Spelling suggestions: "subject:"atatistical pattern recognition"" "subject:"estatistical pattern recognition""
Minimal kernal classifiers for pattern recognition problemsHooper, Richard January 1996 (has links)
No description available.
An approach to high performance image classifier design using a moving window principleHoque, Md. Sanaul January 2001 (has links)
No description available.
Cube technique for Nearest Neighbour(s) searchShehu, Usman Gulumbe January 2002 (has links)
No description available.
A study of the generalized eigenvalue decomposition in discriminant analysisZhu, Manli 12 September 2006 (has links)
No description available.
Predicting Failures and Estimating Duration of Remaining Service Life from Satellite TelemetryLosik, Len, Wahl, Sheila, Owen, Lewis 10 1900 (has links)
International Telemetering Conference Proceedings / October 28-31, 1996 / Town and Country Hotel and Convention Center, San Diego, California / This paper addresses research completed for predicting hardware failures and estimating remaining service life for satellite components using a Failure Prediction Process (FPP). It is a joint paper, presenting initial research completed at the University of California, Berkeley, Center for Extreme Ultraviolet (EUV) Astrophysics using telemetry from the EUV EXPLORER (EUVE) satellite and statistical computation analysis completed by Lockheed Martin. This work was used in identifying suspect "failure precursors." Lockheed Martin completed an exploration into the application of statistical pattern recognition methods to identify FPP events observed visually by the human expert. Both visual and statistical methods were successful in detecting suspect failure precursors. An estimate for remaining service life for each unit was made from the time the suspect failure precursor was identified. It was compared with the actual time the equipment remained operable. The long-term objective of this research is to develop a resident software module which can provide information on FPP events automatically, economically, and with high reliability for long-term management of spacecraft, aircraft, and ground equipment. Based on the detection of a Failure Prediction Process event, an estimate of remaining service life for the unit can be calculated and used as a basis to manage the failure.
Application of Support Vector Machines for Damage Detection in StructuresSharma, Siddharth 05 January 2009 (has links)
Support vector machines (SVMs) are a set of supervised learning methods that have recently been applied for structural damage detection due to their ability to form an accurate boundary from a small amount of training data. During training, they require data from the undamaged and damaged structure. The unavailability of data from the damaged structure is a major challenge in such methods due to the irreversibility of damage. Recent methods create data for the damaged structure from finite element models. In this thesis we propose a new method to derive the dataset representing the damage structure from the dataset measured on the undamaged structure without using a detailed structural finite element model. The basic idea is to reduce the values of a copy of the data from the undamaged structure to create the data representing the damaged structure. The performance of the method in the presence of measurement noise, ambient base excitation, wind loading is investigated. We find that SVMs can be used to detect small amounts of damage in the structure in the presence of noise. The ability of the method to detect damage at different locations in a structure and the effect of measurement location on the sensitivity of the method has been investigated. An online structural health monitoring method has also been proposed to use the SVM boundary, trained on data measured from the damaged structure, as an indicator of the structural health condition.
