# Introduction

he increase in computing power and electronic storage capacity has lead to an exponential increase of digital content available to users in the form of images which form the bases of many applications [1]. Consequently, the search for the relevant information in the large space of image databases has become more challenging. How to manage appropriate extracted outcome is still difficult problem and it is a proper field to make experiment. A typical image retrieval system includes feature extraction usually in conjunction with feature selection [2]. We can depict any image as a collection of color, texture and shape features. While several image retrieval systems rely on only one feature for the extraction of relevant images, but exact collection of relevant features can yield better retrieval performance [3]. The process of determining the combination of features that is most representative of a particular query image is called feature selection.

In case of analyzing real-world maps, the images shown there may not distinctly identify accurate and comprehensible information; rather lots of knowledge may be embedded in the domain in a hidden and unexplored form.


# a) Fuzzy Logic

The logic which works with approximation instead of exact and constant value is called fuzzy logic. The logic has been used from long back to solve various problem domains. The working value of fuzzy logic can be any value in between 0 and 1.Although the fuzzy logic is relatively young theory, the areas of applications are very wide: process control, management and decision making, operations research, economies and, for this paper the most important, pattern recognition and classification. An idea to solve the problem of image classification in fuzzy logic manner as well as comparison of the results of supervised and fuzzy classification was the main motivation of this work. Where, U M fcn, , V = (v 1 ,v 2 ,?,v c ), v i R p is the i th prototype m>1 is the fuzzifier and The objective is to find that U and V which minimize J m The Steps fro FCM Algorithm: 1. Choose: 1 < c < n, 1 < m < , = tolerance, max iteration = N 2. Calculation of membership values as according to the equation ( 2 ) , ( J


# II.


# Fuzzy c-means algorithm
ik c i n k m ik m D u U ?? ? ? ? V j i D D u c k m ik ij ij , 1 1 1 2 ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 2 2 k i ik D v x ? ? ? ? ? ? 3.
Computer the centroids values according to the equation (

(3)

4. Selection of new multiplier fields. 5. Repeat the step 2 until the algorithm has converged.

III.


# Fuzzy matching

Let us consider the fuzzy matching for the mixing images on the input images [10]. .The degree to which the input target images satisfy the conditions of fuzzy rules and conditions .Suppose IMAGE X is defined by rules R1 and IMAGES Y is defined by rules R2.In this case the matching degree will be represented by as follows:

Matching Degree (IMAGE X,R1)=?(IMAGE X)


# Matching Degree (IMAGE Y,R2) =? (IMAGE Y)

Where ? is the fuzzy membership function.

The fuzzy matching determines the actual outcome for fuzzy optimization which is accomplished here by fuzzy matrix. Here is a graphical view of fuzzy matching degree for IMAGE Y as follows: 


# Classification procedure

In our previous work we have done the classification by projecting the maximum classifier without NULL classifier is used. We implied a normal distribution and evaluate the variance and correlation of spectral response during the classification of the unknown pixel.

Here we have had fixed the partitioning as follows:

Let we have a data set X = {x 1 , x 2 ?., x n } R p and A classification of X is a c n matrix U = [U 1 U 2

?U n ] = [u ik ], where U n denotes the k-th column of U.

We have found three classifications efficient and suitable for our research activity. The labeled vectors for these classifications are: 1. N pc = { y R c : y i [0 1] i, y i > 0 i} Possibility Label 2. N fc = {y N pc : y i =1} Fuzzy Label 3. N hc ={y N fc : y i {0 ,1} i } Hard Label

The Fuzzy classification = V.


# Ontology and knowledge base

According to Ehrig (2007), ontology contains core ontology, logical mappings, a knowledge base, and a lexicon [3]. Core ontology, S, is defined as a tuple of five sets: concepts, concept hierarchy or taxonomy, properties, property hierarchy, and concept to property function. 
S = (C, ?c R, ? ,? R) _ two disjointi u u n k m ik n k k m ik i ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 1 1 x v µ 1 IMAGE Y µ 1 IMAGE X Best Degree of match=0.99 Degree of match=0.01 Poor ? ? ? ? ? ? ï?" ? ? ? ? ? ? k N M U M fc k pcn fcn ? ? ? ? U : ? ? C ? ? C ? F 2012

# June

where C and R are two disjoint sets called concepts" and relations" respectively. A relation is also known as a property of a concept. A function represented by ?(r) =< dom(r); ran(r) > where r ? R, domain is dom(r) and range is ran(r). A partial order ?R represents on R, called relation hierarchy, where r 1 ?R r 2 iff dom (r 1 ) ? C dom (r 2 ) and ran (r 1 ) ? C ran (r 2 ). The notation ? C represents a partial order on C, called concept hierarchy or taxonomy". In a taxonomy, if c 1 < C c 2 for c 1 ; c 2 ? C , then c 1 is a sub concept of c 2 , and c 2 is a super concept of c 1 . If c1 < C c 2 and there is no c 3 ? C with c1 < C c 3 < C c 2 , then c1 is a direct sub concept of c 2 , and c 2 is a direct super concept of c 1 denoted by c 1 c 2 . The core ontology formalizes the intentional aspects of a domain. The extensional aspects are provided by knowledge bases, which contain asserts about instances of the concepts and relations. A knowledge base is a structure KB = (C,R, I, C, , R) consisting of
? ? ? VI.

