This Master s thesis proposes a non-assisted procedural method for 3D canyons scenes generation based on techniques of computer graphics, computer vision and graph search algorithm. In order to define all the features to be reproduced in our scenes, we have analyzed several images of real canyons and have categorized them in two canyon features models: a recursive and an ordinary one. The proposed approach manipulates a heightmap, created using Perlin noise, in order to imitate the geological features formation previously analyzed. Several parametrizations are used to guide and constraint the generation of terrains, canyons features, course of river, plain areas, soft slope regions, cliffs and plateaus. This work also uses the Mean Shift algorithm as mechanism of segmentation to define regions of interest. A binary mask, with plain areas, is defined based on a threshold operation by a given data set provided by the Mean Shift algorithm. Thereafter a connected-component labeling algorithm is executed using the previously binary mask. This algorithm finds all plains centroids. Right after that, the Dijkstra s algorithm is performed in order to connect all plain areas, creating a valid path between the centroids. The Dijkstra s algorithm is executed again to define the river s course. Finally, a Gaussian smoothing operation is applied to interpolate the soft slope regions. The combination of all those techniques produces as a result automatically generated feature-rich canyons. / Esta dissertação propõe um método procedural não assistido, baseado em técnicas de computação gráfica, visão computacional e busca em grafos, para a geração de cenários 3D de cânions com foco em jogos digitais. Para definir as características a serem reproduzidas, foram analisadas diversas imagens de cânions reais chegando-se em dois modelos, um comum e outro recursivo. A abordagem proposta manipula um reticulado gerado com ruído de Perlin, moldando assim as características inerentes a essa formação geológica. São levadas em conta as diversas parametrizações necessárias para permitir que o algoritmo construa cânions com curso de rio, áreas de planícies, regiões de encosta suave, estruturas de penhascos e, por fim, planaltos nas regiões mais altas. Para atingir o resultado final, o trabalho utiliza o algoritmo Mean Shift como mecanismo de segmentação, definindo dados e regiões de interesse. Munido dos dados do algoritmo de clusterizacao, é definido um limiar para a criação de uma máscara binária com a definição das planícies. Em um segundo momento, um algoritmo de rotulação de componentes conectados é executado, extraindo-se os centróides de cada planície. Por sua vez, o algoritmo de Dijkstra encaixa-se na definição de rotas que conectam estas planícies. O algoritmo de Dijkstra é, então, executado novamente, tendo por base uma função de custo de inclinação, para definir o curso do rio. Por fim, uma filtragem espacial baseada em um filtro Gaussiano é aplicada para interpolar as regiões de encostas de declive suave. A combinação dessas técnicas gera terrenos com grande variabilidade e com as características inerentes à formação geológica de cânions.

This thesis proposes and implements a system for movement analysis of vehicles on crossroads. It detects and tracks the movement of vehicles in the video, gained from the stationary video camera, which has the view of some crossroad. The trajectories are stored and their number and directions are analysed. The detection was made using cascade classifier. A dataset of 10500 positive and 10500 negative samples has been created to train the classifier. Vehicles are tracked using KCF method. For trajectory clustering, needed by analysis, the Mean Shift method is used. Testing showed, that the overall success of vehicle movement analysis is 92.77%.

