Remote Sensing Textual Image Classification based on Ensemble Learning

Size: px
Start display at page:

Download "Remote Sensing Textual Image Classification based on Ensemble Learning"

Transcription

1 I.J. Image, Graphcs and Sgnal Processng, 2016, 12, Publshed Onlne December 2016 n MECS ( DOI: /jgsp Remote Sensng Textual Image Classfcaton based on Ensemble Learnng Ye zhwe 1, Yang Juan 1, Zhang Xu 1, Hu Zhengbng 2 1 School of Computer Scence, Hube unversty of Technology, Wuhan, Chna 2 School of Educatonal Informaton Technology, Central Chna Normal Unversty, Wuhan, Chna Abstract Remote sensng textual mage classfcaton technology has been the hottest topc n the fled of remote sensng. Texture s the most helpful symbol for mage classfcaton. In common, there are complex terran types and multple texture features are extracted for classfcaton, n addton; there s nose n the remote sensng mages and the sngle classfer s hard to obtan the optmal classfcaton results. Integraton of multple classfers s able to make good use of the characterstcs of dfferent classfers and mprove the classfcaton accuracy n the largest extent. In the paper, based on the dversty measurement of the base classfers, J48 classfer, IBk classfer, sequental mnmal optmzaton (SMO) classfer, Nave Bayes classfer and multlayer perceptron (MLP) classfer are selected for ensemble learnng. In order to evaluate the nfluence of our proposed method, our approach s compared wth the fve base classfers through calculatng the average classfcaton accuracy. Experments on fve UCI data sets and remote sensng mage data sets are performed to testfy the effectveness of the proposed method. Index Terms Remote Sensng, Textual Image Classfcaton, Ensemble Learnng, Baggng. I. INTRODUCTION Remote sensng mage mnng or classfcaton s one of the most mportant methods of extractng land cover nformaton on the Earth [1]. Dfferent from standard alphanumerc mnng, mage mnng or classfcaton s very dffcult because mages data are unstructured [2]. There are two man mage classfcaton technques, unsupervsed mage classfcaton and supervsed mage classfcaton. As for supervsed mage classfcaton, frst, the user selects representatve samples called tranng set for each land cover classes, then a learnng classfer s traned by a set of gven tranng data set whch contans a lot of tranng samples, n the end, the traned classfer wll be utlzed for practcal applcaton. In each tranng samples, there are a low-level feature vector and ts related class label. The traned classfer s able to dstngush unknown low-level feature vectors nto a class whch has been traned. Several classfers lke Maxmum Lkelhood Classfer, Mnmum Dstance Classfer has been used for mage classfcaton [3]. Wth the development of remote sensng technology, the spatal and spectral resoluton of remote sensng mages has been gettng hgher and hgher [4]. It presents new challenges to remote sensng mage classfcaton and requres the development of new data classfcaton methods. Many new classfcaton methods such as spectral nformaton dvergence, object orented paradgm appeared [5]. To a certan extent, these classfers or classfcaton strategy can mprove the classfcaton accuracy; however, dfferent classfers have ther own characterstcs. For dfferent applcatons, the performance of classfcaton s not dentcal [6]. Some of the samples are wrongly classfed by one classfer whle these samples may be correctly labeled by another classfer, whch ndcates that there s complementarty between the classfers. It s dffcult to desgn a powerful model for classfyng remote sensng mage because the model should not only have man dscrmnaton nformaton of remote sensng mage and t should be robust to ts varatons at the same tme. As a result, only mprovng tradtonal methods to acheve robust classfcaton s not always feasble. In 1998, Dun et al. proposed combnng multple classfers to enhance classfcaton performance of a sngle classfer [7]. That s, the combnaton of classfers s able to amend the errors made by a sngle classfer on dstnct parts of the nput space. It s concevable that the performance of combnng multple classfers s better than one of the base classfers used n solaton [8]. The emergence of ensemble learnng provdes a new research dea for solvng the problem of strong correlaton and redundancy exsts n the bands. Hanson et al. frstly proposed the concept of neural network ensemble [9]. They proved that, the generalzaton ablty of learnng systems could be sgnfcantly mproved through the tranng of multple neural networks. In 2011, multple classfers ensemble was appled to face recognton [10]. At the same year, support vector machne (SVM) was used as the base classfer to recognze the facal expresson [11]. As s known, texture s a vtal characterstcs for remote sensng mage nterpretaton. However, texture often changes n orentaton, scale or other vsual appearance thus t s hard to be accurately descrbed by use of a sngle mathematcal model. Generally, several descrptors wll utlzed for classfyng textures, whch may mprove the classfcaton accuracy and lead to classfcaton dffculty n the meantme. In the paper, based on the dversty measurement of the base classfers, J48 classfer, IBk classfer, sequental mnmal optmzaton (SMO) classfer, Nave Bayes

2 22 Remote Sensng Textual Image Classfcaton based on Ensemble Learnng classfer and multlayer perceptron (MLP) classfer are selected for ensemble learnng. These classfers respectvely use C4.5 classfcaton algorthm, Nave Bayes classfcaton algorthm, k-nearest Neghbors (k- NN) classfcaton algorthm, artfcal neural network (ANN) classfcaton algorthm as the base classfer. In order to evaluate the nfluence of our proposed method, our approach s compared wth the fve base classfers through calculatng the average classfcaton accuracy. The remander of ths paper s organzed as follows. Secton 2 brefly revews the ensemble learnng. In Secton 3, the selecton of base classfers and the proposed method s descrbed n detal. The effectveness of the proposed method s demonstrated n Secton 4 by experments on several publc data sets from UCI machne learnng repostory and real remote sensng mages. Fnally, Secton 5 draws the concluson from the expermental results. II. OVERVIEW OF ENSEMBLE LEARNING In a narrow sense, ensemble learnng just uses the same type of learners to learn the same problem. For example, we can put all the learners as support vector machne or neural network classfers. In a broad sense, a varety of learners are appled to solve the problem, whch could be also consdered as ensemble learnng. The followng s the dea of ensemble learnng. In general, when learnng new examples, the dea of ensemble learnng s ntegratng multple ndvdual learners and the result s determned by combnng the results of multple learners n order to acheve better performance than a sngle learner [12]. If consdered the sngle learner as a decson maker, ensemble learnng s consdered as the decson whch s made by a number of decson-makers. Wth combnng k base classfers, M1, M2,, M k, * an mproved composte classfcaton model M s created. A gven data set D1, D2,, D where k D (1 k 1) s devoted to generate classfer M, s used to create k tranng data sets. The ensemble result s a predcton of class based on votes from the base classfers. The flow chart of ensemble learnng s shown n Fg 1. Fg.1. The flow chart of ensemble learnng There are three methods consdered n the theoretcal gudance to ensemble learnng: 1) Sample set reconstructon. 2) Feature level reconstructon. 3) Output varable reconstructon. The way to employ ensemble learnng has two steps usually. The frst step s to obtan ndvdual models through producng several tranng subset. The second step s to use the synthess technology to get the fnal results on the ndvdual output through the thrd methods. Detterch expounded why an ensemble learner s superor to a sngle model n three ways [13]. Usually from a statstcal perspectve, the hypothess space need to be searched s very large, but t s not enough to accurately learn the target hypothess as only a few tranng samples could be used to compare wth real samples n the world, whch causes the results of learnng to be a seres of hypotheses that meet the tranng sets and have approxmaton accuracy. The hypotheses may well meet the tranng sets but not hold a good performance n practce, n consequence the choce of only one classfy wll lead to a bg rsk. Fortunately, t s able to reduce ths rsk by consderng multple hypotheses at the same tme. III. MULTIPLE CLASSIFIERS ENSEMBLE BASED ON BAGGING ALGORITHM A. Selecton of Base Classfers As s dscussed above, an mportant reason for the success of ensemble classfer algorthm s that a group of dfferent base classfers are employed. Dversty among a team of classfers s deemed to be a key ssue n classfer ensemble [14]. However, measurng dversty s not specfc for there s no wdely accepted formal defnton. In 1989, Lttlewood and Mller proposed that dversty has been recognzed as a very mportant characterstc n classfer combnaton [15]. However, there s no rgd defnton of what s drectly perceved as dependence, dversty or orthogonalty of classfers. Many measures of the connecton between two classfer outputs are able to be derved from the statstcal lterature, such as the Q statstcs and the correlaton coeffcent. There are formulas, methods and deas amng at quantfyng dversty when three or more classfers are concerned, but lttle s put on a strct or systematc bass due to lack of a defnton. The general antcpaton s that desgnng the base classfers and the combnaton technology can be helped by dversty measures. In the paper, we measure the dversty between 5 types of supervsed classfer and use the non-parwse dversty measures, such as entropy, Kappa measure, Kohava- Wolpert varance, etc. Let D { D1, D2,, D L } be a set of base classfers. Let { 1, 2,, c } be a set of class labels, n whch x be a vector wth n features to be labeled [16].The entropy measure E s defned as Eq.(1) N 1 1 E mn{ l( x j), L l( x j)} (1) N ( L [ L / 2]) j1

