PDF Archive

Easily share your PDF documents with your contacts, on the Web and Social Networks.

Share a file Manage my documents Convert Recover Search Help Contact



14I18 IJAET0118683 v6 iss6 2439 2447 .pdf



Original filename: 14I18-IJAET0118683_v6_iss6_2439-2447.pdf
Title: Format guide for IJAET
Author: Editor IJAET

This PDF 1.5 document has been generated by Microsoft® Word 2013, and has been sent on pdf-archive.com on 04/07/2014 at 08:01, from IP address 117.211.x.x. The current document download page has been viewed 444 times.
File size: 450 KB (9 pages).
Privacy: public file




Download original PDF file









Document preview


International Journal of Advances in Engineering & Technology, Jan. 2014.
©IJAET
ISSN: 22311963

PROBABILISTIC BASED ROCK TEXTURE CLASSIFICATION
R.Vinoth1, R.Srinivasan2, D.Vimala3, M.M.Arun Prasath4, D.Vinoth5
1AP,

Dept of ECE, Muthayammal College of Engg, Rasipuram, Tamil Nadu, India,
Dept of EEE, Muthayammal College of Engg, Rasipuram, Tamil Nadu, India,
3AP, Dept of EEE, PGP College of Engg & Tech, Namakkal, Tamilnadu, India,
4AP, Dept of ECE, Annapoorana Engineering College, Salem, Tamil Nadu, India,
5AP, Dept of EEE, Muthayammal College of Engg, Rasipuram, Tamil Nadu, India

2AP,

ABSTRACT
The classification of natural images is an essential task in computer vision and pattern recognition applications.
Rock images are the typical example of natural images, and their analysis is of major importance in rock
industries and bedrock investigations. Rocks are mainly classified into three types. They are Igneous,
Metamorphic and Sedimentary. They are further classified into Andesite, Basalt, Amphibolite, Granite, Breccia,
Coal and etc… In this project classification is done in three subdivisions. First the given rock image is classified
into major class. Next it is classified into subclass. Finally the group of coal images is segmented and classified
using Tamura features, Probabilistic Latent Semantic Analysis (PLSA) and Sum of Square Difference classifier.
Rock image classification is based on specific visual descriptors extracted from the images. Using these
descriptors images are divided into classes according to their visual similarity. This project deals with the rock
image classification using two approaches. Firstly the textural features of the rock images are calculated by
applying Tamura features extraction method. The Tamura features are Coarseness, Contrast, Directionality,
Line likeness, Roughness and Regularity, Smoothness and Angular second moments. In next step calculated
Tamura features are applied to Probabilistic Latent Semantic Analysis (PLSA) to generate a topic model. This
topic model is applied to SSD classifier to classify the rock image into one of the major class. Similarly the rock
textures are classified into subclass, and the group of coal images is segmented and classified. This method is
compared with Gray Level Co-occurrence Matrix (GLCM) method and Color Co-occurrence Matrix method.
This method gives a better accuracy when compared to those methods. This technique can readily be applied to
automatically classify the rocks in such fields of rock industries and bedrock investigations.

I.

INTRODUCTION

1.1. Rocks
Rocks are among the most basic things on Earth. They can be found just about everywhere. They do
not seem very exciting at first glance. But there is a lot more in rocks than first meets the eye. There
are several types of rocks, which are very useful in many fields like marble, granite, coal, quartzite
and etc. The classification of rocks is very essential in rock industries and bedrock investigations,
which has been done manually.
1.1.1. Igneous
Igneous rocks form from molten rocks, which are thick, fluid masses of very hot elements and
compounds. There are many different types of igneous rocks. However, they were once melted and
have since cooled down. The two major factors that influence the creation of igneous rocks are the
original rock that was melted and the cooling history of the molten rocks
1.1.2. Metamorphic

