本論文應用隱喻與轉喻之理論觀點，探討台灣客語分類詞諺語之認知語意機制如何運作。首先，根據Kövecses and Radden (1998) 從認知語言學角度所提出的轉喻理論，分析台灣客語分類詞諺語中的轉喻類型。接著，本論文分析台灣客語分類詞諺語中隱喻機制的運作，結果發現普遍而言，隱喻都是以轉喻為基礎。此外，本研究針對Radden (2003) 所提出以轉喻為基礎的隱喻之四種來源分類提出修正。 除了呈現認知語意機制，文化制約之世界普遍性及台灣客家文化之特殊性也在台灣客語分類詞諺語中展現出來。最後，透過Lakoff and Turner (1989) 所提出的生命物種之大鏈隱喻，我們了解諺語所表達的最終概念是以人為中心，而且諺語通常帶有勸世的功能。簡言之，本論文藉由探討認知語意機制如何在台灣客語分類詞諺語中運作，呈現人類認知過程以及展現台灣客家文化。 / This thesis aims to explore how the cognitive mechanisms are operated in the classifiers and measure words in Taiwanese Hakka proverbial expressions, in particular metonymy, the interaction between metaphor and metonymy, idiomaticity, and cultural constraints. Since human conceptual system is fundamentally metaphorical in nature, classifiers, representing conceptual classification of the world, are found to manifest metonymically and metaphorically. First, based on the metonymic relationships proposed by Kövecses and Radden (1998), cases involving metonymy are carefully spelled out. Then, cases involving the interaction between metaphor and metonymy are elaborated. The metaphors activated in these cases are generally grounded in metonymy, which evidences that metaphors generally have a metonymic basis (Radden 2003). Apart from displaying cognitive mechanisms, the classifier/measure word proverbial expressions in Taiwanese Hakka exhibit Taiwanese Hakka-specific cultural constraints and near universality in conceptual metaphors (Kövecses 2002). Cases which are more specific to Taiwanese Hakka are semantically more opaque whereas cases which are more near universal are semantically more transparent (Gibbs 1995). Furthermore, through the GREAT CHAIN METAPHOR proposed by Lakoff and Turner (1989), we know that all the proverbial expressions are ultimately concerned about human beings. Moreover, proverbial expressions tend to carry pragmatic-social functions, conveying exhortations. In brief, the cognitive mechanisms of metonymy as well as the interaction between metaphor and metonymy are pervasively found in classifier/measure word proverbial expressions in Taiwanese Hakka. Through unraveling the conceptual mechanisms associated with classifiers and measure words in Taiwanese Hakka proverbial expressions, this study betters our understanding of human cognition in general and Taiwanese Hakka culture in particular.

This thesis investigates the function of number marking in Persian, within the framework of principles and parameters (P&P), and its relationship to inflectional and derivational number marking. Following the assumption in Distributed Morphology that inflectional and derivational morphology are not distinct, the distribution and properties of number marking in Persian provide evidence for both inflectional and derivational number marking. Assuming the two parameters of number marking (Wiltschko, 2007, 2008), number marking as a functional head and number marking as a modifier, I propose that number marking in Persian is mainly inflectional while number functions as a functional head; moreover, I propose that number marking in Persian can be derivational while number functions as a modifier. This explains that number morphology in Persian is not split to either inflectional or derivational. Rather, following Booij’s (1993, 1995) claim that inflectional morphology can be used contextually as well as inherently, I propose that number morphology in Persian is inflectional while number is a functional head; however, it has inherent residues as a modifier. Considering the functions of inflectional plural morphology in Persian, I argue that the functional category Number Phrase (NumP) is projected in Persian, and number is generated in the head of this functional category. Besides, Persian is a classifier language in which classifiers are in complementary distribution with plural marking. Following Borer’s (2005) discussion of the complementary distribution of plural marking and classifiers in Armenian, I argue that the head of NumP in Persian is either occupied by the plural maker or by full/empty classifiers. Moreover, I show that the presence of bare singulars/plurals in certain syntactic positions in Persian is related to the projection/non-projection of NumP.

