ISSN 2091-2730
Table of Content
Topics Page no
Chief Editor Board
3-4
Message From Associate Editor
5
Research Papers Collection
6-1148
ISSN 2091-2730
CHIEF EDITOR BOARD
1. Dr Chandrasekhar Putcha, Outstanding Professor, University Of California, USA 2. Dr Shashi Kumar Gupta, , Professor,New Zerland
3. Dr Kenneth Derucher, Professor and Former Dean, California State University,Chico, USA 4. Dr Azim Houshyar, Professor, Western Michigan University, Kalamazoo, Michigan, USA
5. Dr Sunil Saigal, Distinguished Professor, New Jersey Institute of Technology, Newark, USA
6. Dr Hota GangaRao, Distinguished Professor and Director, Center for Integration of Composites into Infrastructure, West Virginia University, Morgantown, WV, USA
7. Dr Bilal M. Ayyub, professor and Director, Center for Technology and Systems Management, University of Maryland College Park, Maryland, USA
8. Dr Sarâh BENZIANE, University Of Oran, Associate Professor, Algeria
9. Dr Mohamed Syed Fofanah, Head, Department of Industrial Technology & Director of Studies, Njala University, Sierra Leone
10. Dr Radhakrishna Gopala Pillai, Honorary professor, Institute of Medical Sciences, Kirghistan
11. Dr Ajaya Bhattarai, Tribhuwan University, Professor, Nepal ASSOCIATE EDITOR IN CHIEF
1. Er. Pragyan Bhattarai , Research Engineer and program co-ordinator, Nepal ADVISORY EDITORS
1. Mr Leela Mani Poudyal, Chief Secretary, Nepal government, Nepal 2. Mr Sukdev Bhattarai Khatry, Secretary, Central Government, Nepal 3. Mr Janak shah, Secretary, Central Government, Nepal
4. Mr Mohodatta Timilsina, Executive Secretary, Central Government, Nepal 5. Dr. Manjusha Kulkarni, Asso. Professor, Pune University, India
6. Er. Ranipet Hafeez Basha (Phd Scholar), Vice President, Basha Research Corporation, Kumamoto, Japan
ISSN 2091-2730
Technical Members
1. Miss Rekha Ghimire, Research Microbiologist, Nepal section representative, Nepal
2. Er. A.V. A Bharat Kumar, Research Engineer, India section representative and program co-ordinator, India 3. Er. Amir Juma, Research Engineer ,Uganda section representative, program co-ordinator, Uganda
4. Er. Maharshi Bhaswant, Research scholar( University of southern Queensland), Research Biologist, Australia
ISSN 2091-2730
Message from Associate Editor In Chief
Let me first of all take this opportunity to wish all our readers a very happy, peaceful and prosperous year ahead.
This is the Fifth Issue of the Third Volume of International Journal of Engineering Research and General Science. A total of 143 research articles are published and I sincerely hope that each one of these provides some significant stimulation to a reasonable segment of our community of readers.
In this issue, we have focused mainly on the students innovation and ongoing challenging trends. We also welcome more research oriented ideas in our upcoming Issues.
Author‘s response for this issue was really inspiring for us. We received many papers from many countries in this issue but our technical team and editor members accepted very less number of research papers for the publication. We have provided editors feedback for every rejected as well as accepted paper so that authors can work out in the weakness more and we shall accept the paper in near future. We apologize for the inconvenient caused for rejected Authors but I hope our editor‘s feedback helps you discover more horizons for your research work.
I would like to take this opportunity to thank each and every writer for their contribution and would like to thank entire International Journal of Engineering Research and General Science (IJERGS) technical team and editor member for their hard work for the development of research in the world through IJERGS.
Last, but not the least my special thanks and gratitude needs to go to all our fellow friends and supporters. Your help is greatly appreciated. I hope our reader will find our papers educational and entertaining as well. Our team have done good job however, this issue may possibly have some drawbacks, and therefore, constructive suggestions for further improvement shall be warmly welcomed.
Er. Pragyan Bhattarai,
Associate Editor-in-Chief, P&R,
International Journal of Engineering Research and General Science
E-mail [email protected]
ISSN 2091-2730
Comparative Study of Data Mining Classification Techniques over Soybean Disease by Implementing PCA-GA
Dr. Geraldin B. Dela Cruz
Institute of Engineering, Tarlac College of Agriculture, Philippines, [email protected]
Abstract— Data mining is a relatively new approach in the field of agriculture that can be used in the extraction of knowledge and discovery of patterns and relationships in agricultural data. Classification techniques in data mining are used to discover patterns and knowledge agricultural datasets, however, the accuracy of these classification techniques depends on the quality of data that are used as inputs in the data mining process. In this paper, an efficient data mining methodology based on PCA-GA is applied as a data pre processing technique, to reduce the dimensionality of the soybean dataset. The mechanism draws improvements to classification problems by applying Principal Components Analysis (PCA) and subsequently applying Genetic Algorithm (GA) to further reduce the dimensionality of the dataset, and selecting the best representative subsets, thereby improving the performance of classifiers. Different data mining classification techniques are applied to the resulting reduced dataset and classification metrics are compared. This approach is to asses classification rates on the PCA-GA reduced soybean dataset. The learning and validation experiment was performed using WEKA, a workbench containing implementations of the k-NN, Naïve Bayes, J4.8 and MLP classification algorithms, including the PCA and GA. Classification accuracy was validated using 10-folds cross validation..
Keywords—classification, data mining, genetic algorithm, optimization, PCA-GA, soybean, INTRODUCTION
Databases not only store and provide data but also contain hidden knowledge which can be very important however, human ability to analyze and understand these massive datasets lags far behind his ability to gather and store data in databases. Data in the agricultural domain are robust, comes in different formats, complex, multidimensional and contains noise. Interesting patterns can be mined in this space in discovering knowledge, revealing solutions to specific domain problems [1]. Extraction of the useful set of features is usually unknown from these volumes of data [2], considering every single feature of an input pattern in a large feature set makes classification and knowledge discovery computationally complex. Also, the inclusion of irrelevant or redundant features in the data mining model results in poor predictions and interpretation, high computational cost and high memory usage [3], [4].
In general, it is desired to keep a number of features as discriminating and as small as possible in order to reduce computational time and complexity in the data mining process [5], [6]. This can be addressed by dimensionality reduction [7] method that improves data mining classification and facilitates visualization and data understanding. It is a process that creates a set of features based on transformations or combinations of the original dataset, thereby reducing it into a smaller representative dataset.
The focus of this study is to implement an efficient mechanism based on the combination of Principal Component Analysis (PCA) and Genetic Algorithm (GA) [8], [9] as a data preprocessing method, to reduce the dimensionality of the data by keeping a number of features as small as possible. Apply the k-NN, MLP, NB and J4.8 algorithms in the data mining process and compare the classification results. In so doing, the PCA-GA [10] mechanism is validated as an efficient data reduction method.