# Ontological instance matching algorithom

The operational block of the instance matching integrates ontology alignment, retrieves semantic link clouds of an instance in ontology and measures the terminological and structural similarities to produce matched instance pairs. Pseudo code of the Instance Matching algorithm 


# Experiments with real-world data

For the procedures of image classification was used to gather images from "Google Earth" on the Bangladesh region (Chittagong zone).). It uses this as a case study for implementing feature extraction. The collected images contain some common features such as roads, water, field, agriculture, buildings. The features will be separated based on the pixel intensity value selected for the individual features. It has been chosen as an application area because number of spatial features can be extracted from the map images of Forestry complex.

This image contains three channels recorded in three bands: the first band for green, the second for red and the third for blue. In the figure below, we present a fragment of this image and some statistics for the whole image.

Figure 2 : Forestry Complex Area After performing thresholding [ 5] based on color intensities defined for each and every feature, the features are higlighted with individual colors. Therefore, the highlighted feature area is clearly distinguished from the background. The thresholding process finally extract number of spatial features from the particular region such as road, water, field, building and forest.    
[i]=finput.nextInt(); matrix two[k][j]=finput.nextInt(); (matrixone[m][n]>=matrixtwo[m][n]) System.out.print(" "+matrixone[m][n]);} X.

# Result evaluations for fuzzy cmeans Classification

One way of the result evaluation was through the accuracy assessment. The classification results are compared to the raw image data and the report is created. This process is done during the random sample selection. The idea of the accuracy assessment is: point is highlighted in the sample list and observation [9] was done where it is located on the image.

The following table shows the mean and standard deviation for the classified classes Creation of the membership functions for the output variables is done in the similar manner. Since this is Sugeno-type inference, constant type of output variable fits the best to the given set of outputs (land classes). When the variables have been named and the membership functions have appropriate shapes and names, everything is ready for writing down the rules.

Based on the descriptions of the input (green, red and blue channels) and output variables (water, agriculture, forest, buildings, and roads), the rule statements can be constructed: Rules for image classification procedure in verbose format are as follows:  
IF (GREEN is

# Accuracy assessments by ontological classification

Idea for accuracy assessment of ontological classification results comes from the manner the maximum likelihood accuracy assessment was performed: select random sample areas with known classes and then let fuzzy logic 'say' what these samples are. With 100 random selected samples, results were as following:

Correctly classified samples: 92 Misclassified: 08 Accuracy: 92%

The both experiments and observations clearly showed that Fuzzy Logic classification is better than ontological knowledge base classification for Histo for Spatial Feature Extractions.


# XIV. Discussion and conclusion

This paper aimed for extracting the spatial features for providing a fundamental abstraction for modeling the structure of maps representing various raster images. The central part of this paper is an established procedure that is carried out for spatial Historical Heritages classification. As the work continues, it tries to implement every part of the procedure so as to establish its effectiveness and efficiency. It involved the use of supervised learning, assigning membership functions and discovery of pattern feature phases for successfully classifying an image. In the knowledge base, it must be well known whether selected sample forest area or water area.  
![Fuzzy C-means Algorithm capitalizes image segmentation under consideration of pixels values. It bring the pixels into multiple classes under the value of membership function. Fuzzy C-means Algorithm is formulated as the minimization of the following objective function:(1)](image-2.png "")
1![Fig. 1: The matching degree of fuzzy IV.](image-3.png "Fig. 1 :")
![sets C and R as defined before, _ a set I whose elements are called instance identifiers (or instance for short), _ a function ? C : C? ?"?(I) called concept instantiation, _ A function {? R: R ? ?"?(I2) with (r) (dom(r)) (ran(r)), for all r R. The function ? R is called relation instantiation. Global Journal of Computer Science and Technology Volume XII Issue X Version I Uncertainty Analysis for Spatial Image Extractions in the Context of Ontology and Fuzzy C-Means Algorithm](image-4.png "")
![Algo. InstanceMatch (ABox ab1, ABox ab2, Alignment A) for each insi element of ab1 cloudi=makeCloud(insi,ab1) for each insj element of ab2 cloudj=makeCloud(insj,ab2) if a(c1; c2) elements of A|c1 elements of Block(ins1:type) ^ c2 elements of Block(ins2:type) if Simstruct(cloudi; cloudj) ? imatch=imatch makeAlign(insi; insj) VII.](image-5.png "")
4![Figure 4: Extracted Features from Forestry Area ImageThe extracted features are further threshold for separating them from the background. This has been done by setting the background to all white form, thus displaying the particular feature is.](image-6.png "Figure 4 :")
![Uncertainty Analysis for Spatial Image Extractions in the Context of Ontology and Fuzzy C-Means Algorithm](image-7.png "")
5![Figure 5: Separated Features from Forestry Area Image VIII.](image-8.png "Figure 5 :")
![a1) AND (RED is a1) AND (NIR is a1) THEN (class is water) IF (GREEN is a2) AND (RED is a2) AND (NIR is a2) THEN (class is agriculture) © 2012 Global Journals Inc. (US) Global Journal of Computer Science and Technology Volume XII Issue X Version I](image-9.png "")
:ChannelMeanStandardwater (from 50 samples)DeviationGreen63.1122.12Red32.0118.31Blue47.5824.14Forest (from 75 samples)Green120.5432.31Red43.3527.02( D D D D )Blue Agriculture (from 50 samples) 33.1925.33Green92.1219.31Red98.5835.64Blue69.1121.65Buildings (from 50 samples)Green71.5525.35Red96.2521.98Blue48.5620.28Road (from 75 samples)Green89.3219.22Red64.105.63Blue49.5418.16
			© 2012 Global Journals Inc. (US)
		
