Dissertations / Theses on the topic 'Fuzzy pattern'

The Objective of this thesis is to talk about the usage of Fuzzy Logic in pattern recognition. There are different fuzzy approaches to recognize the pattern and the structure in data. The fuzzy approach that we choose to process the data is completely depends on the type of data. Pattern reorganization as we know involves various mathematical transforms so as to render the pattern or structure with the desired properties such as the identification of a probabilistic model which provides the explaination of the process generating the data clarity seen and so on and so forth. With this basic school of thought we plunge into the world of Fuzzy Logic for the process of pattern recognition. Fuzzy Logic like any other mathematical field has its own set of principles, types, representations, usage so on and so forth. Hence our job primarily would focus to venture the ways in which Fuzzy Logic is applied to pattern recognition and knowledge of the results. That is what will be said in topics to follow. Pattern recognition is the collection of all approaches that understand, represent and process the data as segments and features by using fuzzy sets. The representation and processing depend on the selected fuzzy technique and on the problem to be solved. In the broadest sense, pattern recognition is any form of information processing for which both the input and output are different kind of data, medical records, aerial photos, market trends, library catalogs, galactic positions, fingerprints, psychological profiles, cash flows, chemical constituents, demographic features, stock options, military decisions.. Most pattern recognition techniques involve treating the data as a variable and applying standard processing techniques to it.

Die Arbeit ordnet sich in das Gebiet der unscharfen Klassifikation ein und stellt im Detail eine Weiterführung der Forschung zur Fuzzy-Pattern-Klassifikation dar. Es handelt sich dabei um eine leistungsfähige systemtheoretische Methodik zur klassifikatorischen Modellierung komplexer, hochdimensionaler, technischer oder nichttechnischer Systeme auf der Basis von metrischen Messgrößen und/oder nichtmetrischen Experten-Bewertungen. Die Beschreibung der Unschärfe von Daten, Zuständen und Strukturen wird hierbei durch einen einheitlichen Typ einer Zugehörigkeitsfunktion des Potentialtyps realisiert. Ziel der Betrachtungen ist die weiterführende Nutzung des bestehenden Klassenmodells zur unscharfen Beschreibung nichtkonvexer Objektmengenmorphologien. Ausgehend vom automatischen datengetriebenen Aufbau der konvexen Klassenbeschreibung, deren vorteilhaften Eigenschaften sowie Defiziten wird im Rahmen der Arbeit eine Methodik vorgestellt, die eine Modellierung beliebiger Objektmengenmorphologien erlaubt, ohne das bestehende Klassifikationskonzept zu verlassen. Kerngedanken des Vorgehens sind: 1.) Die Aggregation von Fuzzy-Pattern-Klassen auf der Basis so genannter komplementärer Objekte. 2.) Die sequentielle Verknüpfung von Fuzzy-Pattern-Klassen und komplementären Klassen im Sinne einer unscharfen Mengendifferenz. 3.) Die Strukturierung des Verknüpfungsprozesses durch die Clusteranalyse von Komplementärobjektmengen und damit der Verwendung von Konfigurationen aus komplementären Fuzzy-Pattern-Klassen. Das dabei gewonnene nichtkonvexe Fuzzy-Klassifikationsmodell impliziert eine Vernetzung von Fuzzy-Klassifikatoren in Form von Klassifikatorbäumen. Im Ergebnis entstehen Klassifikatorstrukturen mit hoher Transparenz, die - neben der üblichen zustandsorientierten klassifikatorischen Beschreibung in den Einzelklassifikatoren - zusätzliche Informationen über den Ablauf der Klassifikationsentscheidungen erfassen. Der rechnergestützte Entwurf und die Eigenschaften der entstehenden Klassifikatorstruktur werden an akademischen Teststrukturen und realen Daten demonstriert. Die im Rahmen der Arbeit dargestellte Methodik wird in Zusammenhang mit dem Fuzzy-Pattern-Klassifikationskonzept realisiert, ist jedoch aufgrund ihrer Allgemeingültigkeit auf eine beliebige datenbasierte konvexe Klassenbeschreibung übertragbar
This work contributes to the field of fuzzy classification. It dedicates itself to the subject of "Fuzzy-Pattern-Classification", a versatile method applied for classificatory modeling of complex, high dimensional systems based on metric and nonmetric data, i.e. sensor readings or expert statements. Uncertainties of data, their associated morphology and therewith classificatory states are incorporated in terms of fuzziness using a uniform and convex type of membership function. Based on the properties of the already existing convex Fuzzy-Pattern-Class models and their automatic, data-driven setup a method for modeling nonconvex relations without leaving the present classification concept is introduced. Key points of the elaborated approach are: 1.) The aggregation of Fuzzy-Pattern-Classes with the help of so called complementary objects. 2.) The sequential combination of Fuzzy-Pattern-Classes and complementary Fuzzy-Pattern-Classes in terms of a fuzzy set difference. 3.) A clustering based structuring of complementary Fuzzy-Pattern-Classes and therewith a structuring of the combination process. A result of this structuring process is the representation of the resulting nonconvex fuzzy classification model in terms of a classifier tree. Such a nonconvex Fuzzy-Classifier features high transparency, which allows a structured understanding of the classificatory decision in working mode. Both the automatic data-based design as well as properties of such tree-like fuzzy classifiers will be illustrated with the help of academic and real word data. Even though the proposed method is introduced for a specific type of membership function, the underlying idea may be applied to any convex membership function

Der Gesetzgeber fordert von den Energieversorgungsunternehmen, ¨die Versorgung so sicher und billig wie moeglich zu gestalten¨. Um jederzeit ein stoerungsfreies und kostenguenstiges Angebot von elektrischer Energie zu gewaehrleisten, bedarf es einer moeglichst genauen Prognose der zu erwartenden Belastung im Versorgungsgebiet. Aufbauend auf den Erfahrungen bei der Prognose von Zeitreihen aus den Bereichen Umwelt und Verkehr wird in dieser Arbeit eine kurzfristige Vorhersage der elektrischen Belastung mit Hilfe der ¨Fuzzy-Pattern- Klassifikation¨ dargestellt. Dabei erfolgt die Modellbildung nicht allein auf der Basis der elektrischen Leistung, sondern wird durch zusaetzliche energiewirtschaftlich relevante Informationen, wie z.B. Klimadaten unterstuetzt. Zentraler Gegenstand der Untersuchungen ist die Frage, ob durch den Einsatz ergaenzender Informationen die Genauigkeit der Prognose bei kurzfristigen Vorhersagehorizonten (15 bis 120 Minuten) verbessert werden kann. Die mannigfaltigen Abhaengigkeiten zwischen elektrischer Belastung und ursaechlich wirkenden Einflussgroessen fuehren auf differenzierte Strategien zur Analyse und Prognose des Datenmaterials. Ausfuehrlich werden die Vorstrukturierung der Datenbasis, eine Prototypenmaskierung sowie die dynamische Parametrierung der Prognose erlaeutert und deren Wirksamkeit an realen Daten ueberprueft. Die Einschaetzungen zur Brauchbarkeit der Zusatzinformationen beruhen auf einem Vergleich von Prognoseresultaten der unterschiedlichen Modelle.

