Щоб переглянути інші типи публікацій з цієї теми, перейдіть за посиланням: Invariant pattern recognition.
Дисертації з теми "Invariant pattern recognition"
Оформте джерело за APA, MLA, Chicago, Harvard та іншими стилями
Ознайомтеся з топ-50 дисертацій для дослідження на тему "Invariant pattern recognition".
Біля кожної праці в переліку літератури доступна кнопка «Додати до бібліографії». Скористайтеся нею – і ми автоматично оформимо бібліографічне посилання на обрану працю в потрібному вам стилі цитування: APA, MLA, «Гарвард», «Чикаго», «Ванкувер» тощо.
Також ви можете завантажити повний текст наукової публікації у форматі «.pdf» та прочитати онлайн анотацію до роботи, якщо відповідні параметри наявні в метаданих.
Переглядайте дисертації для різних дисциплін та оформлюйте правильно вашу бібліографію.
1
Elliffe, Martin C. M. "Neural networks for Invariant pattern recognition." Thesis, University of Oxford, 1998. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.302530.
Повний текст джерела
2
Reed, Stuart. "Cascaded linear shift invariant processing in pattern recognition." Thesis, Loughborough University, 2000. https://dspace.lboro.ac.uk/2134/7481.
Повний текст джерелаАнотація:
Image recognition is the process of classifying a pattern in an image into one of a number of stored classes. It is used in such diverse applications as medical screening, quality control in manufacture and military target recognition. An image recognition system is called shift invariant if a shift of the pattern in the input image produces a proportional shift in the output, meaning that both the class and location of the object in the image are identified. The work presented in this thesis considers a cascade of linear shift invariant optical processors, or correlators, separated by fields of point non-lineari ties, called the cascaded correlator. This is introduced as a method of providing parallel, shiftinvariant, non-linear pattern recognition in a system that can learn in the manner of neural networks. It is shown that if a neural network is constrained to give overall shift invariance, the resulting structure is a cascade of correlators, meaning that the cascaded correlator is the only architecture which will provide fully shift invariant pattern recognition. The issues of training of such a non-linear system are discussed in neural network terms, and the non-linear decisions of the system are investigated. By considering digital simulations of a two-stage system, it is shown that the cascaded correlator is superior to linear filtering for both discrimination and tolerance to image distortion. This is shown for theoretical images and in real-world applications based on fault identification in can manufacture. The cascaded correlator has also been proven as an optical system by implementation in a joint transform correlator architecture. By comparing simulated and optical results, the resulting practical errors are analysed and compensated. It is shown that the optical implementation produces results similar to those of the simulated system, meaning that it is possible to provide a highly non-linear decision using robust parallel optical processing techniques.
3
Chan, Lai-Wan. "Adaptive and invariant connectionist models for pattern recognition." Thesis, University of Cambridge, 1989. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.238206.
Повний текст джерела
4
Li, Duwang. "Invariant pattern recognition algorithm using the Hough Transform." PDXScholar, 1989. https://pdxscholar.library.pdx.edu/open_access_etds/3899.
Повний текст джерелаАнотація:
A new algorithm is proposed which uses the Hough Transform to recognize two dimensional objects independent of their orientations, sizes and locations. The binary image of an object is represented by a set of straight lines. Features of the straight lines, namely the lengths and the angles of their normals, their lengths and the end point positions are extracted using the Hough Transform. A data structure for the extracted lines is constructed so that it is efficient to match the features of the lines of one object to those of another object, and determine if one object is a rotated and/or scaled version of the other. Finally a generalized Hough Transform is used to match the end points of the two sets of lines. The simulation experiments show good results for objects with significant linear features .
5
Tonge, Ashwini Kishor. "Object Recognition Using Scale-Invariant Chordiogram." Thesis, University of North Texas, 2017. https://digital.library.unt.edu/ark:/67531/metadc984116/.
Повний текст джерелаАнотація:
This thesis describes an approach for object recognition using the chordiogram shape-based descriptor. Global shape representations are highly susceptible to clutter generated due to the background or other irrelevant objects in real-world images. To overcome the problem, we aim to extract precise object shape using superpixel segmentation, perceptual grouping, and connected components. The employed shape descriptor chordiogram is based on geometric relationships of chords generated from the pairs of boundary points of an object. The chordiogram descriptor applies holistic properties of the shape and also proven suitable for object detection and digit recognition mechanisms. Additionally, it is translation invariant and robust to shape deformations. In spite of such excellent properties, chordiogram is not scale-invariant. To this end, we propose scale invariant chordiogram descriptors and intend to achieve a similar performance before and after applying scale invariance. Our experiments show that we achieve similar performance with and without scale invariance for silhouettes and real world object images. We also show experiments at different scales to confirm that we obtain scale invariance for chordiogram.
6
Ojansivu, V. (Ville). "Blur invariant pattern recognition and registration in the Fourier domain." Doctoral thesis, University of Oulu, 2009. http://urn.fi/urn:isbn:9789514292552.
Повний текст джерелаАнотація:
Abstract
Pattern recognition and registration are integral elements of computer vision, which considers image patterns. This thesis presents novel blur, and combined blur and geometric invariant features for pattern recognition and registration related to images. These global or local features are based on the Fourier transform phase, and are invariant or insensitive to image blurring with a centrally symmetric point spread function which can result, for example, from linear motion or out of focus.
The global features are based on the even powers of the phase-only discrete Fourier spectrum or bispectrum of an image and are invariant to centrally symmetric blur. These global features are used for object recognition and image registration. The features are extended for geometrical invariances up to similarity transformation: shift invariance is obtained using bispectrum, and rotation-scale invariance using log-polar mapping of bispectrum slices. Affine invariance can be achieved as well using rotated sets of the log-log mapped bispectrum slices. The novel invariants are shown to be more robust to additive noise than the earlier blur, and combined blur and geometric invariants based on image moments.
The local features are computed using the short term Fourier transform in local windows around the points of interest. Only the lowest horizontal, vertical, and diagonal frequency coefficients are used, the phase of which is insensitive to centrally symmetric blur. The phases of these four frequency coefficients are quantized and used to form a descriptor code for the local region. When these local descriptors are used for texture classification, they are computed for every pixel, and added up to a histogram which describes the local pattern. There are no earlier textures features which have been claimed to be invariant to blur. The proposed descriptors were superior in the classification of blurred textures compared to a few non-blur invariant state of the art texture classification methods.