This thesis investigates the function of number marking in Persian, within the framework of principles and parameters (P&P), and its relationship to inflectional and derivational number marking. Following the assumption in Distributed Morphology that inflectional and derivational morphology are not distinct, the distribution and properties of number marking in Persian provide evidence for both inflectional and derivational number marking. Assuming the two parameters of number marking (Wiltschko, 2007, 2008), number marking as a functional head and number marking as a modifier, I propose that number marking in Persian is mainly inflectional while number functions as a functional head; moreover, I propose that number marking in Persian can be derivational while number functions as a modifier. This explains that number morphology in Persian is not split to either inflectional or derivational. Rather, following Booij’s (1993, 1995) claim that inflectional morphology can be used contextually as well as inherently, I propose that number morphology in Persian is inflectional while number is a functional head; however, it has inherent residues as a modifier. Considering the functions of inflectional plural morphology in Persian, I argue that the functional category Number Phrase (NumP) is projected in Persian, and number is generated in the head of this functional category. Besides, Persian is a classifier language in which classifiers are in complementary distribution with plural marking. Following Borer’s (2005) discussion of the complementary distribution of plural marking and classifiers in Armenian, I argue that the head of NumP in Persian is either occupied by the plural maker or by full/empty classifiers. Moreover, I show that the presence of bare singulars/plurals in certain syntactic positions in Persian is related to the projection/non-projection of NumP.

A CT-scan of lungs has become ubiquitous as a thoracic diagnostic tool. Thus, using CT-scan images in developing predictive models for tumor types and survival time of patients afflicted with Non-Small Cell Lung Cancer (NSCLC) would provide a novel approach to non-invasive tumor analysis. It can provide an alternative to histopathological techniques such as needle biopsy. Two major tumor analysis problems were addressed in course of this study, tumor type classification and survival time prediction. CT-scan images of 109 patients with NSCLC were used in this study. The first involved classifying tumor types into two major classes of non-small cell lung tumors, Adenocarcinoma and Squamous-cell Carcinoma, each constituting 30% of all lung tumors. In a first of its kind investigation, a large group of 2D and 3D image features, which were hypothesized to be useful, are evaluated for effectiveness in classifying the tumors. Classifiers including decision trees and support vector machines (SVM) were used along with feature selection techniques (wrappers and relief-F) to build models for tumor classification. Results show that over the large feature space for both 2D and 3D features it is possible to predict tumor classes with over 63% accuracy, showing new features may be of help. The accuracy achieved using 2D and 3D features is similar, with 3D easier to use. The tumor classification study was then extended by introducing the Bronchioalveolar Carcinoma (BAC) tumor type. Following up on the hypothesis that Bronchioalveolar Carcinoma is substantially different from other NSCLC tumor types, a two-class problem was created, where an attempt was made to differentiate BAC from the other two tumor types. To make a three-class problem a two-class problem, misclassification amongst Adenocarcinoma and Squamous-cell Carcinoma were ignored. Using the same prediction models as the previous study and just 3D image features, tumor classes were predicted with around 77% accuracy. The final study involved predicting two year survival time in patients suffering from NSCLC. Using a subset of the image features and a handful of clinical features, predictive models were developed to predict two year survival time in 95 NSCLC patients. A support vector machine classifier, naive Bayes classifier and decision tree classifier were used to develop the predictive models. Using the Area Under the Curve (AUC) as a performance metric, different models were developed and analyzed for their effectiveness in predicting survival time. A novel feature selection method to group features based on a correlation measure has been proposed in this work along with feature space reduction using principal component analysis. The parameters for the support vector machine were tuned using grid search. A model based on a combination of image and clinical features, achieved the best performance with an AUC of 0.69, using dimensionality reduction by means of principal component analysis along with grid search to tune the parameters of the SVM classifier. The study showed the effectiveness of a predominantly image feature space in predicting survival time. A comparison of the performance of the models from different classifiers also indicate SVMs consistently outperformed or matched the other two classifiers for this data.

