Dans le domaine de l'analyse de documents on voudrait être capable de traiter automatiquement n'importe quel genre de documents numériques et d'extraire l' information la plus importante. Plus précisément, on voudrait connaître la configuration du document,identifier chacune de ses parties et reconnaître ses contenus, pour faire des requêtes par le contenu du document lui-même mais aussi, parmi des documents différents. Ceci est une problème difficile qui a suscité un nombre important de travaux à différents niveaux. On a développé un ensemble de techniques destinés à pré-traiter les images numériques afin d'augmenter leurs qualités, en réduisant le bruit provenant des systèmes d'acquisition et en minimisant les effets de la dégradation des documents. On trouve aussi, beaucoup de travaux destinés à la segmentation de zones d'intérêts du fond du document. Finalement, depuis les années 60 à aujourd'hui un nombre important des descripteurs on été proposé pour représenter ces zones d'intérêts.<br /><br />Dans ce thèse, nous avons travaillé sur la description des formes et la fusion de classificateurs pour les appliquer à la reconnaissance de graphiques. Dans la reconnaissance de formes, beaucoup d'applications sont confrontées au problème de description de grands ensembles de formes complexes pour les reconnaître, mais aussi pour les identifier dans des grandes bases de données. En plus du nombre important de formes on doit également faire face aux problèmes de similitude des formes ou de variabilité des classes des symboles. Dans ces cas, un point clé dans le processus de la reconnaissance des formes est la définition de descripteurs ayant une grande capacité de discrimination. Malheureusement, un seul descripteur ne suffit pas pour obtenir des résultats satisfaisants et donc, nous devons combiner l'information provenant de différentes sources pour améliorer le comportement global du système de reconnaissance. Cette combinaison est réalisée par un mécanisme de fusion des classificateurs.<br /><br />Par rapport aux descriptions des formes, traditionnellement les symboles graphiques ont été représentés par des descripteurs structurelles, construits à partir d'une représentation vectorielle. Les méthodes de vectorisation sont sensibles aux bruits et aux distorsions des symboles ébauchés. On peut essayer de contourner ce problème en définissant une grammaire de descripteurs ou en construisant des modèles déformables des symboles. Une autre possibilité, celle poursuivie dans ce mémoire, est d'utiliser des descripteurs que n'ont pas besoin d'une représentation vectorielle. Dans le contexte de la description des formes on a proposé un descripteur basé sur la transformation de ridgelets qu'on peut définir comme: multiresolution, polaire, en 2D et qui préserve l'information d'invariance aux similitudes. D'un autre coté, malgré qu'on puisse considérer ce descripteur comme un seul, il nous offre une représentation des formes permettant de la décomposer en groupes de coefficients de ridgelets qui sont chacun définis comme un descripteur. De cette manière, pour chaque descripteur, nous avons entraîné des classifieurs qui sont combinés linéairement en utilisant des règles de combinaison: IN (Indépendant et Normale) et DN (Dépendant et Normal), que minimisent l'erreur de classification pour ces classifieurs par rapport à un ensemble de contraintes. <br /><br />Ces développements théoriques ont été validés à partir d'un ensemble de résultats expérimentaux. Les descripteurs ridgelets décrivent mieux les symboles que d'autres descripteurs plus classiques. Les règles de fusion IN et DN réduisent l'erreur de classification par rapport aux autres méthodes de références. Enfin, la méthode IN appliquée aux descripteurs de ridgelets, en combinaison avec des classificateurs du genre "boosting", aboutie à un taux de reconnaissance d'environ 100% sur la base de données définies au workshop GREC'03.