7
Rahtu, E. (Esa). "A multiscale framework for affine invariant pattern recognition and registration." Doctoral thesis, University of Oulu, 2007. http://urn.fi/urn:isbn:9789514286018.
Повний текст джерелаАнотація:
Abstract
This thesis presents a multiscale framework for the construction of affine invariant pattern recognition and registration methods. The idea in the introduced approach is to extend the given pattern to a set of affine covariant versions, each carrying slightly different information, and then to apply known affine invariants to each of them separately. The key part of the framework is the construction of the affine covariant set, and this is done by combining several scaled representations of the original pattern. The advantages compared to previous approaches include the possibility of many variations and the inclusion of spatial information on the patterns in the features.
The application of the multiscale framework is demonstrated by constructing several new affine invariant methods using different preprocessing techniques, combination schemes, and final recognition and registration approaches. The techniques introduced are briefly described from the perspective of the multiscale framework, and further treatment and properties are presented in the corresponding original publications. The theoretical discussion is supported by several experiments where the new methods are compared to existing approaches.
In this thesis the patterns are assumed to be gray scale images, since this is the main application where affine relations arise. Nevertheless, multiscale methods can also be applied to other kinds of patterns where an affine relation is present.
An additional application of one multiscale based technique in convexity measurements is introduced. The method, called multiscale autoconvolution, can be used to build a convexity measure which is a descriptor of object shape. The proposed measure has two special features compared to existing approaches. It can be applied directly to gray scale images approximating binary objects, and it can be easily modified to produce a number of measures. The new measure is shown to be straightforward to evaluate for a given shape, and it performs well in the applications, as demonstrated by the experiments in the original paper.
8
Woo, Myung Chul. "Biologically-inspired translation, scale, and rotation invariant object recognition models /." Online version of thesis, 2007. http://hdl.handle.net/1850/3933.
Повний текст джерела
9
Mertzanis, Emmanouel Christopher. "A new neural network based approach to position and scale invariant pattern recognition." Thesis, University of York, 1992. http://etheses.whiterose.ac.uk/10897/.
Повний текст джерела
10
Heck, Larry Paul. "A subspace approach to the auomatic design of pattern recognition systems for mechanical system monitoring." Diss., Georgia Institute of Technology, 1991. http://hdl.handle.net/1853/15016.
Повний текст джерела
11
陳浩邦 and Hau-bang Bernard Chan. "Matching patterns of line segments by affine-invariant area features." Thesis, The University of Hong Kong (Pokfulam, Hong Kong), 2002. http://hub.hku.hk/bib/B31225652.
Повний текст джерела
12
Chan, Hau-bang Bernard. "Matching patterns of line segments by affine-invariant area features /." Hong Kong : University of Hong Kong, 2002. http://sunzi.lib.hku.hk/hkuto/record.jsp?B25151319.
Повний текст джерела
13
Li, Yue. "Active Vision through Invariant Representations and Saccade Movements." Ohio University / OhioLINK, 2006. http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1149389174.
Повний текст джерела
14
Cui, Chen. "Adaptive weighted local textural features for illumination, expression and occlusion invariant face recognition." University of Dayton / OhioLINK, 2013. http://rave.ohiolink.edu/etdc/view?acc_num=dayton1374782158.
Повний текст джерела
15
Pratt, John Graham le Maistre. "Application of the Fourier-Mellin transform to translation-, rotation- and scale-invariant plant leaf identification." Thesis, McGill University, 2000. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=33440.
Повний текст джерелаАнотація:
The Fourier-Mellin transform was implemented on a digital computer and applied towards the recognition and differentiation of images of plant leaves regardless of translation, rotation or scale. Translated, rotated and scaled leaf images from seven species of plants were compared: avocado ( Persea americana), trembling aspen (Populus tremuloides), lamb's-quarter (Chenopodium album), linden (Tilla americana), silver maple (Acer saccharinum), plantain (Plantago major) and sumac leaflets (Rhus typhina ). The rate of recognition was high among translated and rotated leaf images for all plant species. The rates of recognition and differentiation were poor, however, among scaled leaf images and between leaves of different species. Improvements to increase the effectiveness of the algorithm are suggested.
16
Nelson, Eric D. "Zoom techniques for achieving scale invariant object tracking in real-time active vision systems /." Online version of the thesis, 2006. https://ritdml.rit.edu/dspace/handle/1850/2620.
Повний текст джерела
17
Heywood, M. I. "A practical framework for training sigma-pi neural networks with an application in rotation invariant pattern recognition." Thesis, University of Essex, 1994. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.241204.
Повний текст джерела
18
Voils, Danny. "Scale Invariant Object Recognition Using Cortical Computational Models and a Robotic Platform." PDXScholar, 2012. https://pdxscholar.library.pdx.edu/open_access_etds/632.
Повний текст джерелаАнотація:
This paper proposes an end-to-end, scale invariant, visual object recognition system, composed of computational components that mimic the cortex in the brain. The system uses a two stage process. The first stage is a filter that extracts scale invariant features from the visual field. The second stage uses inference based spacio-temporal analysis of these features to identify objects in the visual field. The proposed model combines Numenta's Hierarchical Temporal Memory (HTM), with HMAX developed by MIT's Brain and Cognitive Science Department. While these two biologically inspired paradigms are based on what is known about the visual cortex, HTM and HMAX tackle the overall object recognition problem from different directions. Image pyramid based methods like HMAX make explicit use of scale, but have no sense of time. HTM, on the other hand, only indirectly tackles scale, but makes explicit use of time. By combining HTM and HMAX, both scale and time are addressed. In this paper, I show that HTM and HMAX can be combined to make a com- plete cortex inspired object recognition model that explicitly uses both scale and time to recognize objects in temporal sequences of images. Additionally, through experimentation, I examine several variations of HMAX and its
19
Krasilenko, V. G., O. I. Nikolskyy, A. A. Lazarev, D. V. Nikitovich, В. Г. Красіленко, О. І. Нікольський, О. О. Лазарєв, and Д. В. Нікітович. "Simulating optical pattern recognition algorithms for object tracking based on nonlinear models and subtraction of frames." Thesis, Український державний хіміко-технологічний університет, 2015. http://ir.lib.vntu.edu.ua//handle/123456789/23850.
Повний текст джерелаАнотація:
We have proposed and discussed optical pattern recognition algorithms for object
tracking based on nonlinear equivalence models and subtraction of frames. Experimental results of
suggested algorithms in Mathcad and LabVIEW are shown. Application of equivalent functions and
difference of frames gives good results for recognition and tracking moving objects
20
Mathew, Alex. "Rotation Invariant Histogram Features for Object Detection and Tracking in Aerial Imagery." University of Dayton / OhioLINK, 2014. http://rave.ohiolink.edu/etdc/view?acc_num=dayton1397662849.