THE PCA-GA MECHANISM
The PCA-GA algorithm is a combination of two algorithms, PCA, for data preprocessing and reduction and the GA for feature subset selection method, which makes the whole process a hybrid data mining mechanism. The PCA mechanism is a very useful data dimensionality reduction technique, reducing the number of variables to a few interpretable linear combinations of the data. PCA maps the rows and columns of a given matrix into two or three dimensional points to reveal the structure of dataset. The original data are projected into smaller space. Thus data reduction is performed. The data can be represented by a collection of n points in the z- dimensional space, where each axis corresponds to a measured variable. From this space, a line Y1 can be searched such that the dispersion of n points when projected unto this line is a maximum. The derived variable is denoted by the equation in (1):
Y1 = e1x1 + e2x2 + …epxp (1)
Where, ei are coefficients satisfying the condition in (2):
ISSN 2091-2730
∑
j= 1 ze
i2= 1
(2)
After obtaining Y1, the (z-1) – dimensional subspace orthogonal to Y1 is considered and line Y2 is found in this subspace such that the dispersion of points when projected onto this line is also maximum, and is perpendicular to Y1 such that the dispersion of points when they are onto this line is the maximum. Having obtained Y2, a line in the (z-2)-dimensional subspace is considered, which is orthogonal to both Y1 and Y2, such that dispersion of points when projected onto this line is as large as possible. The process can be continued until z mutually orthogonal lines are determined. Each of these lines now defines a derived variable shown in equation (3):
Yi = e1iX1 + e2iX2 + e3iX3 + … + eniXi (3)
where the constants eij are determined by the requirement that the variance of Yi is a maximum, subject to the constraint of orthogonality as well as in (4) :
∑
k= 1 p
e
ik2= 1
(4)
Thus, the Yi obtained in (3) are called principal components of the system. The process produces a list of linear vectors in (5) called principal components.
Y1 = e11X1 + e12X2 + … + e1pXp Y2 = e21X1 + e22X2 + … + e2pXp
…. (5)
Yp = ep1X1 + ep2X2 + … + eppXp
Each of the principal components can be thought of as a linear regression predicting Yi in (3) from X1, X2, …, Xp. There is no intercept, but ei1, ei2, …, eip can be viewed as regression coefficients. The first principal component is the linear combination of X variables that has maximum variance among the linear combinations, accounting for as much variation in the data as possible. The remaining principal components, accounts for as much of the remaining variation as possible, thus the principal components are uncorrelated with each other.
GA Process
Genetic algorithms are search algorithms based on natural genetics. It is an iterative process that operates on a population or a set of candidate solutions, in this case, the principal components generated by the PCA mechanism. Each solution is obtained by means of encoding/decoding mechanism, which enables representing the solution. GA is considered as a function optimizer and performs efficient by searching the best representative sets from candidate sets of solutions (principal components). The interest is in the minimization of a set of variables that can represent a dataset with maximum results. Genetic algorithms consist of three essential elements: a coding of the optimization problem, a mutation operator and crossover. The coding of the optimization problem produces the required discretization of the variable values and makes their simple management in a population of search points possible. A binary coding is ideal because in this way the mutation and crossover operators are simple to implement. Thus, the values of the individuals P1…Pn can be encoded with the binary fixed-point coding in (6).
P1 = b5b4b3b2b1b0
……. (6)
Pn = b5b4b3b2b1b0
The crossover operator, control the recombination of the individuals in order to generate a new, better population of individuals at each iteration step. Before recombining, the individuals must be evaluated by a fitness function (7) for all data structures in the population. The fitness value is then used as the basis for selection in the crossover or mating.
Fitness = countone(Pn) (7)
A typical reproduction operator is crossover (8). Before the crossover, two individuals P1 and P2 are selected as ―parents‖, based on their fitness value. Selection is based on tournament using their fitness. Individuals with higher fitness value, wins the tournament and
ISSN 2091-2730
is chosen to mate. The offspring C1 and C2 are formed so that the left side comes from one parent and the right side from the other.
This produces an interchange of the information stored in each parent. The whole process is reminiscent of genetic exchange in living organisms.
C1 = Mask1 & P1 + Mask2 & P2 (8)
C2 = Mask2 & P1 + Mask1 & P2 Where :
P1 , P2 – parents chromosomes C1, C2 - children chromosomes
Mask1, Mask2 – bit masks (Mask2 = NOT(Mask1))
A favorable interchange can produce an offspring with better genes. When the individuals P1 = b5b4b3b2b1b0 is recombined with the number P2 = a5a4a3a2a1a0. The new individual is then:
Cn = b9b8 ··· biai−1 ··· a0. (9)
Crossover can be interpreted as a variation of optimization. The mutation operator is the simplest. In binary strings, a mutation corresponds to a bit flip. A mutation of the ith bit of the string Cn = b5b4b3b2b1b0 produces a change. Thus a new individual is generated and the fitness is again evaluated.
DATA MINING CLASSIFICATION TECHNIQUES
The book in [11], identifies, presents and describes different data mining classification techniques [12]. Classification [13] is a form of data analysis that can be used to construct a model, which can be used in the future to predict the class label of new datasets. It is a two step process, first is the learning step where the classification algorithm builds the classifier by analyzing a training set made up of database tuples and their associated class labels, using a mapping function in the form of classification rules. In the second step, the accuracy of the classifier is predicted.
Some of the most popular and common classification algorithms are adopted and presented herein, based on their capabilities simplicity and robustness. The k-NN and Naïve Bayes were chosen based on the study in [14], which proves to perform excellent using the WEKA [15] data mining tool, likewise the J4.8 and MLP were used for their exemplary performance in classification problems in different datasets [16].
k-Nearest Neighbor (k-NN)
k-NN is a nearest neighbor algorithm that classifies entities taking a class of the closest associated vectors in the training set via distance metrics. The principle behind this method is to find predefined numbers of training samples closest in the distance to the new point and predict the label from these.
Naïve Bayes
Based on the Bayes rule of conditional probability, it uses all of the attributes contained in the data, and analyses them individually as though they are equally important and independent of each other. It considers each of the attributes separately when classifying a new instance. It assumes that one attribute works independently of the other attributes contained by the sample.
J4.8
A popular tree based machine learner, the J4.8 decision trees algorithm is an open source Java implementation of the C4.5 [17]. It grows a tree and uses divide-and-conquer algorithm. It is a predictive machine-learning model that decides the target value (dependent variable) of a new sample based on various attribute values of the available data. To classify a new item, it creates a decision tree based on the attribute values of the training data. When it encounters a set of items in a training set, it identifies the attribute that discriminates. It uses information gain to tell most about the data instances so that it can classify them the best.
