Note:

This thesis presents several descriptor selection schemes for video content pairwise-matching tasks. Those proposed schemes attempt to leverage two significant properties of videos, temporal correlation and motion information. Aiming to find an efficient and descriptive representation for a video sequence, the concept of descriptor persistency is defined. Those descriptors that satisfy this definition are called persistent descriptors. In order to exploit descriptor persistency, an encoder is proposed. The proposed encoder consists of five main components. First, keyframe labelling is introduced to reduce complexity and ensure a reasonable size of persistent sets. After that, persistent descriptor detection is performed on each group of pictures (GOP) separately. The second component is the standard SIFT descriptor extraction. The third part is to identify persistent descriptors from each GOP, called persistent descriptors extraction. In this stage, three different methods are proposed: The direct method and two approximation approaches. Persistent descriptors selection, which is the fourth stage, is carried out to control the size of the persistent set. For this stage, three selection methods are proposed. All of them attempt to utilize the motion information to select more descriptive descriptors among all the persistent descriptors in the GOP. In order to perform pairwise-matching, in this thesis, a simple but efficient pairwise-matching method is proposed. Experiments are carried out to evaluate the performance of the proposed schemes. The datasets used for performance evaluation are subsets from the categories that describe in [1]. Two metrics de ned in [2], namely false positive rate (FPR) and true positive rate (TPR), are used for the performance evaluation. / Thesis / Master of Applied Science (MASc)

No description available.

No description available.

Orientador: Adjair Antonio do Nascimento / Banca: Rosimeri de Oliveira Vasconcelos / Banca: Claudia Dias Zettermann / Resumo: A Caatinga é um bioma exclusivamente brasileiro que, até pouco tempo, era considerado pobre em biodiversidade. O cachorro-do-mato Cerdocyon thous (Linnaeus, 1766) é um canídeo de porte médio com distribuição em quase todo território brasileiro, sendo o único canídeo com ocorrência registrada no semi-árido Nordestino. O presente estudo teve como objetivos identificar a helmintofauna de Cerdocyon thous provenientes da Caatinga do estado da Paraíba e determinar os indicadores ecológicos de infecção helmíntica, na contribuição favorável para o conhecimento da biodiversidade deste bioma ainda pouco conhecido. Para tal, foram utilizados 58 animais encontrados atropelados em rodovias localizadas nas proximidades do município de Patos, dos quais foram colhidos os helmintos gastrintestinais. Todas os animais estudados estavam parasitados por helmintos. Foram identificados 16 espécies de helmintos, sendo dois trematódeos, um cestódeo, um acantocéfalo e 12 nematódeos. Neste estudo, uma nova espécie denominada Pterygodermatites pluripectinata n. sp. foi descrita, além da descrição de novos registros de hospedeiro e localidade. / Abstract: The 'Caatinga' is a biome exclusive to Brazil, and which was, until very recently, cosidered to be poor in biodiversity. The native dog, Cerdocyon thous (Linnaeus, 1766), is a medium sized canine which is found in almost every region of Brazil. It is the only the only registered canine species to be found in the semi-arid Northeastern region of the country. This study has as its objectives the identification of the helminthfauna common to the Cerdocyon thous found in the Caatinga of the state of Paraiba and to determine the ecological indications of helminthic infection. It is hoped that this will make a favourable addition to the understanding of this little known biome. In this study there were used 58 animals that were found as 'road-kill' on the highways in the municipality of Patos. From these 58 animals the gastroinestinal helminths were colected. All the animals used were found to be infected with helminths. 16 species of helminths were identified. Two being Trematodes, one being a cestode, one being an acantocefalum, and the remaining twelve being nematodes. In this study a new species, Pterygodermatites pluripectinata n. sp. was discoveredand its host and location described. / Mestre

We present a new method to perform reliable matching between different images. This method exploits a projective invariant property between concentric circles and the corresponding projected ellipses to find complete region correspondences centered on interest points. The method matches interest points allowing for a full perspective transformation and exploiting all the available luminance information in the regions. Experiments have been conducted on many different data sets to compare our approach to SIFT local descriptors. The results show the new method offers increased robustness to partial visibility, object rotation in depth, and viewpoint angle change. / Singapore-MIT Alliance (SMA)