3 Remote Sensng Textual Image Classfcaton based on Ensemble Learnng 23 E vares between 0 and 1, where E 0 ndcates no dfference between base classfes and E 1 ndcates the hghest possble dversty. It ndcates hgher possble dversty when the value of entropy s larger. Kohav-Wolpert Varance use a specfc classfer model vx (1 P ( y 1 x ) P ( y 0 x ) ) to express the 2 dversty of the predcted class label y for x across tranng samples, where P( y x) s estmated as an average over dfferent data sets. Averagng over the entre Z, the KW measure of dversty s defned as Eq.(2) N 1 KW ( )( ( )) 2 l z j L l z j (2) NL j 1 Let p be the average accuracy of each classfcaton, 1 N L j.e., p y,, then Kappa measurement s NL j 1 1 defned Eq.(3) N l( x)( L l( x)) 1 (3) 1 K 1 N N( L 1) P(1 P) From Eq.(3), the Kappa value ncreases wth the ncrease of the correlaton between classfers. In ths paper, J48 classfer, IBk classfer, SMO classfer, Nave Bayes classfer and MLP classfer are selected as the base classfers. Frst of all, the fve supervsed classfers are ntroduced brefly. Then, the dversty between these fve classfers s measured by usng the non-parwse dversty measures. 1) J48 Classfer Decson tree learnng construct predctve model as a decson tree, mappng observatons about conclusons about an tem's target value. Ross Qunlan developed the algorthm of decson tree whch called C4.5. C4.5 s an expanson of earler ID3 algorthm [17]. C4.5 has the same way as ID3 to buld decson trees from tranng data by the concept of nformaton entropy. The method of the constructon of decson tree was frst derved from Hunt method, whch ncludes two steps [18]. The frst step s f there s only one class, the node s a leaf node, otherwse t wll enter the next step. The second step s to search for a varable that s to dvde the data nto two or more subsets of data wth hgher purty accordng to the condton of the varable. That s to say, t selects the varable accordng to local optmalty and then returns to the frst step. J48 s an open source Java mplementaton of the C4.5 algorthm n Weka. The classfcaton rules of the C4.5 algorthm are easy to understand and ts accuracy s hgh. Its man drawback s that t needs to scan and sort the data set repeatedly n the process of constructng the decson tree, whch leads to the low effcency of the algorthm. 2) IBk The second classfcaton chosen s k-nn. The nput of k-nn comprses k closest tranng sets n the feature space and the output s a class member. An object s classfed as a majorty of ts neghbors, and the object s assgned to the commonest k nearest neghbor. If k 1, then the object s assgned to the class of that sngle nearest neghbor n a nutshell. The tranng examples, each of whch wth a class label, are vectors n a multdmensonal feature space. In the classfcaton phase, k s a user-defned constant value. The unlabeled vector s classfed by attrbutng the label most frequent among the k tranng samples nearest to that unknown pont. When the tranng samples are a few, t can smply put the tranng set as a reference set. When there are many tranng samples, t can use the exstng selecton or calculate the prototype of the reference sets. k-nn algorthm has strong adaptablty to the tested samples wth more overlappng domans. A commonly used dstance metrc for contnuous varables s Eucldean dstance [19]. The length of the lne segment connectng ponts and j, ( j ) s the Eucldean dstance. In Cartesan coordnates, the dstance (d) from to j, or from j to when ( x 1, x 2,, xn ) and j ( xj 1, x j2,, x jn) are two ponts n Eucldean n- space s gven by Eq.(4) d(, j) ( x x ) ( x x ) ( x x ) j1 2 j2 n jn k-nn s a non parametrc classfcaton technque [20]. It has hgher classfcaton accuracy to unknown and non normal dstrbuton. It has the advantages of ntutve thnkng, hgh feasble degree and clear concept. It drectly uses the relatonshp between the samples, whch can reduce the error probablty of the classfcaton and avod unnecessary trouble. Of course, t s a knd of lazy learnng method. It has the dsadvantages of slow classfcaton speed and strong dependence on sample sze. 3) Sequental Mnmal Optmzaton Classfer SMO s get from the dea of decomposton algorthm to the extreme for solvng the optmzaton problem [21]. It s an teratve algorthm to break the problem nto some smallest possble sub-problems, of whch the most promnent place s that the optmzaton problem of two data ponts can be obtaned analytcally. Therefore, there s no need to take twce plannng optmzaton algorthm as a part of the algorthm. Each step of SMO selects two elements to optmze [22]. The optmal values of two parameters need to be found and updated the correspondng vectors on the premse that other parameters have been fxed. In spte of more teratons needed to converge, there s an ncrease n the number of speed because the operaton of each teraton s very small. Based on the prncple of structural rsk mnmzaton and VC dmenson theory of statstcal learnng theory, Support vector machne (SVM) fnds the best balance between the learnng ablty and the complexty of the model by a certan samples, so as to get the best (4)