Multi Layer Perceptron (MLP)
MLP is a feed forward artificial neural network model that maps sets of input data onto a set of appropriate outputs. It consists of multiple layers of nodes with each layer fully connected to the next one. Each node is a neuron with a nonlinear activation function. It uses a learning technique called back propagation for training the network.
ISSN 2091-2730
MATERIALS AND METHODS
The soybean dataset in [18], was used in the experiment. The data mining software used in the experiment is the WEKA version 3.6.10, which contained the implementations of the classification algorithms presented. A computer with two (2) Gigabytes of memory, with a 32 bit 2.80 Ghz processor, and a proprietary 32 bit Operating System was utilized. The default settings in the data mining software and in the configurations of the algorithms were used.
The dataset was cleaned, encoded and saved as attribute relation file format (arff) file using a text editor, the dataset were loaded in WEKA, the PCA was applied as a data preprocessing method to transform and simplify the dataset into smaller representative sets called principal components. The GA was then applied to the PCA transformed dataset that resulted to an optimized dataset.
Subsequently, each of the machine learning algorithms in WEKA, the J4.8, Naïve Bayes, MLP and k-NN were then applied to the resulting PCA-GA reduced datasets and results were recorded and compared accordingly.
Classification accuracy was validated using 10-fold cross validation. This validation method is a standard method to estimate classification accuracy over unseen data.
RESULTS AND DISCUSSION
This section presents and discusses the results of the experiment after applying the PCA-GA mechanism used in the study. There are two parts; first, the outcome of applying the PCA-GA mechanism that was used in generating the visualization model of the reduced dataset. Second the comparison of classification results of classifiers between the original and PCA-GA reduced dataset.
(a) (b)
Figure 1. Visualization of the Soybean Disease (a) original dataset, (b) PCA-GA reduced dataset
The soybean dataset originally has thirty six (36) attributes. As can be seen in Figure 1b, after applying the PCA-GA mechanism to the soybean dataset, resulted to fifteen (15) feature sets. This reduced dataset is now considered the smaller representative soybean dataset.
As can be observed, the resulting feature sets in Figure 1b are simplified in structure compared to the original dataset in Figure 1a. The reduced dataset is the optimized smaller representative dataset of the original, thru the optimization technique of GA. The presented visualization in Figure 1b, confirms the PCA-GA efficiency as a dimensionality reduction method, this implies that extracting knowledge from this smaller and optimized dataset is more accurate, efficient and faster. Based on analysis of Figure 1b, the features sets are combinations of the original attributes, that resulted in process of pre-processing through the PCA-GA mechanism. Further the figure shows a distinct variation of the feature sets, classifying the different soybean disease.
Classifier Original Dataset PCA-GA Reduced Dataset
Accuracy Time Accuracy Time
k-NN 91.22% 0.00 sec 99.85% 0.00 sec
J4.8 91.51% 0.02 sec 98.68% 0.01 sec
Naïve Bayes 92.97% 0.01 sec 94.44% 0.00 sec
MLP 93.41% 112.0 sec 98.83% 18.20 sec
Table 1. Classification Rates of Soybean Dataset When Applied with Different Classifiers
ISSN 2091-2730
Table 1 shows the comparison of classification rates of various classifiers over the original and PCA-GA reduced soybean dataset. It can be seen that the classification rates between the original and reduced dataset, have noticeable improvements in the J4.8, Naïve Bayes, k-NN and MLP classifiers. Among the classification algorithms tested, the fastest was the k-NN classifier and a significant improvement on accuracy can be observed. Interesting to note also is the speed of the MLP, which is significantly faster on the PCA- GA reduced dataset, the accuracy also significantly increased. Although the MLP improved on its processing time on the reduced dataset, it can be seen that its processing time took longer compared to the other classifiers
Figure 2. Comparison of Accuracy Rates of the PCA-GA Reduced Dataset When Applied with Different Classifiers In comparing the classifiers accuracy rates on the PCA-GA reduced dataset, it can be seen in Figure 2, that the k-NN classifier is the most accurate. Naïve Bayes is the least accurate among the classifiers, with MLP and J4.8 are as nearly accurate as the k-NN.
Generally, the results presented imply that the PCA-GA method significantly improves classification rates, over the soybean disease dataset with the PCA-GA dimensionality reduction method.
ACKNOWLEDGMENT
The researcher would like to extend his gratitude to the Tarlac College of Agriculture for the support of this study.
SUMMARY AND CONCLUSION
Presented in this study, is the PCA-GA hybrid data reduction method and comparison of classification rates of various data mining classification techniques over the soybean dataset. The k-NN, J4.8. Naïve Bayes ad MLP classifiers were applied to the resulting PCA-GA reduced datasets. Based on the results, the PCA-GA mechanism reduced the original dataset into smaller representative datasets. Visualization model was generated based on the result of the data reduction mechanism based on PCA-GA to present a clearer view of its potential as a hybrid method. Results also show that classification accuracy and processing speed improved for all of the classifiers.
The implementation of the algorithm based on PCA as a preprocessing technique and GA as a feature subset selector is efficient in reducing the dimensionality of the soybean dataset. Results imply that all classifiers can be implemented and are efficient in classifying soybean disease using the PCA-GA pre processing mechanism. Classification speed is further improved for all the classifiers used. On the other hand, results of comparison between the classifiers accuracy rates show that the k-NN is the most accurate and the least accurate is the Naïve Bayes. Generally, using the PCA-GA data reduction technique proves to have significant results in characterizing soybean disease using the presented data mining classification techniques. Thus, simplifies the process of extracting knowledge, discovering patterns and relationships and the interpretation of soybean disease.
REFERENCES:
[1] Arora, Rohit and Suman Suman. "Comparative analysis of classification algorithms on different datasets using WEKA."
International Journal of Computer Applications Vol 54, No 13, pp. 21-25, 2012
[2] Raymer, Michael L., William F. Punch, Erik D. Goodman, Leslie Kuhn, and Anil K. Jain. "Dimensionality reduction using genetic algorithms." Evolutionary Computation, IEEE Transactions Vol 4, No 2, pp. 164-171, 2000.
90.00%
91.00%
92.00%
93.00%
94.00%
95.00%
96.00%
97.00%
98.00%
99.00%
100.00%
k-NN J4.8 Naïve Bayes MLP
ISSN 2091-2730
[3] Qu, Guangzhi, Salim Hariri, and Mazin Yousif. "A new dependency and correlation analysis for features." Knowledge and Data Engineering, IEEE Transactions Vol 17, No. 9, pp. 1199-1207, 2005.
[4] Janecek, Andreas, Wilfried N. Gansterer, Michael Demel, and Gerhard Ecker. "On the relationship between feature selection and classification accuracy‖. Journal of Machine Learning Research-Proceedings Track 4, Antwerp, Belgium, pp. 90-105, 2008.