La symétrie est une propriété géométrique importante en perception visuelle qui traduit notre perception des correspondances entre les différents objets ou formes présents dans une scène. Elle est utilisée comme élément caractéristique dans de nombreuses applications de la vision par ordinateur (comme par exemple la détection, la segmentation ou la reconnaissance d'objets) mais également comme une caractéristique formelle en sciences de l'art (ou en analyse esthétique). D’importants progrès ont été réalisés ces dernières décennies pour la détection de la symétrie dans les images mais il reste encore de nombreux verrous à lever. Dans cette thèse, nous nous intéressons à la détection des symétries de réflexion, dans des images réelles, à l'échelle globale. Nos principales contributions concernent les étapes d'extraction de caractéristiques et de représentation globale des axes de symétrie. Nous proposons d'abord une nouvelle méthode d'extraction de segments de contours à l'aide de bancs de filtres de Gabor logarithmiques et une mesure de symétrie intersegments basée sur des caractéristiques locales de forme, de texture et de couleur. Cette méthode a remporté la première place à la dernière compétition internationale de symétrie pour la détection mono- et multi-axes. Notre deuxième contribution concerne une nouvelle méthode de représentation des axes de symétrie dans un espace linéaire-directionnel. Les propriétés de symétrie sont représentées sous la forme d'une densité de probabilité qui peut être estimée, de manière non-paramétrique, par une méthode à noyauxbasée sur la distribution de Von Mises-Fisher. Nous montrons que la détection des axes dominants peut ensuite être réalisée à partir d'un algorithme de type "mean-shift” associé à une distance adaptée. Nous introduisons également une nouvelle base d'images pour la détection de symétrie mono-axe dans des photographies professionnelles issue de la base à grande échelle AVA (Aestetic Visual Analysis). Nos différentes contributions obtiennent des résultats meilleurs que les algorithmes de l'état de l'art, évalués sur toutes les bases disponibles publiquement, spécialement dans le cas multi-axes. Nous concluons que les propriétés de symétrie peuvent être utilisées comme des caractéristiques visuelles de niveau sémantique intermédiaire pour l'analyse et la compréhension de photographies. / Symmetry is a fundamental principle of the visual perception to feel the equally distributed weights within foreground objects inside an image. It is used as a significant visual feature through various computer vision applications (i.e. object detection and segmentation), plus as an important composition measure in art domain (i.e. aesthetic analysis). The development of symmetry detection has been improved rapidly since last century. In this thesis, we mainly aim to propose new approaches to detect reflection symmetry inside real-world images in a global scale. In particular, our main contributions concern feature extraction and globalrepresentation of symmetry axes. First, we propose a novel approach that detects global salient edges inside an image using Log-Gabor filter banks, and defines symmetry oriented similarity through textural and color around these edges. This method wins a recent symmetry competition worldwide in single and multiple cases.Second, we introduce a weighted kernel density estimator to represent linear and directional symmetrical candidates in a continuous way, then propose a joint Gaussian-vonMises distance inside the mean-shift algorithm, to select the relevant symmetry axis candidates along side with their symmetrical densities. In addition, we introduce a new challenging dataset of single symmetry axes inside artistic photographies extracted from the large-scale Aesthetic Visual Analysis (AVA) dataset. The proposed contributions obtain superior results against state-of-art algorithms among all public datasets, especially multiple cases in a global scale. We conclude that the spatial and context information of each candidate axis inside an image can be used as a local or global symmetry measure for further image analysis and scene understanding purposes.

Tese de Mestrado Integrado. Engenharia Electrotécnica e de Computadores (Área de Especialização de Energia). Faculdade de Engenharia. Universidade do Porto. 2012

De nombreuses applications en vision par ordinateur nécessitent la détection et le suivi des objets en mouvement dans une séquence d'images. La plupart des méthodes existantes ne donnent de bons résultats que pour des séquences avec des fonds peu changeants, ou si le fond et les objets sont rigides. Le but de cette thèse est de détecter et suivre les objets mobiles dans des séquences (telles que des séquences de conducteurs) ayant un fond dynamique, avec de forts changements d'illumination, de faibles contrastes et éventuellement acquises par une caméra mobile. Cette thèse est décomposée en deux parties. Dans la première, une méthode de détection est proposée. Elle repose sur la définition d'une grille de points décrits par leur mouvement et leur photométrie. Ces points sont ensuite regroupés en "clusters en mouvement" avec un algorithme mean shift à noyau variable et une méthode de sélection automatique de la taille des noyaux. La deuxième partie propose une méthode de suivi combinant des distributions de couleur et de mouvement, la prédiction de l'objet et des observations extérieures (pouvant être les clusters en mouvement) dans une fonction d'énergie minimisée par coupe minimale/flot maximal dans un graphe. Les algorithmes de détection et de suivi sont validés sur différentes séquences aux contenus dynamiques complexes.