4 24 Remote Sensng Textual Image Classfcaton based on Ensemble Learnng promoton ablty. Compared wth the tradtonal artfcal neural network, SVM has the advantages of smple structure. It ncreases a lot n the generalzaton performance and solves the local optmum problem whch could not be avoded n the neural network. SVM can solve the problems of small collectve samples, hgh dmenson and nonlnear. It has a lot of specal propertes to ensure that the generalzaton ablty n learnng perod s better. At the same tme, t also averts the problem of dmenson. 4) Nave Bayes Nave Bayes classfer s a smple probablstc classfer based on Bayes' theorem, whch has a strong ndependent assumpton [23]. Bayes' theorem s based on the pror probablty of a gven class known, and then t uses the Bayes formula to calculate the posteror probablty. Fnally, the class that has the largest posteror probablty s selected as the class of object. In the theory, Nave Bayes s a condtonal probablty model: suppose the sample space of experment E to be S, represented by B1, B2,..., B n representng n features. A s a event of E and PA ( ) 0. For each of possble results or classes B, the nstance probabltes are P( B ) 0,( 1,2,..., n). Usng Bayes' theorem, the condtonal probablty s able to be calculated as Eq.(5) P( A B) P( B) P( B A), 1,2,..., n n P( A B ) P( B ) j1 A clear dstncton between Nave Bayes and other learnng methods s that t does not explctly search possble hypothess space. Nave Bayes algorthm takes less tme and consders the logc relatvely smple. Nave Bayes algorthm also has a hgh degree of feasblty and the characterstcs of logc and hgh stablty. 5) Multlayer Perceptron j An artfcal neural network s a smulaton of bologcal neural network system whch are used to evaluate or approxmate functons that can depend on a large number of smple computng unts connected n some form to form a network [24]. In the stage of network learnng, network acheves the correspondence between nput samples and correct sample by adjustng the weghts. The neural network has a strong ablty to dentfy and classfy the nput samples, whch s to fnd out the segmentaton regons each of whch belongs to a class meetng the classfcaton requrements through sample space n fact. A MLP s a feedforward ANN model, whch d M can be regarded as a mappng F : R R. What makes a MLP dfferent from other neural network s that a number of neurons use a nonlnear actvaton functon. Learnng happens n the perceptron by alterng the weght between neurons after each tranng sample s processed, based on the comparson between the output of the error and the expected results. j (5) Compared wth other algorthms, neural network has the advantages of hgh capacty of nose data, and t has a very good performance for the classfcaton of the tranng data. Dfferent numbers and types of classfers are used to measure ther dversty. The results are shown n TABLE I. Table 1. Kappa Measurement of Dfferent Classfers number classfers Kappa (J48,IBk) (J48,SMO) (J48,Bayes) (J48,MLP) (IBk,SMO) (IBk,Bayes) (IBk,MLP) (SMO,Bayes) (SMO,MLP) (Bayes,MLP) (J48,IBk,SMO) (J48,IBk,Bayes) (J48,IBk,MLP) (J48,SMO,Bayes) (J48,SMO,MLP) (J48,Bayes,MLP) (IBk,SMO,Bayes) (IBk,SMO,MLP) (IBk,Bayes,MLP) (SMO,Bayes,MLP) (J48,IBk,SMO,Bayes) (J48,IBk,SMO,MLP) (J48,IBk,Bayes,MLP) (J48,SMO,Bayes,MLP) (IBk,SMO,Bayes,MLP) (J48,SMO,MLP,IBk,Bayes) It can be seen from TABLE I, f we choose two classfers from all classfers, the best choce are IBk classfer and SMO classfer. Smlarly, f we choose three classfers from all classfers, the best choces are J48 classfer, IBk classfer and SMO classfer. If we choose four classfers from all classfers, the best choces are J48 classfer, IBk classfer, SMO classfer and Bayes classfer. If we want to get better results, we need to choose the fve classfers. Therefore, J48 classfer, IBk classfer, SMO classfer, Nave Bayes classfer and MLP classfer are chosen to conduct ensemble learnng.

5 Remote Sensng Textual Image Classfcaton based on Ensemble Learnng 25 B. Multple Classfers Ensemble based on Baggng Algorthm Baggng algorthm s a knd of ensemble learnng method mprovng the classfcaton by combnng classfcatons wth randomly selectng tranng sets, whch proposed by Breman n 1994 [25]. If there s a tranng set of sze m, t s practcable to draw m random nstances from t wth replacement. The m nstances are able to be learned, and ths process can be duplcated several tmes. Some duplcates and omssons are contaned n the nstance compared to the ntal tranng set, snce the draw s wth replacement. Through the process, each cycle results n a classfer. Accordng to the constructon of several classfers, the forecast of each classfer wll be a vote to nfluence the fnal forecast. Algorthm: Baggng algorthm Input: 1. D {( x1; y1),( x2; y2),...,( xm, ym)}, a set of m tranng tuples; 2. T, the number of models n the ensemble; 3. L, a classfcaton learnng scheme (J48, IBk, SMO, Nave Bayes and MLP). 4. Output: The ensemble --- a composte model, H( x) arg max l( y h ( x)) yy T t1 t. When the value n the parentheses s a true proposton, the sum s 1. Otherwse, the sum s Method: 6. for t 1 to T do 7. create bootstrap sample, D Bootstrap ( D) t, by samplng D wth replacement; 8. use Dt and the learnng scheme L to derve a model, h t ; 9. endfor 10. To use the ensemble to classfy a tuple, X : 11. Let each of the T models classfy X and return the majorty vote; Gven a tranng set D {( x1; y1),( x2; y2),...,( xm, ym)} of sze m, baggng algorthm generates T new tranng sets D, each of sze m, by samplng from D congruously. t For each sample set, the probablty s 1 (1 1/ m) m. For large m, the unque examples wll be 11/ e 63.2 and the rest wll be duplcates. The T models are ftted usng the above T knds of samples whch known as a bootstrap sample and combned by castng votes. It has the correct * classfcaton rate r max P( x) PX ( x). Based on baggng algorthm, the probablty of correct classfcaton can be as Eq.(6) r max P ( x ) P ( x ) [ I ( ( x ) ) P ( x )] P ( x ) A xc X C ' A X It can be seen from the correct rate that the result of baggng algorthm s better than the results obtaned by a sngle predcton functon. IV. SIMULATION RESULTS AND DISCUSSION A. Experments for Publc Data Sets In order to evaluate the performance of multple classfers ensemble based on baggng algorthm, fve publc data sets from UCI machne learnng repostory named Image segment, german_credt, hepatts, onosphere and soybean are used n ths part. For example, Image segmentaton data set has 19 contnuous attrbutes, 210 tranng samples and 2100 test samples. It was randomly selected nstances from a database of 7 outdoor mages and each nstance s a 3x3 regon. The mages were segmented to create a classfcaton for each pxel. The classes of the Image segmentaton data set are brckface, cement, folage, sky, path, wndow and grass. The general nformaton of other data sets, such as the number of nstances, the number of attrbutes and the number of classes are shown n TABLE II. Table 2. General Informaton of Publc Data Sets From UCI( Data Sets Instance Attrbute Class segment german_cred t hepatts onosphere soybean Before calculatng the classfcaton accuracy, some approaches are chosen to data cleanng as a process. It not only ensures the degree of unformty and accuracy of the data set, but also makes the data set more conducve to the mplementaton of the mnng process by changng the nternal structure and content of the data fle. The data preprocessng not only mproves the qualty of the data sample set but also mproves the qualty of the data mnng algorthm and reduces the runnng tme. For neural network backpropagaton algorthm, normalzaton helps speed up the learnng phase after normalzng the nput values for each attrbute. If usng a dstance-based method, normalzaton can help prevent attrbutes wth orgnally large ranges from overweghtng attrbutes wth orgnally smaller ranges. Consderng the classfers chosen, we use mn-max normalzaton to preprocess the data. In the attrbutes of Image segmentaton data set, the values of regon-centrod-col, regon-centrod-row, regon-pxel-count, short-lne- (6)