2439

Vol. 6, Issue 6, pp. 2439-2447

International Journal of Advances in Engineering & Technology, Jan. 2014.
©IJAET
ISSN: 22311963
Metamorphic rocks are formed mainly in the lithosphere or crust and upper mantle, wherever there is
high pressure and high temperature. Metamorphic rocks record how temperature and pressure affected
an area when it was forming. The rocks provide clues to their transformation into metamorphic rocks
1.1.3. Sedimentary
Sedimentary rocks form only on the surface of the Earth. Sedimentary rocks form in two main ways,
from clastic material (pieces of other rocks or fragments of skeletons) that are cemented together, and
by chemical mechanisms including precipitation and evaporation. There are many environments
associated with sedimentary rock formation including oceans, lakes, deserts, rivers, beaches, and
glaciers. They may form at all types of plate boundaries, but the thickest sedimentary rock
accumulations occur at convergent plate boundaries. Fossils are associated with sedimentary rocks.

Figure 1.1.The Rock Cycle

1.2 Scope Of The Thesis
Rocks are used in various industries and various places. For example marbles are used in fancy
flooring and countertops, sculptures and carvings, decor on front porch and etc… Slate is widely used
as roofing material. Shale is used as filler for paint. Obsidian can be used for jewelry and knife and
sandstone as Glassware. Coals are used for generation of electricity, which is a major requirement
nowadays. So the classification of rocks and coals is very essential in rock and coal industries. The
rock images can efficiently be classified using the method proposed in this thesis. This can be applied
in rock industries to classify the rocks automatically. And also the automatic classification of rocks is
very essential in bedrock investigations, coal mines and planning for oil reservoirs.

II.

TEXTURE ANALYSIS

2.1. Texture
In many machine vision and image processing algorithms, simplifying assumptions are made about
the uniformity of intensities in local image regions. However, images of real objects often do not
exhibit regions of uniform intensities. For example, the image of a wooden surface is not uniform but
contains variations of intensities which form certain repeated patterns called visual texture. The
patterns can be the result of physical surface properties such as roughness or oriented strands which
often have a tactile quality, or they could be the result of reflectance differences such as the color on a
surface. To enhance the image we are using some of the techniques like Morphological filtering,
spatial domain filtering, wavelet based filtering etc..

2.2 Texture Classification
Classification refers to as assigning a physical object or incident into one of a set of predefined
categories to measure the parameters. In texture classification the goal is to assign an unknown
sample image to one of a set of known texture classes. Texture classification is one of the four
problem domains in the field of texture analysis. The other three are texture segmentation
(partitioning of an image into regions which have homogeneous properties with respect to texture;
supervised texture segmentation with a priori knowledge of textures to be separated simplifies to

2440

Vol. 6, Issue 6, pp. 2439-2447

International Journal of Advances in Engineering & Technology, Jan. 2014.
©IJAET
ISSN: 22311963
texture classification), texture synthesis (the goal is to build a model of image texture, which can then
be used for generating the texture) and shape from texture (a 2D image is considered to be a
projection of a 3D scene and apparent texture distortions in the 2D image are used to estimate surface
orientations in the 3D scene).

2.3. Texture Features
In the case of image classification, features extracted from images are employed. The extracted
features used in image classification relate to the colors, textures and shapes occurring in the images.
There are many different feature extraction methods that were introduced and used for texture
classification problems. Most of these methods that were popularly used in recent years were
statistical and signal processing methods.
2.3.1. GLCM
Grey Level Co-occurrence Matrices (GLCM) is an old feature extraction for texture classification that
was proposed by Haralick et al. It has been widely used on many texture classification applications
and remained to be an important feature extraction method in the domain of texture classification. It is
a statistical method that computes the relationship between pixel pairs in the image. In the
conventional method, textural features will be calculated from the generated GLCMs, e.g. contrast,
correlation, energy, entropy and homogeneity
2.3.2. Local Binary Patterns (LBP)
The original LBP is proposed by Ojala and Pietikainen back in 1999. The original LBP calculates a
value that reflects the relationship within a 3 × 3 neighborhood through a threshold neighborhood that
is multiplied with the respective binomial weights. Since the LBP is used to calculate local features, it
is often used for texture segmentation problems. It has yet to be a very popular method in texture
classification problem.
2.3.3. Gabor Filters
Gabor filters is a popular signal processing method, which is also known as the Gabor wavelets. The
Gabor filters are defined by a few parameters, including the radial center frequency, orientation and
standard deviation The Gabor filters can be used by defining a set of radial center frequencies and
orientations which may vary but usually cover 180° in terms of direction to cover all possible
orientations. Due to the large feature size produced by signal processing methods, the Gabor filters
requires to be downsized to prevent the “curse of dimensionality”. Principal Component Analysis
(PCA) is one of the popular methods to downsize the feature space. Gabor filter