[5] Gerardo, Bobby D., Jaewan Lee, Inho Ra, and Sangyong Byun. "Association rule discovery in data mining by implementing principal component analysis." In Artificial Intelligence and Simulation, pp. 50-60. Springer Berlin Heidelberg, 2005.
[6] Diepeveen, D. & Armstrong, L. ―Identifying key crop performance traits using data mining‖. IAALD AFITA WCCA2008, World Conference on Agricultural Information and IT, 1-21, 2008
[7] Burges, Christopher JC. ―Dimension reduction: A guided tour, Machine Learning‖. Foundations and Trends in Machine Learning, Vol. 2, No. 4, pp. 275-365, 2009.
[8] Yang, Jihoon, and Vasant Honavar. "Feature subset selection using a genetic algorithm." In Feature extraction, construction and selection, pp. 117-136. Springer US, 1998.
[9] Goldberg, David E., and John H. Holland. "Genetic algorithms and machine learning." Machine learning Vol 3, No. 2 pp. 95- 99, 1988.
[10] Cruz, Geraldin B. Dela, Bobby D. Gerardo, and Bartolome T. Tanguilig III. "An Improved Data Mining Mechanism Based on PCA-GA for Agricultural Crops Characterization." International Journal of Computer and Communication Engineering Vol 3, No. 3, pp. 221-225, 2014
[11] Han, Jiawei, Micheline Kamber. Data Mining: Concepts and Techniques, 2nd Edition, Morgan Kaufmann, pp 285-289, 2006 [12] Phyu, Thair Nu. "Survey of classification techniques in data mining." In Proceedings of the International MultiConference of
Engineers and Computer Scientists, vol. 1, pp. 18-20. 2009.
[13] Kotsiantis, Sotiris B., I. Zaharakis, and P. Pintelas.. ―Supervised Machine Learning : A review of Classification Techniques‖.
Informatica 31, pp 246-268, 2007
[14] Wahbeh, A. H., Q. A. Al-Radaideh, M. N. Al-Kabi, and E. M. Al-Shawakfa. "A comparison study between Data Mining Tools over some classification methods. IJACSA, Special Issue on Artificial Intelligence." SAI Publisher Vol 2, No 8 (): 18- 26, 2010.
[15] Hall, Mark, Eibe Frank, Geoffrey Holmes, Bernhard Pfahringer, Peter Reutemann, and Ian H. Witten. "The WEKA data mining software: an update." ACM SIGKDD explorations newsletter Vol 11, No. 1, pp. 10-18, 2009.
[16] Beniwal, Sunita, and Jitender Arora. "Classification and feature selection techniques in data mining." International Journal of Engineering Research & Technology (IJERT) Vol 1, No. 6, 2012
.
[17] Quinlan, J. Ross. C4. 5: Programs for Machine Learning, Vol.1. Morgan Kaufmann. 1993
[18] Bache, Kevin, and Moshe Lichman. "UCI machine learning repository, 2013." URL http://archive. ics. uci. edu/ml (1990): 92.
ISSN 2091-2730
Strength of Sway Frame Infilled with Cement Stabilized Laterite Block
M.E. Ephraim1, L.N. Uzoewulu2 and T.C. Nwofor3
1, 2 Department of Civil Engineering, Rivers State University of Science and Technology, Port Harcourt, Rivers State, Nigeria
3Department of Civil Engineering, University of Port Harcourt, Rivers State, Nigeria
ABSTRACT- The experimental results of the strength and failure mechanism of sway frames with cement stabilized laterite block infill are presented in this study. The engineering properties of cement stabilized laterite block produced from local laterite from Rumueme, Port Harcourt, Nigeria and cement content of zero, two, four, six and eight percent are first investigated to produce blocks at optimum moisture content and then testing for strength property carried out. It was established that four percent cement stabilized laterite block, compacted at its optimum moisture content and dry density produced the best combination of physical and mechanical properties as the tensile, shear and boned strengths for the four percent cement stabilized laterite blocks meet the minimum Code specifications for sandcrete and brick blocks. This agrees with the findings of Nigerian Building and Road Research Institute (NBRRI), from a similar experiment conducted in Kano State, Nigeria. The strength of 1.13N/mm2 achieved at 28days is higher than the Nigerian Industrial Standard (NIS) specification of IN/mm2 for cement stabilized laterite block. The strength and failure mechanism of a one-quarter scale model reinforced concrete frame infilled with four percent cement stabilized laterite block followed the trend established for sandcrete and brick infills with the sway resistance capacity of the frame increasing by about 300%. The experimental collapse load of the frame agrees with the theoretical collapse load to about 1%. From the foregoing a wider application of cement stabilized laterite block as infill will lead to cost effective housing delivery as laterite is very available and relatively cheap.
1.0 INTRODUCTION
The need for decent, functional and affordable housing for the citizenry has formed the thrust of government policies since the 1970s. However, the issue of high cost of basic building materials and hence affordability of housing by the masses still remains topical. In an attempt to address this problem, the federal Government of Nigeria through the Nigerian Building and Road Research Institute (NBRRI) has embarked on the research, development and application of Cement Stabilized Laterite Blocks (CSLB) for affordable housing. In driving the research and development of cement stabilized laterite blocks, NBRRI embarked on collaborative research with the Ahmadu Bello University Zaria, Nigeria, using the black cotton soils of Kano.
On the basis of the studies, it was established that the optimum cement content, satisfying standards for strength, durability and economy of CSLB was put averagely at four percent [1]. The compressive strength values obtained in these tests ranged from 1.50-1.68 N/mm2. It is obvious that the optimum cement will depend on the type of larterite, varying from higher values for more sandy to lower values for the clayey lateritic soils. In view of the above, NBRRI has recommended that the required cement for local laterites should be determined from laboratory tests on trial mixes before production of CSLB. Secondly, CSLB is basically used as walling units and therefore may function as infill for structural frame. Although structural frames are generally designed to resist all lateral loads ignoring the contribution of infill, structural cracks are frequently observed on walls of buildings [2], [3]. The fact is that some loads are transferred to the infill, which must therefore, have some capacity to absorb the induced stresses and deformations [4], [5], [6]. The situation throws the second challenge, namely the need to
ISSN 2091-2730
investigate the structural load carrying capacity and failure mechanism of CSLB inflled frames. In this regard, technical literature reveals a dire paucity in available research studies.
2.0 THEORETICAL AND CONCEPTUAL FRAMEWORK OF STUDY
The study basically based on structural modeling which is an important tool in the analysis, design, and testing of prototype structures. Experiments are carried out on the model that is a reduced scale semblance of the prototype from where the behavior of the prototype is predicted. The theory of dimensional analysis that involves similitude requirements is employed in the sizing and experimentation of the model to predict the prototype behavior.