As the number of surveillance cameras and mobile platforms with cameras increases, automated detection and tracking of objects on these systems gain importance. There are various tracking methods designed for stationary or moving cameras. For stationary cameras, correspondence based tracking methods along with background subtraction have various advantages such as enabling detection of object entry and exit in a scene. They also provide robust tracking when the camera is static. However, they fail when the camera is moving. Conversely, histogram based methods such as mean shift enables object tracking on moving camera cases. Though, with mean shift object&rsquo / s entry and exit cannot be detected automatically which means a new object&rsquo / s manual initialization is required. In this thesis, we propose a dual-mode object tracking system which combines the benefits of correspondence based tracking and mean shift tracking. For each frame, a reliability measure based on background update rate is calculated. Interquartile Range is used for finding outliers on this measure and camera movement is detected. If the camera is stationary, correspondence based tracking is used and when camera is moving, the system switches to the mean shift tracking mode until the reliability of correspondence based tracking is sufficient according to the reliability measure. The results demonstrate that, in stationary camera mode, new objects can be detected automatically by correspondence based tracking along with background subtraction. When the camera starts to move, generation of false objects by correspondence based tracking is prevented by switching to mean shift tracking mode and handing over the correct bounding boxes with a seamless operation which enables continuous tracking.

Road extraction from satellite or aerial images is a popular issue in remote sensing. Extracted road maps or networks can be used in various applications. Normally, maps for roads are available in geographic information systems (GIS), however these informations are not being produced automatically. Generally they are formed with the aid of human. Road extraction algorithms are trying to detect the roads from satellite or aerial images with the minimum in-teraction of human. Aim of this thesis is to analyze a previously defined algorithm about road extraction and to present alternatives and possible improvements to this algorithm. The base-line algorithm and proposed alternative algorithm and steps are based on mean-shift segmen-tation procedure. Proposed alternative methods are generally based on structural features of the image. Firstly, fundamental definitions of applied algorithms and methods are explained, mathematical definitions and visual examples are given for better understanding. Then, the chosen baseline algorithm and its alternatives are explained in detail. After the presentation of alternative methods, experimental results and inferences which are obtained during the implementation and analysis of mentioned algorithms and methods are presented.

Due to increasing demand on deployable surveillance systems in recent years, object tracking and activity recognition are receiving considerable attention in the research community. This thesis contributes to both the tracking and the activity recognition components of a surveillance system. In particular, for the tracking component, we propose two different approaches for tracking objects in video acquired by mobile cameras, each of which uses a different object shape representation. The first approach tracks the centroids of the objects in Forward Looking Infrared Imagery (FLIR) and is suitable for tracking objects that appear small in airborne video. The second approach tracks the complete contours of the objects, and is suitable for higher level vision problems, such as activity recognition, identification and classification. Using the contours tracked by the contour tracker, we propose a novel representation, called the action sketch, for recognizing human activities. Object Tracking in Airborne Imagery: Images obtained from an airborne vehicle generally appear small and can be represented by geometric shapes such as circle or rectangle. After detecting the object position in the first frame, the proposed object tracker models the intensity and the local standard deviation of the object region defined by the shape model. It then tracks the objects by computing the mean-shift vector that minimizes the distance between the kernel distribution for the hypothesized object and its prior. In cases when the ego-motion of the sensor causes the object to move more than the operational limits of the tracking module, a multi-resolution global motion compensation using the Gabor responses of consecutive frames is performed. The experiments performed on the AMCOM FLIR data set show the robustness of the proposed method, which combines automatic model update and global motion compensation into one framework. Contour Tracker: Contour tracking is performed by evolving an initial contour toward the correct object boundaries based on discriminant analysis, which is formulated as a variational calculus problem. Once the contour is initialized, the method generates an online shape model for the object along with the color and the texture priors for both the object and the background regions. A priori texture and color PDFs of the regions are then fused based on the discrimination properties of the features between the object and the background models. The models are then used to compute the posteriori contour likelihood and the evolution is obtained by the Maximum a Posteriori Estimation process, which updates the contour in the gradient ascent direction of the proposed energy functional. During occlusion, the online shape model is used to complete the missing object region. The proposed energy functional unifies commonly used boundary and region based contour approaches into a single framework through a support region defined around the hypothesized object contour. We tested the robustness of the proposed contour tracker using several real sequences and have verified qualitatively that the contours of the objects are perfectly tracked. Behavior Analysis: We propose a novel approach to represent human actions by modeling the dynamics (motion) and the structure (shape) of the objects in video. Both the motion and the shape are modeled using a compact representation, which is called the “action sketch”. An action sketch is a view invariant representation obtained by analyzing important changes that occur during the motion of the objects. When an actor performs an action in 3D, the points on the actor generate space-time trajectories in four dimensions (x, y, z,t). Projection of the world to the imaging coordinates converts the space-time trajectories into the spatiotemporal trajectories in three dimensions (x, y,t). A set of spatio-temporal trajectories constitute a 3D volume, which we call an “action volume”. This volume can be treated as a 3D object in the (x, y,t) space. The action sketch is generated from the action volume by analyzing the differential geometric surface properties, such as peaks, pits, valleys and ridges. These properties reflect the changes in the speed, the motion direction and the shape of the performing actor. We perform action recognition by computing a view invariant distance measure between the sketch generated from the input video and the set of known sketches in the database. Experimental results are provided for twenty eight actions.