The purpose of this paper is to analyze the performance of the Histograms of Oriented Gradients (HOG) as descriptors for traffic signs recognition. The test dataset consists of speed limit traffic signs because of their high inter-class similarities.   HOG features of speed limit signs, which were extracted from different traffic scenes, were computed and a Gentle AdaBoost classifier was invoked to evaluate the different features. The performance of HOG was tested with a dataset consisting of 1727 Swedish speed signs images. Different numbers of HOG features per descriptor, ranging from 36 features up 396 features, were computed for each traffic sign in the benchmark testing. The results show that HOG features perform high classification rate as the Gentle AdaBoost classification rate was 99.42%, and they are suitable to real time traffic sign recognition. However, it is found that changing the number of orientation bins has insignificant effect on the classification rate. In addition to this, HOG descriptors are not robust with respect to sign orientation.

L'accés massiu a les càmeres digitals, els ordinadors personals i a Internet, ha propiciat la creació de grans volums de dades en format digital. En aquest context, cada vegada adquireixen major rellevància totes aquelles eines dissenyades per organitzar la informació i facilitar la seva cerca.Les imatges són un cas particular de dades que requereixen tècniques específiques de descripció i indexació. L'àrea de la visió per computador encarregada de l'estudi d'aquestes tècniques rep el nom de Recuperació d'Imatges per Contingut, en anglès Content-Based Image Retrieval (CBIR). Els sistemes de CBIR no utilitzen descripcions basades en text sinó que es basen en característiques extretes de les pròpies imatges. En contrast a les més de 6000 llengües parlades en el món, les descripcions basades en característiques visuals representen una via d'expressió universal.La intensa recerca en el camp dels sistemes de CBIR s'ha aplicat en àrees de coneixement molt diverses. Així doncs s'han desenvolupat aplicacions de CBIR relacionades amb la medicina, la protecció de la propietat intel·lectual, el periodisme, el disseny gràfic, la cerca d'informació en Internet, la preservació dels patrimoni cultural, etc. Un dels punts importants d'una aplicació de CBIR resideix en el disseny de les funcions de l'usuari. L'usuari és l'encarregat de formular les consultes a partir de les quals es fa la cerca de les imatges. Nosaltres hem centrat l'atenció en aquells sistemes en què la consulta es formula a partir d'una representació pictòrica. Hem plantejat una taxonomia dels sistemes de consulta en composada per quatre paradigmes diferents: Consulta-segons-Selecció, Consulta-segons-Composició-Icònica, Consulta-segons-Esboç i Consulta-segons-Il·lustració. Cada paradigma incorpora un nivell diferent en el potencial expressiu de l'usuari. Des de la simple selecció d'una imatge, fins a la creació d'una il·lustració en color, l'usuari és qui pren el control de les dades d'entrada del sistema. Al llarg dels capítols d'aquesta tesi hem analitzat la influència que cada paradigma de consulta exerceix en els processos interns d'un sistema de CBIR. D'aquesta manera també hem proposat un conjunt de contribucions que hem exemplificat des d'un punt de vista pràctic mitjançant una aplicació final.

No description available.

In today’s fast paced world, as the number of stations of television programming offered is increasing rapidly, time accessible to watch them remains same or decreasing. Sports videos are typically lengthy and they appeal to a massive crowd. Though sports video is lengthy, most of the viewer’s desire to watch specific segments of the video which are fascinating, like a home-run in a baseball or goal in soccer i.e., users prefer to watch highlights to save time. When associated to the entire span of the video, these segments form only a minor share. Hence these videos need to be summarized for effective presentation and data management. This thesis explores the ability to extract highlights automatically using MPEG-7 features and hidden Markov model (HMM), so that viewing time can be reduced. Video is first segmented into scene shots, in which the detection of the shot is the fundamental task. After the video is segmented into shots, extraction of key frames allows a suitable representation of the whole shot. Feature extraction is crucial processing step in the classification, video indexing and retrieval system. Frame features such as color, motion, texture, edges are extracted from the key frames. A baseball highlight contains certain types of scene shots and these shots follow a particular transition pattern. The shots are classified as close-up, out-field, base and audience. I first try to identify the type of the shot using low level features extracted from the key frames of each shot. For the identification of the highlight I use the hidden Markov model using the transition pattern of the shots in time domain. Experimental results suggest that with reasonable accuracy highlights can be extracted from the video.