Figure 1: Infilled Frame Model
Considering the infilled frame structure in Figure 1, the following steps are considered to stimulate the physical parameters of the model.
i. The diagonal tensile stress of the CSLB infill wall depends on the loading Q, span L, the thickness t and the modulus of elasticity E and may he represented im as;
(, Q, I, t, E) = 0 (1)
ii. Taking, the modulus E and span L as dimensionally independent variables [7], [8], the above equation can he expressed as
G(/E, Q/EL2, t/L) = 0 (2)
where
Q
t L
ISSN 2091-2730
= (Q/EL2, t/L) (3)
Hence
(Q/EL2)P = (Q/EL2)m
Qp = Qm SE S2L, Where SE = 1.
iii. For the same material in model and prototype, the scale factor for material SE = 1, and the prototype load QP = QM SL2. Where SL is the linear scale factor.
Therefore the failure load of the prototype can be computed by multiplying that of the model by the square of the linear factor, and based on material scale factor, the model material is the same as the prototype material hence the stress is the same in model and prototype. The success of this theory reduces cost of experimentation, enhances safety and ease with which the behavior of large prototype structures are analyzed and predicted.
3.0 EXPERIMENTAL PROCEDURE
The method of research was based on structural modeling and experimental test in the laboratory. The determination of basic properties of local laterite produced is done in accordance with BS 1377, 1995- Method of Test for Civil Engineering Soils and BS 3921, (1995) and Method of Test for clay bricks and blocks, [9], [10].
The structural modeling of a previously designed frame was achieved on the assumption of a uniform scale factor for materials (laterite, cement, water, sand) and a linear scale factor of 1:4 was adopted for determination of deflections and dimensions.
3.1 Materials
The basic materials used for this study were as follows; water, local laterite from Rumueme borrow pit in Rivers State Nigeria, cement, and sand. They are described briefly in the sub-headings below.
3.1.1 Equipment
These include wooden mold conforming to 1/4 scale model dimensions of 75x37.5x37.5mm, equivalent to the prototype dimensions of 300x150x150mm, hydraulic jack, strain and dial gages, and proving ring.
ISSN 2091-2730
3.1.2 Water
Water conforming to BS 3148 (1995); obtained from the university water supply network was used for the experiment.
3.1.3 Laterite
The laterite was obtained from a borrow pit located between Agip and Ada George Roads at Reumueme in Port Harcourt.
Laterite for the construction of Ada George Road was obtained from this borrow pit. The preparation of the laterite was in accordance with BS 1377 (1995) with respect to sampling the optimum moisture content (OMC), the maximum dry density (MDD), sieve analysis, and index properties determination.
3.1.4 Cement
Portland cement complying with BS 12 (1995) obtained from Eastern Bulkcem Limited, was used in the experiment.
3.1.5 Aggregate
Aggregate conforming to fine gravel on the particle size curve was used in the production of concrete. The aggregate conformed to BS 882 (1995).
3.2 Laterite Block Production
Production of model blocks was by ramming into a wooden mold with internal dimensions 75x37.5x37.5 mm and pressing down to ensure no voids or honeycomb on block when extruded. Curing was in accordance with BS.3578. (1995). The freshly molded block were cured for 7days by covering with polythene sheets to prevent any loss of water. After 7days, the polythene sheet covering was removed. The strength of the blocks was tested at 7,14 and 28 days. The water absorption test was done at 28days for each sample at ½ hour, 1hour, 2hours and 24hours in accordance with BS 3921
3.3 Construction of Model Frame and Test Setup
The model frame was constructed after analysis of a prototype frame to determine the size of reinforcement and load capacity. Basically the prototype column section was controlled by the prototype block size of 300 x 150x150 mm. The details are set out below in Tables 1.
ISSN 2091-2730
Table 1: Prototype and Model Frame Dimensions
Member Prototype (mm) 1/4 Model (mm)
Frame (Full Size) 3000x3000 750x750
Column Section 150x150 37.x37.5
Beam Section 300x150 75x37.5
Block Section 300x150x150 75x37.5x37.5
Mortar thickness 12 3
Reinforcement Y 12 Y3 (BRC wire)
Stirrup R6 R1.5 (binding wire)
Binding wire R1.5 R0.375 (fly screen wire)
Aggregate 12(1/2‘‘) 3 (1/8‘‘)
Foundation Depth 900 x 600 225x150
Four different models were constructed at the premises of the structural engineering laboratory, Rivers State University of science and Technology Port Harcourt, Rivers State, Nigeria. Each frame foundation was made very rigid to avoid over turning effect during testing. The frame starter reinforcement was fixed and cast in-situ with the foundation. The entire foundation was covered and flushed with the ground level and compacted to represent true site situation. The experiment was carried out in dry season when the ground was dry and strong enough to resist the racky load.. The frames were loaded using a hydraulic jack as shown in Figure 2. The laboratory stanchions provided the reactant action for the jack. The strain gages tagged GI, G2,G3 and G4 are 100mm gages. Gages G1 and G3 measured strain on the compression diagonal while G5 and G4 measured strain on the tension diagonal. The dial gage, G5 measured displacement in the sway direction. The hydraulic jack connected to the proving ring assembly, exerted horizontal force on the frame. The load was applied in steps of 0.50KN and maintained for 5 minutes to allow for stabilization of the frame under the load and observation of cracks which may develop. The load at which crack appeared on the infill was recorded as the collapse load.
3.4 Loading Scheme
The horizontal sway load capacity of the prototype and model frame was calculated as 23.81KN and 1.48KN respectively and the loading scheme for the frame was carried out in accordance with BS 5628: part: 1995. The load values were read on the proving ring. The readings of strain gages G1, G2, G3,G4 and G5 were recorded at each loading cycle, until the infill frame showed cracks on the infill panel. The load at which the cracks appeared on the infill was recorded as the maximum load capacity of the frame. The strain gages G1, G2, G3 and G4 recorded internal strains on the infill, while dial gage G5 recorded the horizontal away of the frame. The frame deflection, stress/strain profiles were plotted to show the infill behavior during loading up to failure.
ISSN 2091-2730
4.0 RESULTS AND DISCUSSION
This chapter presents the result obtained in the experimental investigation conducted in this study. The results are analyzed with the Table and graphs provided. The discussion of the results is done with comparison with the code provisions and the result of similar researches conducted by researchers in other locations. The graph of deflections and stress- strain behavior of the infill frame are shown, and deflection values calculated. Model horizontal load was applied to the frames and deflections/strains induced on the frames were measured. The mode of failure of each frame was noted and photographed.
The average resistance of the infill in the three frames with infill was used in calculating the stresses on captured as would be seen in the figures below.