COORDENAÇÃO DE APERFEIÇOAMENTO DO PESSOAL DE ENSINO SUPERIOR
Esta dissertação investiga a utilização de sistemas Neuro- Fuzzy Hierárquicos BSP (Binary Space Partitioning) para classificação de padrões e para extração de regras fuzzy em bases de dados. O objetivo do trabalho foi criar modelos específicos para classificação de registros a partir do modelo Neuro-Fuzzy Hierárquico BSP que é capaz de gerar sua própria estrutura automaticamente e extrair regras fuzzy, lingüisticamente interpretáveis, que explicam a estrutura dos dados. O princípio da tarefa de classificação de padrões é descobrir relacionamentos entre os dados com a intenção de prever a classe de um padrão desconhecido. O trabalho consistiu fundamentalmente de quatro partes: um estudo sobre os principais métodos de classificação de padrões; análise do sistema Neuro-Fuzzy Hierárquico BSP (NFHB) original na tarefa de classificação; definição e implementação de dois sistemas NFHB específicos para classificação de padrões; e o estudo de casos. No estudo sobre os métodos de classificação foi feito um levantamento bibliográfico da área, resultando em um "survey" onde foram apresentadas as principais técnicas utilizadas para esta tarefa. Entre as principais técnicas destacaram-se: os métodos estatísticos, algoritmos genéticos, árvores de decisão fuzzy, redes neurais, e os sistemas neuro-fuzzy. Na análise do sistema NFHB na classificação de dados levou- se em consideração as peculiaridades do modelo, que possui: aprendizado da estrutura, particionamento recursivo do espaço de entrada, aceita maior número de entradas que os outros sistemas neuro-fuzzy, além de regras fuzzy recursivas. O sistema NFHB, entretanto, não é um modelo exatamente desenvolvido para classificação de padrões. O modelo NFHB original possui apenas uma saída e para utilizá- lo como um classificador é necessário criar um critério de faixa de valores (janelas) para representar as classes. Assim sendo, decidiu-se criar novos modelos que suprissem essa deficiência. Foram definidos dois novos sistemas NFHB para classificação de padrões: NFHB-Invertido e NFHB-Class. O primeiro utiliza a arquitetura do modelo NFHB original no aprendizado e em seguida a inversão da mesma para a validação dos resultados. A inversão do sistema consistiu de um meio de adaptar o novo sistema à tarefa específica de classificação, pois passou-se a ter o número de saídas do sistema igual ao número de classes ao invés do critério de faixa de valores utilizado no modelo NFHB original. Já o sistema NFHB-Class utilizou, tanto para a fase de aprendizado, quanto para a fase de validação, o modelo NFHB original invertido. Ambos os sistemas criados possuem o número de saídas igual ao número de classes dos padrões, o que representou um grande diferencial em relação ao modelo NFHB original. Além do objetivo de classificação de padrões, o sistema NFHB-Class foi capaz de extrair conhecimento em forma de regras fuzzy interpretáveis. Essas regras são expressas da seguinte maneira: SE x é A e y é B então padrão pertence à classe Z. Realizou-se um amplo estudo de casos, abrangendo diversas bases de dados Benchmark para a tarefa de classificação, tais como: Iris Dataset, Wine Data, Pima Indians Diabetes Database, Bupa Liver Disorders e Heart Disease, e foram feitas comparações com diversos modelos e algoritmos de classificação de padrões. Os resultados encontrados com os modelos NFHB-Invertido e NFHB-Class mostraram-se, na maioria dos casos, superiores ou iguais aos melhores resultados encontrados pelos outros modelos e algoritmos aos quais foram comparados.O desempenho dos modelos NFHB-Invertido e NFHB-Class em relação ao tempo de processamento também se mostrou muito bom. Para todas as bases de dados descritas no estudo de casos (capítulo 8), os modelos convergiram para uma ótima solução de classificação, além da extração das regras fuzzy, em
This dissertation investigates the use of Neuro-Fuzzy Hierarchical BSP (Binary Space Partitioning) systems for pattern classification and extraction of fuzzy rules in databases. The objective of this work was to create specific models for the classification of registers based on the Neuro-Fuzzy BSP model that is able to create its structure automatically and to extract linguistic rules that explain the data structure. The task of pattern classification is to find relationships between data with the intention of forecasting the class of an unknown pattern. The work consisted of four parts: study about the main methods of the pattern classification; evaluation of the original Neuro-Fuzzy Hierarchical BSP system (NFHB) in pattern classification; definition and implementation of two NFHB systems dedicated to pattern classification; and case studies. The study about classification methods resulted in a survey on the area, where the main techniques used for pattern classification are described. The main techniques are: statistic methods, genetic algorithms, decision trees, neural networks, and neuro-fuzzy systems. The evaluation of the NFHB system in pattern classification took in to consideration the particularities of the model which has: ability to create its own structure; recursive space partitioning; ability to deal with more inputs than other neuro-fuzzy system; and recursive fuzzy rules. The original NFHB system, however, is unsuited for pattern classification. The original NFHB model has only one output and its use in classification problems makes it necessary to create a criterion of band value (windows) in order to represent the classes. Therefore, it was decided to create new models that could overcome this deficiency. Two new NFHB systems were developed for pattern classification: NFHB-Invertido and NFHB-Class. The first one creates its structure using the same learning algorithm of the original NFHB system. After the structure has been created, it is inverted (see chapter 5) for the generalization process. The inversion of the structure provides the system with the number of outputs equal to the number of classes in the database. The second system, the NFHB-Class uses an inverted version of the original basic NFHB cell in both phases, learning and validation. Both systems proposed have the number of outputs equal to the number of the pattern classes, what means a great differential in relation to the original NFHB model. Besides the pattern classification objective, the NFHB- Class system was able to extract knowledge in form of interpretable fuzzy rules. These rules are expressed by this way: If x is A and y is B then the pattern belongs to Z class. The two models developed have been tested in many case studies, including Benchmark databases for classification task, such as: Iris Dataset, Wine Data, Pima Indians Diabetes Database, Bupa Liver Disorders and Heart Disease, where comparison has been made with several traditional models and algorithms of pattern classification. The results found with NFHB-Invertido and NFHB-Class models, in all cases, showed to be superior or equal to the best results found by the others models and algorithms for pattern classification. The performance of the NFHB- Invertido and NFHB-Class models in terms of time-processing were also very good. For all databases described in the case studies (chapter 8), the models converged to an optimal classification solution, besides the fuzzy rules extraction, in a time-processing inferior to a minute.
Esta disertación investiga el uso de sistemas Neuro- Fuzzy Herárquicos BSP (Binary Space Partitioning) en problemas de clasificación de padrones y de extracción de reglas fuzzy en bases de datos. El objetivo de este trabajo fue crear modelos específicos para clasificación de registros a partir del modelo Neuro-Fuzzy Jerárquico BSP que es capaz de generar automáticamente su propia extructura y extraer reglas fuzzy, lingüisticamente interpretables, que explican la extructura de los datos. El principio de la clasificación de padrones es descubrir relaciones entre los datos con la intención de prever la clase de un padrón desconocido. El trabajo está constituido por cuatro partes: un estudio sobre los principales métodos de clasificación de padrones; análisis del sistema Neuro-Fuzzy Jerárquico BSP (NFHB) original en la clasificación; definición e implementación de dos sistemas NFHB específicos para clasificación de padrones; y el estudio de casos. En el estudio de los métodos de clasificación se realizó un levatamiento bibliográfico, creando un "survey" donde se presentan las principales técnicas utilizadas. Entre las principales técnicas se destacan: los métodos estadísticos, algoritmos genéticos, árboles de decisión fuzzy, redes neurales, y los sistemas neuro-fuzzy. En el análisis del sistema NFHB para clasificación de datos se tuvieron en cuenta las peculiaridades del modelo, que posee : aprendizaje de la extructura, particionamiento recursivo del espacio de entrada, acepta mayor número de entradas que los otros sistemas neuro-fuzzy, además de reglas fuzzy recursivas. El sistema NFHB, sin embargo, no es un modelo exactamente desarrollado para clasificación de padrones. El modelo NFHB original posee apenas una salida y para utilizarlo conmo un clasificador fue necesario crear un criterio de intervalos de valores (ventanas) para representar las clases. Así, se decidió crear nuevos modelos que supriman esta deficiencia. Se definieron dos nuevos sistemas NFHB para clasificación de padrones: NFHB- Invertido y NFHB-Clas. El primero utiliza la arquitectura del modelo NFHB original en el aprendizaje y en seguida la inversión de la arquitectura para la validación de los resultados. La inversión del sistema es un medio para adaptar el nuevo sistema, específicamente a la clasificación, ya que el sistema pasó a tener número de salidas igual al número de clases, al contrario del criterio de intervalo de valores utilizado en el modelo NFHB original. En el sistema NFHB-Clas se utilizó, tanto para la fase de aprendizajeo, cuanto para la fase de validación, el modelo NFHB original invertido. Ambos sistemas poseen el número de salidas igual al número de clases de los padrones, lo que representa una gran diferencia en relación al modelo NFHB original. Además del objetivo de clasificación de padrones, el sistema NFHB-Clas fue capaz de extraer conocimento en forma de reglas fuzzy interpretables. Esas reglas se expresan de la siguiente manera: Si x es A e y es B entonces el padrón pertenece a la clase Z. Se realizó un amplio estudio de casos, utilizando diversas bases de datos Benchmark para la clasificación, tales como: Iris Dataset, Wine Data, Pima Indians Diabetes Database, Bupa Liver Disorders y Heart Disease. Los resultados se compararon con diversos modelos y algoritmos de clasificación de padrones. Los resultados encontrados con los modelos NFHB-Invertido y NFHB-Clas se mostraron, en la mayoría de los casos, superiores o iguales a los mejores resultados encontrados por los otros modelos y algoritmos con los cuales fueron comparados. El desempeño de los modelos NFHB-Invertido y NFHB-Clas en relación al tiempo de procesamiento tambiém se mostró muy bien. Para todas las bases de datos descritas en el estudio de casos (capítulo 8), los modelos convergieron para una solución óptima, además de la extracción de las reglas fuzzy, con tiemp

In this thesis we develop a novel network model that extends the traditional artificial neural network (ANN) model to include oscillatory behaviour. This model is able to correctly classify combinations of previously learned input patterns by grouping features that belong to the same category. This grouping process is termed segmentation and we show how synchrony of oscillations is the necessary missing component of ANNs to be able to perform this segmentation. Using this model we go on to show that top-down modulatory feedback is necessary to enable separation of multiple objects in a scene and segmentation of their individual features. This type of feedback is distinctly different than recurrency and is what enables the rich dynamics between the nodes of our network. Additionally, we show how our model's dynamics avoid the combinatorial explosion in required training repetitions of traditional feed-forward classification networks. In these networks, relations between objects must explicitly be learned. In contrast, the dynamics of modulatory feedback allow us to defer calculation of these relations until run-time, thus creating a more robust system. We call our model Fuzzy Oscillations, and it achieves good results when compared to existing models. However, oscillatory neural network models successful in achieving segmentation are a relatively recent development. We thus feel that our model is a contribution to the field of oscillatory neural networks.