Turbulent exchange of scalars is analyzed for a year of turbulence data, above and within an Araucaria Forest in southern Brazil. Sensible and latent heat fluxes, as well as CO2 fluxes are calculated by the eddy covariance method and analyzed in terms of the seasonal variation of their daily cycles. Diurnal and nocturnal vertical velocity spectra, sensible and latent heat flux cospectra andCO2 flux cospectra are also calculated using the multiresolution decomposition in both levels. Seasonal variations in spectra and cospectra are analyzed, as well as nocturnal and diurnal differences and the differences between the levels. Emphasis is given to the mechanisms controlling the connection between the two levels and its implications on scalar flux quantification. All scalars show a clear seasonality along the year. CO2 fluxes are negative during the day and positive at night above the canopy. In the lower level, CO2 fluxes are mostly positive along the day, with the same magnitude observed in the upper level during the night, indicating that nocturnal emission is dominated by lower levels of the forest, originated predominantly close to the floor. This fact is attributed to periods with low turbulence intensity when CO2 respired accumulates near the surface and gets transported horizontally, being captured by the lower system but missed by the upper one. Sensible heat fluxes show opposite signal in both levels, being more intense above the canopy, with a small contribution from the lower level to the total flux due the low penetration of radiation within the canopy. Latent heat fluxes are mostly positive in both levels, with the larger values above the canopy, indicating that the canopy is responsible for most of the forest evapotranspiration. Analysis of multiresolution spectra and cospectra shows that, above the canopy, the nocturnal time scale of turbulent exchange is smaller than during the day. Between the two levels, the opposite occurs, and the time scale within the canopy shifts towards the larger scales. There is a connection between the two levels, and not only in the highly turbulent nights, but also intermittently in more stable conditions. During these events of connection, the turbulence time scales above and within the canopy become very similar, and the differences in scalar concentration vanish. When calm periods precede highly turbulent ones, a good portion of scalars that accumulated at the canopy during the earlier calm period gets transferred by intermittent events. We show that this process is systematic enough to affect the overall fluxes when the turbulence history is taken into account, and the effect of this process is more intense for CO2 fluxes than for energy fluxes. / As trocas turbulentas de escalares são analisadas, para um ano de dados, acima e no interior de uma Floresta de Araucária localizada no centro-sul do estado do Paraná. Fluxos de calor sensível, calor latente e CO2 são calculados pelo método da covariância de vórtices e analisados de acordo com as variações sazonais dos seus ciclos diários. São calculados também espectros da velocidade vertical e coespectros dos fluxos de calor sensível, calor latente e CO2, para os períodos diurno e noturno, separadamente, através da decomposição em multirresolução, analisando as suas variações sazonais e as diferenças entre o período diurno e noturno e entre os níveis de análise. É dada ênfase aos mecanismos que controlam a conexão entre os níveis e as implicações nas quantificações dos fluxos de escalares. Todos os escalares mostram uma clara variação sazonal dos fluxos ao longo do ano. Fluxos de CO2 são negativos durante o dia e positivos durante a noite no nível superior. No interior do dossel, os fluxos são predominantemente positivos durante todo o dia, tendo, à noite, magnitude semelhante ao nível superior, indicando que a respiração noturna é dominada pelos níveis mais baixos da floresta, sendo predominantemente originados próximo ao solo. Este fato é normalmente atribuído aos períodos de pouca turbulência onde o CO2 respirado acumula próximo à superfície e é transportado horizontalmente, sendo registrado no nível inferior mas não acima do dossel. Fluxos de calor sensível possuem, em geral, sinal contrário em ambos os níveis, sendo mais intenso no nível superior, com uma contribuição muito pequena do interior do dossel para o fluxo total devido à pouca penetração de radiação no interior da floresta. Fluxos de calor latente foram predominantemente positivos em ambos os níveis, tendo maior intensidade no nível superior, mostrando que a copa é responsável pela maior parte da evapotranspiração da floresta. A análise dos espectros e coespectros da decomposição em multirresolução mostra que, acima do dossel, as trocas turbulentas tendem a ocorrer em escalas temporais menores à noite do que durante o dia. Entre os dois níveis ocorre o contrário, e, no interior do dossel, as escalas temporais das trocas turbulentas movem-se em direção às maiores escalas temporais em relação ao nível superior. Há um acoplamento dinâmico entre os dois níveis da floresta, e não somente durante as noites altamente turbulentas, ocorrendo também intermitentemente em condições mais estáveis. Durante os eventos de acoplamento, as escalas de tempo da turbulência entre o interior da floresta e o topo mostram-se similares, assim como as diferenças na concentração de escalares entre os dois níveis deixam de existir. Quando períodos calmos precedem períodos intensamente turbulentos, uma grande porção dos escalares acumulados no nível do dossel durante os períodos de pouca atividade turbulenta pode ser transferida por estes eventos intermitentes. Foi mostrado que este processo é sistemático o suficiente para afetar os fluxos médios quando a história da turbulência é considerada, e o efeito deste processo é mais intenso para os fluxos de CO2 do que para os fluxos de energia.