4.1 Failure Mechanism of the Frames
The failure mechanisms of the frames are shown in the figures below. Figure 2 shows the collapsed mechanisms of the rigid frame (MF0) with hinge formation on the joint of the columns, while figure 3 to 5 shows the collapse mechanisms of infilled frames MFI, MF2, and MF3, respectively with the crack formations at critical sections
Figure 2: Failure Mode of MF0 Frame Figure 3: Failure Mode of MF1 Frame
Figure 4: Failure Mode of MF2 Frame Figure 5: Failure Mode of MF3 Frame
ISSN 2091-2730
The failure of the rigid frame was observed to be by formation of hinges at the upper and lower portions of the two columns, wile the failure of the infill frame was by formation of diagonal cracks starting from the load application point running parallel to the loading diagonal. The crack width ranged between 2 and 3mm. Separation of infill panel from the column near the application of the load as well as, separation of the infill panel from the ground beam was also observed. There were also micro cracks initiating from point of application of the load mainly due to corner crushing mode of the infilled frame.
Generally, it was observed that the infill increased the resistance of the frame as would be seen from the readings and the reduced and corresponding calculation from the proving ring and dial gauge.
4.2 Critical Loads on Model Frame.
The critical loads are the loads at which first cracks are initiated on the frames without further increase on the proving ring readings. These loads are shown on the Table 2 below. These loads are used to predict the prototype sway load for the infilled frame.
Table 2: Critical Sway Loads
Frame type Height (mm) Span (mm) Infill Type Load (kN)
MF0 ― 750 ― 750
MF1 ― ― CSLB infill 5.00
MF2 ― ― ― 3.50
MF3 ― ― ― 4.50
From the foregoing the critical load on the prototype model can be predicted as follows:
Critical load on frame MF0 load through experimental modeling = 1.5KN Predicted load from analysis =1.49KN
Accuracy of predication = 1.5/1.49 = 1.0067
Critical prototype Load, = 1.5S2 (from similitude requirement ) = 1.5x42 = 24kN
It is possible to determine the prototype infill frame sway loads based on the model result by applying the scale factor as also seen in Table 2.
4.3 Deflection of Infilled Frame Model
The deflection of the frames was monitored by dial gage seen in the test set-up in the sway direction. Generally it was observed that the introduction of the cement stabilized Laterite Block with acted as an infill panel reduced the horizontal defection of the frame by 67% as can be seen in the comparison with the load-displacement profile of the rigid frame model
ISSN 2091-2730
tagged MF0. The MF0 model which is without infill, was swayed through 18.5mm before collapse by a horizontal load of 1.5kN, as recorded by the dial gage. The infilled frame models tagged MF1, MF2 and MF3 were displaced through, 11.6mm 3.6mm and 3.08mm respectively by horizontal load of 4.4, 3.5 and 5kN acting on the frame respectively. The load- displacement diagram is shown in Figure 6 where it is clearly seen that the introduction of infill reduced the sway of the frames at increased load before collapse by stiffening the frame.
Figure 6: Load-Displacement Profile for Structural Models MF0, MF1, MF2 and MF3
4.4 Streets-Strain Characteristics of Infilled Frame Model
The estimation of the linear strain in the infill was monitored at increased load and measured as points G1, G2, G3, and G4, using strain gages and the values recorded. Diagonal cracks and separation of infill from the frames through a length of about 600mm on each frame were observed, as well as separation from the foundation through the entire length of 750mm as shown in Figure 2 to 5. This could be attributed to the frames bearing pressure on the infill while being acted upon by the external horizontal load and the foundation providing the counter thrust to the horizontal force. The end action was that frame and infill were crushed under the two opposing force leading to the collapse of the frame and infill. Strain gages G1 G3 measured strain in same direction but opposite to strain gages G2, G4 which were placed on the other side of the infill panel measuring strain in same direction. The net strain is given by [19] as G0 = G1 + G2 – G3+ G4 with the stress-strain profile given in Figure 7.
0 1 2 3 4 5 6
0 5 10 15 20 25 30
Loa d ( kN )
Deflection (mm)
MF0 MF1 MF2 MF3
ISSN 2091-2730
Figure 7: Stress-Strain Profile of Infilled Frame Models, MFI, MF2, MF3
Crack formation on the infill was visible and it is attributed to the limit of shell buckling, which is the point at which failure of the interior web of the infill starts [8],[11]. This behavior is attributed to the failure of each member of the composite even though it still contributes to the overall resistance of the composite frame till collapse mechanisms, thus confirming that the high in-plane rigidity of the masonry wall significantly stiffens the otherwise relatively flexible frame, while the ductile frame contains the brittle masonry after cracking, up to critical sway load. The tensile stress which corresponds to the maximum tensile strain [12], is obtained as 0.09N/mm and it compares favorably with the codes specification [9], [10] for sandcete blocks and clay bricks.
5.0 CONCLUSION
This study the strength of sway frame using cement stabilized literite block unit as structural infill has been presented. From the analysis of the results obtained in the study, the following conclusions are arrived at.
1. Four percent Cement Stabilized Leterite Blocks, produced at optimum moisture content and maximum dry density presents the best combination of physical and mechanical properties. The strength of 1.13 N/mm2 achieved in
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
0 0.05 0.1 0.15 0.2
Str ess (kN /mm
2)
Strain (mm/mm)
MF1 MF2 MF3
ISSN 2091-2730
28days is within the Nigerian Industrial Standards (NIS) specification for cement stabilized laterite blocks. This also confirms the applicability of four percent cement content recommended by NBRRI to Niger Delta laterite soil.
2. The strength and failure mechanism of a ¼ scale model reinforced concrete frame, with the four percent CSLB as infill, followed the trend established for sandcrete and burnt brick block infills and the sway resistance capacity increased from 1.5kN to 4.5kN, suggesting an average of 300%
3. The experimental failure load compares favorably with the theoretical failure load from the model collapse load.
4. The tensile, shear and bond stresses for the CSLB are within the code specification for infill material. The values obtained for four percent cement stabilized infill constitute respectively, 0.09N/mm2, 0.22N/mm2 and 0.21Nmm2. 5. The CSLB is cost effective and competitive over sandcrete and burnt brick in terms of strength; thus a wider use of
CSLB as infill will lead to cost effective housing delivery in the country.
5.1 RECOMMENDATIONS
1. The CSLB from Rumueme laterite is recommended for the state mass housing project.
2. The effort of Nigerian Building and Road Research Instituate should cover the entire nation by collaborating with all the universities in research on the local laterite, to establish the economic cement mix for production of cement stabilized laterite block for the locality in order to help local people in housing.
3. Further research on the carrying capacity of CSLB infill and resistance to both racky and gravity loads is recommended.
REFERENCES:
1) Nigerian Building and Road Research Institute (NBRRI) (1988). Ten years of Building and Road Research pp13-30.
2) Riddington J.R. (1994). Composite behavior of walls interacting with flexural members. Ph.D, Thesis, University of Southampton.