6 26 Remote Sensng Textual Image Classfcaton based on Ensemble Learnng densty-5, short-lne-densty-2, vedge-mean, vegde-sd, hedge-mean, hedge-sd, ntensty-mean, rawred-mean, rawblue-mean, rawgreen-mean, exred-mean, exbluemean, exgreen-mean, value-mean, saturaton-mean and hue-mean are [1,254], [11,251], [9,9], [0,0.333], [0,0.222], [0,29.222], [0, ], [0,44.722], [0, ], [0, ], [0, ], [0, ], [0, ], [ ,9.889],[ ,82],[ ,24.667], [0, ], [0,1], [-3.044,2.912], respectvely. It s clear that the 19 numerc attrbutes are not n the same range, so they need to be unfed to a certan extent. The experments use Normalze, whch s an unsupervsed flter n Weka. By mn-max normalzaton, suppose that mn A and max A are the mnmum and maxmum values of an attrbute, A. The value v s mapped by the value v of A n the range [0,1] by mnmax normalzaton, computng as Eq.(7) Fg.3. Vsualzaton of the german_credt data set usng a scatter-plot matrx wth part of attrbutes v mn A v' max mn A A (7) Fg.2, Fg, 3, Fg, 4, Fg.5 and Fg. 6 are the vsualzaton of the above fve data sets usng a scatterplot matrx. Fg.4. Vsualzaton of the hepatts data set usng a scatter-plot matrx wth part of attrbutes Fg.2. Vsualzaton of the Image Segmentaton data set wth part of attrbutes Fg.5. Vsualzaton of the onosphere data set usng a scatter-plot matrx wth part of attrbutes

7 Remote Sensng Textual Image Classfcaton based on Ensemble Learnng 27 Fg.6. Vsualzaton of the soybean data set usng a scatter-plot matrx wth part of attrbutes In the experment, C4.5 algorthm, k-nn algorthm, SVM algorthm, Nave Bayes algorthm, ANN algorthm and ensemble learnng algorthm are used to classfy the test data sets separately. For fve publc datasets, TABLE III shows average classfcaton accuracy of usng the J48 classfer, IBk classfer, SMO classfer, Nave Bayes classfer, MLP classfer and baggng classfer. Table 3. Comparson Of Classfcaton Accuracy For Fve Publc Data Sets Data Sets J48 IBk segment german_ credt SM O Nav e Baye s MLP Baggng performance of the base classfers are dfferent. There s no one knd of classfer has absolute advantage. Ths s also the purpose of ths experment. Based on the dfference and nformaton complementarty between the base classfers, t combnes dfferent classfers wth baggng algorthm and gves full play to the advantages of each base classfer. From the expermental data of TABLE III, the average classfcaton accuracy of ensemble learnng algorthm based on the above fve knds of classfers s hgher than the average classfcaton accuracy usng one of the base classfers separately. It can be seen that the results of the fve data sets usng baggng algorthms for classfcaton respectvely are 97.2, 76.1, 85.8, 92.0, B. Experments for Remote Sensng Image Data Sets In order to further llustrate the performance of our method on real remote sensng mages, we selected some real remote sensng mages as tranng data. There are 684 nstances whch are dvded nto four classes, ncludng resdent, paddy feld water and vegetaton area. Some tranng samples of each class are shown n Fg.7, Fg.8, Fg.9 and Fg.10. Fg.7. Resdent tranng samples Fg.8. Paddy feld tranng samples hepatts onosphere Fg.9. Water tranng sample. soybean It can be seen from TABLE III, for Image segment data set, the average classfcaton accuracy of IBk classfer s sgnfcantly hgher than the other four knds of base classfers, reachng 97.1; as for german_credt data set, the average classfcaton accuracy of Nave Bayes reaches 75.6; for hepatts data set and soybean data set, the average classfcaton accuracy of SMO classfer s hgher than the other four knds of base classfers, reachng 85.1 and 93.8 respectvely; n the experment of onosphere data set, the average classfcaton accuracy of J48 classfer reaches 91.4, however the average classfcaton accuracy of IBk algorthm, SMO algorthm, Nave Bayes algorthm and MLP algorthm are 86.3, 88.6, 82.6, 91.1, respectvely. It can be seen that, for dfferent data sets, the results of the classfcaton accuracy are dfferent, because the Fg.10. Vegetaton tranng samples The same as fve publc data sets, remote sensng mage data set uses mn-max normalzaton to deal wth the orgnal data of 22 attrbutes, ncludng varance, skewness, promnence, energy, absolute value and texture energy of each order. Fg.11 shows vsualzaton of the remote sensng mage data set usng a scatter-plot matrx. Then C4.5 algorthm, k-nn algorthm, SVM algorthm, Nave Bayes algorthm, ANN algorthm and ensemble learnng algorthm are used to classfy remote sensng mage data set separately. Table IV shows average classfcaton accuracy of usng the J48 classfer, IBk classfer, SMO classfer, Nave Bayes classfer, MLP classfer and baggng classfer for remote sensng mage data set.

8 28 Remote Sensng Textual Image Classfcaton based on Ensemble Learnng REFERENCES Fg.11. Vsualzaton of the remote sensng mage data set usng a scatter-plot matrx wth part of attrbutes Table 4. Comparson of Classfcaton Accuracy for Remote Sensng Image Data Set Data Sets J48 IBk SMO Remote Sensng Image Nave Bayes MLP Baggng From TABLE IV, the average classfcaton accuracy of usng the J48 classfer, IBk classfer, SMO classfer, Nave Bayes classfer, MLP classfer and baggng classfer for remote sensng mage data set are 81.2, 78.6, 86.5, 85.1, 86.9, successvely. As for the result of baggng algorthm, t rses to 89.1, whch s nearly 2 hgher than MLP, whch performs best as a sngle classfer n base classfers. It may be deduced that as for texture mages classfcaton, ensemble learnng s a promsng approach whch could acqure the satsfed results n practce. V. CONCLUSION In order to mprove the classfcaton accuracy of remote sensng mage, our method uses ensemble learnng to combne the classfers of J48, IBk, sequental mnmal optmzaton, Nave Bayes and multlayer perceptron, whch classfy the data sets by straght votng. At last, fve set of publc data and real remote sensng mages are selected to verfy the results. The expermental results show that multple classfer ensemble can effectvely mprove the classfcaton accuracy of textural remote sensng mages. However, n the paper, classfers are ntegrated wth the sample mode, n the future, some better way would be employed. ACKNOWLEDGMENT Ths work s funded by the Natonal Natural Scence Foundaton of Chna under Grant No and funded by State Key Laboratory of Geo-Informaton Engneerng, No. SKLGIE2014-M-3-3. [1] Ghasseman H. A revew of remote sensng mage fuson methods[j]. Informaton Fuson, 2016, 32(PA): [2] Tsa C F. Image mnng by spectral features: A case study of scenery mage classfcaton[j]. Expert Systems wth Applcatons, 2007, 32(1): [3] Goel S, Gaur M, Jan E. Nature Inspred Algorthms n Remote Sensng Image Classfcaton[J]. Proceda Computer Scence, 2015, 57: [4] Xu M, Zhang L, Du B. An Image-Based Endmember Bundle Extracton Algorthm Usng Both Spatal and Spectral Informaton[J]. IEEE Journal of Selected Topcs n Appled Earth Observatons & Remote Sensng, 2015, 8(6): [5] Rutherford V. Platt, Lauren Rapoza. An Evaluaton of an Object-Orented Paradgm for Land Use/Land Cover Classfcaton[J]. Professonal Geographer, 2008, 60(1): [6] Wolpert, D H. The supervsed learnng no-free-lunch theorem [C]. Proceedngs of the 6th Onlne World Conference on Soft Computng n Industral Applcatons, [7] Kttler J, Hatef M, Dun R P W, et al. On combnng classfers[j]. IEEE Transactons on Pattern Analyss & Machne Intellgence, 1998, 20(3): [8] Doan H T, Foody G M. Increasng soft classfcaton accuracy through the use of an ensemble of classfers [J]. Internatonal Journal of Remote Sensng, 2007, 28(20): [9] Hansen L K, Salamon P. Neural network ensembles[j]. Pattern Analyss & Machne Intellgence IEEE Transactons on, 1990, 12(10): [10] Le Z, Lao S, Petka&#x, et al. Face Recognton by Explorng Informaton Jontly n Space, Scale and Orentaton[J]. IEEE Transactons on Image Processng, 2011, 20(1): [11] Mountraks G, Im J, Ogole C. Support vector machnes n remote sensng: A revew[j]. Isprs Journal of Photogrammetry & Remote Sensng, 2011, 66(3): [12] Rokach L. Ensemble-based classfers[j]. Artfcal Intellgence Revew, 2010, 33(1-2):1-39. [13] Detterch T G. Ensemble Methods n Machne Learnng[C]// Internatonal Workshop on Multple Classfer Systems. Sprnger-Verlag, 2000:1-15. [14] Detterch T G. An expermental comparson of three methods for constructng ensembles of decson trees: baggng, boostng, and randomzaton. Machne Learnng, 2000,40(2): [15] Lttlewood B, Mller D R. Conceptual modelng of concdent falures n multverson software[j]. IEEE Transactons on Software Engneerng, 1989, 15(12): [16] Kuncheva L, Whtaker C J, Measures of dversty n classfer ensembles and ther relatonshp wth the ensemble accuracy [J]. Machne Learnng, 2003, 51(2): [17] Qunlan J R. Improved use of contnuous attrbutes n C4.5[J]. Journal of Artfcal Intellgence Research, 1996, 4(1): [18] Hunt E B, Marn J, Stone P J. Experments n nducton.[j]. Amercan Journal of Psychology, 1967, 80(4): [19] Luxburg U V. A tutoral on spectral clusterng[j]. Statstcs & Computng, 2007, 17(17): [20] Yang J F. A Novel Template Reducton K-Nearest Neghbor Classfcaton Method Based on Weghted