Τα τεχνητά νευρωνικά δίκτυα είναι μια μορφή τεχνητής νοημοσύνης, τα οποία αποτελούνται από ένα σύνολο απλών, διασυνδεδεμένων και προσαρμοστικών μονάδων, οι οποίες συνιστούν ένα παράλληλο πολύπλοκο υπολογιστικό μοντέλο. Μέχρι σήμερα έχουν εφαρμοστεί επιτυχημένα σε ένα ευρύ φάσμα περιοχών για την επίλυση προβλημάτων ταξινόμησης ή πρόβλεψης, όπως η βιολογία, η ιατρική, η γεολογία, η φυσική κ.ά. Σε αυτήν την εργασία θα ασχοληθούμε με την εκπαίδευση τεχνητών νευρωνικών δικτύων ανά πρότυπο εισόδου. Αυτή η προσέγγιση θεωρείται κατεξοχήν κατάλληλη για περιπτώσεις όπου η εκπαίδευση διαθέτει σημαντικό χρόνο και απαιτεί μεγάλο αποθηκευτικό χώρο, όπως συμβαίνει συχνά όταν έχουμε μεγάλα σύνολα προτύπων ή/και δίκτυα. Μέχρι σήμερα έχουν προταθεί πολλοί αλγόριθμοι εκπαίδευσης νευρωνικών δικτύων, καλύπτοντας ο ένας τα κενά του άλλου, σχεδιασμένοι ώστε να επιλύουν τα προβλήματα που παλιότερα ήταν δύσκολο να επιλυθούν. Στόχος της εργασίας είναι η εκτενής ανάλυση και αξιολόγηση των αλγορίθμων εκπαίδευσης καθώς και η ικανότητα γενίκευσης των εκπαιδευόμενων δικτύων σε μια ποικιλία προβλημάτων από τους τομείς τις ιατρικής και της βιοπληροφορικής. Επίσης επηρεασμένοι από τη δυνατότητα για την επίτευξη καλύτερης απόδοσης θα μελετήσουμε την συμβολή των νευρωνικών δικτύων στη μηχανική μάθηση. Συγκεκριμένα θα αποτιμήσουμε τη συνεισφορά των νευρωνικών δικτύων στη δημιουργία αξιόπιστων συστημάτων αποφάσεων χρησιμοποιώντας τεχνικές συνδυασμού ταξινομητών. Τέλος, θα μελετήσουμε τις δυνατότητες συνδυασμού τους με διάφορες άλλες κατηγορίες ταξινομητών μηχανικής μάθησης για την ανάπτυξη ισχυρότερων υβριδικών συστημάτων εξαγωγής πληροφορίας. / Literature review corroborates that artificial neural networks are being successfully applied in a variety of regression and classification problems. Due of their ability to exploit the tolerance for imprecision and uncertainty in real-world problems and their robustness and parallelism, artificial neural networks have been increasingly used in many applications. It is well-known that the procedure of training a neural network is highly consistent with unconstrained optimization theory and many attempts have been made to speed up this process. In particular, various algorithms motivated from numerical optimization theory have been applied for accelerating neural network training. Moreover, commonly known heuristics approaches such as momentum or variable learning rate lead to a significant improvement. In this work we compare the performance of classical gradient descent methods and examine the effect of incorporating into them a variable learning rate and an adaptive nonmonotone strategy. We perform a large scale study on the behavior of the presented algorithms and identify their possible advantages. Additionally, we propose two modifications of two well-known second order algorithms aiming to overcome the limitations of the original methods.