Повний текст джерела
21
Фенько, Альона Дмитрівна. "Web-система аналізу медичних зображень". Bachelor's thesis, КПІ ім. Ігоря Сікорського, 2021. https://ela.kpi.ua/handle/123456789/44037.
Повний текст джерелаАнотація:
Дана дипломна робота присвячена розробці web-системи по обробці медичних зображень.
Метою роботи є створення web-системи для обробки медичних зображень та виявлення і класифікація метастазів раку молочної залози на повних слайд-зображеннях гістологічних зрізів лімфатичних вузлів.
Для досягнення мети були вирішені наступні задачі:
1.Проведено аналіз нових та вже існуючих алгоритмів для автоматичного виявлення і класифікації метастазів раку молочної залози на повних слайд-зображеннях гістологічних зрізів лімфатичних вузлів.
2.Проведено порівняльний аналіз основних архітектур нейронних мережей.
3.Навчання нейронної мережі.
4.Розробка інтерфейсу.
5.Тестування роботи веб-системи.This thesis is devoted to the development of a web-system for the processing of medical images. The purpose of the work is to create a web-system for processing medical images and to identify and classify breast cancer metastases in full slide-image of histological sections of the lymph nodes. To achieve the goal, the following tasks were solved: 1. An analysis of new and existing algorithms for the automatic detection and classification of breast cancer metastases is performed on complete slide images of histological sections of lymph nodes. 2.A comparative analysis of the basic architectures of neural networks is carried out. 3.Teaching the neural network. 4.Development of the interface. 5.Testing the operation of the web system.
Данная дипломная работа посвящена разработке web-системы по обработке медицинских изображений. Целью работы является создание web-системы для обработки медицинских изображений и обнаружения и классификация метастазов рака молочной железы на полных слайд-изображениях гистологических срезов лимфатических узлов. Для достижения цели были решены следующие задачи: 1. Проведен анализ новых и уже существующих алгоритмов для автоматического обнаружения и классификации метастазов рака молочной железы на полных слайд-изображениях гистологических срезов лимфатических узлов. 2.Проведен сравнительный анализ основных архитектур нейронных сетей. 3.Обучение нейронной сети. 4.Разработка интерфейса. 5.Тестирование работы веб-системы.
22
Shao, Yuan. "Higher order spectra invariants for shape pattern recognition." Ohio : Ohio University, 2000. http://www.ohiolink.edu/etd/view.cgi?ohiou1179949998.
Повний текст джерела
23
Hanson, Adam. "Character recognition of optically blurred textual images using moment invariants /." Online version of thesis, 1993. http://hdl.handle.net/1850/11748.
Повний текст джерела
24
Yuen, P. C. "Multi-scale representation and recognition of three dimensional surfaces using geometric invariants." Thesis, University of Surrey, 2001. http://epubs.surrey.ac.uk/979/.
Повний текст джерела
25
Hoang, Thai V. "Image Representations for Pattern Recognition." Phd thesis, Université Nancy II, 2011. http://tel.archives-ouvertes.fr/tel-00714651.
Повний текст джерелаАнотація:
La pertinence d'une application de traitement de signal relève notamment du choix d'une "représentation adéquate''. Par exemple, pour la reconnaissance de formes, la représentation doit mettre en évidence les propriétés salientes d'un signal; en débruitage, permettre de séparer le signal du bruit; ou encore en compression, de synthétiser fidèlement le signal d'entrée à l'aide d'un nombre réduit de coefficients. Bien que les finalités de ces quelques traitements soient distinctes, il apparait clairement que le choix de la représentation impacte sur les performances obtenues. La représentation d'un signal implique la conception d'un ensemble génératif de signaux élémentaires, aussi appelé dictionnaire ou atomes, utilisé pour décomposer ce signal. Pendant de nombreuses années, la conception de dictionnaire a suscité un vif intérêt des chercheurs dans des domaines applicatifs variés: la transformée de Fourier a été employée pour résoudre l'équation de la chaleur; celle de Radon pour les problèmes de reconstruction; la transformée en ondelette a été introduite pour des signaux monodimensionnels présentant un nombre fini de discontinuités; la transformée en contourlet a été conçue pour représenter efficacement les signaux bidimensionnels composées de régions d'intensité homogène, à frontières lisses, etc. Jusqu'à présent, les dictionnaires existants peuvent être regroupés en deux familles d'approches: celles s'appuyant sur des modèles mathématiques de données et celles concernant l'ensemble de réalisations des données. Les dictionnaires de la première famille sont caractérisés par une formulation analytique. Les coefficients obtenus dans de telles représentations d'un signal correspondent à une transformée du signal, qui peuvent parfois être implémentée rapidement. Les dictionnaires de la seconde famille, qui sont fréquemment des dictionnaires surcomplets, offrent une grande flexibilité et permettent d'être adaptés aux traitements de données spécifiques. Ils sont le fruit de travaux plus récents pour lesquels les dictionnaires sont générés à partir des données en vue de la représentation de ces dernières. L'existence d'une multitude de dictionnaires conduit naturellement au problème de la sélection du meilleur d'entre eux pour la représentation de signaux dans un cadre applicatif donné. Ce choix doit être effectué en vertu des spécificités bénéfiques validées par les applications envisagées. En d'autres termes, c'est l'usage qui conduit à privilégier un dictionnaire. Dans ce manuscrit, trois types de dictionnaire, correspondant à autant de types de transformées/représentations, sont étudiés en vue de leur utilisation en analyse d'images et en reconnaissance de formes. Ces dictionnaires sont la transformée de Radon, les moments basés sur le disque unitaire et les représentations parcimonieuses. Les deux premiers dictionnaires sont employés pour la reconnaissance de formes invariantes tandis que la représentation parcimonieuse l'est pour des problèmes de débruitage, de séparation des sources d'information et de classification. Cette thèse présentent des contributions théoriques validées par de nombreux résultats expérimentaux. Concernant la transformée de Radon, des pistes sont proposées afin d'obtenir des descripteurs de formes invariants, et conduisent à définir deux descripteurs invariants aux rotations, l'échelle et la translation. Concernant les moments basés sur le disque unitaire, nous formalisons les stratégies conduisant à l'obtention de moments orthogonaux. C'est ainsi que quatre moments harmoniques polaires génériques et des stratégies pour leurs calculs rapides sont introduits. Enfin, concernant les représentations parcimonieuses, nous proposons et validons un formalisme de représentation permettant de combiner les trois critères suivant : la parcimonie, l'erreur de reconstruction ainsi que le pouvoir discriminant en classification.