9 Remote Sensng Textual Image Classfcaton based on Ensemble Learnng 29 Dstance[J]. Danz Yu Xnx Xuebao/journal of Electroncs & Informaton Technology, 2011, 33(10): [21] Chen P H, Fan R E, Ln C J. A study on SMO-type decomposton methods for support vector machnes.[j]. IEEE Transactons on Neural Networks, 2006, 17(4): [22] Karatzoglou A, Smola A, Hornk K, et al. kernlab - An S4 Package for Kernel Methods n R[J]. Journal of Statstcal Software, 2004, 11(09): [23] Hameg S, Lazr M, Ameur S. Usng nave Bayes classfer for classfcaton of convectve ranfall ntenstes based on spectral characterstcs retreved from SEVIRI[J]. Journal of Earth System Scence, 2016:1-11. [24] Roy M, Routaray D, Ghosh S, et al. Ensemble of Multlayer Perceptrons for Change Detecton n Remotely Sensed Images[J]. IEEE Geoscence & Remote Sensng Letters, 2014, 11(11): [25] Wolpert D H, Macready W G. An Effcent Method To Estmate Baggng's Generalzaton Error[C]// Santa Fe Insttute, 1999: Authors Profles Ye Zhwe, Born n Hube Chna. He s an assocate professor n school of computer scence, Hube Unversty of technology, Wuhan Chna. Hs major research nterests nclude mage processng, swarm ntellgence and machne learnng. How to cte ths paper: Ye zhwe, Yang Juan, Zhang Xu, Hu Zhengbng,"Remote Sensng Textual Image Classfcaton based on Ensemble Learnng", Internatonal Journal of Image, Graphcs and Sgnal Processng(IJIGSP), Vol.8, No.12, pp.21-29, 2016.DOI: /jgsp

The Research of Support Vector Machine in Agricultural Data Classification

The Research of Support Vector Machine in Agricultural Data Classification The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining the Optimal Bandwidth Based on Multi-criterion Fusion Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

An Image Fusion Approach Based on Segmentation Region

An Image Fusion Approach Based on Segmentation Region Rong Wang, L-Qun Gao, Shu Yang, Yu-Hua Cha, and Yan-Chun Lu An Image Fuson Approach Based On Segmentaton Regon An Image Fuson Approach Based on Segmentaton Regon Rong Wang, L-Qun Gao, Shu Yang 3, Yu-Hua

More information

Classifying Acoustic Transient Signals Using Artificial Intelligence

Classifying Acoustic Transient Signals Using Artificial Intelligence Classfyng Acoustc Transent Sgnals Usng Artfcal Intellgence Steve Sutton, Unversty of North Carolna At Wlmngton (suttons@charter.net) Greg Huff, Unversty of North Carolna At Wlmngton (jgh7476@uncwl.edu)

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng

More information

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

Journal of Chemical and Pharmaceutical Research, 2014, 6(6): Research Article. A selective ensemble classification method on microarray data

Journal of Chemical and Pharmaceutical Research, 2014, 6(6): Research Article. A selective ensemble classification method on microarray data Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(6):2860-2866 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A selectve ensemble classfcaton method on mcroarray

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and

More information

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET 1 BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET TZU-CHENG CHUANG School of Electrcal and Computer Engneerng, Purdue Unversty, West Lafayette, Indana 47907 SAUL B. GELFAND School

More information

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach Angle Estmaton and Correcton of Hand Wrtten, Textual and Large areas of Non-Textual Document Images: A Novel Approach D.R.Ramesh Babu Pyush M Kumat Mahesh D Dhannawat PES Insttute of Technology Research

More information

Lecture 5: Multilayer Perceptrons

Lecture 5: Multilayer Perceptrons Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

A Lazy Ensemble Learning Method to Classification

A Lazy Ensemble Learning Method to Classification IJCSI Internatonal Journal of Computer Scence Issues, Vol. 7, Issue 5, September 2010 ISSN (Onlne): 1694-0814 344 A Lazy Ensemble Learnng Method to Classfcaton Haleh Homayoun 1, Sattar Hashem 2 and Al

More information

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1 4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:

More information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

Edge Detection in Noisy Images Using the Support Vector Machines

Edge Detection in Noisy Images Using the Support Vector Machines Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona

More information

Detection of an Object by using Principal Component Analysis

Detection of an Object by using Principal Component Analysis Detecton of an Object by usng Prncpal Component Analyss 1. G. Nagaven, 2. Dr. T. Sreenvasulu Reddy 1. M.Tech, Department of EEE, SVUCE, Trupath, Inda. 2. Assoc. Professor, Department of ECE, SVUCE, Trupath,

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

The Study of Remote Sensing Image Classification Based on Support Vector Machine

The Study of Remote Sensing Image Classification Based on Support Vector Machine Sensors & Transducers 03 by IFSA http://www.sensorsportal.com The Study of Remote Sensng Image Classfcaton Based on Support Vector Machne, ZHANG Jan-Hua Key Research Insttute of Yellow Rver Cvlzaton and