Introduction: A computer-based classification system is proposed for the characterization of cervical intervertebral disc degeneration from saggital magnetic resonance images. Materials and methods: Cervical intervertebral discs from saggital magnetic resonance images where assessed by an experienced orthopaedist as normal or degenerated (narrowed) employing Matsumoto’s classification scheme. The digital images where enhanced and the intervertebral discs which comprised the regions of interest were segmented. First and second order statistics textural features extracted from thirty-four discs (16 normal and 16 degenerated) were used in order to design and test the classification system. In addition textural features were calculated employing Laws TEM images. The existence of statistically significant differences between the textural features values that were generated from normal and degenerated discs was verified employing the Student’s paired t-test. A subset with the most discriminating features (p<0.01) was selected and the Exhaustive Search and Leave-One-Out methods were used to find the best features combination and validate the classification accuracy of the system. The proposed system used the Least Squares Minimum Distance Classifier in combination with four textural features with comprised the best features combination in order to classify the discs as normal or degenerated. Results: The overall classification accuracy was 93.8% misdiagnosing 2 discs. In addition the system’s sensitivity in detecting a narrow disc was 93.8% and its specificity was also 93.8%. Conclusion: Further investigation and the use of a larger sample for validation could make the proposed system a trustworthy and useful tool to the physicians for the evaluation of degenerative disc disease in the cervical spine. / Σκοπός: Η στένωση των μεσοσπονδύλιων δίσκων της αυχενικής μοίρας, ως κύρια έκφραση εκφυλιστικής νόσου, είναι μια από τις σημαντικότερες αιτίες πρόκλησης πόνου στην περιοχή του αυχένα. Στην κλινική πράξη η αξιολόγηση της στένωσης γίνεται μέσω μέτρησης του μεσοσπονδύλιου διαστήματος, σε διάφορες απεικονίσεις της αυχενικής μοίρας του ασθενούς. Στην παρούσα εργασία προτείνεται μια υπολογιστική μέθοδος ανάλυσης εικόνας, για την αυτοματοποιημένη εκτίμηση της στένωσης από εικόνες μαγνητικής τομογραφίας. Υλικό και Μέθοδος: Μελετήθηκαν 34 μεσοσπονδύλιοι δίσκοι από οβελιαίες τομές μαγνητικής τομογραφίας της αυχενικής μοίρας, οι οποίες ελήφθησαν με χρήση Τ2 ακολουθίας. Η στένωση των μεσοσπονδύλιων δίσκων αξιολογήθηκε από έμπειρο ορθοπαιδικό βάσει της κλίμακας Matsumoto. Οι δίσκοι χωρίστηκαν σε δύο κατηγορίες: (α) 16 φυσιολογικοί και (β) 16 δίσκοι που παρουσίαζαν στένωση. Με χρήση διαδραστικού περιβάλλοντος επεξεργασίας εικάνας καθορίστηκε το περίγραμμα των μεσοσπονδύλιων δίσκων οι οποίοι αποτελούν τις προς ανάλυση περιοχές ενδιαφέροντος (Π.Ε.). Σε κάθε Π.Ε. εφαρμόστηκαν αλγόριθμοι εξαγωγής χαρακτηριστικών υφής. Συγκεκριμένα υπολογίστικαν χαρακτηριστικά υφής από στατιστικά πρώτης και δεύτερης τάξης καθώς και χαρακτηριστικά από τα μέτρα ενέργειας υφλης κατλα Laws. Τα παραπάνω χαρακτηριστικά, ποσοτικοποιούν διαγνωστικές πληροφορίες της έντασης του σήματος της Π.Ε. και συσχετίζονται με τη βιοχημική σύσταση των απεικονιζόμενων δομών. Τα εξαχθέντα χαρακτηριστικά υφής αξιοποιήθηκαν για τη σχεδίαση του ταξινομητή ελάχιστης απόστασης ελαχίστων τετραγώνων, ο οποίος χρησιμοποιήθηκε για το διαχωρισμό μεταξύ φυσιολογικών δίσκων και δίσκων που παρουσίαζαν στένωση (εκφυλισμένων). Αποτελέσματα: Η ακρίβεια της ταξινόμησης φυσιολογικών και εκφυλισμένων μεσοσπονδύλιων δίσκων ανήλθε σε 93.8%. Η ευαισθησία καθώς και η ειδικότητα της μεθόδου, σε ότι αφορά την ανίχνευση εκφυλισμένων δίσκων, είναι επίσης 93.8%. Συμπέρασμα: Με δεδομένο το μικρό μέγεθος του δείγματος που χρησιμοποιήθηκε για το σχεδιασμό της μεθόδου, απαιτούνται περετέρω εργασίες πιστοποίησης της ακρίβειας ταξινόμησης, προκειμένου η μέθοδος αυτή να αξιοποιηθεί από ακτινολόγους και ορθοπαιδικους, ως βοηθητικό διαγνωστικό εργαλείο.