Les motifs graduels visent à décrire des co-variations au sein des données et sont de la forme plus l'âge est important, plus le salaire est élevé. Ces motifs ont fait l'objet de nombreux travaux en fouille de données ces dernières années, du point de vue des définitions que peuvent avoir de tels motifs et d'un point de vue algorithmique pour les extraire efficacement. Ces définitions et algorithmes considèrent qu'il est possible d'ordonner de manière stricte les valeurs (par exemple l'âge, le salaire). Or, dans de nombreux champs applicatifs, il est difficile voire impossible d'ordonner de cette manière. Par exemple, quand l'on considère l'expression de gènes, dire que l'expression d'un gène est plus importante que l'expression d'un autre gène quand leurs expressions ne diffèrent qu'à la dixième décimale n'a pas de sens d'un point de vue biologique. Ainsi, nous proposons dans cette thèse une approche fondée sur les ordres flous. Les algorithmes étant très consommateurs tant en mémoire qu'en temps de calcul, nous proposons des optimisations d'une part du stockage des degrés flous et d'autre part de calcul parallélisé. Les expérimentations que nous avons menées sur des bases de données synthétiques et réelles montrent l'intérêt de notre approche
Gradual patterns aim at describing co-variations of data such as the older, the higher the salary. They have been more and more studied from the data mining point of view in recent years, leading to several ways of defining their meaning and and several algorithms to automatically extract them.They consider that data can be ordered regarding the values taken on the attributes (e.g. the age and the salary).However, in many application domains, it is hardly possible to consider that data values are crisply ordered. For instance, when considering gene expression, it is not true, from the biological point of view, to say that Gene 1 is more expressed than Gene 2 if the levels of expression only differ from the tenth decimal. This thesis thus considers fuzzy orderings and propose both formal definitions and algorithms to extract gradual patterns considering fuzzy orderings. As these algorithms are both time and memory consuming, we propose some optimizations based on an efficient storage of the fuzzy ordering informationcoupled with parallel algorithms. Experimental results run on synthetic and real database show the interest or our proposal

La logique floue est aujourd'hui universellement admise comme discipline ayant fait ses preuves à l'intersection des mathématiques, de l'informatique, des sciences cognitives et de l'Intelligence Artificielle. En termes formels, la logique floue est une extension de la logique classique ayant pour but de mesurer la flexibilité du raisonnement humain, et permettant la modélisation des imperfections des données, en particulier, les deux imperfections les plus fréquentes : l'imprécision et l'incertitude. En outre, la logique floue ignore le principe du tiers exclu et celui de non-contradiction.Nous n'allons pas, dans ce court résumé de la thèse, reprendre et définir tous les concepts de cet outil devenu désormais classique : fonction d'appartenance, degré d'appartenance, variable linguistique, opérateurs flous, fuzzyfication, défuzzication, raisonnement approximatif … L'un des concepts de base de cette logique est la notion de possibilité qui permet de modéliser la fonction d'appartenance d'un concept. La possibilité d'un événement diffère de sa probabilité dans la mesure où elle n'est pas intimement liée à celle de l'événement contraire. Ainsi, par exemple, si la probabilité qu'il pleuve demain est de 0,6, alors la probabilité qu'il ne pleuve pas doit être égale à 0,4 tandis que les possibilités qu'il pleuve demain ou non peuvent toutes les deux être égales à 1 (ou encore deux autres valeurs dont la somme peut dépasser 1).Dans le domaine de l'informatique, l'apprentissage non supervisé (ou « clustering ») est une méthode d'apprentissage automatique quasi-autonome. Il s'agit pour un algorithme de diviser un groupe de données, en sous-groupes de manière que les données considérées comme les plus similaires soient associées au sein d'un groupe homogène. Une fois que l'algorithme propose ces différents regroupements, le rôle de l'expert ou du groupe d'experts est alors de nommer chaque groupe, éventuellement diviser certains ou de regrouper certains, afin de créer des classes. Les classes deviennent réelles une fois que l'algorithme a fonctionné et que l'expert les a nommées.Encore une fois, notre travail, comme tous les travaux du domaine, vise à adapter les modèles traditionnelles d'apprentissage et/ou de raisonnement à l'imprécision du monde réel. L'analyse des sentiments à partir de ressources textuelles et les images dans le cadre de l'agriculture de précision nous ont permis d'illustrer nos hypothèses. L'introduction par le biais de notre travail du concept de motifs flous est sans aucun doute une contribution majeure.Ce travail a donné lieu à trois contributions majeures :
Standard (type-1) fuzzy sets were introduced to mimic human reasoning in its use of approximate information and uncertainty to generate decisions. Since knowledge can be expressed in a natural way by using fuzzy sets, many decision problems can be greatly simpli_ed. However, standard type-1 fuzzy sets have limitations when it comes to modelinghuman decision making.When Zadeh introduced the idea of higher types of fuzzy sets called type-n fuzzy sets andtype-2 fuzzy sets, the objective was to solve problems associated with modeling uncertainty using crisp membership functions of type-1 fuzzy sets. The extra dimension presented by type-2 fuzzy sets provides more design freedom and exibility than type-1 fuzzy sets. The ability of FLS to be hybridized with other methods extended the usage of Fuzzy LogicSystems (FLS) in many application domains. In architecture and software engineering the concept of patterns was introduced as a way of creating general repeatable solutions to commonly occurring problems in the respective_elds. In software engineering for example, the design pattern is not a _nished design that can be transformed directly into code. It is a description or template on how to solve a problem that can be used in many di_erent situations. This thesis introduces the novel concept of fuzzy patterns in T2 FLS. Micro-blogs and social media platforms are now considered among the most popular forms of online communication. Through a platform like TwitterTM much information reecting people's opinions and attitudes is published and shared among users on a daily basis. This has brought great opportunities to companies interested in tracking and monitoring the reputation of their brands and businesses, and to policy makers and politicians to support their assessment of public opinions about their policies or political issues. Thisresearch demonstrates the importance of the neutral category in sentiment polarity analysis, it then introduces the concept of fuzzy patterns in sentiment polarity analysis. The xvii Interval Type-2 Fuzzy Set (IT2 FS), were proposed by reference [Men07c] to model words. This is because it is characterized by its Footprint Of Uncertainty (FOU). The FOU providesa potential to capture word uncertainties. The use of IT2 FS in polarity sentiment classi_cation is demonstrated. The importance of the neutral category is demonstrated in both supervised and unsupervised learning methods. In the _nal section the concept of fuzzy patterns in contrast

Stastical pattern recognition is the most successful approach to automatic speech and speaker recognition (ASASR). Of all the statistical pattern recognition techniques, the hidden Markov model (HMM) is the most important. The Gaussian mixture model (GMM) and vector quantisation (VQ) are also effective techniques, especially for speaker recognition and in conjunction with HMMs. for speech recognition. However, the performance of these techniques degrades rapidly in the context of insufficient training data and in the presence of noise or distortion. Fuzzy approaches with their adjustable parameters can reduce such degradation. Fuzzy set theory is one of the most, successful approaches in pattern recognition, where, based on the idea of a fuzzy membership function, fuzzy C'-means (FCM) clustering and noise clustering (NC) are the most, important techniques. To establish fuzzy approaches to ASASR, the following basic problems are solved. First, a time-dependent fuzzy membership function is defined for the HMM. Second, a general distance is proposed to obtain a relationship between modelling and clustering techniques. Third, fuzzy entropy (FE) clustering is proposed to relate fuzzy models to statistical models. Finally, fuzzy membership functions are proposed as discriminant functions in decison making. The following models are proposed: 1) the FE-HMM. NC-FE-HMM. FE-GMM. NC-FEGMM. FE-VQ and NC-FE-VQ in the FE approach. 2) the FCM-HMM. NC-FCM-HMM. FCM-GMM and NC-FCM-GMM in the FCM approach, and 3) the hard HMM and GMM as the special models of both FE and FCM approaches. Finally, a fuzzy approach to speaker verification and a further extension using possibility theory are also proposed. The evaluation experiments performed on the TI46, ANDOSL and YOHO corpora showbetter results for all of the proposed techniques in comparison with the non-fuzzy baseline techniques.