More information

Modular PCA Face Recognition Based on Weighted Average

Modular PCA Face Recognition Based on Weighted Average odern Appled Scence odular PCA Face Recognton Based on Weghted Average Chengmao Han (Correspondng author) Department of athematcs, Lny Normal Unversty Lny 76005, Chna E-mal: hanchengmao@163.com Abstract

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

More information

Yan et al. / J Zhejiang Univ-Sci C (Comput & Electron) in press 1. Improving Naive Bayes classifier by dividing its decision regions *

Yan et al. / J Zhejiang Univ-Sci C (Comput & Electron) in press 1. Improving Naive Bayes classifier by dividing its decision regions * Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press 1 Journal of Zhejang Unversty-SCIENCE C (Computers & Electroncs) ISSN 1869-1951 (Prnt); ISSN 1869-196X (Onlne) www.zju.edu.cn/jzus; www.sprngerlnk.com

More information

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1 A New Feature of Unformty of Image Texture Drectons Concdng wth the Human Eyes Percepton Xng-Jan He, De-Shuang Huang, Yue Zhang, Tat-Mng Lo 2, and Mchael R. Lyu 3 Intellgent Computng Lab, Insttute of Intellgent

More information

Available online at Available online at Advanced in Control Engineering and Information Science

Available online at   Available online at   Advanced in Control Engineering and Information Science Avalable onlne at wwwscencedrectcom Avalable onlne at wwwscencedrectcom Proceda Proceda Engneerng Engneerng 00 (2011) 15000 000 (2011) 1642 1646 Proceda Engneerng wwwelsevercom/locate/proceda Advanced

More information

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero

More information

BioTechnology. An Indian Journal FULL PAPER. Trade Science Inc.

BioTechnology. An Indian Journal FULL PAPER. Trade Science Inc. [Type text] [Type text] [Type text] ISSN : 0974-74 Volume 0 Issue BoTechnology 04 An Indan Journal FULL PAPER BTAIJ 0() 04 [684-689] Revew on Chna s sports ndustry fnancng market based on market -orented

More information

SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE

SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE Dorna Purcaru Faculty of Automaton, Computers and Electroncs Unersty of Craoa 13 Al. I. Cuza Street, Craoa RO-1100 ROMANIA E-mal: dpurcaru@electroncs.uc.ro

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton

More information

A fast algorithm for color image segmentation

A fast algorithm for color image segmentation Unersty of Wollongong Research Onlne Faculty of Informatcs - Papers (Arche) Faculty of Engneerng and Informaton Scences 006 A fast algorthm for color mage segmentaton L. Dong Unersty of Wollongong, lju@uow.edu.au

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned

More information

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION SHI-LIANG SUN, HONG-LEI SHI Department of Computer Scence and Technology, East Chna Normal Unversty 500 Dongchuan Road, Shangha 200241, P. R. Chna E-MAIL: slsun@cs.ecnu.edu.cn,

More information

Machine Learning 9. week

Machine Learning 9. week Machne Learnng 9. week Mappng Concept Radal Bass Functons (RBF) RBF Networks 1 Mappng It s probably the best scenaro for the classfcaton of two dataset s to separate them lnearly. As you see n the below

More information

Fuzzy Logic Based RS Image Classification Using Maximum Likelihood and Mahalanobis Distance Classifiers

Fuzzy Logic Based RS Image Classification Using Maximum Likelihood and Mahalanobis Distance Classifiers Research Artcle Internatonal Journal of Current Engneerng and Technology ISSN 77-46 3 INPRESSCO. All Rghts Reserved. Avalable at http://npressco.com/category/jcet Fuzzy Logc Based RS Image Usng Maxmum

More information

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images Internatonal Journal of Informaton and Electroncs Engneerng Vol. 5 No. 6 November 015 Usng Fuzzy Logc to Enhance the Large Sze Remote Sensng Images Trung Nguyen Tu Huy Ngo Hoang and Thoa Vu Van Abstract

More information

Announcements. Supervised Learning

Announcements. Supervised Learning Announcements See Chapter 5 of Duda, Hart, and Stork. Tutoral by Burge lnked to on web page. Supervsed Learnng Classfcaton wth labeled eamples. Images vectors n hgh-d space. Supervsed Learnng Labeled eamples

More information

An Improved Image Segmentation Algorithm Based on the Otsu Method

An Improved Image Segmentation Algorithm Based on the Otsu Method 3th ACIS Internatonal Conference on Software Engneerng, Artfcal Intellgence, Networkng arallel/dstrbuted Computng An Improved Image Segmentaton Algorthm Based on the Otsu Method Mengxng Huang, enjao Yu,

More information

CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION

CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION 48 CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION 3.1 INTRODUCTION The raw mcroarray data s bascally an mage wth dfferent colors ndcatng hybrdzaton (Xue

More information

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for

More information

Face Recognition Based on SVM and 2DPCA

Face Recognition Based on SVM and 2DPCA Vol. 4, o. 3, September, 2011 Face Recognton Based on SVM and 2DPCA Tha Hoang Le, Len Bu Faculty of Informaton Technology, HCMC Unversty of Scence Faculty of Informaton Scences and Engneerng, Unversty

More information

CS 534: Computer Vision Model Fitting

CS 534: Computer Vision Model Fitting CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust

More information

ALEXNET FEATURE EXTRACTION AND MULTI-KERNEL LEARNING FOR OBJECT- ORIENTED CLASSIFICATION

ALEXNET FEATURE EXTRACTION AND MULTI-KERNEL LEARNING FOR OBJECT- ORIENTED CLASSIFICATION ALEXNET FEATURE EXTRACTION AND MULTI-KERNEL LEARNING FOR OBJECT- ORIENTED CLASSIFICATION Lng Dng 1, Hongy L 2, *, Changmao Hu 2, We Zhang 2, Shumn Wang 1 1 Insttute of Earthquake Forecastng, Chna Earthquake

More information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,

More information

Discriminative Dictionary Learning with Pairwise Constraints

Discriminative Dictionary Learning with Pairwise Constraints Dscrmnatve Dctonary Learnng wth Parwse Constrants Humn Guo Zhuoln Jang LARRY S. DAVIS UNIVERSITY OF MARYLAND Nov. 6 th, Outlne Introducton/motvaton Dctonary Learnng Dscrmnatve Dctonary Learnng wth Parwse

More information

Face Recognition Method Based on Within-class Clustering SVM

Face Recognition Method Based on Within-class Clustering SVM Face Recognton Method Based on Wthn-class Clusterng SVM Yan Wu, Xao Yao and Yng Xa Department of Computer Scence and Engneerng Tong Unversty Shangha, Chna Abstract - A face recognton method based on Wthn-class

More information

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour 6.854 Advanced Algorthms Petar Maymounkov Problem Set 11 (November 23, 2005) Wth: Benjamn Rossman, Oren Wemann, and Pouya Kheradpour Problem 1. We reduce vertex cover to MAX-SAT wth weghts, such that the

More information

PERFORMANCE EVALUATION FOR SCENE MATCHING ALGORITHMS BY SVM

PERFORMANCE EVALUATION FOR SCENE MATCHING ALGORITHMS BY SVM PERFORMACE EVALUAIO FOR SCEE MACHIG ALGORIHMS BY SVM Zhaohu Yang a, b, *, Yngyng Chen a, Shaomng Zhang a a he Research Center of Remote Sensng and Geomatc, ongj Unversty, Shangha 200092, Chna - yzhac@63.com