III.

PROBLEM STATEMENT

In the rock industry, the visual inspection of products is essential because the color and texture
properties of rock often vary greatly, even within the same rock type. Therefore, when rock plates are
manufactured, it is important that the plates used, such as in flooring, share common visual properties.
In addition, visual inspection is necessary in the quality control of rock products. Traditionally, rock
products have been manually classified into different categories on the basis of their visual similarity.
However, in recent years the rock and stone industry has adopted computer vision and pattern

2441

Vol. 6, Issue 6, pp. 2439-2447

International Journal of Advances in Engineering & Technology, Jan. 2014.
©IJAET
ISSN: 22311963
recognition tools for use in rock image inspection and classification. Compared to manual inspection
and classification, the use of automated image analysis provides several benefits. Manual inspection
carried out by people is, as might be affected by human factors. These factors include personal
preferences, fatigue, and the concentration levels of the individual performing the inspection task.
Therefore, inspection is a subjective task, dependent on the personal inclinations of the individual
inspector, with individuals often arriving at different judgments. By contrast, automated inspection by
computer with a camera system performs both inspection and classification tasks dependably and
consistently. Another drawback of manual inspection is the amount of manual labor expended on each
task. Additionally, in the field of rock science, the development of digital imaging has made it
possible to store and manage images of the rock material in digital form. One typical application area
of rock imaging is bedrock investigation which is utilized in many areas from coal mining to
geological research. In coal mining the various types of coal would be accumulated or gathered in
same place.

3.1. Gray level co-occurrence matrix method
In this method the given rock image is classified into major class and sub class. These classifications
are as follows.
3.1.1 Classification into Major Class
This section describes a method to classify the given rock image into one of the three types of rocks
called igneous, metamorphic and sedimentary. The flowchart of classification process is shown
below.

Figure 3.1 Flowchart of rock image classification process using GLCM.

3.1.1.1 Feature Extraction using GLCM Matrix
The analysis of surface of the rock images is done to extract the prominent features. The feature
extraction method considered here is ‘Gray level co-occurrence matrix’. GLCM matrix is also known
as spatial gray level dependency (SGLD) matrix.
Statistical methods use second order statistics to model the relationships between pixels within the
region by constructing Spatial Gray Level Dependency (SGLD) matrices. A SGLD matrix is the joint
probability occurrence of gray levels i and j for two pixels with a defined spatial relationship in an
image. The spatial relationship is defined in terms of distance d and angle θ. If the texture is coarse
and distance d is small compared to the size of the texture elements, the pairs of points at distance d
should have similar gray levels.
Conversely, for a fine texture, if distance d is comparable to the texture size, then the gray levels of
points separated by distance d should often be quite different, so that the values in the SGLD matrix
should be spread out relatively uniformly.
Hence, a good way to analyse texture coarseness would be, for various values of distance d, some
measure of scatter of the SGLD matrix around the main diagonal. Similarly, if the texture has some
direction, i.e. is coarser in one direction than another, then the degree of spread of the values about the

2442

Vol. 6, Issue 6, pp. 2439-2447

International Journal of Advances in Engineering & Technology, Jan. 2014.
©IJAET
ISSN: 22311963
main diagonal in the SGLD matrix should vary with the direction d. Thus texture directionality can be
analyzed by comparing spread measures of SGLD matrices constructed at various distances d. From
SGLD matrices, a variety of features may be extracted. The original investigation into SGLD features
was pioneered by Haralick et al.