This thesis investigates the function of number marking in Persian, within the framework of principles and parameters (P&P), and its relationship to inflectional and derivational number marking. Following the assumption in Distributed Morphology that inflectional and derivational morphology are not distinct, the distribution and properties of number marking in Persian provide evidence for both inflectional and derivational number marking. Assuming the two parameters of number marking (Wiltschko, 2007, 2008), number marking as a functional head and number marking as a modifier, I propose that number marking in Persian is mainly inflectional while number functions as a functional head; moreover, I propose that number marking in Persian can be derivational while number functions as a modifier. This explains that number morphology in Persian is not split to either inflectional or derivational. Rather, following Booij’s (1993, 1995) claim that inflectional morphology can be used contextually as well as inherently, I propose that number morphology in Persian is inflectional while number is a functional head; however, it has inherent residues as a modifier. Considering the functions of inflectional plural morphology in Persian, I argue that the functional category Number Phrase (NumP) is projected in Persian, and number is generated in the head of this functional category. Besides, Persian is a classifier language in which classifiers are in complementary distribution with plural marking. Following Borer’s (2005) discussion of the complementary distribution of plural marking and classifiers in Armenian, I argue that the head of NumP in Persian is either occupied by the plural maker or by full/empty classifiers. Moreover, I show that the presence of bare singulars/plurals in certain syntactic positions in Persian is related to the projection/non-projection of NumP.

This thesis presents a complete self-calibrating illumination intensity-invariant colour classification system. We extend a novel fuzzy colour processing tech- nique called Fuzzy Colour Contrast Fusion (FCCF) by combining it with a Heuristic- assisted Genetic Algorithm (HAGA) for automatic fine-tuning of colour descriptors. Furthermore, we have improved FCCF’s efficiency by processing colour channels at varying colour depths in search for the optimal ones. In line with this, we intro- duce a reduced colour depth representation of a colour image while maintaining efficient colour sensitivity that suffices for accurate real-time colour-based object recognition. We call the algorithm Variable Colour Depth (VCD) and we propose a technique for building and searching a VCD look-up table (LUT). The first part of this work investigates the effects of applying fuzzy colour contrast rules to vary- ing colour depths as we extract the optimal rule combination for any given target colour exposed under changing illumination intensities. The second part introduces the HAGA-based parameter-optimisation for automatically constructing accurate colour classifiers. Our results show that for all cases, the VCD algorithm, combined with HAGA for parameter optimisation improve colour classification via a pie-slice colour classifier.For 6 different target colours, the hybrid algorithm was able to yield 17.63% higher overall accuracy as compared to the pure fuzzy approach. Fur- thermore, it was able to reduce LUT storage space by 78.06% as compared to the full-colour depth LUT.