Analyzing symbols in architectural floor plans via traditional computer vision and deep learning approachesRezvanifar, Alireza 13 December 2021 (has links)
Architectural floor plans are scale-accurate 2D drawings of one level of a building, seen from above, which convey structural and semantic information related to rooms, walls, symbols, textual data, etc. They consist of lines, curves, symbols, and textual markings, showing the relationships between rooms and all physical features, required for the proper construction or renovation of the building. First, this thesis provides a thorough study of state-of-the-art on symbol spotting methods for architectural drawings, an application domain providing the document image analysis and graphic recognition communities with an interesting set of challenges linked to the sheer complexity and density of embedded information, that have yet to be resolved. Second, we propose a hybrid method that capitalizes on strengths of both vector-based and pixel-based symbol spotting techniques. In the description phase, the salient geometric constituents of a symbol are extracted by a variety of vectorization techniques, including a proposed voting-based algorithm for finding partial ellipses. This enables us to better handle local shape irregularities and boundary discontinuities, as well as partial occlusion and overlap. In the matching phase, the spatial relationship between the geometric primitives is encoded via a primitive-aware proximity graph. A statistical approach is then used to rapidly yield a coarse localization of symbols within the plan. Localization is further refined with a pixel-based step implementing a modified cross-correlation function. Experimental results on the public SESYD synthetic dataset and real-world images demonstrate that our approach clearly outperforms other popular symbol spotting approaches. Traditional on-the-fly symbol spotting methods are unable to address the semantic challenge of graphical notation variability, i.e. low intra-class symbol similarity, an issue that is particularly important in architectural floor plan analysis. The presence of occlusion and clutter, characteristic of real-world plans, along with a varying graphical symbol complexity from almost trivial to highly complex, also pose challenges to existing spotting methods. Third, we address all the above issues by leveraging recent advances in deep learning-based neural networks and adapting an object detection framework based on the YOLO (You Only Look Once) architecture. We propose a training strategy based on tiles, avoiding many issues particular to deep learning-based object detection networks related to the relatively small size of symbols compared to entire floor plans, aspect ratios, and data augmentation. Experimental results demonstrate that our method successfully detects architectural symbols with low intra-class similarity and of variable graphical complexity, even in the presence of heavy occlusion and clutter. / Graduate
Investigation Of Damage Detection Methodologies For Structural Health MonitoringGul, Mustafa 01 January 2009 (has links)
Structural Health Monitoring (SHM) is employed to track and evaluate damage and deterioration during regular operation as well as after extreme events for aerospace, mechanical and civil structures. A complete SHM system incorporates performance metrics, sensing, signal processing, data analysis, transmission and management for decision-making purposes. Damage detection in the context of SHM can be successful by employing a collection of robust and practical damage detection methodologies that can be used to identify, locate and quantify damage or, in general terms, changes in observable behavior. In this study, different damage detection methods are investigated for global condition assessment of structures. First, different parametric and non-parametric approaches are re-visited and further improved for damage detection using vibration data. Modal flexibility, modal curvature and un-scaled flexibility based on the dynamic properties that are obtained using Complex Mode Indicator Function (CMIF) are used as parametric damage features. Second, statistical pattern recognition approaches using time series modeling in conjunction with outlier detection are investigated as a non-parametric damage detection technique. Third, a novel methodology using ARX models (Auto-Regressive models with eXogenous output) is proposed for damage identification. By using this new methodology, it is shown that damage can be detected, located and quantified without the need of external loading information. Next, laboratory studies are conducted on different test structures with a number of different damage scenarios for the evaluation of the techniques in a comparative fashion. Finally, application of the methodologies to real life data is also presented along with the capabilities and limitations of each approach in light of analysis results of the laboratory and real life data.