Figure 3.2 GLCM matrix

But for simplicity of classification process some of the 14 features are considered. Those features are,
i. Angular second moment
ii. Entropy
iii. Contrast
iv. Correlation
v. Inverse difference moment
The definitions and formulae of those features are discussed as follows.
(i. i) ANGULAR SECOND MOMENT
Angular Second Moment, also called energy and uniformity, is a measure of textural uniformity of an
image. Energy reaches its highest value when gray level distribution has either a constant or a periodic
form. A homogenous image contains very few dominant gray tone transitions, and therefore the P
matrix for this image will have fewer entries of larger magnitude resulting in large value for energy
feature. In contrast, if the P matrix contains a large number of small entries, the energy feature will
have smaller value.
Angular Second Moment
.... (3.1)

(i.ii) ENTROPY
Entropy measures the disorder of an image and it achieves its largest value when all elements in P
matrix are equal [3]. When the image is not texturally uniform many GLCM elements have a very
small value, which implies that entropy is very large. Therefore, entropy is inversely proportional to
GLCM energy.
Entropy

  P(i, j )*log( P(i, j ))

.... (3.2)

i, j

(i.iii) CONTRAST
Contrast is a difference moment of the P and it measures the amount of local variations in an image.
The contrast texture feature will give higher values for the areas with larger differences between
pixels within each Gray-Tone Spatial-Dependence Matrices. This is similar to the variance but is
calculated a little differently. The contrast 2D scatter plot and image as would be expected is similar
to variance. Until contrast and variance are used in the classification methods, the extent to which
these are similar cannot be further described here.

  i  j P(i, j )
2

Contrast

…(3.3)

i, j

2443

Vol. 6, Issue 6, pp. 2439-2447

International Journal of Advances in Engineering & Technology, Jan. 2014.
©IJAET
ISSN: 22311963
(i.iv) CORRELATION
Correlation is a measure of image linearity. It shows how the pixel values in the images are correlated.
The correlation feature describes whether the variations in the pixel values are linear or non- linear.



Correlation

(i  i )( j   j ) P(i, j )

....(3.4)

 i j

i, j

Where µ is mean and σ is standard deviation.
(i.v) INVERSE DIFFERENCE MOMENT

Inverse difference moment measures image homogeneity. This parameter achieves its largest value
when most of the occurrences in GLCM are concentrated near the main diagonal. IDM is inversely
proportional to GLCM contrast.

IV.

TOPIC MODEL GENERATION USING PROBABILISTIC LATENT SEMANTIC
ANALYSIS (PLSA)

PLSA is a novel statistical technique for the analysis of two mode and co-occurrence data, which has
applications in information retrieval and filtering, natural language processing, machine learning from
text, and in related areas. This PLSA method is based on a mixture decomposition derived from a
latent class model.
The PLSA model was originally developed for topic discovery in a text corpus, where each document
is represented by its word frequency. The core of PLSA model is to map high dimensional word
distribution vector of a document to a lower dimensional topic vector. Therefore, PLSA introduces a
latent topic variable between the document d i  {d1 ....d n } and the word w j  {w1 ....wm } . Then the
PLSA model is given by the following generative scheme,
1. select a document d i with probability p(d i )
2. pick a latent topic z k with probability p( z k / d i )
3. Generate a word w j with probability p( w j / z k ) .
The model is graphically shown in the fig