Made available in DSpace on 2015-03-26T13:53:29Z (GMT). No. of bitstreams: 1 texto completo.pdf: 3809983 bytes, checksum: 1e1c4e38e746cb96d9590dfed817a6ba (MD5) Previous issue date: 2013-02-28 / In areas of the field of Training Barriers in southern Bahia and northern Espírito Santo is a kind of distinctive environment in terms of vegetation and soil characteristics, known regionally Muçununga. They are mainly found on Spodosols, ecologically unique adaptations depending on the conditions of nutritional poverty of the soil, with species tolerant to extremes of excess and lack of water, and still represent an ecosystem associated with the Atlantic Forest. This study aimed to: 1) evaluate the performance of the Maximum Likelihood classifiers and Support Vector Machine and the contribution of different compositions of multispectral bands, the vegetation index (NDVI), and principal component analysis of sensor images TM/Landsat5 for separating features of muçunungas. 2) Assess the supervised classifications for separating features of muçununga, with reference to the visual classification. 3) evaluate the distribution pattern, size, frequency and density of muçunungas in the region. We acquired nine scenes from the satellite / sensor Sensor TM/Landsat5-TM (Thematic Mapper) orbit 215, paragraphs 71, 72 and 73 days 29/05/2006; orbit 216, paragraphs 71, 72 and 73 on 19 / 07/2007 and orbit 215, paragraphs 71, 72 and 73 day 27/08/2007 with spatial resolution of 30 meters, which range from the municipality of Linhares-up Prado ES-BA divided into three areas: the coastal region of Spirit Santo, Bahia coastal region and the interior region of both states. Moreover, were provided by the company Fibria Cellulose three mosaics of RapidEye satellite images with radiometric and geometric corrections. The classes of land use were established according to the prior knowledge of the area. We collected 75 samples for each training class studied and validation samples were obtained from the database of land use provided by Fibria Cellulose. We performed a combination of ten bands, six bands of the sensor TM/Landsat5 (lanes 1, 2, 3, 4, 5 and 7), one index vegatação NDVI and the first three principal components. For supervised classification algorithm MaxVer, the bands were grouped in combinations of one to 10 bands for a total of 1023 combinations processed using ArcGIS10.1. A supervised classification algorithm of Support Vector Machine was held at Envi software, testing the ten best combinations MaxVer generated in each of the three areas studied. The evaluation of the accuracy of the classifications was performed from the confusion matrices that were obtained by crossing thematic maps derived classification. The region was divided into ranges of 10 km long from the coast towards the interior of the states of Bahia and Espírito Santo. In these bands, and analyzed the percentage area muçunungas. Were also analyzed four indentations in the region east / west, aiming to analyze the distribution, frequency and density of muçunungas in the region. From the results of ratings, performance classifier Support Vector Machine (SVM) can be considered satisfactory. However, MaxVer classifier obtained better results for the three regions analyzed. The areas classified as Muçununga merged with other classes such as pasture, eucalypt forest, body of water due to the spectral characteristics of vegetation enclave. The thematic map produced by supervised classification of the best combination in each region, using the algorithm Maxver reached Kappa index of 0.91 in the Holy Spirit, 0.90 and 0.81 in the Bahia region of the interior. The combination of one or two bands inferior results obtained while the best results were combinations with 6-8 bands. The use of normalized difference vegetation index (NDVI) has promoted improvement in Kappa, but the increase is most notable in the classification obtained from the combination of these with the visible bands. The use of principal components representative did not represent an increase in the accuracy of the ratings except for those who had only the visible bands. The distribution of muçunungas presented a heterogeneous pattern, it increases its occurrence as it departed from the coast up to 40 km. From kilometer 50 to 80 there was a decrease explained by the approach of Crystalline Basement. 