[pt] Este trabalho apresenta um estudo comparativo de algumas representações em multi-resolução para linhas poligonais. São estudados as estruturas strip tree, arc tree e box tree, e suas variantes, comparando as velocidades de construção, de percorrimento (drawing), de operações de interseção e seleção (pick); e custo de armazenagem em memória. A comparação é feita usando bancos de dados cartográficos reais disponíveis na Internet. O objetivo do estudo é identificar técnicas e algoritmos adequados para exploração interativa de grandes bancos de dados cartográficos. / [en] This work presents a comparative study of some multiresolution representations for polygonal lines. We study the strip tree, the arc tree and the box tree data structures, comparing their performance for construction, drawing, intersection and selection; and memory storage costs. The comparation uses actual databases available in the Internet. The goal of this study is to identify techniques and algorithms for interactive exploration of large cartographic data bases.

Submitted by Maike Costa (maiksebas@gmail.com) on 2017-05-24T12:51:24Z No. of bitstreams: 1 arquivototal.pdf: 2705219 bytes, checksum: cb06011410866630871a99b2788703f1 (MD5) / Made available in DSpace on 2017-05-24T12:51:24Z (GMT). No. of bitstreams: 1 arquivototal.pdf: 2705219 bytes, checksum: cb06011410866630871a99b2788703f1 (MD5) Previous issue date: 2015-10-26 / The three-phase induction motors are the most important way for electromechanical conversion, being present in almost every industrial process. Because of their importance, it is crucial that these devices have a proper predictive maintenance, avoiding lost production and operational accidents in the event of unexpected failures. In this scenario several researchers have conducted studies to detect incipient faults. About the sensing method used, the most common techniques are related to measuring the vibration levels using accelerometers and measuring electrical motor signals. The choice of technique involves factors such as invasiveness, drive motor type and the sensitivity to failure. The purpose of this work involves the development of a hybrid system that uses data collected by vibration and current sensors for fault detection in induction motors, earlier and more efficiently. The current and vibration signals were processed in the frequency domain and through the multiresolution analysis, serving as inputs of a fuzzy logic system, allowing to increase the sensitivity and efficiency in fault detection techniques in relation to the individual. The unbalance failure was investigated on a workbench with the motor coupled to a propeller and broken bars with another bench with DC motor to apply the load, the best methods being chosen in each case. The system was developed in Matlab software and was validated with correct detection for both failures, being able to detect the unbalance failure on the shaft or propeller as broken bars in different load conditions. / O motor de indução trifásico é o principal meio de conversão eletromecânica existente, estando presente em praticamente todos os processos industriais. Devido à sua importância, é fundamental que estes equipamentos tenham uma correta manutenção preditiva, evitando perda de produção e acidentes operacionais em caso de falhas inesperadas. Diante deste cenário vários pesquisadores têm realizado estudos para detecção de falhas incipientes. Quanto ao método sensor utilizado, as técnicas mais comuns estão relacionadas a medição dos níveis de vibração utilizando acelerômetros e medição de sinais elétricos do motor. A escolha da técnica envolve fatores como a invasividade, tipo de acionamento do motor e a sensibilidade à falha. A proposta deste trabalho envolve o desenvolvimento de um sistema híbrido que utilize dados coletados por sensores de vibração e de corrente para detecção de falhas incipientes em motores de indução trifásicos de maneira mais precoce e eficiente. Os sinais de corrente e de vibração foram processados no domínio da frequência pela transformada de Fourier e através da análise multiresolução, servindo como entrada para sistemas de lógica Fuzzy, permitindo que se aumente a eficiência na detecção da falha em relação às técnicas individuais. Foi investigada a falha de desbalanceamento em uma bancada com o motor acoplado a uma hélice e barras quebradas em outra bancada com motor de corrente contínua acoplado para aplicar a carga, sendo escolhidos os melhores métodos em cada caso. O sistema foi desenvolvido no software Matlab e foi validado através de diagnósticos corretos para ambas as falhas, sendo capaz de detectar a falha de desbalanceamento tanto na hélice quanto no eixo e de barras quebradas em diferentes condições de carga.