3) Stafford, S. B. (1974). The composite behavior of infilled frames. Symposium on tall buildings, University of Southampton.
4) Weeks, G.A. and Mainstone, R.J. (1970). Influence of a builging frame on racking stiffness and strengths of brick wall.
Second International Brick Masonry Conference, Stoke on Trent, pp. 165-171.
5) Nwofor, T.C.(2012). Shear resistance of reinforced concrete infilled frames. International Journal of Applied Science and Technology, 2(5), 148-163.
6) Nwofor, T.C. and Chinwah, J.G. (2012). Finite Element Modeling of Shear Strength of Infilled Frames with openings.
International Journal of Engineering and Technology, 2(6), 992- 1001.
ISSN 2091-2730
7) Ephraim, M.E. (1999). Modeling techniques and instrumentation in laboratories. Unpublished lecture note, University of Science and technology port Harcourt Nigerian, pp.2-19
8) Sabnis, G.M., Harris, H.G., White, R.N., and Saeed Mirza, M. (1983). Structural Modelling and Experimental Techniques.
Prentice-Hall, Inc., Englewood Cliffs, N.J. 07632, USA
9) BS 3921 (1995). Method Of Test Of Soil for Civil Engineering purposes, British Standard Institution London, pp.5-39 10) BS 1377 (1995). Clay Bricks and Blocks, British Standard Instauration. London
11) Nwofor, T.C. (2012). Numerical micro-modeling of masonry infilled frames. Archives of Applied Science Research, 4(2), 764-771.
12) Hetenyi, M. ―Beams On Elastic Foundations‖. Scientific Journal service, University of Michigan Studies, Vo1. XVI. 1946
ISSN 2091-2730
Texture Filter based Medical Images Segmentation for Cancer Disease
Dr. Sana'a khudayer Jadwa Computer Unit, College of Medicine
Al-Nahrain University Baghdad- Iraq
E-mail: [email protected]
Abstract— Medical Image processing is one of the most challenging topics in research field. In medical field, CT (Computed Tomography) scan imaging and MRI (magnetic resonance imaging) are the most important for image based visual diagnostics, but applying segmentation to these images is very tedious and requires an adjusting approach. The main objective of medical image segmentation is to extract and characterize anatomical structures with respect to some input features or expert knowledge. In this paper we have formulated a simple, general, fast, and user-friendly approach to the problem of medical image segmentation based on texture filter. In this method, the experimental results show that the segmentation results are visually satisfactory of medical image texture segmentation.
Keywords
— Medical image processing, image texture, image segmentation, texture analysis,
medical imaging, , Medical image Analysis ,texture filter.1. INTRODUCTION
Medical imaging application plays an indispensable role by automating or facilitating the delineation of anatomical structures.
Medical image segmentation is a challenging task due to the various characteristics of the images, which leads to the complexity of segmentation. [1]. In computer vision, Image segmentation is known as a process of partitioning an image into several segments also known as super pixels. The important goal of image segmentation is to simplify or change the representation of an image into form that is more meaningful and is easy for analysis [2]. Segmentation is an important process in the analysis of MR (Medical Resonance) Images for medical diagnosis. It divides the MR image into different types of classes and groups the homogeneous pixels into clusters.
This is used in medical diagnosis in many ways, detecting brain tumor, tissue analysis, bone fractures and similar problems [3].
Segmentation of medical images involves three main image-related problems. The image may contain noise that can alter the intensity of a pixel such that its classification becomes uncertain. Also, the images can exhibit intensity nonuniformity where the intensity level of a tissue class varies gradually over the extent of the image. Third, the images have finite pixel size are subject to partial volume averaging where individual pixel volumes contain a mixture of tissue classes so that the intensity of a pixel in the image may not be consistent with any single tissue class[4].An image texture can be defined as the local spatial variations in pixel intensities and orientation. In order to recognize objects and scenes in computer vision, it is essential to be able to partition an image into meaningful regions with respect to texture characteristics. Texture segmentation has a wide range of applications like content based image retrieval, medical diagnosis, analysis of satellite or aerial images, surface defect detection and terrain classification for mobile robot navigation[5]. This paper produce texture segmentation method for medical images. The organization of the rest of this paper is as follows. Section 2 highlights the related works. Section 3 introduces image texture analysis . Section 4 describes the proposed method. Section 5 present the experimental results and section 6 concludes the paper.
2.RELATED WORKS
Medical image segmentation is a challenging task due to the various characteristics of the images, which leads to the complexity of segmentation Eldman and Maria [6] introduced an automatic method of medical image segmentation used in the study of the Central Nervous System (CNS) by multilevel thresholding based on histogram difference. V. Grau*, A. U. J. Mewes [7] Presented a method to combine the watershed transform and atlas registration, through the use of markers. This new algorithm applied to two challenging
ISSN 2091-2730
applications: knee cartilage and gray matter/white matter segmentation in MR images. Numerical validation of the results is provided, demonstrating the strength of the algorithm for medical image segmentation. Ch.Hima Bindu,QISCET, Ongole [8] Employed an optimized Otsu method based on improved thresholding algorithm for medical image segmentation ,the experimental results show that the new optimized method dramatically reduces the operating time and increases the separability factor in medical image segmentation while ensures the final image segmentation quality. Seongjai Kim and Hyeona Lim [9] proposed the background subtraction (MBS) in order to minimize difficulties arising in the application of segmentation methods to medical imagery. Ebrahim and Dehmeshki[10]developed a method that requires the definition of a speed function that controls curve evolution. The image intensity gradient and the curvature are utilized together to determine the speed and direction of the propagation. Although level set methods are highly effective in segmenting image, but. they are sometimes unable to exactly detect objects in images with low- contrast boundaries. In this method hybrid speed functions are used for an implicit active contour (level set) method which is capable of segmenting images with low-contrast boundaries.
3.IMAGE TEXTURE ANALYSIS
The regular repetition of an element or pattern on a surface it is called as texture. It is used to identify different textured and non- textured regions in an image, to classify/segment different texture regions in an image, to extract boundaries between major texture regions ,figure(1) illustrate three examples of image texture[11]:
Figure(1): Different examples of image texture
Texture is a difficult concept to represent, the identification of specific textures in an image is achieved primarily by modeling texture as a two-dimensional gray level variation. The relative brightness of pairs of pixels is computed such that degree of contrast, regularity, coarseness and directionality[11]. Texture analysis refers to the characterization of regions in an image by their texture content. Texture analysis attempts to quantify intuitive qualities described by terms such as rough, silky, or bumpy in the context of an image. In this case, the roughness or bumpiness refers to variations in the brightness values or gray levels[12]. Texture analysis of an image gives distributed arrangements of the intensity of the pixel in an image[13].An image texture can be defined as the local spatial variations in pixel intensities and orientation. In order to recognize objects and scenes in computer vision, it is essential to be able to partition an image into meaningful regions with respect to texture characteristics[14].Texture analysis is used in a variety of applications, including remote sensing, automated inspection, and medical image processing. Texture analysis can be used to find the texture boundaries, called texture segmentation[12]. All texture functions operate in a similar way. They define a neighborhood around the pixel of interest calculate the statistic for that neighborhood and then use the computed statistic value as the value of the pixel of interest in the output image. The example that shown in Figure( 2) illustrates how the range filtering function operates on a simple matrix. In this example, the value of element B (2, 4) is calculated from A (2, 4). Range filtering function use m by n pixels, in this example 3 × 3, neighborhood around the pixels[12].