2254 were mapped areas muçunungas across the region, from the RapidEye satellite image. So are approximately 1.08% of the mapped region with presence of muçunungas. Despite the low percentage of muçunungas, the amount of which is representative of the region and most of them are smaller than five hectares. So do efforts are required to know the space of distruibuição muçunungas in order to generate information for the conservation of the remnants. / Em áreas de domínio da Formação Barreiras na região sul da Bahia e norte do Espírito Santo ocorre um tipo de ambiente diferenciado em termos de vegetação e características edáficas, denominado regionalmente de Muçununga. São encontradas principalmente sobre Espodossolos, ecologicamente únicas em função das adaptações às condições de pobreza nutricional do solo, com espécies tolerantes a extremos de excessos e falta de água, e ainda representam um dos ecossistemas associados à Mata Atlântica. Esse trabalho teve como objetivos: 1) avaliar o desempenho dos classificadores da Máxima Verossimilhança e Support Vector Machine e a contribuição de diferentes composições de bandas multiespectrais, do índice de vegetação da diferença normalizada (NDVI), e da análise de componentes principais das imagens do sensor TM/Landsat5 para a separação de feições das muçunungas. 2) Avaliar as classificações supervisionadas para a separação de feições de muçununga, tendo como referência a classificação visual. 3) Avaliar o padrão de distribuição, o tamanho, frequência e densidade das muçunungas na região. Foram adquiridas nove cenas do satélite/sensor Sensor TM/Landsat5-TM (Thematic Mapper) da órbita 215, pontos 71, 72 e 73 do dia 29/05/2006; da órbita 216, pontos 71, 72 e 73 do dia 19/07/2007 e da órbita 215, pontos 71, 72 e 73 do dia 27/08/2007 com resolução espacial de 30 metros, que englobam desde o município de Linhares- ES até Prado-BA divididas em três áreas: região litorânea do Espírito Santo, região litorânea da Bahia e região do interior de ambos os estados. Além disso, foram disponibilizados pela empresa Fibria Celulose três mosaicos de imagens do satélite Rapideye com as devidas correções radiométricas e geométricas. As classes de uso do solo foram estabelecidas de acordo com o conhecimento prévio da área. Foram coletadas 75 amostras de treinamento para cada classe estudada e as amostras de validação foram obtidas a partir de base de dados de uso do solo fornecida pela Fibria Celulose. Foi realizada a combinação de dez bandas: seis bandas do Sensor TM/Landsat5 (bandas 1, 2, 3, 4, 5 e 7), uma do índice de vegatação NDVI e as três primeiras componentes principais. Para a classificação supervisionada do algoritmo MaxVer, as bandas foram agrupadas em combinações de uma até 10 bandas perfazendo um total de 1023 combinações, processadas no software ArcGIS10.1. A classificação supervisionada do algoritimo Support Vector Machine foi realizada no software Envi, testando as dez melhores combinações geradas no MaxVer de cada uma das três áreas estudadas. A avaliação da exatidão das classificações foi realizada a partir das matrizes de confusão que foram obtidas pelo cruzamento dos mapas temáticos, resultantes da classificação. A região foi dividida em faixas de 10 km de extensão a partir do litoral em direção ao interior dos estados da Bahia e do Espírito Santo. Nestas faixas, analisaram a percentagem e a área de muçunungas. Também foram analisados quatro recortes da região na direção leste/oeste, com objetivo de analisar a distribuição, a frequência e a densidade das muçunungas na região. A partir dos resultados das classificações, o desempenho do classificador Support Vector Machine (SVM) pode ser considerado satisfatório. Entretanto, classificador MaxVer obteve melhores resultados para as três regiões analisadas. As áreas classificadas como Muçununga confundiram-se com outras classes tais como pastagem, eucalipto, mata, corpo d agua devido às características espectrais desse enclave vegetacional. O mapa temático produzido pela classificação supervisionada da melhor combinação em cada região, utilizando o algoritmo Maxver, atingiu índice Kappa de 0,91 no Espírito Santo; 0,90 na Bahia e 0,81 na região do interior. A combinação de uma e duas bandas obtiveram resultados inferiores enquanto que os melhores resultados foram com combinações de 6 a 8 bandas. O uso do índice de vegetação normalizada (NDVI) promoveu melhora no índice Kappa, mas o incremento é mais notável na classificação obtida a partir da combinação destes com as bandas do visível. A utilização das componentes principais não representou aumento representativo na acurácia das classificações, exceto para aquelas que só apresentavam bandas do visível. A distribuição das muçunungas apresentou um padrão heterogêneo, pois aumenta sua ocorrência à medida que se distanciava do litoral até 40 km. A partir do kilometro 50 até aos 80 houve um decréscimo explicado pela aproximação do Embasamento Cristalino. Foram mapeadas 2254 áreas de muçunungas em toda a região, a partir da imagem satélite Rapideye. Assim, são cerca de 9,08% da região mapeada com presença de muçunungas. Apesar da percentagem baixa de muçunungas, a quantidade delas é representativa na região e a maioria delas são menores que cinco hectares. Assim, fazem-se necessários esforços para se conhecer a distruibuição espacial das muçunungas, no sentido de gerar subsídios para a conservação dos remanescentes.