Satellite images increasingly become a major data source for monitoring changes in the natural environment. A main task in the analysis of satellite images is concerned with the modelling of land use classes by reducing uncertainty during a classification process. In the approach presented in this paper uncertainty is perceived to be due to the vagueness of geographical categories caused by either the complexity of the category (like 'urban area') or by the use of the category in several application contexts. Two circumstances of use of an extended set theoretical concept (fuzzy sets) are discussed. First, two algorithms from the ISODATA class are used to determine the grades of membership to three a priori defined classes (woodland, suburban area, urban area) of a LANDSAT TM satellite image of Vienna, Austria. The results are visualized by a RGB image of the grades of membership to each category. Second, a measure of fuzziness is employed on the results. The measure relies on the concept of distance between a seUcategory and its complement. The so determined vagueness provide more information on the uncertainty of the different categories and may improve further information processing tasks. (authors' abstract)
Series: Discussion Papers of the Institute for Economic Geography and GIScience

Fast and accurate recognition of the Statistical Control Chart Patterns (SPCCP) is significant for supervising manufacturing processes to accomplish better control and to make high value products. SPCCP can display eight kinds of patterns: normal, stratification, systematic, increasing trend, decreasing trend, up shift, down shift and cyclic. With the exception of the natural pattern, all other patterns indicate that the supervised manufacturing process is not performing properly and actions need to be taken to correct the problems. This research proposes new approaches, neural networks and neural-fuzzy systems, to the (SPCCP) recognition. This dissertation also investigates the use of features extracted from statistical analysis for simple patterns, and wavelet analysis for concurrent patterns as the components of the input vectors. Results based on simulated data show that the proposed approaches perform better than conventional approaches. Our work concluded that the extracted features improve the performance of the proposed recognizer systems.

In this thesis, Genetic Granular Cognitive Fuzzy Neural Networks (GGCFNN), combining genetic algorithms (GA) and granular cognitive fuzzy neural networks (GCFNN), is proposed for pattern recognition problems. According to cognitive patterns, biological neural networks in the human brain can recognize different patterns. Since GA and neural networks represent two learning methods based on biological science, it is indispensable and valuable to investigate how biological neural networks and artificial neural networks recognize different patterns. The new GGCFNN, based on granular computing, soft computing and cognitive science, is used in the pattern recognition problems. The hybrid forward-wave-backward-wave learning algorithm, as a main learning technology in GCFNN, is used to enhance learning quality. GA optimizes parameters to make GGCFNN get better learning results. Both pattern recognition results generated by human persons and those by GGCFNN are analyzed in terms of computer science and cognitive science.

The thesis begins by discussing the Modern Paradigm, which, it is argued, forms the underpinning to much contemporary management thinking. This in turn is seen as having its own foundation, the Cartesian-Newtonian Synthesis. It is argued that this does not form an adequate basis for strategic management, and the thesis then draws upon four main streams of thinking: complexity theory, fuzzy logic, the debate on power in organizations, and critical theory. The material developed from these four streams is integrated, thereby developing a number of principles for strategic management. The context within which most of the case studies are set is also outlined, by reviewing the recent history and present situation in local government. We then turn to the practical implications. When teaching strategic management and change, a frequent response from managers is that they are comfortable with the rational planning approach, which they find straightforward in approach, and its tools and techniques readily usable. But when we get on to all this other stuff ... what does it mean, and how is it used? This relates also to my own experience as a manager, particularly in local government. The practical implications are important, and this whole thesis can be seen as an action research programme, with practical interventions enriching the theoretical perspective, which in turn has fed back into practice. This discussion begins by considering methodology, and identifies three interlinked methods - action research, action learning and whole systems intervention. These are related to critical theory, and it is argued that these approaches provide a practice based upon the theoretical themes developed earlier. This is followed by a discussion of action research, exploring one case study in some depth, chosen because it helps to illustrate both the strengths and the potential limitations of a critical approach to action research. The work is assessed, and its implication for contemporary management are considered, drawing also upon a current action research project concerned with the roles of trade unions in the regions of Europe. The thesis then turns to what can best be seen as an extended action research project concerned specifically with whole systems intervention. It examines the extent to which this can be developed and undertaken on the basis of the principles developed in the thesis. Five case studies are presented in which Search Conferences and/or ColourFlow Dialogue have been used. Reflecting the original remit of the thesis, these case studies have a common link in local government. Two involve local authorities directly; one concerns local government politicians and their political party, and one involves an area of local authority activity being moved into independent Trust status. The fifth has a more tenuous link with local government: it is a voluntary body which receives significant funding from Councils, but is otherwise independent; it is included because it was the first such exercise undertaken, and brought with it significant personal learning. Finally, the thesis reviews the findings, considers their implications, and draws conclusions. Thus the purpose of this thesis is both to present an approach to strategic management and organizational development which is richer than those premised on the Modern Paradigm, and to argue that this is more than a set of interesting or provocative ideas - it is an approach which can be put into practice.

Thesis (Ph. D.)--University of Cincinnati, 2006..
Title from electronic thesis title page (viewed Jan.27, 2007). Includes abstract. Keywords: Fuzzy Hamming Distance, artificial intelligence, fuzzy, image retrieval system Includes bibliographical references.

Classification of terrain cover from satellite radar imagery represents an area of considerable current interest and research. Most satellite sensors used for land applications are of the imaging type. They record data in a variety of spectral channels and at a variety of ground resolutions. Spectral pattern recognition refers to classification procedures utilizing pixel-by-pixel spectral information as the basis for automated land cover classification. A number of methods have been developed in the past to classify pixels [resolution cells] from multispectral imagery to a priori given land cover categories. Their ability to provide land cover information with high classification accuracies is significant for work where accurate and reliable thematic information is needed. The current trend towards the use of more spectral bands on satellite instruments, such as visible and infrared imaging spectrometers, and finer pixel and grey level resolutions will offer more precise possibilities for accurate identification. But as the complexity of the data grows, so too does the need for more powerful tools to analyse them. It is the major objective of this study to analyse the capabilities and applicability of the neural pattern recognition system, called fuzzy ARTMAP, to generate high quality classifications of urban land cover using remotely sensed images. Fuzzy ARTMAP synthesizes fuzzy logic and Adaptive Resonance Theory (ART) by exploiting the formal similarity between the computations of fuzzy subsethood and the dynamics of category choice, search and learning. The paper describes design features, system dynamics and simulation algorithms of this learning system, which is trained and tested for classification (8 a priori given classes) of a multispectral image of a Landsat-5 Thematic Mapper scene (270 x 360 pixels) from the City of Vienna on a pixel-by-pixel basis. Fuzzy ARTMAP performance is compared with that of an error-based learning system based upon the multi-layer perceptron, and the Gaussian maximum likelihood classifier as conventional statistical benchmark on the same database. Both neural classifiers outperform the conventional classifier in terms of classification accuracy. Fuzzy ARTMAP leads to out-of-sample classification accuracies, very close to maximum performance, while the multi-layer perceptron - like the conventional classifier - shows difficulties to distinguish between some land use categories. (authors' abstract)
Series: Discussion Papers of the Institute for Economic Geography and GIScience

Este trabalho propõe uma alternativa para o problema de classificação de espécies de abelhas a partir da implementação de um algoritmo com base na Morfométria Geométrica e estudo das Formas dos marcos anatômicos das imagens obtidas pelas asas das abelhas. O algoritmo implementado para este propósito se baseia no algoritmo dos k-Vizinho mais Próximos (do inglês, kNN) e na Lógica Fuzzy kNN (Fuzzy k-Nearest Neighbor) aplicados a dados analisados e selecionados de pontos bidimensionais referentes as características geradas por marcos anatômicos. O estudo apresentado envolve métodos de seleção e ordenação de marcos anatômicos para a utilização no algoritmo por meio da implementação de um método matemático que utiliza o calculo dos marcos anatômicos mais significativos (que são representados por marcos matemáticos) e a formulação da Ordem de Significância onde cada elemento representa variáveis de entrada para a Fuzzy kNN. O conhecimento envolvido neste trabalho inclui uma perspectiva sobre a seleção de características não supervisionada como agrupamentos e mineração de dados, analise de pré-processamento dos dados, abordagens estatísticas para estimação e predição, estudo da Forma, Analise de Procrustes e Morfométria Geométrica sobre os dados e o tópico principal que envolve uma modificação do algoritmo dos k- Vizinhos mais Próximos e a aplicação da Fuzzy kNN para o problema. Os resultados mostram que a classificação entre amostras de abelhas no seu próprio grupo apresentam acuracia de 90%, dependendo da espécie. As classificações realizadas entre as espécies de abelhas alcançaram acuracia de 97%.
This work presents a proposal to solve the bees classification problem by implementing an algorithm based on Geometrics Morphometrics and the Shape analysis of landmarks generated from bees wings images. The algorithm is based on the K-Nearest Neighbor (K-Nearest Neighbor) algorithm and Fuzzy Logic KNN applied to the analysis and selection of two-dimensional data points relating to landmarks. This work is part of the Architecture Reference Model for Automatic identification and Taxonomic Classification System of Stingless Bee using the Wing Morphometry. The study includes selection and ordering methods for landmarks used in the algorithm by developing a mathematical model to represent the significance order, generating the most significant mathematical landmarks as input variables for Fuzzy Logic kNN. The main objective of this work is to develop a classification system for bee species. The knowledge involved in the development of this work include an overview of feature selection, unsupervised clustering and data mining, analysis of data pre-processing, statistical approaches for estimation and prediction, study of Shape, Procrustes Analysis on data that comes from Geometric Morphometry and the modification of the k-Nearest Neighbors algorithm and the Fuzzy Logic kNN. The results show that the classification in bee samples of the same species presents a accuracy above 90%, depending on the specie in analysis. The classification done between the bees species reach accuracies of 97%.