More information

Smoothing Spline ANOVA for variable screening

Smoothing Spline ANOVA for variable screening Smoothng Splne ANOVA for varable screenng a useful tool for metamodels tranng and mult-objectve optmzaton L. Rcco, E. Rgon, A. Turco Outlne RSM Introducton Possble couplng Test case MOO MOO wth Game Theory

More information

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation Internatonal Conference on Logstcs Engneerng, Management and Computer Scence (LEMCS 5) Maxmum Varance Combned wth Adaptve Genetc Algorthm for Infrared Image Segmentaton Huxuan Fu College of Automaton Harbn

More information

IMAGE FUSION TECHNIQUES

IMAGE FUSION TECHNIQUES Int. J. Chem. Sc.: 14(S3), 2016, 812-816 ISSN 0972-768X www.sadgurupublcatons.com IMAGE FUSION TECHNIQUES A Short Note P. SUBRAMANIAN *, M. SOWNDARIYA, S. SWATHI and SAINTA MONICA ECE Department, Aarupada

More information

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana

More information

Audio Content Classification Method Research Based on Two-step Strategy

Audio Content Classification Method Research Based on Two-step Strategy (IJACSA) Internatonal Journal of Advanced Computer Scence and Applcatons, Audo Content Classfcaton Method Research Based on Two-step Strategy Sume Lang Department of Computer Scence and Technology Chongqng

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z.

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z. TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS Muradalyev AZ Azerbajan Scentfc-Research and Desgn-Prospectng Insttute of Energetc AZ1012, Ave HZardab-94 E-mal:aydn_murad@yahoocom Importance of

More information

Unsupervised Learning

Unsupervised Learning Pattern Recognton Lecture 8 Outlne Introducton Unsupervsed Learnng Parametrc VS Non-Parametrc Approach Mxture of Denstes Maxmum-Lkelhood Estmates Clusterng Prof. Danel Yeung School of Computer Scence and

More information

Network Intrusion Detection Based on PSO-SVM

Network Intrusion Detection Based on PSO-SVM TELKOMNIKA Indonesan Journal of Electrcal Engneerng Vol.1, No., February 014, pp. 150 ~ 1508 DOI: http://dx.do.org/10.11591/telkomnka.v1.386 150 Network Intruson Detecton Based on PSO-SVM Changsheng Xang*

More information

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures A Novel Adaptve Descrptor Algorthm for Ternary Pattern Textures Fahuan Hu 1,2, Guopng Lu 1 *, Zengwen Dong 1 1.School of Mechancal & Electrcal Engneerng, Nanchang Unversty, Nanchang, 330031, Chna; 2. School

More information

Robust visual tracking based on Informative random fern

Robust visual tracking based on Informative random fern 5th Internatonal Conference on Computer Scences and Automaton Engneerng (ICCSAE 205) Robust vsual trackng based on Informatve random fern Hao Dong, a, Ru Wang, b School of Instrumentaton Scence and Opto-electroncs

More information

Human Face Recognition Using Generalized. Kernel Fisher Discriminant

Human Face Recognition Using Generalized. Kernel Fisher Discriminant Human Face Recognton Usng Generalzed Kernel Fsher Dscrmnant ng-yu Sun,2 De-Shuang Huang Ln Guo. Insttute of Intellgent Machnes, Chnese Academy of Scences, P.O.ox 30, Hefe, Anhu, Chna. 2. Department of

More information

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines A Modfed Medan Flter for the Removal of Impulse Nose Based on the Support Vector Machnes H. GOMEZ-MORENO, S. MALDONADO-BASCON, F. LOPEZ-FERRERAS, M. UTRILLA- MANSO AND P. GIL-JIMENEZ Departamento de Teoría

More information

MULTISPECTRAL REMOTE SENSING IMAGE CLASSIFICATION WITH MULTIPLE FEATURES

MULTISPECTRAL REMOTE SENSING IMAGE CLASSIFICATION WITH MULTIPLE FEATURES MULISPECRAL REMOE SESIG IMAGE CLASSIFICAIO WIH MULIPLE FEAURES QIA YI, PIG GUO, Image Processng and Pattern Recognton Laboratory, Bejng ormal Unversty, Bejng 00875, Chna School of Computer Scence and echnology,

More information

IMAGE FUSION BASED ON EXTENSIONS OF INDEPENDENT COMPONENT ANALYSIS

IMAGE FUSION BASED ON EXTENSIONS OF INDEPENDENT COMPONENT ANALYSIS IMAGE FUSION BASED ON EXTENSIONS OF INDEPENDENT COMPONENT ANALYSIS M Chen a, *, Yngchun Fu b, Deren L c, Qanqng Qn c a College of Educaton Technology, Captal Normal Unversty, Bejng 00037,Chna - (merc@hotmal.com)

More information

Local Quaternary Patterns and Feature Local Quaternary Patterns

Local Quaternary Patterns and Feature Local Quaternary Patterns Local Quaternary Patterns and Feature Local Quaternary Patterns Jayu Gu and Chengjun Lu The Department of Computer Scence, New Jersey Insttute of Technology, Newark, NJ 0102, USA Abstract - Ths paper presents

More information

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices Steps for Computng the Dssmlarty, Entropy, Herfndahl-Hrschman and Accessblty (Gravty wth Competton) Indces I. Dssmlarty Index Measurement: The followng formula can be used to measure the evenness between

More information

Learning-Based Top-N Selection Query Evaluation over Relational Databases

Learning-Based Top-N Selection Query Evaluation over Relational Databases Learnng-Based Top-N Selecton Query Evaluaton over Relatonal Databases Lang Zhu *, Wey Meng ** * School of Mathematcs and Computer Scence, Hebe Unversty, Baodng, Hebe 071002, Chna, zhu@mal.hbu.edu.cn **

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System Fuzzy Modelng of the Complexty vs. Accuracy Trade-off n a Sequental Two-Stage Mult-Classfer System MARK LAST 1 Department of Informaton Systems Engneerng Ben-Guron Unversty of the Negev Beer-Sheva 84105

More information

THE PATH PLANNING ALGORITHM AND SIMULATION FOR MOBILE ROBOT

THE PATH PLANNING ALGORITHM AND SIMULATION FOR MOBILE ROBOT Journal of Theoretcal and Appled Informaton Technology 30 th Aprl 013. Vol. 50 No.3 005-013 JATIT & LLS. All rghts reserved. ISSN: 199-8645 www.jatt.org E-ISSN: 1817-3195 THE PATH PLANNING ALGORITHM AND

More information

Lecture 13: High-dimensional Images

Lecture 13: High-dimensional Images Lec : Hgh-dmensonal Images Grayscale Images Lecture : Hgh-dmensonal Images Math 90 Prof. Todd Wttman The Ctadel A grayscale mage s an nteger-valued D matrx. An 8-bt mage takes on values between 0 and 55.