D

Z

W

Figure 4.1 PLSA model

This PLSA model very well suits for the generation of topic model from the Tamura features of the
rock images. Here the three types of rock are considered as documents. The latent topic is the sub
classes of rocks such as andesite, basalt, gabbro, coal, granite, amphibolites, and gneiss. The set of
Tamura features calculated for the query image and the training rock images are considered as words.
This is depicted as follows.
Document
class of rock image
Latent topic
subclass of rock image
Word
feature of rock image.
As a result one obtain an observation pair ( d i , w j ) while the latent topic variable is discarded.
This generative model can be expressed by the following probabilistic model

p( w j , d i )  p( d i ) p( w j / d i )

… (4.1)

K

Where

p( w j / d i )   p( w j / z k ) p( z k / d i )

… (4.2)

k 1

The unobservable probability distribution p( z k / d i ) and p( w j / z k ) are learnt from the complete
dataset using expectation maximization (EM) algorithm []. The log likelihood of the complete dataset
is

2444

Vol. 6, Issue 6, pp. 2439-2447

International Journal of Advances in Engineering & Technology, Jan. 2014.
©IJAET
ISSN: 22311963

L   n(d i , w j ) log p(d i , w j )
N

M

… (4.3)

K

 n(di , w j ) log  p(w j / zk ) p( zk / di )
i 1 j 1

… (4.4)

k 1

Where n(d i , w j ) is value of the visual word occurred in the word image matrix ‘n’. Each row in the
matrix represents an image. The first row corresponds to the features of the query image and
remaining rows are corresponds to the reference images i.e. training sequences.

V.

RESULT
TABLE 6.4 SSD values for major class (CCM METHOD)

Rock
name

Metamorphic
(Amphibolite)

Sedimentary
(Breccia)

5.5696e+019

1.7095e+019

Basalt

1.5428e+019 1.2975e+020

4.2738e+016

Gabbro

3.0965e+018 3.2527e+019

3.4743e+019

Granite

2.1000e+019 8.2970e+018

7.5988e+019

Scoria

9.6426e+017 4.2003e+019

2.6179e+019

Peridotite

5.2353e+017 6.7019e+019

1.1635e+019

Pumice

8.1885e+017 7.0021e+019

1.0431e+019

Andesite

Rock image

Igneous
(Andesite)
0

On comparing the results the Scoria rocks gives the maximum Igneous output of 9.6426e+017 and
Basalt rock gives the maximum Sedimentary output of 4.2738e+016 and the Granite rock gives the
maximum output of 8.2970e+018.

VI.

CONCLUSION

With the rapid development of digital imaging tools, imaging applications have been adopted in many
areas in which inspection and monitoring have been done manually. The application area of this thesis
is an example of this change. Formerly, rock and coal samples were inspected manually in the rock
industry as well as in geological research. It was not recently fair that imaging and image processing
methods made it feasible to start developing automatic approaches for the visual inspection and
recognition of rocks and coals. Compared to several other goods and materials that are inspected by
computer vision systems on the production line, inspection of rock material is significant and more
demanding analysis task.
In this thesis the methods and techniques have been developed for the classification of natural rock
images and segmentation of group of coal images. In image classification, visual descriptors extracted
from the images are used to describe image content. In this project, the visual properties of rock and

2445

Vol. 6, Issue 6, pp. 2439-2447

International Journal of Advances in Engineering & Technology, Jan. 2014.
©IJAET
ISSN: 22311963
coal images are extracted by using Tamura features. In addition, a better topic model generation
process known as PLSA is considered. In GLCM and CCM methods the features are extracted and
directly applied to classifier. But here the PLSA model measures the probability of co occurrence of
particular features being with various types of rock images. Since this is computationally effective and
more efficient generative model this technique improves the classification accuracy compared to
conventional texture methods.
The method introduced in this thesis is directly applicable to practical rock image classification and
coal image segmentation problems. They can, therefore, be used whenever rock image classification
systems are being constructed. The accuracy obtained in this probabilistic based method can be
further improved by using better classifier called random forest classifier.

VII.

FUTURE WORKS

In future we can enhance the image to extract the features by using the latest filtering methods like
wavelet filtering, spatial domain filtering by using median filters and classify the samples by using
support vector machine classifier.