This dissertation introduces advanced artificial intelligence based algorithm for detecting and classifying faults on the power system transmission line. The proposed algorithm is aimed at substituting classical relays susceptible to possible performance deterioration during variable power system operating and fault conditions. The new concept relies on a principle of pattern recognition and detects the existence of the fault, identifies fault type, and estimates the transmission line faulted section. The approach utilizes self-organized, Adaptive Resonance Theory (ART) neural network, combined with fuzzy decision rule for interpretation of neural network outputs. Neural network learns the mapping between inputs and desired outputs through processing a set of example cases. Training of the neural network is based on the combined use of unsupervised and supervised learning methods. During training, a set of input events is transformed into a set of prototypes of typical input events. During application, real events are classified based on the interpretation of their matching to the prototypes through fuzzy decision rule. This study introduces several enhancements to the original version of the ART algorithm: suitable preprocessing of neural network inputs, improvement in the concept of supervised learning, fuzzyfication of neural network outputs, and utilization of on-line learning. A selected model of an actual power network is used to simulate extensive sets of scenarios covering a variety of power system operating conditions as well as fault and disturbance events. Simulation results show improved recognition capabilities compared to a previous version of ART neural network algorithm, Multilayer Perceptron (MLP) neural network algorithm, and impedance based distance relay. Simulation results also show exceptional robustness of the novel ART algorithm for all operating conditions and events studied, as well as superior classification capabilities compared to the other solutions. Consequently, it is demonstrated that the proposed ART solution may be used for accurate, high-speed distinction among faulted and unfaulted events, and estimation of fault type and fault section.

Orientador: Lee Luan Ling
Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação
Made available in DSpace on 2018-08-17T04:05:46Z (GMT). No. of bitstreams: 1 Rodrigues_RicardoNagel_M.pdf: 1212348 bytes, checksum: c1f563f073688d610f7a7013a642214c (MD5) Previous issue date: 2006
Resumo: Neste trabalho, apresentamos um novo método para fazer a fusão de dois sistemas biométricos unimodais. O objetivo é gerar um sistema biométrico multimodal que apresente menores taxas de erro, maior robustez e maior segurança. O método proposto pode ser usada para integrar qualquer tipo de modalidade biométrica. Desta forma, fazemos uma descrição geral do sistema multimodal proposto, sem restrição quanto ao tipo de tecnologias biométricas que serão combinadas. Após esta descrição geral, três sistemas de reconhecimento biométrico (baseados na face, impressão digital e dinâmica da digitação) são apresentados e a metodologia é testada combinando-se a face com a impressão digital e a face com a dinâmica da digitação. A fusão dos dados biométricos é feita no nível de comparação, sendo que uma das principais inovações do método proposto é que, além dos índices de similaridade, o módulo de fusão recebe também um índice de confiabilidade das amostras coletadas e um parâmetro que indica a segurança dos sistemas unimodais. Estes dados são processados através de um sistema de inferência nebuloso, produzindo um único valor de saída que é usado para decidir se o usuário é genuíno ou impostor. Novos procedimentos de testes, que simulam condições de operação adversas, foram adotados e mostraram que o método de fusão biométrica proposto apresenta vantagens quando comparado com a fusão através da regra da soma
Abstract: In this work, we present a new method for fusing two unimodal biometric systems. The objective is to create a multimodal biometric system with low error rates, high robustness and security. The proposed method can be used to combine any two kinds of biometric modalities. We make a general description of the proposed method, with no restrictions about which biometric technologies will be combined. After this general description, three biometric recognition systems (based on face, fingerprint and keystroke dynamics) are introduced and the fusion method is tested by combining face with fingerprint and face with keystroke dynamics. The biometric data fusion is performed at the matching score level. One of the main novelties of the proposed method is that, besides similarity scores, the fusion module also receives as input a sample reliability index and a parameter that indicate the security level of the unimodal biometric systems. This set of data is processed by a fuzzy inference system, producing one single output score that is used to decide if the user is either genuine or impostor. Novel test procedures, that simulate adverse operational conditions, have indicated that the proposed biometric fusion method presents some advantages when compared with the fusion using the sum rule
Mestrado
Telecomunicações e Telemática
Mestre em Engenharia Elétrica

Clustering algorithms are widely used in pattern recognition and data mining applications. Due to their computational efficiency, partitional clustering algorithms are better suited for applications with large datasets than hierarchical clustering algorithms. K-means is among the most popular partitional clustering algorithm, but has a major shortcoming: it is extremely sensitive to the choice of initial centers used to seed the algorithm. Unless k-means is carefully initialized, it converges to an inferior local optimum and results in poor quality partitions. Developing improved method for selecting initial centers for k-means is an active area of research. Genetic algorithms (GAs) have been successfully used to evolve a good set of initial centers. Among the most promising GA-based methods are those that exchange neighboring centers between candidate partitions in their crossover operations. K-means is best suited to work when datasets have well-separated non-overlapping clusters. Fuzzy c-means (FCM) is a popular variant of k-means that is designed for applications when clusters are less well-defined. Rather than assigning each point to a unique cluster, FCM determines the degree to which each point belongs to a cluster. Like k-means, FCM is also extremely sensitive to the choice of initial centers. Building on GA-based methods for initial center selection for k-means, this dissertation developed an evolutionary program for center selection in FCM called FCMGA. The proposed algorithm utilized region-based crossover and other mechanisms to improve the GA. To evaluate the effectiveness of FCMGA, three independent experiments were conducted using real and simulated datasets. The results from the experiments demonstrate the effectiveness and consistency of the proposed algorithm in identifying better quality solutions than extant methods. Moreover, the results confirmed the effectiveness of region-based crossover in enhancing the search process for the GA and the convergence speed of FCM. Taken together, findings in these experiments illustrate that FCMGA was successful in solving the problem of initial center selection in partitional clustering algorithms.

Diversas abordagens vêm sendo empregadas para a descrição de texturas, entre elas a teoria dos conjuntos fuzzy e lógica fuzzy. O Local Fuzzy Pattern (LFP) é um descritor de texturas diferente dos demais métodos baseados em sistemas fuzzy, por não utilizar regras linguísticas e sim números fuzzy que são usados na codificação de um padrão local de escala de cinza. Resultados anteriores indicaram o LFP como um descritor eficaz para a classificação de texturas a partir de amostras rotacionadas ou não. Este trabalho propõe uma análise mais abrangente sobre sua viabilidade para aplicação em cada um desses problemas, além de propor uma modificação a este descritor, adaptando-o para a captura de padrões em multiresolução, o Sampled LFP. A avaliação da performance do LFP e do Sampled LFP para o problema de classificação de texturas foi feita através da aplicação de uma série de testes envolvendo amostras de imagens rotacionadas ou não das bases de imagens Outex, álbum de Brodatz e VisTex, onde a sensibilidade obtida por esses descritores foi comparada com um descritor de referência, a variante do Local Binary Pattern (LBP) melhor indicada para o teste em execução. Os resultados apontaram o LFP como um descritor não indicado para aplicações que trabalhem exclusivamente com amostras não rotacionadas, visto que o LBP mostrou maior eficácia para este tipo de problema. Já para a análise de amostras rotacionadas, o Sampled LFP se mostrou o melhor descritor entre os comparados. Todavia, foi verificado que o Sampled LFP somente supera o LBP para resoluções de análise maiores ou iguais a 32x32 pixels e que o primeiro descritor é mais sensível ao número de amostras usadas em seu treinamento do que o segundo, sendo, portanto, um descritor indicado para o problema de classificação de amostras rotacionadas, onde seja possível trabalhar com imagens a partir de 32x32 pixels e que o número de amostras utilizadas para treinamento seja maximizado.
Several approaches have been employed for describing textures, including the fuzzy sets theory and fuzzy logic. The Local Fuzzy Pattern is a texture descriptor different from other methods based on fuzzy systems, which use linguistic rules to codify a texture. Instead, fuzzy numbers are applied in order to encode a local grayscale pattern. Previous results indicated the LFP as an effective descriptor employed to characterize statically oriented and rotated textures samples. This paper proposes a more comprehensive analysis of its feasibility for use in each of these problems, besides proposing a modification to this descriptor, adapting it to capture patterns in multiresolution, the Sampled LFP. The LFP and Sampled LFP performance evaluation when applied to the problem of texture classification was conducted by applying a series of tests involving images samples, rotated or not, from image databases such as Outex, the Brodatz album and Vistex, where the sensitivity obtained by these descriptors were compared with a reference descriptor, the variant Local Binary Pattern (LBP) best suited to running the test. The results indicated the LFP as a descriptor not suitable for applications who work exclusively with non-rotated samples, since the LBP showed greater efficacy for this problem kind. As for rotated samples analysis, the Sampled LFP proved the best descriptor among those compared. However, it was determined that the Sampled LFP only overcomes the LBP when the analysis resolutions are greater or equal to 32x32 pixels, besides that, the first descriptor is more sensitive to the number of training samples than the latter, therefore, this descriptor is indicated for the problem of rotated samples classification, where it is possible to work with resolution from 32x32 pixels while maximizing the number of samples used for training.