26
Mazetti, Cristina Mônica Dornelas. "Metodologia para extração de características invariantes à rotação em imagens de impressões digitais." Universidade de São Paulo, 2006. http://www.teses.usp.br/teses/disponiveis/18/18133/tde-03032007-085126/.
Повний текст джерелаАнотація:
O objetivo deste trabalho é apresentar algoritmos aplicados para extração de características invariantes à rotação em imagens de impressões digitais. No pré-processamento da imagem utiliza-se detecção de bordas pelo detector de Canny tendo como resultado uma imagem binarizada e afinada. Na extração das minúcias a metodologia adotada é o número de cruzamentos (CN), que extrai os aspectos locais, tais como, as minúcias fim de linha e bifurcações. A direção das cristas locais não é utilizada porque nas imagens rotacionadas a condição de permanência das propriedades biométricas não são satisfeitas. A comparação das impressões digitais utiliza os vetores gerados pela extração de minúcias considerando a posição (x,y) da minúcia armazenada em um vetor por tipo de minúcia (um vetor para crista final e outro vetor para crista bifurcada) e calculando a distância Euclidiana dessa posição (x,y) ao centro de massa da distribuição de minúcias para cada tipo de minúcia. Assim, as duas imagens são similares quando a distância Euclidiana entre os vetores de cada imagem e por tipo de minúcia forem mínimas. São discutidas as limitações de outros trabalhos existentes envolvendo rotação, translação e distorção da imagem de impressão digital, mostrando que os poucos trabalhos que tratam o problema possuem resultados não satisfatórios. Os maiores problemas ocorridos foram a extração de minúcias espúrias, mas foram resolvidos com os métodos sugeridos por Dixon (1979), tendo resultados satisfatórios em duas metodologias. No método média, a precisão para encontrar uma imagem foi de 100%, duas imagens 97,32%, três imagens 92,35%, quatro imagens 86,41% e cinco imagens 71,86%. E no método normal, a precisão para encontrar uma imagem foi de 100%, duas imagens 99,20%, três imagens 96,95%, quatro imagens 94,00% e cinco imagens 76,43%.The objective of this research is to present algorithms that can be applied in fingerprints images in order to extract certain features, which are invariant to an likely rotation in the given image. In the preprocessing stage, the Canny border detector is used, resulting in a binary, fine tuned image. For the minutiae extraction, the crossing number method is used, which extracts local aspects such as minutiae endings and bifurcations. The direction of local ridges is ignored because, in rotated images, the permanence conditions of the biometric attributes are not fulfilled. The process of matching fingerprints uses two arrays (one for ridge endings and the other for bifurcations), which are generated by the extraction of the minutiae, considering the (x,y) coordinate of the given minutiae stored in the arrays, and calculating its Euclidian distance relating to the center of mass of the minutiae distribution, for each of its types (ending or bifurcation). Thus, both images are similar when the Euclidian distance between the arrays of each image, distinct by the type of each minutiae, is minimal. The limitations of other pieces of research works concerning fingerprint image rotation, translation and distortion are discussed, indicating that the only few ones that deal with these kinds of problems give unsatisfactory results.
27
Koukoulas, Triantafillos. "Invariance analysis and performance assessment of pattern recognition filters for hybrid optical/digital correlator configurations." Thesis, University of Sussex, 2003. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.271772.
Повний текст джерела
28
Batog, Guillaume. "Problèmes classiques en vision par ordinateur et en géométrie algorithmique revisités via la géométrie des droites." Phd thesis, Université Nancy II, 2011. http://tel.archives-ouvertes.fr/tel-00653043.
Повний текст джерелаАнотація:
Systématiser: tel est le leitmotiv des résultats de cette thèse portant sur trois domaines d'étude en vision et en géométrie algorithmique. Dans le premier, nous étendons toute la machinerie du modèle sténopé des appareils photos classiques à un ensemble d'appareils photo (deux fentes, à balayage, oblique, une fente) jusqu'à présent étudiés séparément suivant différentes approches. Dans le deuxième, nous généralisons avec peu d'effort aux convexes de $\R^3$ l'étude des épinglages de droites ou de boules, menée différemment selon la nature des objets considérés. Dans le troisième, nous tentons de dégager une approche systématique pour élaborer des stratégies d'évaluation polynomiale de prédicats géométriques, les méthodes actuelles étant bien souvent spécifiques à chaque prédicat étudié. De tels objectifs ne peuvent être atteints sans un certain investissement mathématique dans l'étude des congruences linéaires de droites, des propriétés différentielles des ensembles de tangentes à des convexes et de la théorie des invariants algébriques, respectivement. Ces outils ou leurs utilisations reposent sur la géométrie des droites de $\p^3(\R)$, construite dans la seconde moitié du XIX\ieme{} siècle mais pas complètement assimilée en géométrie algorithmique et dont nous proposons une synthèse adaptée aux besoins de la communauté.
29
Maceček, Aleš. "Umělá neuronová síť RCE." Master's thesis, Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií, 2013. http://www.nusl.cz/ntk/nusl-220065.
Повний текст джерелаАнотація:
This paper is focused on an artificial neural network RCE, especially describing the topology, properties and learning algorithm of the network. This paper describes program uTeachRCE developed for learning the RCE network and program RCEin3D, which is created to visualize the RCE network in 3D space. The RCE network is compared with a multilayer neural network with a learning algorithm backpropagation in the practical application of recognition letters. For a descriptions of the letters were chosen moments invariant to rotation, translation and scaling image.
30
gundam, madhuri, and Madhuri Gundam. "Automatic Classification of Fish in Underwater Video; Pattern Matching - Affine Invariance and Beyond." ScholarWorks@UNO, 2015. http://scholarworks.uno.edu/td/1976.
Повний текст джерелаАнотація:
Underwater video is used by marine biologists to observe, identify, and quantify living marine resources. Video sequences are typically analyzed manually, which is a time consuming and laborious process. Automating this process will significantly save time and cost. This work proposes a technique for automatic fish classification in underwater video. The steps involved are background subtracting, fish region tracking and classification using features. The background processing is used to separate moving objects from their surrounding environment. Tracking associates multiple views of the same fish in consecutive frames. This step is especially important since recognizing and classifying one or a few of the views as a species of interest may allow labeling the sequence as that particular species. Shape features are extracted using Fourier descriptors from each object and are presented to nearest neighbor classifier for classification. Finally, the nearest neighbor classifier results are combined using a probabilistic-like framework to classify an entire sequence.
