- Universiti Teknologi Malaysia
Transcription
- Universiti Teknologi Malaysia
PENGECAMAN IMEJ KAPAL TERBANG MENGGUNAKAN TEKNIK LMI" TMI DAN KMI BAGI KAEDAH SOM NURULLIZ AH BINTI MAHIDIN Laporan Projek ini dikemukakan Sebagaimemenuhisebahagiandaripadasyarat penganugerahan ljazah SarjanaMuda Sains(SainsKomputer) F'akultiSainsKomputer dan SistemMaklumat Universiti TeknologiMalaysia MEr,2008 I ABSTRAK Mata merupakansuatuanugerahyangtidak ternilai di manamelaluinyakita dapatmelihat apayangberlakudi sekelilingkita. Mata bertindaksebagaikamera iaitu denganmenangkapsuatuimej untuk dihantardandiprosesoleh otak. Maka giat menciptasatumesinbaruyangberkait tidak hairanlahsekiranya,parapenyelidik rapatdengansistempenglihatanmanusia.Bagi mesinbaruini terdapatdua(2) fasa yangpentingiaitu fasapengekstrakan imej danfasapengelasan.Bagi melakukan fasapengekstrakan imej teknik momentak varianiaitu LegendreMomentsInvariant (LMD, Tchebichef MomentsInvariant(TMI), danKrawtchoukMomentInvariant (KMD digunakanmanakalakaedahrangkaianneuraliaitu Self-OrganisingMap (SOM)digunakanuntuk membuatpengelasan imej danpengecaman terhadapobjek domainyangdipilih iaitu kapalterbang.Hasil yangdiperolehidaripadakajian menunjukkanbahawateknik KMI merupakanteknik yangterbaikdalammembuat pengekstrakan imej ke atasimej kapalterbang.Peratusralat mutlakyangdiperolehi daripadakajian ini ke ataskapalterbangberjenisAirbus A320-214ialahkurang daripada0.0794iaitu mempunyaiketepatansebanyak99.92peratus.Teknik ini juga sesuaidigunakanuntuk membuatanalisismengenaiciri-ciri intrakelasdaninterkelas. Selainitu, bagipengelasan imej didapatikaedahSOM dank-NearestNeighbour(kNN) sesuaidigunakanberdasarkan nilai Percentageof CorrectClassffication(PCC) yangtinggi bagi ketiga-tigasampeliaitu lebih daripada70 peratus. v1 , ABSRACT Eyesarespecialgifts that aregivenby our Creator.Throughthemwe can comprehendwhat happenedin the environment.Eyesact asa camerathat takesthe particularof imageobjectsand sendit to the brainto be processed.Thus,researchers areenthusiasticto invent a new machinethat mimickedhumanvision system.There aretwo (2) significantphasesin the associated systems;featureextractionand classificationphase.In this research,three(3) momentinvarianttechniquesare adoptedto performthe featureextractionof the aircraftimages,namelyLegendre MomentInvariant(LMI), TchebichefMomentInvariant(TMI), andKrawtchouk MomentInvariant(KMD. Self-Organising Map (SOM)is usedto classi$the resultingfeaturevectors. The resultfrom the resealchshowsthat KMI techniqueis the bestfeatureextractiontechniquewhencomparedto the LMI andTMI. The percentage of absoluteerrorfor the featureextractionof an aircraftmodelAirbus 4320-214is leastthan 0.0794which is its accuracyis 99.92percent.Thusthe KMI featuresvectorsareutilizedfor intraclass andinterclass'analysis. Besides,SOMand k-NearestNeighbour(k-Itil.l)techniquearethe bestclassificationtechnique (PCC). ThePCCfor the accordingto percentage of correctclassification classification of samplel,2,and 3 aremorethan7} percent. 60 RUJUKAN Biehl, M., Ghosh,A. and Hammer, B (2007). Dynamics and GeneralizationAbility of LVQ Algorithms.Journal of MachineLearningResearch8. pages:323-360. Chee,W. C., Raveendran,R., and Mukundan,R. (2004).Translationand Scale Invariants of LegendreMoments. TheJournal of the Pattern RecognitionSociety. PatternRecognition37, ll9-129. chowdhury, N., and Saha,D. (2005).unsupervisedText classificationUsing Kohonen'sSelf-OrganizingNetwork. 