ABSTRACT POSSIBILISTIC INTERPRETATION OF MISTUNING IN BLADED DISKS BY FUZZY ALGEBRA Karatas, Hamit Ç
aglar M.S., Department of Mechanical Engineering Supervisor: Prof. Dr. H. Nevzat Ö
zgü
ven Co-supervisor: Asst. Prof. Dr. Ender Cigeroglu September 2012, 103 pages This study aims to define the possibilistic interpretation of mistuning and examine the way of determining the worst case situations and assessing reliability value to that case by using possibilistic methods. Furthermore, in this study, benefits of using possibilistic interpretation of mistuning in comparison to probabilistic interpretation of mistuning are investigated. For the possibilistic analysis of mistuned structures, uncertain mistuning parameters are modeled as fuzzy variables possessing possibility distributions. In this study, alpha-cut representations of fuzzy numbers are used which makes fuzzy variables to be represented by interval numbers at each and every confidence level. The solution of fuzzy equations of motion is governed by fuzzy algebra methods. The bounds of the solution of the fuzzy equation of motion, i.e. fuzzy vibration responses of the mistuned structure, are determined by the extension principle of fuzzy functions. The performance of the method for possibilistic interpretation of mistuning is investigated by comparing it to the probabilistic methods both computational and accuracy wise. For the comparison study, two different optimization tools &ndash
genetic algorithm as the global optimization tool and constrained nonlinear minimization method as the gradient based optimization tool- are utilized in possibilistic analysis and they are compared to solutions of probabilistic methods resulted from Monte-Carlo method. The performances of all of the methods are tested on both a cyclically symmetric lumped parameter model and a realistic reduced order finite element model.

Orientador: Peter Sussner
Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Matemática Estatística e Computação Científica
Made available in DSpace on 2018-08-26T15:06:30Z (GMT). No. of bitstreams: 1 Bueno_FelipeRoberto_M.pdf: 1499339 bytes, checksum: 85b58d8b856fafa47974349e80c1729e (MD5) Previous issue date: 2015
Resumo: Perceptrons morfológicos (MPs) pertencem à classe de redes neurais morfológicas (MNNs). Estas redes representam uma classe de redes neurais artificiais que executam operações de morfologia matemática (MM) em cada nó, possivelmente seguido pela aplicação de uma função de ativação. Vale ressaltar que a morfologia matemática foi concebida como uma teoria para processamento e análise de objetos (imagens ou sinais), por meio de outros objetos chamados elementos estruturantes. Embora inicialmente desenvolvida para o processamento de imagens binárias e posteriormente estendida para o processamento de imagens em tons de cinza, a morfologia matemática pode ser conduzida de modo mais geral em uma estrutura de reticulados completos. Originalmente, as redes neurais morfológicas empregavam somente determinadas operações da morfologia matemática em tons de cinza, denominadas de erosão e dilatação em tons de cinza, segundo a abordagem umbra. Estas operações podem ser expressas em termos de produtos máximo aditivo e mínimo aditivo, definidos por meio de operações entre vetores ou matrizes, da álgebra minimax. Recentemente, as operações da morfologia matemática fuzzy surgiram como funções de agregação das redes neurais morfológicas. Neste caso, falamos em redes neurais morfológicas fuzzy. Perceptrons híbridos lineares/morfológicos fuzzy foram inicialmente projetados como uma generalização dos perceptrons lineares/morfológicos existentes, ou seja, os perceptrons lineares/morfológicos fuzzy podem ser definidos por uma combinação convexa de uma parte morfológica fuzzy e uma parte linear. Nesta dissertação de mestrado, introduzimos uma rede neural artificial alimentada adiante, representando um perceptron híbrido linear/morfológico fuzzy chamado F-DELP (do inglês fuzzy dilation/erosion/linear perceptron), que ainda não foi considerado na literatura de redes neurais. Seguindo as ideias de Pessoa e Maragos, aplicamos uma suavização adequada para superar a não-diferenciabilidade dos operadores de dilatação e erosão fuzzy utilizados no modelo F-DELP. Em seguida, o treinamento é realizado por intermédio de um algoritmo de retropropagação de erro tradicional. Desta forma, aplicamos o modelo F-DELP em alguns problemas de classificação conhecidos e comparamos seus resultados com os produzidos por outros classificadores
Abstract: Morphological perceptrons (MPs) belong to the class of morphological neural networks (MNNs). These MNNs represent a class of artificial neural networks that perform operations of mathematical morphology (MM) at every node, possibly followed by the application of an activation function. Recall that mathematical morphology was conceived as a theory for processing and analyzing objects (images or signals), by means of other objects called structuring elements. Although initially developed for binary image processing and later extended to gray-scale image processing, mathematical morphology can be conducted very generally in a complete lattice setting. Originally, morphological neural networks only employed certain operations of gray-scale mathematical morphology, namely gray-scale erosion and dilation according to the umbra approach. These operations can be expressed in terms of (additive maximum and additive minimum) matrix-vector products in minimax algebra. It was not until recently that operations of fuzzy mathematical morphology emerged as aggregation functions of morphological neural networks. In this case, we speak of fuzzy morphological neural networks. Hybrid fuzzy morphological/linear perceptrons was initially designed by generalizing existing morphological/linear perceptrons, in other words, fuzzy morphological/linear perceptrons can be defined by a convex combination of a fuzzy morphological part and a linear part. In this master's thesis, we introduce a feedforward artificial neural network representing a hybrid fuzzy morphological/linear perceptron called fuzzy dilation/erosion/linear perceptron (F-DELP), which has not yet been considered in the literature. Following Pessoa's and Maragos' ideas, we apply an appropriate smoothing to overcome the non-differentiability of the fuzzy dilation and erosion operators employed in the proposed F-DELP models. Then, training is achieved using a traditional backpropagation algorithm. Finally, we apply the F-DELP model to some well-known classification problems and compare the results with the ones produced by other classifiers
Mestrado
Matematica Aplicada
Mestre em Matemática Aplicada