Detecção de danos estruturais usando analise de series temporais e atuadores e sensores piezeletricos / Structural damage detection using time series analysis and piezoelectries actuators and sensorsSilva, Samuel da 14 February 2008 (has links)
Orientadores: Milton Dias Junior e Vicente Lopes Junior / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecanica / Made available in DSpace on 2018-08-10T04:58:56Z (GMT). No. of bitstreams: 1 Silva_Samuelda_D.pdf: 9025537 bytes, checksum: ac86884d08ba00adbf77aeac335e7acc (MD5) Previous issue date: 2008 / Resumo: A contribuição deste trabalho foi desenvolver uma metodologia para detecção e localização de danos considerando apenas respostas de deslocamento ou aceleração e medidas obtidas por atuadores e sensores piezelétricos (PZTs) distribuídos e colados em estruturas flexíveis. Modelos de filtros discretos do tipo auto-regressivos, como AR-ARX, ARMA e ARMAX, são usados para extrair um indicador de danos a partir dos erros de predição linear destes filtros. Investiga-se também o uso de séries discretas de Wiener/Volterra escritas com filtros de Kautz para obtenção de erros de predição não-lineares. Para classificar os erros de predição (lineares ou não-lineares) nas classes ¿sem dano¿ ou ¿com dano¿ comparou-se o uso de ferramentas não-supervisionadas de classificação de padrões estatísticos, como agrupamento fuzzy e controle estatístico de processos. Testes numéricos e experimentais foram realizados e os resultados alcançados com a metodologia desenvolvida apresentaram vantagens em relação aos métodos convencionais que são discutidas no decorrer do trabalho / Abstract: This work proposes a novel approach to detect and locate incipient damage in structures by using only acceleration responses and coupled piezoelectric actuators and sensors. Though the major focus in smart damage detection is given by on the monitoring of the electrical impedance in the frequency domain, the current contribution applies a novel technique based on time series analysis. Regressive models, such as AR-ARX, ARMA and ARMAX, are employed to extract a feature index using the linear prediction errors. The use of nonlinear prediction by using discrete-time Wiener/Volterra models expanded by Kautz filter is also investigated. In order to decide correctly whether damage exists or not, a set of unsurpervised statistical pattern recognition techniques, namely the fuzzy clustering and the statistical process control, are implemented. Several numerical and experimental tests are performed to illustrate and compare the methodology developed with classical approaches. The efficacy of the approach is demonstrated through these tests / Doutorado / Mecanica dos Sólidos e Projeto Mecanico / Doutor em Engenharia Mecânica
Reconhecimento de PadrÃes AtravÃs de AnÃlises EstatÃsticas e Fractais Aplicadas a Dados de Ensaios NÃo-Destrutivos / Pattern Recognition by Statistical Fluctuation and Fractal Analyses Applied to Nondestructive Testing DataFrancisco EstÃnio da Silva 19 December 2011 (has links)
FundaÃÃo de Amparo Ã Pesquisa do Estado do CearÃ / Conselho Nacional de Desenvolvimento CientÃfico e TecnolÃgico / In this work a procedure is studied for pattern classification related to different types of data, namely: (1) signals obtained from ultrasonic testing ( pulse-echo technique) and magnetic signals obtained from BarkhÃusen noise in samples of ferritic-pearlitic carbon steel tubes which, due to temperature effects, have shown microstructural changes as consequence of the total or partial transformation of the pearlite into spherodite; (2) images built from TOFD ultrasonic testing and 8 bit digital radiographic images obtained from carbon steel 1020 sheets, with different welding defects. From the data obtained, images have been considered with the defects as lack of fusion, lack of penetration, porosity and images without defect. For this aim, non-conventional mathematical techniques have been used for the preprocessing of the data, namely, the statistical analyses, Hurst analysis (RSA) and detrended fluctuation analysis (DFA), and fractal analyses, box counting analysis (BCA) and minimal cover analysis (MCA). The curves obtained with the initial mathematical treatment, discrete functions of the temporal window width, have been handled with the supervised and non-supervised pattern recognition techniques known as principal component analysis and Karhunen-LoÃve (KL) transformation analysis respectively. With respect to the magnetic signals, the KL classifier has been shown to be very efficient when applied to DFA obtained from the magnetic flux, with a success rate around 94%. On the other hand, for the magnetic noise signals we have not obtained an acceptable success rate independently of the pre-processing used. However, when were considered the curves obtained by concatenating all curves of the