ISSN 2091-2730
Figure(2): Range filtering function 4.PROPOSED METHOD
In this paper the texture segmentation for medical image based on applying a range filter is proposed. At first the color ima ge is read from a database that contain a collection of medical images, then these images are converted to grayscale image, later the range filter is applied in order to extract the texture content of medical image. The texture segmentation process diagram is illustrated in figure(3) as shown below:
Figure(3): Block diagram for texture image segmentation 4.1IMAGE DATABASE
The starting point of this work was the creation of a database with six medical images modalities having different sizes that is collect from the web. The database consist of three groups : brain cancer , kidney cancer and stomach cancer images. Figure (4) show the database images.
Figure(4): Different medical images
ISSN 2091-2730
4.2 RGB IMAGE CONVERSION
The colored medical image is converted to gray scale image by converts RGB values to grayscale values by forming a weighted sum of the R, G, and B components using equation(1):
………..(1) 4.3RANGE FILTERING
Filtering is perhaps the most fundamental operation of image processing. The term filtering can be defined as the value of the filtered image at a given location. It is a function of the values of the input image in a small neighborhood of the same location. Filter operators can be used to sharpen or blur images, to selectively suppress image noise, to detect and enhance edges, or to alter the contrast of the image. The filters use the local statistical variations in an image to reveal the edges and its histogram [15].
5.EXPERIMENTAL RESULTS
The proposed algorithm is applied on the medical images of cancer disease, at first the color image is reading from the database then it converted to grayscale as shown in figure(5):
Figure (5 ): Colored Image conversion.
Then the range filter is applied on grayscale image to obtain the texture segmented medical image as show in figure ( 6 ):
Figure (6 ): Texture segmented medical image
The same steps applied for the rest image given the result as shown in figure ( 7 ):
ISSN 2091-2730
Texture segmented image Grayscale image
Original image
Figure(7): Results of texture segmented for medical images
ISSN 2091-2730
6.CONCLUSION
In this paper we have considered the method of texture filter for an effective segmentation of medical images.
The method uses the range filter to achieve robust and accurate segmentation results which are visually satisfactory. The whole process is autonomous and requires no supervision, which is one of the advantages of the proposed method. The method guarantees best segmentation of textures in poor-quality images also. The resulting figure show the efficiency, simplicity and robustness of medical image texture segmentation.7.REFERENCES
[1] Ch. Hima Bindu1 and K. Satya Prasad2," An Efficient Medical Image Segmentation Using Conventional OTSU Method", International Journal of Advanced Science and Technology Vol. 38, January, 2012.
[2] Dilpreet Kaur and Yadwinder Kaur," Intelligent Medical Image Segmentation Using FCM, GA and PSO", International Journal of Computer Science and Information Technologies,Vol. 5 (5), 2014.
[3] R. Venkateswaran1and S. Muthukumar," Genetic Approach on Medical Image Segmentation by Generalized Spatial Fuzzy C- Means Algorithm", IEEE International Conference on Computational Intelligence and Computing Research,2010.
[4] Daniel J. Withy and Zoltan J.Koles,"A Review of Medical Image Segmentation: Methods and Available Software", International Journal of Bioelectromagnetism Vol. 10,No.3, pp125-148. 2008.
[5] Saka.Kezia, Dr.I.Santi Prabha and Dr.V.VijayaKumar," A New Texture Segmentation Approach for Medical Images", International Journal of Scientific & Engineering Research Volume 4, Issue 1, ISSN 2229-5518 , January-2013.
[6] Eldman de Oliveira Nunes and Maria Gabriela Pérez," Medical Image Segmentation by Multilevel Thresholding Based on Histogram Difference", 17th International Conference on Systems, Signals and Image Processing, 2010.
[7]
V. Grau*, A. U. J. Mewes, M. Alcañiz, Kikinis, and S. K. Warfield " Improved Watershed Transform for Medical ImageSEGMENTATION USING PRIOR INFORMATION",IEEETRANSACTIONSONMEDICALIMAGING,VOL.23,NO.4,APRIL2004
.
[8] Ch.Hima Bindu,QISCET, Ongole," AN IMPROVED MEDICAL IMAGE SEGMENTATION ALGORITHM USING OTSU METHOD", International Journal of Recent Trends in Engineering, Vol 2, No. 3, November 2009.[9] Seongjai Kim and Hyeona Lim" Method of Background Subtraction for Medical Image Segmentation", The work is supported in part by NSF grants DMS- 0312223 & DMS-0609815, 2005.
[10] Ebrahim doost, Y.; Dehmeshki, J. ; Ellis, T.S. and Firooz bakht, M,".Medical Image Segmentation Using Active Contours and a Level Set Model: Application to Pulmonary Embolism (PE) Segmentation", IEEE Fourth International Conference on Digital
Society, 269-273, Feb. 2010.
[11] Vaijinath V. Bhosle and Vrushsen P. Pawar " Texture Segmentation: Different Methods",International Journal of Soft computing and Engineering (IJSCE) ISSN: 2231-2307, Volume-3, Issue-5, November 2013 .
[12] Matheel Emaduldeen Abdulmnim," Segmenting the Dermatological Diseases Images by Developing the Range Operator", Iraqi Journal of Science, Vol 55, No.3B, pp:1376-1382, 2014.
[13] Asheesh Kumar, Apurva Mohan Gupta, Naresh Rameshrao Pimplikar and Natarajn P," TEXTURE SEGMENTATION IN MEDICAL IMAGING FOR RED SPOT BLOTCHES ANALYSIS IN HUMAN BODY",International Journal of Advanced Research in Computer and Communication Engineering ,Vol. 3, Issue 3, March 2014.
[14] Saka.Kezia, Dr.I.Santi Prabha and Dr.V.VijayaKumar," A New Texture Segmentation Approach for Medical Images", International Journal of Scientific & Engineering Research Volume 4, Issue 1, January-2013.
[15] Amir Rajaei, LalithaRangarajan and ElhamDallalzadeh," MEDICAL IMAGE TEXTURE SEGMENTATION USINGRANGE FILTER", Computer Science & Information Technology,2009.