Cette thèse a pour objet la mesure de déformations sans contact lors d'un essai de traction à l'aide de la méthode de corrélation d'images numériques DIC (Digital Image Correlation). Cette technologie utilise le repérage d'un motif aléatoire de tachetures pour mesurer avec précision les déplacements sur une surface donnée d'un objet subissant une déformation. Plus précisément, un algorithme DIC plus efficace a été formulé, appliqué et validé. La présente thèse comporte cinq parties consacrées au développement et à la validation du nouvel algorithme DIC: (a) la formulation mathématique et la programmation, (b) la vérification numérique, (c) la validation expérimentale, par essai de traction, en comparant les mesures DIC à celles obtenues par des jauges de déformation, (d) l'étude d'un procédé d'atomisation novateur pour générer de façon reproductible le motif de tachetures pour un repérage plus exact, et (e) l'analyse des sources d'erreur dans les mesures DIC. Plus précisément, l'algorithme DIC a servi à analyser, à titre d'exemple d'application, les propriétés mécaniques du polyméthyl métacrylate utilisé pour la reconstruction du squelette. Avec l'algorithme DIC, les images d'un objet sont acquises pendant la déformation de celui-ci. On applique ensuite des techniques d'optimisation non linéaire pour suivre le motif de tachetures à la surface des objets subissant une déformation en traction avant et après le déplacement. Ce procédé d'optimisation demande un choix de valeurs de déplacement initiales. Plus l'estimation de ces valeurs de déplacement initiales est juste, plus il y a de chances que la convergence du processus d'optimisation soit efficace. Ainsi, cette thèse présente une technique de traitement novatrice reposant sur une logique floue incluant aussi l'approximation des valeurs initiales du déplacement pour démarrer un processus itératif d'optimisation, ayant pour résultat une reproduction plus exacte et efficace des déplacements et des déformations. La formulation mathématique du nouvel algorithme a été développée et ensuite mise en œuvre avec succès dans le langage de programmation MATLAB. La vérification de l'algorithme a été faite à l'aide d'images de synthèse simu­lant des déplacements de corps rigides et des déformations de traction uniformes. Plus particulièrement, les images de déplacement simulaient (1) des déplacements de 0, 1 - 1 pixel en translation, (2) des angles de rotation de 0, 5 - 5°, et (3) de grandes déformations en traction de l'ordre de 5000 à 300000µE déformation, respectivement. Les processus de vérification ont démontré que le taux d'exactitude du nouvel algorithme DIC est supérieur à 99% en ce qui concerne les mesures des différents types et niveaux de déplacements simulés. Une validation expérimentale a été menée afin d'examiner l'efficacité de la nouvelle tech­nique dans des conditions réalistes. Des échantillons de PMMA normalisés, respectant la norme ASTM F3087, ont été produits, inspectés et soumis à une charge de traction jus­qu'à la rupture. La déformation de la surface des échantillons a été mesurée au moyen (a) du nouvel algorithme DIC, et (b) des techniques utilisant des jauges de déformation de type rosette. La force maximale moyenne et la limite de résistance mécanique des quatre échantillons étaient de 880 ± 110 N et 49 ± 7 MPa, respectivement. La limite moyenne de déformation mesurée par la jauge de déformation et provenant de l'algorithme DIC étaient de 15750±2570 et 19890±3790 µs déformation, respectivement. Des déformations d'un tel ordre sont courantes pour les matériaux polymériques, et jusqu'à maintenant, la technique DIC n'n’était pas développée pour faire des mesures de déformations aussi importantes. On a constaté que l'erreur relative de la mesure DIC, par rapport à la technique de la jauge de déformation, s'élevait à 26 ± 8%. Par ailleurs, le module de Young moyen et le coefficient de Poisson moyen mesurés en utilisant des jauges de déformations étaient de 3, 78 ± 0, 07 G Pa et 0, 37 ± 0, 02, alors qu'ils étaient de 3, 16 ± 0, 61 GPa et 0, 37 ± 0, 08, respectivement lorsque mesurés avec l'algorithme DIC. L'écart croissant entre les mesures de déformation DIC et celles obtenues au moyen de jauges de déformation est probablement lié à la dis­torsion graduelle du motif de tachetures à la surface des échantillons de traction. Par la suite, on a introduit un facteur de correction de 1, 27 afin de corriger l'erreur systématique dans les mesures de déformation provenant de l'algorithme DIC. La limite de déformation des mesures DIC a été rajustée à 15712±357 µs déformation avec un taux d'erreur moyen relatif de -0, 5 ± 7, 1 %, comparé aux déformations mesurées par la jauge de déformation. Le module de Young moyen et le coefficient moyen de Poisson de l'algorithme DIC et des mesures obtenues par la jauge de déformation ont par ailleurs été rajustés à 3, 8 ± 0, 4 GPa et 0, 368 ± 0, 025, respectivement. Au moyen d'un procédé d'atomisation, des taches de peinture ont été générées de façon reproductible sur la surface d'un objet. Une approche expérimentale de planification facto­rielle a été utilisée pour étudier le motif de tachetures (répartition et gradient de l'échelle des tons de gris) pour mesurer l'exactitude de l'algorithme DIC. Plus particulièrement, neuf motifs de tachetures différents ont été générés au moyen du procédé d'atomisation et testés pour la translation et la rotation de corps rigides. Les résultats ont révélé que l'erreur moyenne relative parmi les neuf motifs de tachetures variait de 1, 1 ± 0, 3% à -6, 5 ± 3, 6%. Le motif de tachetures préféré, lequel se démarquait par une large gamme de taches claires et de valeurs de tons de gris, a produit une erreur relative de 1, 1 ± 0, 3%. Une analyse des erreurs et des sources d'erreurs relatives de la mesure de l'algorithme DIC a été menée. Ti-ois catégories de sources d'erreurs, incluant l'algorithme lui-même, les paramètres du processus (taille des sous-ensembles, nombre de pixels calculés) et l'en­vironnement physique (uniformité des échantillons, motifs de tachetures, effet thermique de la caméra CCD et distorsion de la lentille, erreur de non-linéarité dans le circuit de la jauge de déformation) ont fait l'objet d'une étude et de discussions. Enfin, des solutions ont été amenées afin d'aider à réduire les erreurs systématiques et aléatoires en lien avec les trois catégories de sources d'erreurs susmentionnées. Pour terminer, un nouvel algorithme DIC permettant une approximation plus juste de l'estimation initiale, entraînant par conséquent une convergence efficace et précise de l'op­timisation a été développé, programmé, mis en oeuvre et vérifié avec succès pour ce qui est des déformations importantes. La validation expérimentale a fait ressortir une erreur systé­matique inattendue des mesures DIC lorsque comparées aux mesures obtenues au moyen de la technique des jauges de déformation. Plus l'échantillon se déformait, plus l'erreur augmentait proportionnellement. Par conséquent, la distorsion graduelle des tachetures sur la surface de l'objet était probablement la cause de l'erreur. L'erreur étant systéma­tique, elle a été corrigée. Le procédé d'atomisation a permis de générer des tachetures de façon reproductible sur la surface d'un objet. Grâce aux mesures DIC, le comportement mécanique des polymères soumis à des déformations importantes, comme le polyméthyl métacrylate servant à la reconstruction du squelette, peut être étudié et une fois maîtrisé, servir à l'élaboration de matériaux plus efficaces.
Abstract : The present thesis is focused on the non-contact and efficient strain measurement using the Digital Image Correlation (DIC) method, which employs the tracking of random speckle pattern for accurate measurement of displacements on a surface of an object undergoing deformation. Specifically, a more efficient DIC algorithm was successfully developed, implemented, and validated. This thesis consists of five parts related to the novel DIC algorithm: (a) the development and implementation, (b) the numerical verification, (c) the experimental validation, for tensile loading, by comparing to the deformation measurements using the strain gauge technique, (d) the investigation of a novel atomization process to reproducibly generate the speckle pattern for accurate tracking, and (e) the analysis of the error sources in the DIC measurements. Specifically, the DIC algorithm was used to exemplarily examine the mechanical properties of polymethyl methacrylate (PMMA) used in skeletal reconstruction. In the DIC algorithm, images of an object are captured as it deforms. Nonlinear optimization techniques are then used to correlate the speckle on the surface of the objects before and after the displacement. This optimization process includes a choice of suitable initial displacement values. The more accurate the estimation of these initial displacement values are, the more likely and the more efficient the convergence of the optimization process is. The thesis introduced a novel, fuzzy logics based processing technique, approximation of the initial values of the displacement for initializing iterative optimization, which more accurately and efficiently renders the displacements and deformations as results. The mathematical formulation of the novel algorithm was developed and then successfully implemented into MATLAB programming language. The algorithmic verification was performed using computer-generated images simulating rigid body displacements and uniform tensile deformations. Specifically, the rigid motion images simulated (1) displacements of 0.1-1 pixel for the rigid body translation, (2) rotation angles of 0.5-5 ̊ for rigid body rotation and (3) large tensile deformations of 5000-300000µɛ, respectively. The verification processes showed that the accuracy of the novel DIC algorithm, for the simulated displacement types and levels above 99%. The experimental validation was conducted to examine the effectiveness of the novel technique under realistic testing conditions. Normalized PMMA specimens, in accordance to ASTM F3087, were produced, inspected and subjected to tensile loading until failure. The deformation of the specimen surface was measured using (a) the novel DIC, and (b) strain gauge rosette techniques. The mean maximum force and ultimate strength of four specimens were 882.2±108.3 N and 49.3±6.2 MPa, respectively. The mean ultimate deformation from the gauge and DIC groups were 15746±2567µɛ and 19887±3790µɛ, respectively. These large deformations are common in polymeric materials, and the DIC technique has thus far not been investigated for large deformation. The relative mean error of the DIC measurement, in reference to those of the strain gauge technique, was found to be up to 26.0±7.1%. Accordingly, the mean Young's modulus and Poisson's ratio of strain gauge measurement were 3.78±0.07 GPa and 0.374±0.02, and of the DIC measurements were 3.16±0.61 GPa and 0.373±0.08, respectively. The increasing difference of the DIC strain measurements relative to those of the strain gauge technique is likely related to the gradual distortion of the speckle pattern on the surface of the tensile specimen. Subsequently, a Correction Factor (CF) of 1.27 was introduced to correct for the systematic error in the deformation measurements of the DIC group. The corrected ultimate deformation of the DIC measurements became 15712±357µɛ with the relative mean error of -0.5±7.1%, if compared to those measurements of the strain gauge techniques. Correspondingly, the mean Young's Modulus and Poisson's ratio of the DIC and of the strain gauge measurements became 3.8±0.4 GPa and 0.368±0.025, respectively. Using an atomization process, paint speckles were reproducibly generated on the surface of an object. A factorial design of experiments was used to investigate the speckle pattern (grey value distribution and gradient) for the DIC measurement accuracy. Specifically, nine different speckle patterns were generated using the atomization process and tested for rigid body translation and rotation. The results showed the relative mean errors among the nine speckle patterns varied from 1.1±0.3% to -6.5±3.6%. The preferred speckle pattern, which was characterized by a wide range of sharp speckle and of grey values, produced a mean error of 1.1±0.3%. The analysis of errors and relating sources in the DIC measurement was conducted. Three categories of sources including algorithmic sources, processing parameters sources (subset size, number of pixels computed) and physical environment sources (specimen uniformity, speckle pattern, self-heating effect of the CCD camera and lens distortion of the camera, non-linearity error in strain gauge circuit) were investigated and discussed. Finally, the solutions were provided in order to help reduce the systematic and random errors relating to the aforementioned three categories of sources for errors. In conclusion, a novel DIC algorithm for a more accurate approximation of the initial guess and accordingly for an efficient and accurate convergence of the optimization was successfully formulated, developed, implemented and verified for relatively large deformations. The experimental validation surprisingly showed a systematic error of the DIC measurements, if compared to the measurements of the strain gauge technique. The larger the deformation applied to the specimen, the larger the error gradually became. Therefore, the gradual distortion of the speckles on the surface of the object was likely the underlying cause of the error. The error was systematic and therefore corrected. The atomization process allowed generating reproducible speckles on the surface of an object. Using the DIC measurements, the mechanical behavior of polymers, undergoing large deformations, such as polymethyl methacrylate used in skeletal reconstruction can be investigated and, once understood, the knowledge gained can help develop more effective materials.