The majority of the existing pattern matching techniques focus on affine invariance, mainly because rotation, scale, translation and shear are common image transformations. However, in some situations, other transformations may be modeled as a small deformation on top of an affine transformation. The proposed algorithm complements the existing Fourier transform-based pattern matching methods in such a situation. First, the spatial domain pattern is decomposed into non-overlapping concentric circular rings with centers at the middle of the pattern. The Fourier transforms of the rings are computed, and are then mapped to polar domain. The algorithm assumes that the individual rings are rotated with respect to each other. The variable angles of rotation provide information about the directional features of the pattern. This angle of rotation is determined starting from the Fourier transform of the outermost ring and moving inwards to the innermost ring. Two different approaches, one using dynamic programming algorithm and second using a greedy algorithm, are used to determine the directional features of the pattern.
31
Costa, José Alfredo Ferreira. "Sistema de reconhecimento de padrões visuais invariante a transformações geométricas utilizando redes neurais artificiais de múltiplas camadas." Universidade de São Paulo, 1996. http://www.teses.usp.br/teses/disponiveis/18/18133/tde-23012018-135451/.
Повний текст джерелаАнотація:
As áreas de visão computacional e redes neurais artificiais (RNAs) e suas aplicações, tiveram um enorme progresso em pesquisa e aplicações práticas nos últimos anos. Sistemas de inspeção visual automática têm despertado muita atenção na indústria pois provêem meios econômicos, eficientes e precisos de obtenção de controle de qualidade. Porém, apesar do grande avanço tecnológico, a maioria dos sistemas existentes, com exceção de alguns poucos experimentais, são especializados e foram projetados para inspecionar um único objeto ou peça, de tipo previamente conhecido, e em posição, orientação e distância em relação à câmara altamente restritas. Este trabalho descreve um sistema de reconhecimento de imagens contendo múltiplos objetos de classes aleatórias e tolerante a ruído. Um estágio de pré-processamento filtra parte do ruído e segmenta regiões conectadas da imagem (RCI). A classificação dos padrões é feita com redes neurais de múltiplas camadas a partir de atributos invariantes calculados sobre as RCis. No final do processo temos uma listagem dos objetos contidos na cena, suas posições e orientações, os quais podem servir de entrada a um sistema de entendimento da cena, de mais alto nível, ou para outras máquinas, como um manipulador automático. Outros parâmetros podem ser utilizados para normalizar, em escala, orientação e posição, os padrões contidos na imagem, para efeito de comparações com imagens e parâmetros dos objetos previamente armazenados em bancos de dados. Dois métodos de treinamento de RNAs foram testados, o gradiente conjugado e o Levenberg-Marquardt, em conjunção com simulated annealing, para diferentes condições de erro e conjuntos de atributos. Imagens reais e sintéticas foram utilizadas para efeitos de testes de classificação correta e rejeição de padrões espúrios. Resultados são apresentados e comentados, destacando a capacidade de generalização do sistema mesmo com elevada degradação da imagem pelo ruído. Uma das vantagens do tipo de RNA empregado é a velocidade de execução, que permite ao sistema ser integrado a uma linha de montagem industrial. O sistema foi projetado com a utilização de recursos acessíveis e de baixo custo, sendo executado em computadores pessoais, e podendo ser facilmente adaptado para o uso em pequenas e médias empresas.Computer vision (CV) and artificial neural networks (ANN) are important research fields of artificial intelligence. Visual pattern recognition (VPR) and object recognition (2 or 3-D) are central tasks in a high level computer vision system. Despite the great development in the recent years, most of the current automatic visual inspection systems work with only one kind of pattern at time which has pose highly restricted. This dissertation describes a system designed to recognize patterns and objects in a digital image which have unknown number object types and poses. Such image, which is also degraded by noise, serve as input for the system. After gray level change and filtering, the pixel connected regions (CR) are codified, and the remained noise is eliminated. lnvariant features, i.e., moment invariants, serve as inputs for artificial neural networks that perform pattern classification. An interpretation module decode the net\'s outputs and increases the correct assignment by testing the net\'s higher outputs values. After all identified patterns were classified, we have an object listing of the scene, their positions and other information, which can be the input for a higher level scene understanding system, that may check for objects relations and could send information for humans or for other machines. Two ANN learning methods were adopted for training the networks, the conjugate gradient and the Levenberg-Marquardt Algoritms, both in conjuction with siumlated annealing, for different error conditions and feature sets. Sinthetic and real images were utilized for testing the net\'s correct class assignments and rejections. Results are presented as well as comments focusing the system\'s generalization capability despite noise, geometrical transformations, object shadows and other degradations over the images. One of the advantages of the ANN employed is the low execution time allowing the system to be integrated to an assembly industry line. The system runs on low cost personal computers, therefore it can be easily adapted for the Brazilian reality and can even be used by little companies and industries.
32
Vass, György. "Réseaux de neurones multicouches appliqués à la reconnaissance invariante des formes planes." Valenciennes, 1998. https://ged.uphf.fr/nuxeo/site/esupversions/2a32a089-db0c-4e22-8403-0399fd02eb59.
