega, James Z. Wang, and Manuel Canton-Garbin, “Fuzzy Content-Based Image Retrieval for Oceanic
Sensing” IEEE Transactions On Geo-Science And Remote Sensing, VOL. 52, NO. 9, September 2014
8. Jose A. Piedra-Fernandez, James Z.Wang, and Manuel Canton-Garbin, “Feature Selection in AVHRR Ocean Satellite Images by Means of
Filter Methods Fuzzy", IEEE Transactions On GeoScience And Remote Sensing, 2010.
9. Eva Vidal- Fernandez , Jose A. Piedra Fernandez, Manuel Canton-Garbin, “OBIA System for Identifying Mesoscale Oceanic Structures in
SeaWiFS and MODIS-Aqua Images", IEEE Journal Of Selected Topics In Applied Earth Observations And Remote Sensing, Vol. 8, No.
3, March 2015.
10. V.K.Panchal, Parminder Singh, Navdeep Kaur and Harish Kundra, “Biogeography based Satellite Image Classification”, International
Journal of Computer Science and Information Security IJCSIS, Vol. 6, No. 2, pp. 269-274, November 2009.
11. R. Datta, D. Joshi, J. Li, and J. Z. Wang, “Image retrieval: Ideas, influences,and trends of the new age,” ACM Comput. Surveys, vol. 40,
no. 2,pp. 5:1–5:60, Apr. 2008.
12. Saad M. Darwish and Raad A. Ali, “Observations on Using Type-2 Fuzzy Logic for Reducing Semantic Gap in Content–Based Image
Retrieval System” International Journal of Computer Theory and Engineering, Vol. 7, No. 1, February 2015
2-Aug-16 35/37

References
13. Patricia Melin, Claudia I. Gonzalez, Juan R. Castro, Olivia Mendoza, and Oscar Castillo, “Edge Detection Method for Image Processing
based on Generalized Type-2 Fuzzy Logic” IEEE Transactions on Fuzzy Systems,TFS-2013-0313.R2
14. Zhengyu Huang, Kwang Y. Lee, and Robert M. Edwards, “Fuzzy Logic Control Application in Nuclear Power Plant”15th Triennial World
Congress, Barcelona, Spain 2002
15. Maowen Nie, and Woei Wan Tan, “Analytical Structure and Characteristics of Symmetric Karnik–Mendel Type-Reduced Interval Type-2
Fuzzy PI and PD Controllers” IEEE Transactions on Fuzzy Systems, VOL. 20, NO. 3, June 2012
2-Aug-16 36/37

Satellite image classification and content base image retrieval using type-2 fuzzy logic

More Related Content

What's hot (17)

Similar to Satellite image classification and content base image retrieval using type-2 fuzzy logic (20)

Recently uploaded (20)

Satellite image classification and content base image retrieval using type-2 fuzzy logic