<p style="text-align: justify;">In this doctoral dissertation several integral transforms are discussed.The first one is the Short time Fourier transform (STFT). We present continuity theorems for the STFT and its adjoint on the test function space <em>K</em>₁(ℝⁿ) and the topological tensor product <em>K</em>₁(ℝⁿ) &otimes; <em>U</em>(<strong>ℂ</strong>ⁿ), where <em>U</em>(<strong>ℂ</strong>ⁿ) is the space of entirerapidly decreasing functions in any horizontal band of&nbsp;<strong>ℂ</strong>ⁿ. We then use such continuity results to develop a framework for the STFT on K&#39;₁(ℝⁿ). Also, we devote one section to the characterization of <em>K</em>&rsquo;₁(ℝⁿ) and related spaces via modulation spaces. We also obtain various Tauberian theorems for the short-time Fourier transform.</p><p style="text-align: justify;">Part of the thesis is dedicated to the ridgelet and the Radon transform. We define and study the ridgelet transform of (Lizorkin) distributions and we show that the ridgelet transform and the ridgelet synthesis operator can be extended as continuous mappings <em>R</em>_{<em>&psi;&nbsp;</em>}: <em>S</em>&rsquo;₀(ℝⁿ) &rarr; <em>S</em>&rsquo;(<strong>Y</strong>ⁿ⁺¹) and <em>R^t</em>_{<span style="vertical-align: sub;">&psi;</span>}: <em>S</em>&rsquo;(<strong>Y</strong>ⁿ⁺¹) &rarr; <em>S</em>&rsquo;₀(ℝⁿ). We then use our results to develop a distributional framework for the ridgelet transform that is, we treat the ridgelet transform on <em>S</em>&rsquo;₀(ℝⁿ) via a duality approach. Then, the continuity theorems for the ridgelet transform are applied to discuss the continuity of the Radon transform on these spaces and their duals. Finally, we deal with some Abelian and Tauberian theorems relating the quasiasymptotic behavior of distributions with the quasiasymptotics of the its Radon and ridgelet transform.</p><p style="text-align: justify;">The last chapter is dedicated to the MRA of M-exponential distributions. We study the convergence of multiresolution expansions in various test function and distribution spaces and we discuss the pointwise convergence of multiresolution expansions to the distributional point values of a distribution. We also provide a characterization of the quasiasymptotic behavior in terms of multiresolution expansions and give an MRA sufficient condition for the existence of &alpha;-density points of positive measures.</p> / <p>U ovoj doktorskoj disertaciji razmotreno je nekoliko integralnih transformacija. Prva je short time Fourier transform (STFT). Date su i dokazane teoreme o neprekidnosti STFT i njena sinteza na prostoru test funkcije <em>K</em>₁(ℝⁿ) i na prostoru <em>K</em>₁(ℝⁿ) &otimes; <em>U</em>(ℂⁿ), gde je&nbsp;<em>U</em>(ℂⁿ) prostor od celih brzo opadajućih funkcija u proizvoljnom horizontalnom opsegu na ℂⁿ. Onda, ovi rezultati neprekidnosti su iskori&scaron;teni za razvijanje teorije STFT na prostoru <em>K</em>&rsquo;₁(ℝⁿ). Jedno poglavlje je posvećeno karakterizaciji&nbsp;<em>K</em>&rsquo;₁(ℝⁿ) sa srodnih modulaciskih prostora. Dokazani su i različiti Tauberovi rezultata za STFT. Deo teze je posvećen na ridglet i Radon transformacije. Ridgelet transformacija je definisana na (Lizorkin) distribucije i pokazano je da ridgelet transformacija i njen operator sinteze mogu da se pro&scaron;ire kako neprekidna preslikava <em>R</em>_&psi; : <em>S</em>&rsquo;₀(ℝⁿ) &rarr; <em>S</em>&rsquo;(<strong>Y</strong>ⁿ⁺¹) and <em>R</em>^t_&Psi;: <em>S</em>&rsquo;(<strong>Y</strong>ⁿ⁺¹) &rarr; <em>S</em>&rsquo;₀(ℝⁿ).&nbsp;Ridgelet transformacija na <em>S</em>&rsquo;₀(ℝⁿ) je data preko dualnog pristupa. Na&scaron;e teoreme neprekidnosti ridgelet transformacije su primenjene u dokazivanju neprekidnosti Radonove transformacije na Lizorkin test prostorima i njihovim dualima. Na kraju, dajemo Abelovih i Tauberovih teorema koji daju veze izmedju kvaziasimptotike distribucija i kvaziasimptotike rigdelet i Radonovog transfomaciju.<br />Zadnje poglavje je posveceno multirezolucijskog analizu M - eksponencijalnih distrubucije. Proucavamo konvergenciju multirezolucijkog razvoja u razlicitih prostori test funkcije i distribucije i razmotrena je tackasta konvergencija multirezolucijkog razvoju u tacku u distributivnog smislu. Obezbedjena je i karakterizacija kvaziasimptotike u pogled multirezolucijskog razvoju i dat dovoljni uslov za postojanje &alpha;-tacka gustine za pozitivne mere.</p>

No description available.