Este trabalho desenvolve uma metodologia para a automação da análise morfológica de imagens de cromossomos humanos irradiados no reator nuclear IEA-R1 (localizado no Instituto de Pesquisas Energéticas e Nucleares, IPEN, em São Paulo, Brasil) e, portanto, sujeitos a aberrações morfológicas. Esta metodologia se propõe a auxiliar na identificação, caracterização e classificação de cromossomos pelo profissional citogeneticista. O desenvolvimento da metodologia inclui a elaboração de um aplicativo baseado em técnicas de inteligência artificial utilizando Lógica Fuzzy e técnicas de processamento de imagens. O aplicativo desenvolvido foi denominado de CHRIMAN e é composto de módulos que contêm etapas metodológicas que suprem aspectos importantes para a obtenção de uma análise automatizada. A primeira etapa é a padronização dos procedimentos de aquisição das imagens digitais bidimensionais de metáfases através do acoplamento de uma câmera fotográfica digital comercial comum à ocular do microscópio utilizado na análise metafásica convencional. A segunda etapa é relativa ao tratamento das imagens obtidas através da aplicação de filtros digitais, armazenamento e organização das informações tanto do conteúdo da imagem em si, como das características extraídas e selecionadas, para posterior utilização nos algoritmos de reconhecimento de padrões. A terceira etapa consiste na utilização do banco de imagens digitalizadas e informações extraídas e armazenadas para a identificação dos cromossomos, sua caracterização, contagem e posterior classificação. O acerto no reconhecimento das imagens cromossômicas é de 93,9%. Esta classificação é baseada nos padrões encontrados classicamente em Buckton [1973], e possibilita o auxílio ao geneticista no procedimento de análise dos cromossomos com diminuição do tempo de análise e criando condições para a inclusão deste método num sistema mais amplo de avaliação de danos causados às células pela exposição à radiação ionizante.
This work has implemented a methodology for automation of images analysis of chromosomes of human cells irradiated at IEA-R1 nuclear reactor (located at IPEN, São Paulo, Brazil), and therefore subject to morphological aberrations. This methodology intends to be a tool for helping cytogeneticists on identification, characterization and classification of chromosomal metaphasic analysis. The methodology development has included the creation of a software application based on artificial intelligence techniques using Fuzzy Logic combined with image processing techniques. The developed application was named CHRIMAN and is composed of modules that contain the methodological steps which are important requirements in order to achieve an automated analysis. The first step is the standardization of the bi-dimensional digital image acquisition procedure through coupling a simple digital camera to the ocular of the conventional metaphasic analysis microscope. Second step is related to the image treatment achieved through digital filters application; storing and organization of information obtained both from image content itself, and from selected extracted features, for further use on pattern recognition algorithms. The third step consists on characterizing, counting and classification of stored digital images and extracted features information. The accuracy in the recognition of chromosome images is 93.9%. This classification is based on classical standards obtained at Buckton [1973], and enables support to geneticist on chromosomic analysis procedure, decreasing analysis time, and creating conditions to include this method on a broader evaluation system on human cell damage due to ionizing radiation exposure.

Orientador: Jurandir Zullo Junior
Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Agricola
Made available in DSpace on 2018-08-05T08:07:53Z (GMT). No. of bitstreams: 1 Antunes_JoaoFranciscoGoncalves_M.pdf: 7524504 bytes, checksum: e36a3c933615dc4ef031bf119f6c09ff (MD5) Previous issue date: 2005
Resumo: A estimativa precisa com antecedência à época da colheita de áreas plantadas com culturas agrícolas, como a soja, é de fundamental importância para a economia brasileira. A previsão do escoamento e comercialização da produção agrícola é estratégica para o Brasil, pois estão diretamente relacionados com o planejamento, custos e preço. Com o recente avanço tecnológico na obtenção de dados por sensoriamento remoto orbital é possível melhorar a previsão de safras, diminuindo cada vez mais o nível de subjetividade. Embora designadas para fins meteorológicos, as imagens AVHRR-NOAA de elevada repetitividade temporal, têm sido utilizadas para o monitoramento agrícola. Porém, a sua baixa resolução espacial faz com que possa ocorrer a mistura espectral das classes de cobertura do solo dentro de um mesmo pixel e isso pode acarretar problemas de imprecisão na estimativa de área plantada de uma cultura agrícola. O objetivo principal do trabalho foi desenvolver uma metodologia de classificação automática com a aplicação de lógica fuzzy para o reconhecimento de padrões em imagens AVHRR-NOAA, utilizando índices de vegetação para estimar a área plantada de soja no nível sub-pixel. Para oito municípios produtores de soja da região oeste do Estado do Paraná, foi possível obter a estimativa de área no final de janeiro de 2004, com antecedência em relação à época da colheita, ao contrário dos levantamentos oficiais que se estendem até o final da safra, além de utilizarem dados subjetivos vindos do campo. As estimativas de área de soja baseadas em classificação fuzzy mostraram-se altamente correlacionadas com as estimativas de área de referência obtidas a partir da máscara de soja e por expansão direta, sendo um indicativo de boa precisão. E também apresentaram alta correlação, balizadas com as estimativas oficiais da SEAB/DERAL e do IBGE. Em ambas comparações, o nível de erro relativo geral foi aceitável. O sistema desenvolvido para processamento e geração de produtos das imagens AVHRR-NOAA mostrou-se uma ferramenta fundamental de infra-estrutura, por aliar automação e precisão a metodologia do trabalho
Abstract: An early accurate estimation of agricultural crop areas, such as soybean, is fundamental for the Brazilian economy. The draining forecast and the estimation of agricultural production commercialization are strategic to Brazil, since they are directly related to planning, costs and price. Recent technological progress of data acquisition from orbital remote sensing makes possible to improve harvest forecast, reducing more and more the level of subjectivity. Although designed for meteorological aims, the AVHRR-NOAA images of high temporal resolution, have been used for the crop monitoring. However, its low spatial resolution might cause the spectral mixture of the different land cover classes within the same pixel and it can lead to accuracy problems on crop area estimation. The main objective of the work was to develop an automatic classification methodology with the application of fuzzy logic for pattern recognition in AVHRR-NOAA images, using vegetation indices to estimate the soybean crop areas at sub-pixel level. For eight soybean producer counties in the West region of the Paraná State, it was possible to obtain the crop area estimation at the end of january 2004, prior to the harvest period, on the contrary of the official surveys that extend until the end of the harvest, besides using subjective data collected on the field. The soybean crop area estimation based on fuzzy classification showed to be highly correlated with the reference area estimation obtained from the soybean mask and by direct expansion, being an indicative of good accuracy. And also presented high correlation, marked out with the official estimations from SEAB/DERAL and IBGE. In both comparisons, the level of general relative error was acceptable. The system developed for processing and products generation of AVHRR-NOAA images had proved to be a fundamental infrastructure tool, due to its capacity to combine automation and accuracy to the work methodology
Mestrado
Planejamento e Desenvolvimento Rural Sustentável
Mestre em Engenharia Agrícola

Esta dissertação apresenta um sistema de indução de classificadores fuzzy. Ao invés de utilizar a abordagem tradicional de sistemas fuzzy baseados em regras, foi utilizado o modelo de Árvore de Padrões Fuzzy(APF), que é um modelo hierárquico, com uma estrutura baseada em árvores que possuem como nós internos operadores lógicos fuzzy e as folhas são compostas pela associação de termos fuzzy com os atributos de entrada. O classificador foi obtido sintetizando uma árvore para cada classe, esta árvore será uma descrição lógica da classe o que permite analisar e interpretar como é feita a classificação. O método de aprendizado originalmente concebido para a APF foi substituído pela Programação Genética Cartesiana com o intuito de explorar melhor o espaço de busca. O classificador APF foi comparado com as Máquinas de Vetores de Suporte, K-Vizinhos mais próximos, florestas aleatórias e outros métodos Fuzzy-Genéticos em diversas bases de dados do UCI Machine Learning Repository e observou-se que o classificador APF apresenta resultados competitivos. Ele também foi comparado com o método de aprendizado original e obteve resultados comparáveis com árvores mais compactas e com um menor número de avaliações.
This work presents a system for induction of fuzzy classifiers. Instead of the traditional fuzzy based rules, it was used a model called Fuzzy Pattern Trees (FPT), which is a hierarchical tree-based model, having as internal nodes, fuzzy logical operators and the leaves are composed of a combination of fuzzy terms with the input attributes. The classifier was obtained by creating a tree for each class, this tree will be a logic class description which allows the interpretation of the results. The learning method originally designed for FPT was replaced by Cartesian Genetic Programming in order to provide a better exploration of the search space. The FPT classifier was compared against Support Vector Machines, K Nearest Neighbour, Random Forests and others Fuzzy-Genetics methods on several datasets from the UCI Machine Learning Repository and it presented competitive results. It was also compared with Fuzzy Pattern trees generated by the former learning method and presented comparable results with smaller trees and a lower number of functions evaluations.