		
			

				


* 
	
		Six Challenges for the Semantic Web
		
			VBenjamins
		
		
			JContreras
		
		
			OCorcho
		
		
			AG_Omez-P_Erez
		
	
	
		AIS SIGSEMIS Bulletin
		
			1
			
			2004
		
	




* 
	
		Weaving the Web: The Original Design and Ultimate Destiny of the World Wide Web
		
			TBerners-Lee
		
		
			MFischetti
		
		
			MDertouzos
		
		
			1999
			Harper
			San Francisco
		
	




* 
	
		Results of the Ontology Alignment Evaluation Initiative
		
			CCaracciolo
		
		
			JEuzenat
		
		
			LHollink
		
		
			RIchise
		
		
			AIsaac
		
		
			VMalais_E
		
		
			CMeilicke
		
		
			JPane
		
		
			PShvaiko
		
		
			Stuckenschmidt
		
		
			O_Sv_Ab-Zamazal
		
		
			VSv_Atek
		
	
	
		Proceedings of Ontology Matching Work-shop of the 7 th International Semantic Web Conference
				Ontology Matching Work-shop of the 7 th International Semantic Web ConferenceKarlsruhe, Germany, 73{119
		
			2008. 2008
		
	




* 
	
		A hierarchical fuzzy classification approach for highresolution multispectral data Over urban areas
		
			AShackelford
		
		
			CHDavis
		
	
	
		IEEE Trans. Geo. Rem. Sen
		
			41
			9
			
			2003
		
	




* 
	
		Texture analysis and classification of ERS SAR images for map updating of urban areas June in the Netherlands
		
			RobJDecker
		
	
	
		IEEE. Trans. Geo. Rem. Sen
		
			41
			9
			
			2003
		
	




* 
	
		Automatic road extraction using fuzzy concepts. IGARSS'98
		
			BSolaiman
		
		
			RFiset
		
		
			FCavayas
		
		
			1998. 2004
			Seattle, USA; Sunnyvale, CA
		
	




* 
	
		Enhancing a Database Management System for GIS with Fuzzy Set Methodologies
		
			EStefanakis
		
		
			TSellis
		
	
	
		Proceedings of the 19th International Cartographic Conference
				the 19th International Cartographic ConferenceOttawa, Canada
		
			August 1999
		
	




* 
	
		Introduction to neural networks and their use in remote sensing
		
			MTörmä
		
	
	
		Photogrammetric Journal of Finland
		
			13
			1
			
			1992
		
	




* 
	
		Automated Road Extraction from Satellite Imagery Using Hybrid Genetic Algorithms and Cluster Analysis
		
			HLiu
		
		
			JLi
		
		
			MAChapman
		
	
	
		Journal of Environmental Informatics
		
			1
			2
			
			2003
		
	




* 
	
		Bu_ering Fuzzy Maps in GIS
		
			HGuesgen
		
		
			JHertzberg
		
		
			RLobb
		
		
			AMantler
		
	
	
		Spatial Cognition and Computation
		
			2003
		
	




* 
	
		
			WKPratt
		
		Digital Image Processing
				
			John Wiley & Sons. NY
			1991
		
	




* 
	
		Idrisi 32 Release 2, Guide to GIS and Image Processing, Volume I-II
		
			REastman
		
		
			May 2001
		
		
			Clark University
		
	




* 
	
		Multi-resolution, object-oriented fuzzy analysis of remote sensing data for GIS-ready information
		
			UBenz
		
		
			PHofmann
		
		
			GWillhauck
		
		
			ILingenfelder
		
		
			MHeynen
		
	
	
		Journal of Photogrammetry & Remote Sensing
		
			2004