Повний текст джерелаАнотація:
Dans ce mémoire, nous avons proposé l'utilisation des descripteurs invariants et des réseaux de neurones multicouches pour la reconnaissance des formes contours planes. Le réseau de neurones est constitué d'une ou de plusieurs couches. L'algorithme de rétropropagation permet d'effectuer l'apprentissage. Nous avons montré que dans le cas d'un réseau de neurones à une seule couche cachée le choix du nombre des neurones peut être réalisé en utilisant un critère qui ne fait intervenir que les sorties des neurones cachés et les classes d'appartenance des observations. En fonction des formes planes considérées, nous avons proposé différents types de descripteurs invariants possédant des propriétés telles que la stabilité et la complétude. Dans une application telle que la reconnaissance des formes étoilées, l'utilisation des descripteurs de Fourier basés sur une représentation radiale permet au réseau de neurones d'apprendre les formes indépendamment des transformations qu'ils ont pu subir. Par ailleurs, nous avons montré que les invariants complets et stables peuvent être utilisés efficacement dans le cadre d'un apprentissage en mode non supervisé ou auto-associatif. Les résultats obtenus sont très encourageants et nous ont permis de voir la capacité d'un réseau de neurone à générer un codage interne. Cependant, dans le cas où les contours ne sont connus que partiellement, nous avons proposé une description locale invariante au groupe des similitudes et une architecture neuronale modulaire. Les résultats obtenus sont très encourageants. Nous pensons qu'ils peuvent être améliorés en choisissant un autre algorithme d'approximation polygonaleIn this thesis we propose the use of invariant descriptors and multilayer neural networks for the recognition of planar contour patterns. The neural network architecture may contain one or more layers. For training, the back-propagation algorithm is used. We have shown that in case of a single hidden layer work, the choice of the number of neurons can be realized by using a criterion where the only parameters involved are the hidden neuron outputs and the correct classes of the observations. In function of the planar shapes considered we propose different kind of invariant descriptors having the properties of stability and completeness. For applications like recognition of star-shaped patterns, the use of Fourier descriptors based on a radial representation permit the neural network to learn the patterns independently of the transformations they could undergo. Furthermore, we have shown that a complete and stable set of invariant descriptors can be effectively used in a nonsupervised, auto-associative type of training procedure. The obtained results are encouraging and allowed us to reveal the capability of the neural network to generate an internal coding scheme. However, for the case of contours only known partially we propose a local description technique which is invariant with the respect to the group of similarities, together with a modular neural architecture. The obtained results are encouraging. We believe that they can be improved by choosing a different polygonal approximation algorithm
33
Ernst, Jan [Verfasser]. "The Trace Model for Spatial Invariance with Applications in Structured Pattern Recognition, Image Patch Matching and Incremental Visual Tracking / Jan Ernst." Aachen : Shaker, 2014. http://d-nb.info/1060622025/34.
Повний текст джерела
34
Horák, Karel. "Aplikace metod rozpoznávání obrazu v defektoskopii." Doctoral thesis, Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií, 2008. http://www.nusl.cz/ntk/nusl-233438.
Повний текст джерелаАнотація:
A lot of production lines contain camera inspection systems that increase quality of production. Therefore this presented work deals with applications of computer image processing methods in defectoscopy. Concretely the thesis is concerned with defects evaluation of glass bottles in food operations by the help of visual system BTCAM612, which is in existing configuration installed inland and in several foreign countries. The system is developed in conjunction with developer company CAMEA Ltd. from Brno and it is its sole ownership. The whole process of bottles inspection is described in sequence. First of all it is the hardware acquisition of images of three main controlled parts of bottles – neck, bottom and side. Next chapters are concentrated on image processing and features classification. The features are obtained from image by methods based on detection of in-homogeneities on glass material. Essential part of work is focused on filtration of synthetic patterns from bottles bottoms using function of complex invariants. These patterns are occurred especially in many plants in eastern countries, where marketplace with inspection systems and generally with quality inspection of industrial lines is expanded lately.
35
López, Guillermo Ángel Pérez. "AFORAPRO: reconhecimento de objetos invariante sob transformações afins." Universidade de São Paulo, 2011. http://www.teses.usp.br/teses/disponiveis/3/3142/tde-31052011-155411/.
Повний текст джерелаАнотація:
Reconhecimento de objetos é uma aplicação básica da área de processamento de imagens e visão computacional. O procedimento comum do reconhecimento consiste em achar ocorrências de uma imagem modelo numa outra imagem a ser analisada. Consequentemente, se as imagens apresentarem mudanças no ponto de vista da câmera o algoritmo normalmente falha. A invariância a pontos de vista é uma qualidade que permite reconhecer um objeto, mesmo que este apresente distorções resultantes de uma transformação em perspectiva causada pela mudança do ponto de vista. Uma abordagem baseada na simulação de pontos de vista, chamada ASIFT, tem sido recentemente proposta no entorno desta problemática. O ASIFT é invariante a pontos de vista, no entanto falha na presença de padrões repetitivos e baixo contraste. O objetivo de nosso trabalho é utilizar uma variante da técnica de simulação de pontos de vista em combinação com a técnica de extração dos coeficientes de Fourier de projeções radiais e circulares (FORAPRO), para propor um algoritmo invariante a pontos de vista, e robusto a padrões repetitivos e baixo contraste. De maneira geral, a nossa proposta resume-se nas seguintes fases: (a) Distorcemos a imagem, variando os parâmetros de inclinação e rotação da câmera, para gerar alguns modelos e conseguir a invariância a deformações em perspectiva, (b) utilizamos cada como modelo a ser procurado na imagem, para escolher o que melhor case, (c) realizamos o casamento de padrões. As duas últimas fases do processo baseiam-se em características invariantes por rotação, escala, brilho e contraste extraídas pelos coeficientes de Fourier. Nossa proposta, que chamamos AFORAPRO, foi testada com 350 imagens que continham diversidade nos requerimentos, e demonstrou ser invariante a pontos de vista e ter ótimo desempenho na presença de padrões repetitivos e baixo contraste.Object recognition is a basic application from the domain of image processing and computer vision. The common process recognition consists of finding occurrences of an image query in another image to be analyzed A. Consequently, if the images changes viewpoint in the camera it will normally result in the algorithm failure. The invariance viewpoints are qualities that permit recognition of an object, even if this present distortion resultant of a transformation of perspective is caused by the change in viewpoint. An approach based on viewpoint simulation, called ASIFT, has recently been proposed surrounding this issue. The ASIFT algorithm is invariant viewpoints; however there are flaws in the presence of repetitive patterns and low contrast. The objective of our work is to use a variant of this technique of viewpoint simulating, in combination with the technique of extraction of the Coefficients of Fourier Projections Radials and Circulars (FORAPRO), and to propose an algorithm of invariant viewpoints and robust repetitive patterns and low contrast. In general, our proposal summarizes the following stages: (a) We distort the image, varying the parameters of inclination and rotation of the camera, to produce some models and achieve perspective invariance deformation, (b) use as the model to be search in the image, to choose the that match best, (c) realize the template matching. The two last stages of process are based on invariant features by images rotation, scale, brightness and contrast extracted by Fourier coefficients. Our approach, that we call AFORAPRO, was tested with 350 images that contained diversity in applications, and demonstrated to have invariant viewpoints, and to have excellent performance in the presence of patterns repetitive and low contrast.
36
MATOS, Caio Eduardo Falcão. "Diagnóstico de câncer de mama em imagens mamográficas através de características locais e invariantes." Universidade Federal do Maranhão, 2017. http://tedebc.ufma.br:8080/jspui/handle/tede/1324.