In dieser Arbeit werden Algorithmen für die Berechnung elektronischer Korrelations- und Anregungsenergien mittels der Coupled-Cluster Methode auf adaptiven Gittern entwickelt und implementiert. Die jeweiligen Funktionen und Operatoren werden adaptiv durch Multiskalenanalyse dargestellt, was eine Basissatz unabängige Beschreibung mit kontrollierter numerischer Genauigkeit ermöglicht. Gleichungen für die Coupled-Cluster Methode werden in einem verallgemeinerten Rahmen, unabhängig von virtuellen Orbitalen und globalen Basissätzen, neu formuliert. Hierzu werden die amplitudengewichteten Anregungen in virtuelle Orbitale ersetzt durch Anregungen in n-Elektronenfunktionen, welche durch Gleichungen im n-Elektronen Ortsraum bestimmt sind. Die erhaltenen Gleichungen können, analog zur Basissatz abh¨angigen Form, mit leicht angepasster Interpretation diagrammatisch dargestellt werden. Aufgrund des singulären Coulomb Potentials werden die Arbeitsgleichungen mit einem explizit korrelierten Ansatz regularisiert. Coupled-Cluster singles mit genäherten doubles (CC2) und ähnliche Modelle werden, für geschlossenschalige Systeme und in regularisierter Form, in die MADNESS Bibliothek (eine allgemeine Bibliothek zur Darstellung von Funktionen und Operatoren mittels Multiskalenanalyse) implementiert. Mit der vorgestellten Methode können elektronische CC2 Paarkorrelationsenergien und Anregungsenergien mit bestimmter numerischer Genauigkeit unabhängig von globalen Basissätzen berechnet werden, was anhand von kleinen Molekülen verifiziert wird / In this work algorithms for the computation of electronic correlation and excitation energies with the Coupled-Cluster method on adaptive grids are developed and implemented. The corresponding functions and operators are adaptively represented with multiresolution analysis allowing a basis-set independent description with controlled numerical accuracy. Equations for the coupled-cluster model are reformulated in a generalized framework independent of virtual orbitals and global basis-sets. For this, the amplitude weighted excitations into virtuals are replaced by excitations into n-electron functions which are determined by projected equations in the n-electron position space. The resulting equations can be represented diagrammatically analogous to basis-set dependent approaches with slightly adjusted rules of interpretation. Due to the singular Coulomb potential, the working equations are regularized with an explicitly correlated ansatz. Coupled-cluster singles with approximate doubles (CC2) and similar models are implemented for closed-shell systems and in regularized form into the MADNESS library (a general library for the representation of functions and operators with multiresolution analysis). With the presented approach electronic CC2 pair-correlation energies and excitation energies can be computed with definite numerical accuracy and without dependence on global basis sets, which is verified on small molecules.

This thesis explores the use of single-photon avalanche diodes (SPADs) for visible light communications (VLC). The high sensitivity of SPADs can potentially enhance the performance of VLC receivers. However, a SPAD-based system has challenges that need to be addressed before it can be considered as a viable option for VLC. The first challenge is the susceptibility of SPAD-based receivers to variations in ambient light. The high sensitivity of SPADs is advantageous for signal detection, but also makes SPADs vulnerable to variations in ambient light. In this thesis, the performance of a SPAD-based receiver is investigated under changing lighting conditions. Analytical expressions to quantify performance are derived, and an experiment is conducted to gain further understanding of system performance. It is shown that a SPAD-based receiver is highly sensitive to illumination changes when on-off keying (OOK) is employed, and that pulse-position modulation (PPM) is a preferred modulation scheme as it is more robust. The second challenge is broadcasting to SPAD-based receivers with different capabilities. A traditional broadcasting scheme is time-sharing, whereby a transmitter sends data to receivers in an alternating manner. Broadcasting to SPAD-based receivers is challenging as receivers may have diverse capabilities. In this thesis, a new multiresolution modulation scheme is proposed, which can potentially improve system performance over the traditional timesharing approach. The performance of the proposed scheme is analyzed, and a proof-of-concept experiment is performed to demonstrate its viability.