REFERENCES
[1]. Automatic Rock Detection and Classification in Natural Scenes, 2006 Heather Dunlop.
[2]. Color and Texture Based Classification of Rock Images Using Classifier Combinations, 2006,
LeenaLepisto.
[3]. Correlated PLSA for Image Clustering, 2010, PengLi ,Jian Cheng , Zechao Li , and Hanqing Lu.
[4]. Effect image retrieval based on hidden concept discovery in image database, 2007, Zhang, R.F., Zhang,
Z.F.
[5]. Evaluation of texture methods for image analysis, Mona Sharma, MarkosMarkou, Sameer Singh.
[6]. Finding textures by textual descriptions, visual examples, and relevance feedbacks, 2003, Hsin-ChihLin
,Chih-Yi Chiu , Shi-Nine Yang.
[7]. Image categorization via robust PLSA, 2010, Lu, Z.W, Peng, Y.X., Horace, H.S.Ip.
[8]. Improving the maximum-likelihood co-occurrence classifier: a study on classification of inhomogeneous
rock images, 2006 P.Paclik, S.Verzakov, R.P.W.Duin.
[9]. Macroscopic Rock Texture Image Classification Using a Hierarchical Neuro Fuzzy Class Method, 2010
Laercio B. Goncalves and Fabiana R. Leta.
[10]. Probabilistic Latent Semantic Analysis, Uncertainty in Artificial Intelligence, UAI'99, Stockholm, Thomas
Hofmann.
[11]. Recent development in the use of co–occurrence matrix for texture recognition, 2000, R.F. Walker,
P.T. Jackway, I.D. Longstaff.
[12]. Recent trends in texture classification Jing Yi Tou, Yong HaurTay, Phooi Yee Lau.
[13]. Rock image classification using color features in Gabor space, 2005 LeenaLepisto, IivariKunttu, Ari Visa.
[14]. Rock Image Classification Using Non-Homogenous Textures and Spectral Imaging, 2003, LeenaLepisto,
IivariKunttu, JormaAutio, and Ari Visa.
[15]. Rock image classification based on k- nearest neighbor voting, 2006,L. Lepisto, I. Kuntu, A, Visa.
[16]. Texture features for image classification. Robert M. Haralicket al.
[17]. Rock Textures Classification Based on Textural and Spectral Features Tossaporn Kachanubal, and
Somkait Udomhunsakul 2008.

AUTHORS BIOGRAPHY
R. Vinoth was born in Tamilnadu, India in 1985. He received B.E from Mohamed Sathak
Engg College, Kilakarai, Ramanathapuram, India in the year 2007 and M.E. (VLSI
DESIGN) from Muthayammal Engg College, Rasipuram, India in the year 2009. His area of
interest includes image processing, Signal Processing. He is having 5 years of teaching
experience in the department of Electronics and Communication Engg. He published few
research papers in international journals and presented few papers in national and
international conferences.

2446

Vol. 6, Issue 6, pp. 2439-2447

International Journal of Advances in Engineering & Technology, Jan. 2014.
©IJAET
ISSN: 22311963
R. Srinivasan was born in Tamilnadu, India in 1986. He received B.E from Muthayammal
Engg College, Rasipuram, India in the year 2008 and M.E. (Power Electronics and Drives)
from Paavai Engg College, Rasipuram, India in the year 20012. His area of interest includes
image processing, Machine Drives. He is having 2 years of teaching experience in the
department of Electrical and Electronics Engg. He published few research papers in
international journals and presented few papers in national and international conferences.

2447

Vol. 6, Issue 6, pp. 2439-2447


Related documents


PDF Document 14i18 ijaet0118683 v6 iss6 2439 2447
PDF Document ijetr2101
PDF Document 13n13 ijaet0313544 revised
PDF Document 11 553 937 1 sm
PDF Document hcs 341 week 3 learning team job description matrix
PDF Document 22n13 ijaet0313536 revised


Related keywords