pre-processing was obtained a consistent average success rate of 85%. As far as the rate of success of the PCA classifier is concerned, an excellent success of 96% has been reached for concatenated curves of selected data of magnetic noise only. As far as the analyses of the backscattered ultrasonic signals is concerned, it was not possible to classify the different stages of the microstructural degradation by using KL or PCA independently of the pre-processing used. As far as the analyses of the D-scan images are concerned, by applying PCA a rate of success of 81% has been obtained with MCA data, 73% has been obtained by concatenating all curves from the different fractal and statistical analyses and around 85% when concatenating the best individual results (DFA and MCA). On the other hand, considering the KL classifier, high success rates have been verified for the training stage, between 96% and 99%, and a maximum success rate (100%), when concatenating all analyses. With respect to the testing results, the best success rate which has been reached was approximately 77%, when concatenating all the curves obtained from the statistical and fractal pre-processing. For the digitalized radiographic images, relevant individual rates of success (between 70% and 90%) for the training set (consisting of all data) have been obtained for the classifier KL only, and a 100% success rate, when concatenating all the curves obtained from the pre-processing of the images. / Neste trabalho estudou-se uma metodologia de classificaÃÃo de padrÃes relacionados a dois tipos de dados: (1) sinais obtidos atravÃs dos ensaios ultrassÃnicos (tÃcnica pulso-eco) e sinais magnÃticos obtidos atravÃs de ruÃdo Barkhausen realizados em amostras de tubos de aÃo carbono ferrÃtico-perlÃtico que devido aos efeitos da temperatura de trabalho apresentaram mudanÃas microestruturais decorrentes da transformaÃÃo parcial ou total da perlita em esferoiditas; e (2) imagens construÃdas a partir de ensaios ultrassÃnicos (tÃcnica TOFD) e imagens radiogrÃficas digitais de chapas de aÃo carbono 1020 soldadas, obtidas com resoluÃÃo de 8bits, nas quais foram inseridos diversos tipos de defeitos de soldagem. Dos dados gerados, foram estudadas as imagens com os defeitos de falta de fusÃo (FF), falta de penetraÃÃo (FP), porosidade (PO) e uma classe designada como sem defeito (SD). Para tanto, utilizaram-se de tÃcnicas matemÃticas nÃo convencionais no prÃ-processamentos dos dados conhecidas como anÃlises estatÃsticas de Hurst (RSA) e flutuaÃÃo sem tendÃncia (DFA) e as anÃlises fractais de contagem de caixas (BCA) e de mÃnima cobertura (MCA). Em seguida as curvas obtidas desse tratamento matemÃtico inicial, funÃÃes discretas da largura da janela temporal, foram utilizadas na alimentaÃÃo das tÃcnicas de reconhecimento de padrÃes nÃo supervisionada e supervisionada conhecidas, respectivamente, como anÃlise de componentes principais (PCA) e anÃlise da transformaÃÃo de Karhunen-LoÃve (KL). Em relaÃÃo aos estudos dos sinais magnÃticos, o classificador KL mostrou-se eficiente quando aplicado Ãs DFA do fluxo magnÃtico, com uma taxa de sucesso em torno de 94%. JÃ para os sinais do ruÃdo magnÃtico nÃo se obteve uma taxa de sucesso aceitÃvel, independente do prÃ-processamento utilizado. Entretanto quando todas as curvas de todas as anÃlises, dos dois tipos de sinais magnÃticos (ruÃdo e fluxo), foram concatenadas, obteve-se uma taxa mÃdia de sucesso consistente de aproximadamente 85%. No tocante Ãs taxas de sucesso do classificador PCA, somente para o ruÃdo magnÃtico e considerando todas as curvas concatenadas para um grupo de dados selecionados, conseguiu-se uma taxa de sucesso de 96%. A respeito das anÃlises dos sinais ultrassÃnicos retroespalhados, tambÃm nÃo foi possÃvel classificar, nem com a KL e nem com a PCA, os diferentes estÃgios de degradaÃÃo microestrutural, independemente do prÃ-processamento utilizado. No tocante Ãs analises das imagens D-scan, obteve-se com a PCA, taxas de sucesso de 81% considerando apenas os dados das MCA, 73% quando as curvas de todas as anÃlises estatÃsticas e fractais foram concatenadas, e em torno de 85%, quando apenas as curvas das melhores anÃlises (DFA e MCA) foram concatenadas. JÃ considerando o classificador KL, verificaram-se taxas de sucesso na etapa de treinamento, entre 96% e 99%, e mÃxima taxa de sucesso (100%) no caso dos vetores de todas as anÃlises concatenados. Em relaÃÃo aos resultados dos testes, a melhor taxa de sucesso alcanÃada foi aproximadamente de 77% quando se concatenaram todas as curvas oriundas dos prÃ-processamentos estatÃsticos e fractais. Com respeito Ãs imagens radiogrÃficas digitalizadas somente com o classificador KL (na etapa de treinamento, com 100% dos vetores) obtiveram-se taxas de sucesso individuais entre 70 e 90% de acertos e 100% de sucesso na classificaÃÃo quando se concatenaram as curvas de todos os prÃ-processamentos das imagens.
Page generated in 0.2198 seconds