Made available in DSpace on 2014-06-11T19:29:40Z (GMT). No. of bitstreams: 0 Previous issue date: 2011-08-05Bitstream added on 2014-06-13T19:38:57Z : No. of bitstreams: 1 kami_gjc_me_sjrp.pdf: 1342570 bytes, checksum: 240c6d6b92fda1861dfbed94c9213a10 (MD5) / As técnicas para identificação biométrica têm evoluído cada vez mais devido à necessidade que os seres humanos têm de identificar as pessoas em tempo real e de forma precisa para permitir o acesso a determinados recursos, como por exemplo, as aplicações e serviços WEB. O reconhecimento facial é uma técnica biométrica que apresenta várias vantagens em relação às demais, tais como: uso de equipamentos simples e baratos para a obtenção das amostras e a possibilidade de se realizar o reconhecimento em sigilo e à distância. O reconhecimento de faces a partir de vídeo é uma tendência recente na área de Biometria. Esta dissertação tem por objetivo principal comparar diferentes técnicas de reconhecimento facial a partir de vídeo para determinar as que apresentam um melhor compromisso entre tempo de processamento e precisão. Outro objetivo é a incorporação dessas melhores técnicas no sistema de autenticação biométrica em ambientes de E-Learning, proposto em um trabalho anterior. Foi comparado o classificador vizinho mais próximo usando as medidas de distância Euclidiana e Mahalanobis com os seguintes classificadores: Redes Neurais MLP e SOM, K Vizinhos mais Próximos, Classificador Bayesiano, Máquinas de Vetores de Suporte (SVM) e Floresta de Caminhos Ótimos (OPF). Também foi avaliada a técnica de Modelos Ocultos de Markov (HMM). Nos experimentos realizados com a base Recogna Video Database, criada especialmente para uso neste trabalho, e Honda/UCSD Video Database, os classificadores apresentaram os melhores resultados em termos de precisão, com destaque para o classificador SVM da biblioteca SVM Torch. A técnica HMM, que incorpora informações temporais, apresentou resultados melhores do que as funções de distância, em termos de precisão, mas inferiores aos classificadores / The biometric identification techniques have evolved increasingly due to the need that humans have to identify people in real time to allow access to certain resources, such as applications and Web services. Facial recognition is a biometric technique that has several advantages over others. Some of these advantages are the use of simple and cheap equipment to obtain the samples and the ability to perform the recognition in covert mode. The face recognition from video is a recent approach in the area of Biometrics. The work in this dissertation aims at comparing different techniques for face recognition from video in order to find the best rates on processing time and accuracy. Another goal is the incorporation of these techniques in the biometric authentication system for E-Learning environments, proposed in an earlier work. We have compared the nearest neighbor classifier using the Euclidean and Mahalanobis distance measures with some other classifiers, such as neural networks (MLP and SOM), k-nearest neighbor, Bayesian classifier, Support Vector Machines (SVM), and Optimum Path Forest (OPF). We have also evaluated the Hidden Markov Model (HMM) approach, as a way of using the temporal information. In the experiments with Recogna Video Database, created especially for this study, and Honda/UCSD Video Database, the classifiers obtained the best accuracy, especially the SVM classifier from the SVM Torch library. HMM, which takes into account temporal information, presented better performance than the distance metrics, but worse than the classifiers