On considère des méthodes de Galerkin stochastiques pour des systèmes hyperboliques faisant intervenir des données en entrée incertaines de lois de distribution connues paramétrées par des variables aléatoires. On s'intéresse à des problèmes où un choc apparaît presque sûrement en temps fini. Dans ce cas, la solution peut développer des discontinuités dans les domaines spatial et stochastique. On utilise un schéma de Volumes Finis pour la discrétisation spatiale et une projection de Galerkin basée sur une approximation polynomiale par morceaux pour la discrétisation stochastique. On propose un solveur de type Roe avec correcteur entropique pour le système de Galerkin, utilisant une technique originale pour approcher la valeur absolue de la matrice de Roe et une adaptation du correcteur entropique de Dubois et Mehlmann. La méthode proposée reste coûteuse car une discrétisation stochastique très fine est nécessaire pour représenter la solution au voisinage des discontinuités. Il est donc nécessaire de faire appel à des stratégies adaptatives. Comme les discontinuités sont localisées en espace et évoluent en temps, on propose des représentations stochastiques dépendant de l'espace et du temps. On formule cette méthodologie dans un contexte multi-résolution basé sur le concept d'arbres binaires pour décrire la discrétisation stochastique. Les étapes d'enrichissement et d'élagage adaptatifs sont réalisées en utilisant des critères d'analyse multi-résolution. Dans le cas multidimensionnel, une anisotropie de la procédure adaptative est proposée. La méthodologie est évaluée sur le système des équations d'Euler dans un tube à choc et sur l'équation de Burgers en une et deux dimensions stochastiques / This work is concerned with stochastic Galerkin methods for hyperbolic systems involving uncertain data with known distribution functions parametrized by random variables. We are interested in problems where a shock appears almost surely in finite time. In this case, the solution exhibits discontinuities in the spatial and in the stochastic domains. A Finite Volume scheme is used for the spatial discretization and a Galerkin projection based on piecewise poynomial approximation is used for the stochastic discretization. A Roe-type solver with an entropy correction is proposed for the Galerkin system, using an original technique to approximate the absolute value of the Roe matrix and an adaptation of the Dubois and Mehlman entropy corrector. Although this method deals with complex situations, it remains costly because a very fine stochastic discretization is needed to represent the solution in the vicinity of discontinuities. This fact calls for adaptive strategies. As discontinuities are localized in space and time, stochastic representations depending on space and time are proposed. This methodology is formulated in a multiresolution context based on the concept of binary trees for the stochastic discretization. The adaptive enrichment and coarsening steps are based on multiresolution analysis criteria. In the multidimensional case, an anisotropy of the adaptive procedure is proposed. The method is tested on the Euler equations in a shock tube and on the Burgers equation in one and two stochastic dimensions

Este trabalho apresenta um estudo sobre transformada Wavelet aplicada à qualidade da energia elétrica com o intuito de detectar, localizar e classificar eventuais distúrbios que ocorrem no sistema elétrico. Inicialmente é apresentada uma introdução sobre qualidade da energia, mostrando fatos, evoluções e explicando o conceito dos principais fenômenos que interferem na qualidade da energia do sistema elétrico brasileiro, devido, principalmente, à grande demanda de aparelhos eletrônicos produzidos atualmente. Em seguida é mostrada uma revisão dos principais métodos e modelos aplicados atualmente no mundo a respeito do assunto. A transformada Wavelet vem como uma grande ajuda nesta área de análise de sinais, já que é capaz de extrair simultaneamente informações de tempo e freqüência, diferentemente da transformada de Fourier. A simulação dos diversos distúrbios ocorridos no sistema foi realizada através do software ATP (Alternative Transients Program), cujas características seguem corretamente um sistema de distribuição real da concessionária CPFL. Os distúrbios de tensão gerados e analisados foram detectados e localizados através da técnica de Análise Multiresolução e, posteriormente, classificados, utilizando para isto o método da Curva de Desvio Padrão / This dissertation presents a study of Wavelet transform applied to power quality in order to detect, locate and classify disturbances that may occur in the power system. Initially an introduction of power quality is presented, showing facts, evolutions and explaining the concept of the main phenomena that interfere the on power quality of the brazilian power system, due to, mainly, a great demand for electronic devices produced nowadays. A revision of the main methods and models currently applied in the world regarding this subject is also show. The Wavelet transform comes as a great support in the area of signal assessment, as it can extract information about time and frequency simultaneously, differently from the Fourier transform. The simulation of the diverse disturbances occurred in the system was accomplished through ATP software (Alternative Transients Program), whose characteristics correctly follow a system of real distribution of CPFL eletric utility. The generated and analyzed voltage disturbances were detected and located by Multiresolution Analysis technique and later classified by the method of the Standard Deviation