Повний текст джерелаАнотація:
Submitted by Maria Aparecida (cidazen@gmail.com) on 2017-04-27T13:46:48Z
No. of bitstreams: 1
Caio Eduardo Falcão Matos.pdf: 1884390 bytes, checksum: a5489f8f52a87e6c5958458ed5470488 (MD5)Made available in DSpace on 2017-04-27T13:46:48Z (GMT). No. of bitstreams: 1 Caio Eduardo Falcão Matos.pdf: 1884390 bytes, checksum: a5489f8f52a87e6c5958458ed5470488 (MD5) Previous issue date: 2017-02-08
Breast cancer is one of the leading causes of death among women over the world. The high mortality rates that cancers achieves across the world highlight the importance of developing and investigating the means for the early detection and diagnosis of this disease. Computer Detection and Diagnosis Systems (Computer Assisted Detection / Diagnosis) have been used and proposed as a way to help health professionals. This work proposes a new methodology for discriminating patterns of malignancy and benignity of masses in mammography images by analysis of local characteristics. To do so, it is proposed a combined methodology of feature detectors and descriptors with a model of data representation for an analysis. The goal is to capture both texture and geometry in areas of mammograms. We use the SIFT, SURF and ORB detectors, and the descriptors HOG, LBP, BRIEF and Haar Wavelet. The generated characteristics are coded by a bag of features model to provide a new representation of the data and therefore decrease a dimensionality of the space of characteristics. Finally, this new representation is classified using three approaches: Support Vector Machine, Random Forest, and Adaptive Boosting to differentiate as malignant and benign masses. The methodology provides promising results for the diagnosis of malignant and benign mass encouraging that as local characteristics generated by descriptors and detectors produce a satisfactory a discriminating set.
O câncer de mama é apontado como uma das principais causas de morte entre as mulheres. As altas taxas de mortalidade e registros de ocorrência desse câncer em todo o mundo evidenciam a importância do desenvolvimento e investigação de meios para a detecção e diagnóstico precoce dessa doença. Sistemas de Detecção e Diagnóstico auxiliados por computador (Computer Aided Detection/Diagnosis) vêm sendo usados e propostos como forma de auxílio aos profissionais de saúde. Este trabalho propõe uma metodologia para discriminação de padrões de malignidade e benignidade de massas em imagens de mamografia através da análise de características locais. Para tanto, a metodologia combina detectores e descritores de características locais com um modelo de representação de dados para a análise, tanto de textura quanto de geometria em regiões extraídas das mamografias. São utilizados os detectores SIFT, SURF e ORB, e descritores HOG, LBP, BRIEF e Haar Wavelet. Com as características geradas é aplicado o modelo Bag of Features em uma etapa de representação que objetiva prover nova representação dos dados e por conseguinte diminuir a dimensionalidade do espaço de características. Por fim, esta nova representação é classificada utilizando três abordagens: Máquina de Vetores de Suporte, Random Forests e Adaptive Boosting visando diferenciar as massas malignas e benignas. A metodologia contém resultados promissores para o diagnóstico de massas malignas e benignas fomentando que as características locais geradas pelos descritores e detectores produzem um conjunto descriminate satisfatório.
37
Bruna, Joan. "Representations en Scattering pour la Reconaissance." Phd thesis, Ecole Polytechnique X, 2013. http://pastel.archives-ouvertes.fr/pastel-00905109.
Повний текст джерелаАнотація:
Ma thèse étudie le problème de la reconnaissance des objets et des textures. Dans ce cadre, il est nécessaire de construire des représentations de signaux avec des propriétés d'invariance et de stabilité qui ne sont pas satisfaites par des approches linéaires. Les opérateurs de Scattering itèrent des décompositions en ondelettes et rectifications avec des modules complexes. Ces opérateurs définissent une transformée non-linéaire avec des propriétés remarquables ; en particulier, elle est localement invariante par translation et Lipschitz continue par rapport à l'action des difféomorphismes. De plus, les opérateurs de Scattering définissent une représentation des processus stationnaires qui capture les moments d'ordre supérieur, et qui peut être estimée avec faible variance à partir d'un petit nombre de réalisations. Dans cette thèse, nous obtenons des nouvelles propriétés mathématiques de la représentation en scattering, et nous montrons leur efficacité pour la reconnaissance des objets et textures. Grâce à sa continuité Lipschitz par rapport à l'action des difféomorphismes, la transformée en scattering est capable de linéariser les petites déformations. Cette propriété peut être exploitée en pratique avec un classificateur génératif affine, qui nous permet d'obtenir l'état de l'art sur la reconnaissance des chiffres manuscrites. Nous étudions ensuite les représentations en Scattering des textures dans le cadre des images et du son. Nous montrons leur capacité à discriminer des phénomènes non-gaussiens avec des estimateurs à faible variance, ce qui nous permet d'obtenir de l'état de l'art pour la reconnaissance des textures. Finalement, nous nous intéressons aux propriétés du Scattering pour l'analyse multifractale. Nous introduisons une renormalisation des coéfficients en Scattering qui permet d'identifier de façon efficace plusieurs paramètres multifractales; en particulier, nous obtenons une nouvelle caractérisation de l'intermittence à partir des coefficients de Scattering ré-normalisés, qui peuvent s'estimer de façon consistante.
38
Minařík, Martin. "Strukturální metody identifikace objektů pro řízení průmyslového robotu." Doctoral thesis, Vysoké učení technické v Brně. Fakulta strojního inženýrství, 2009. http://www.nusl.cz/ntk/nusl-233840.
Повний текст джерелаАнотація:
This PhD thesis deals with the use of structural methods of objects identification for industrial robots operation. First, the present state of knowledge in the field is described, i.e. the whole process of objects recognition with the aid of common methods of the syntactic analysis. The main disadvantage of these methods is that is impossible to recognize objects whose digitalized image is corrupted in some ways (due to excessive noise or image disturbances), objects are therefore deformed. Further, other methods for the recognition of deformed objects are described. These methods use structural description of objects for object recognition, i.e. methods which determine the distance between attribute descriptions of images. The core part of this PhD thesis begins in Chapter 5, where deformation grammars, capable of description of all possible object deformations, are described. The only complication in the analysis is the ambiguity of the deformation grammar, which lowers the effectiveness of the analysis. Further, PhD thesis deals with the selection and modification of a proper parser, which is able to analyze a deformation grammar effectively. Three parsers are described: the modified Earley parser, the modified Tomita parser and the modified hybrid LRE(k) parser. As for the modified Earley’s parser, ways of its effective implementation are described. One of the necessary parts of the object recognition is providing the invariances, which this PhD thesis covers in detail, too. Finally, the results of described algorithms are mentioned (successfulness and speed of deformed objects recognition) and suggested testing environment and implemented algorithms are described. In conclusion, all determined possibilities of deformation grammars and their results are summarized.