7 15-718. DassaultFalcon 10 (2007).Retrievedon September6,2007 from http ://www.airliners.neVopen.fi lel 0576193lL. Deghan,M. and Faez,K. (1977).Farsi Handwrittencharacter Recognitionwith MomentInvariants,IEEE, Pages:507-510,Retrievedon September5,2007 from http://ieeexplore.ieee.or gl ieI3I 496l I 13653I 00628387.pdf. Fu, B., Zhou,J.,Li,Y., Zhang,G., and W*g, C. (2007).ImageAnalysisby Modified LegendreMoments. TheJournal of the Pattern RecognitionSociety.Pattern Recognition40. 691-704. Ghorbel,F., Derrode,S., Dhahbi, S., and Mezhoud,R. Reconstructingwith Geometric Moments.Retrievedon August 15,2007 from http ://www. fresnel.frlperso/derrode/publi/Acida2005.pdf. 6l Haddadnia,L,Faez, K. andMoallem,p. (2001).NeuralNetworkBasedFace Recognitionwith Moment Invariants.IEEE, pages:l0lg-1021. Retrievedon September 5,2007 fromhttp:llieeexplore.ieee.org/ie15175g4l20726100g5g21i.pdf. Honkela,T., Leinoeni,T., Lonka,K., andRaike,A. setf-organising Ma/s and constructive Learning. 339-343.Retrievedon August 14,2007 from http://www.ifip.or.atlcon200O/iceut 10-04.pdf. Hopfield Net (2007,9 August).Retrievedon August 12,2007 from http://en.wikipedia. org/. fr-MeansAlgorithm (2008,March 17).Retrievedon March 20" 200g from http://en.wikipedia.org/. ft-NearestNeighbourAlgorithm (2008,March 6). Retrievedon March 20, 2008 from http :II en.wikipedia.org/. Kamgar-Parsi,B,, and Jain,A. K. AutomaticAircraft RecognitionToward Using Human Similarity Measurein a Recognition System. Computer Vision pattern Recognition,1999.IEEE ComputerSocietyConferenceon. June,23-251999.Fort Collns,CO: IEEE. 1999.268-273. Kulkarni, A.D., Yap, A. C. and Byars,P. (1990).Neural NetworlrsforInvariant Object Recognition.2S-32.Retrievedon August 14,2007 from http://ieeexplore.ieee.or gl iel 512781269 5I 00082135.pdf. Lee, H-j. and Cho, S. (2006).Application of LVQ to NoveltyDetectionusingoutlier training data. PattemRecognitionLetters27.pages: 1572-1579. LVQ (2007, 11November).Retrievedon March 16,2008 from http://en.wikipedia.orgl. 62 Mezghani, N., Mitiche,A., andCheriet,M. (2002).On-lineRecognition of Handwritten Arabic Characters Usinga KohonenNeuralNetwork.Proceedings of theEight InternationalWorkshopon Frontiersin HandwritingRecognition(IWFHR'72), Retrievedon August 12,2007from : http://ieeexplore,ieee.org/ie15/8011/2213g/01030958,pdf . t Mukundan,R., ong, s. H., andLee,p.A. (200r).ImageAnalysisby Tchebichef Moments.IEEE Transactions on ImageProcessing.l0(9).I357-l364.Retrieved on August12,2007fromhttp:llieeexplore. ieee.orgliel5/83/20391/0094IgS9.pdf. Mukundan,R. andRamakrishnan K. R. (1998).MomentFunctionin ImageAnalysis. Singapura: World ScientificpublishingCo.pte.Ltd. Negnevitsky, M. (2005).ArtificialIntelligentA Guideto IntelligentSystem.(2nded.) EdinburgGate:Pearson Education Limited. NeuralGas(2007,23December). Retrievedon March16,200gfrom http:II en.wikipedia. org/. NeuralNetworkDocumentation (2007).Retrievedon August 14,2007from http://www.wolfram.com/. Perelomov,r.,Azcarraga, A. p. , Tan,J. andchua, T. s. (2002).usingstructuredself_ organizingMapsin NewsIntegrationwebsites.citeseer.IST. Putehsaad,ShahrulNizamyaakob,andAbu HassanAbdullah.(2005). Features Extractionof BinaryImagesUsingMomentInvariants.