More information

Performance Assessment and Fault Diagnosis for Hydraulic Pump Based on WPT and SOM

Performance Assessment and Fault Diagnosis for Hydraulic Pump Based on WPT and SOM Performance Assessment and Fault Dagnoss for Hydraulc Pump Based on WPT and SOM Be Jkun, Lu Chen and Wang Zl PERFORMANCE ASSESSMENT AND FAULT DIAGNOSIS FOR HYDRAULIC PUMP BASED ON WPT AND SOM. Be Jkun,

More information

Associative Based Classification Algorithm For Diabetes Disease Prediction

Associative Based Classification Algorithm For Diabetes Disease Prediction Internatonal Journal of Engneerng Trends and Technology (IJETT) Volume-41 Number-3 - November 016 Assocatve Based Classfcaton Algorthm For Dabetes Dsease Predcton 1 N. Gnana Deepka, Y.surekha, 3 G.Laltha

More information

Object-Based Techniques for Image Retrieval

Object-Based Techniques for Image Retrieval 54 Zhang, Gao, & Luo Chapter VII Object-Based Technques for Image Retreval Y. J. Zhang, Tsnghua Unversty, Chna Y. Y. Gao, Tsnghua Unversty, Chna Y. Luo, Tsnghua Unversty, Chna ABSTRACT To overcome the

More information

International Conference on Applied Science and Engineering Innovation (ASEI 2015)

International Conference on Applied Science and Engineering Innovation (ASEI 2015) Internatonal Conference on Appled Scence and Engneerng Innovaton (ASEI 205) Desgn and Implementaton of Novel Agrcultural Remote Sensng Image Classfcaton Framework through Deep Neural Network and Mult-

More information

HU Sheng-neng* Resources and Electric Power,Zhengzhou ,China

HU Sheng-neng* Resources and Electric Power,Zhengzhou ,China do:10.21311/002.31.6.09 Applcaton of new neural network technology n traffc volume predcton Abstract HU Sheng-neng* 1 School of Cvl Engneerng &Communcaton, North Chna Unversty of Water Resources and Electrc

More information

Recognition of Handwritten Numerals Using a Combined Classifier with Hybrid Features

Recognition of Handwritten Numerals Using a Combined Classifier with Hybrid Features Recognton of Handwrtten Numerals Usng a Combned Classfer wth Hybrd Features Kyoung Mn Km 1,4, Joong Jo Park 2, Young G Song 3, In Cheol Km 1, and Chng Y. Suen 1 1 Centre for Pattern Recognton and Machne

More information

Comparison Study of Textural Descriptors for Training Neural Network Classifiers

Comparison Study of Textural Descriptors for Training Neural Network Classifiers Comparson Study of Textural Descrptors for Tranng Neural Network Classfers G.D. MAGOULAS (1) S.A. KARKANIS (1) D.A. KARRAS () and M.N. VRAHATIS (3) (1) Department of Informatcs Unversty of Athens GR-157.84

More information

Face Recognition University at Buffalo CSE666 Lecture Slides Resources:

Face Recognition University at Buffalo CSE666 Lecture Slides Resources: Face Recognton Unversty at Buffalo CSE666 Lecture Sldes Resources: http://www.face-rec.org/algorthms/ Overvew of face recognton algorthms Correlaton - Pxel based correspondence between two face mages Structural

More information

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation Intellgent Informaton Management, 013, 5, 191-195 Publshed Onlne November 013 (http://www.scrp.org/journal/m) http://dx.do.org/10.36/m.013.5601 Qualty Improvement Algorthm for Tetrahedral Mesh Based on

More information

Biostatistics 615/815

Biostatistics 615/815 The E-M Algorthm Bostatstcs 615/815 Lecture 17 Last Lecture: The Smplex Method General method for optmzaton Makes few assumptons about functon Crawls towards mnmum Some recommendatons Multple startng ponts

More information

Title: A Novel Protocol for Accuracy Assessment in Classification of Very High Resolution Images

Title: A Novel Protocol for Accuracy Assessment in Classification of Very High Resolution Images 2009 IEEE. Personal use of ths materal s permtted. Permsson from IEEE must be obtaned for all other uses, n any current or future meda, ncludng reprntng/republshng ths materal for advertsng or promotonal

More information

Range images. Range image registration. Examples of sampling patterns. Range images and range surfaces

Range images. Range image registration. Examples of sampling patterns. Range images and range surfaces Range mages For many structured lght scanners, the range data forms a hghly regular pattern known as a range mage. he samplng pattern s determned by the specfc scanner. Range mage regstraton 1 Examples

More information

Machine Learning. Topic 6: Clustering

Machine Learning. Topic 6: Clustering Machne Learnng Topc 6: lusterng lusterng Groupng data nto (hopefully useful) sets. Thngs on the left Thngs on the rght Applcatons of lusterng Hypothess Generaton lusters mght suggest natural groups. Hypothess

More information

Histogram of Template for Pedestrian Detection

Histogram of Template for Pedestrian Detection PAPER IEICE TRANS. FUNDAMENTALS/COMMUN./ELECTRON./INF. & SYST., VOL. E85-A/B/C/D, No. xx JANUARY 20xx Hstogram of Template for Pedestran Detecton Shaopeng Tang, Non Member, Satosh Goto Fellow Summary In

More information

Detection of hand grasping an object from complex background based on machine learning co-occurrence of local image feature

Detection of hand grasping an object from complex background based on machine learning co-occurrence of local image feature Detecton of hand graspng an object from complex background based on machne learnng co-occurrence of local mage feature Shnya Moroka, Yasuhro Hramoto, Nobutaka Shmada, Tadash Matsuo, Yoshak Shra Rtsumekan

More information

Applying EM Algorithm for Segmentation of Textured Images

Applying EM Algorithm for Segmentation of Textured Images Proceedngs of the World Congress on Engneerng 2007 Vol I Applyng EM Algorthm for Segmentaton of Textured Images Dr. K Revathy, Dept. of Computer Scence, Unversty of Kerala, Inda Roshn V. S., ER&DCI Insttute

More information

Implementation Naïve Bayes Algorithm for Student Classification Based on Graduation Status

Implementation Naïve Bayes Algorithm for Student Classification Based on Graduation Status Internatonal Journal of Appled Busness and Informaton Systems ISSN: 2597-8993 Vol 1, No 2, September 2017, pp. 6-12 6 Implementaton Naïve Bayes Algorthm for Student Classfcaton Based on Graduaton Status

More information

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan

More information

Review of approximation techniques

Review of approximation techniques CHAPTER 2 Revew of appromaton technques 2. Introducton Optmzaton problems n engneerng desgn are characterzed by the followng assocated features: the objectve functon and constrants are mplct functons evaluated

More information

Relevance Assignment and Fusion of Multiple Learning Methods Applied to Remote Sensing Image Analysis

Relevance Assignment and Fusion of Multiple Learning Methods Applied to Remote Sensing Image Analysis Assgnment and Fuson of Multple Learnng Methods Appled to Remote Sensng Image Analyss Peter Bajcsy, We-Wen Feng and Praveen Kumar Natonal Center for Supercomputng Applcaton (NCSA), Unversty of Illnos at

More information

TN348: Openlab Module - Colocalization

TN348: Openlab Module - Colocalization TN348: Openlab Module - Colocalzaton Topc The Colocalzaton module provdes the faclty to vsualze and quantfy colocalzaton between pars of mages. The Colocalzaton wndow contans a prevew of the two mages

More information

CSCI 5417 Information Retrieval Systems Jim Martin!

CSCI 5417 Information Retrieval Systems Jim Martin! CSCI 5417 Informaton Retreval Systems Jm Martn! Lecture 11 9/29/2011 Today 9/29 Classfcaton Naïve Bayes classfcaton Ungram LM 1 Where we are... Bascs of ad hoc retreval Indexng Term weghtng/scorng Cosne

More information