39
Zhang, David Xin. "Invariant pattern recognition and neural networks." 1997. http://catalog.hathitrust.org/api/volumes/oclc/37622254.html.
Повний текст джерела
40
Chen, Zhi Cheng, and 陳志誠. "Invariant pattern recognition based on bispectra." Thesis, 1995. http://ndltd.ncl.edu.tw/handle/54169092849596588885.
Повний текст джерела
41
Sun, Pei-Chin, and 孫珮琴. "Pattern recognition by invariant morphological feature extraction." Thesis, 1996. http://ndltd.ncl.edu.tw/handle/00438398059753139881.
Повний текст джерелаАнотація:
碩士大同工學院
電機工程學系
84
Pattern recognition plays an important role in many applications. Many methods about how to extract features of an object in scence regardless of itspositions, rotations, and scales have been proposed in the literature. In thisthesis, we develop a feature extraction method based on mathematical morphology. At first, the concepts of multiscale decomposition andpecstrum(pattern spectrum) has been received. Then, a new feature extraction method called multiscale pecstrum is developed. By investigating the noiseimmunity of the multiscalepecstrum, we extend the proposed method of multiscale pecstrum in spatial domain to frequency domain. Finally, experimental results are given to show thehigh performance.
42
Sun, Pei-Qin, and 孫珮琴. "Pattern recognition by invariant morphological feature extraction." Thesis, 1996. http://ndltd.ncl.edu.tw/handle/75658801283038613422.
Повний текст джерела
43
WU, WEN-JIE, and 吳文傑. "Invariant pattern recognition using higher-order neural networks." Thesis, 1993. http://ndltd.ncl.edu.tw/handle/04087011820093994541.
Повний текст джерела
44
ZENG, YUAN-XIAN, and 曾元顯. "Associative mapping and translation, rotation, scale-invariant pattern recognition." Thesis, 1989. http://ndltd.ncl.edu.tw/handle/11872579950912127601.
Повний текст джерела
45
Shun-ching, Ke, and 柯順清. "Development of Position and Scale-invariant Pattern Recognition Systems." Thesis, 1999. http://ndltd.ncl.edu.tw/handle/39950137540155197392.
Повний текст джерелаАнотація:
碩士國立臺灣科技大學
機械工程系
87
The purpose of this thesis is to develop a position/scaling-invariant pattern recognition system. Two basic methods are proposed based on global features and local features. Both methods are embedded in the traditional Hamming network where pattern recognition activities are performed. For the first method, the global feature of the pattern is input into a feature extraction network to calculate its exact position. An associate memory network is employed to perform static mapping of the input feature position and the weighting matrix of the Hamming network so that a proper position/scaling-invariant property can be obtained. For the second method, local features of the input pattern are assumed to be available for the proposed system. Three schemes are suggested for the encoding of the input local features. Each one of which utilizes properties such as feature contribution and weights on feature connectivity. Simulation results show that both methods give position/scaling-invariant properties with reasonable complexity in terms of the encoding date size, feature order, and robustness.
46
CHEN, PO-CHENG, and 陳柏誠. "Two-dimensional pattern recognition by invariant algebraic feature extraction." Thesis, 1992. http://ndltd.ncl.edu.tw/handle/91022547792968596635.
Повний текст джерела
47
Lee, Cheng-Long, and 李錦榮. "Translation, Rotation, and Scaling Invariant Pattern Recognition by Fuzzy Neural Networks." Thesis, 1994. http://ndltd.ncl.edu.tw/handle/73768232434002506852.
Повний текст джерелаАнотація:
碩士國立交通大學
控制工程系
82
A hybrid pattern recognition system is described in this study which can recognize patterns with translation, rotation, scaling, and a combination of them. The system consists of two parts, i.e., the preprocessor and the fuzzy neuron classifier. The preprocessor, implemented by simple mapping techniques, generates the normalized input image. Next, the fuzzy neuron classifier associates the normalized input patterns with prototype patterns by making an optimal match of the weighted grade membership functions. A single layer gradient type algorithm is used to adapt the membership functions to the patterns to be recognized. Some numerical examples are provided to demonstrate the performance of the proposed technique.
48
Li, Jin Rong, and 李錦榮. "Translation, rotation, and scaling invariant pattern recognition by fuzzy neural networks." Thesis, 1994. http://ndltd.ncl.edu.tw/handle/01702677518212263247.
Повний текст джерела
49
Shih, Jau-Ling, and 石昭玲. "2-D Invariant Pattern Recognition Using a Backpropogation Network Improved by Distributed Associative Memory." Thesis, 1994. http://ndltd.ncl.edu.tw/handle/47410554866730966852.
Повний текст джерелаАнотація:
碩士國立成功大學
電機工程研究所
82
In this paper, a system included image preprocessing and neural networks is proposed. The various function units of the image processing are used to obtain an invariant image representation in the beginning of the system. The space of the neural networks weights can be reduced by using the reduction of the feature dimension before the preprocessed feature applied to the networks.Then, several kinds of the neural models are proposed for pattern recognition : (1) distributed associative memory (DAM), (2) backpropagation network(BPN), (3)DAM combined with BPN, and(4)BPN with the associative memory as initial weights. In the case of (3), this hierarchical networks consist of two levels of neural networks. In the low level, a DAM receives the output vectors of image preprocessing functions to create a system which recognizes pattern regardless of changes in scale or rotation. The higher level is a two- layers BPN which recives the recalled information from the memorized database of the lower level. This neural networks use a BPN after the DAM can raise the recognition ratio in comparison with a DAM, and be faster than a BPN.In the case of (4), the training of the BPN speeds up much because this neural networks use a associative memory of a DAM as initial weights of the first layer of te BPN. Experiment results show that the system can recognize all the patterns correctly when the percentage of the white noises is under under 20% for the case (3) and (4).
50
Wang, Sha-Wei, and 王學偉. "Investigation on shift-, rotational- and size-invariant opti- cal pattern recognition and related filters." Thesis, 1993. http://ndltd.ncl.edu.tw/handle/54676972990248223030.
Повний текст джерела