This thesis focuses on the comprehensive and thorough studies on the geospatial processing of airborne (full) waveform lidar data, including waveform modeling, direct georeferencing, and precise georeferencing with self-calibration.<div><br></div><div>Both parametric and nonparametric approaches of waveform decomposition are studied. The traditional parametric approach assumes that the returned waveforms follow a Gaussian mixture model where each component is a Gaussian. However, many real examples show that the waveform components can be neither Gaussian nor symmetric. To address the problem, this thesis proposes a nonparametric mixture model to represent lidar waveforms without any constraints on the shape of the waveform components. To decompose the waveforms, a fuzzy mean-shift algorithm is then developed. This approach has the following properties: 1) it does not assume that the waveforms follow any parametric or functional distributions; 2) the waveform decomposition is treated as a fuzzy data clustering problem and the number of components is determined during the process of decomposition; 3) neither peak selection nor noise floor filtering prior to the decomposition is needed; and 4) the range measurement is not affected by the process of noise filtering. In addition, the fuzzy mean-shift approach is about three times faster than the conventional expectationmaximization algorithm and tends to lead to fewer artifacts in the resultant digital elevation model. <br></div><div><br></div><div>This thesis also develops a framework and methodology of self-calibration that simultaneously determines the waveform geospatial position and boresight angles. Besides using the flight trajectory and plane attitude recorded by the onboard GPS receiver and inertial measurement unit, the framework makes use of the publically accessible digital elevation models as control over the study area. Compared to the conventional calibration and georeferencing method, the new development has minimum requirements on ground truth: no extra ground control, no planar objects, and no overlap flight strips are needed. Furthermore, it can also solve the problem of clock synchronization and boresight calibration simultaneously. Through a developed two-stage optimization strategy, the self-calibration approach can resolve both the time synchronization bias and boresight misalignment angles to achieve a stable and correct solution. As a result, a consistency of 0.8662 meter is achieved between the waveform derived digital elevation model and the reference one without systematic trend. Such experiments demonstrate the developed method is a necessary and more economic alternative to the conventional, high demanding georeferencing and calibration approach, especially when no or limited ground control is available.<br></div>

Les images satellitaires ont trouvées une large application pour l'analyse des ressources naturelles et des activités humaines. Les images à haute résolution, e.g., SPOT5, sont très nombreuses. Ceci donne un grand intérêt afin de développer de nouveaux aspects théoriques et des outils pour la fouille d'images. L'objectif de la thèse est la fouille non-supervisée d'images et inclut trois parties principales. Dans la première partie nous démontrons le contenu d'images à haute résolution. Nous décrivons les zones d'images par les caractéristiques texturelles et géométriques. Les algorithmes de clustering sont présentés dans la deuxième partie. Une étude de critères de validité et de mesures d'information est donnée pour estimer la qualité de clustering. Un nouveau critère basé sur la Longueur de Description Minimale (LDM) est proposé pour estimer le nombre optimal de clusters. Par ailleurs, nous proposons un nouveau algorithme hiérarchique basé sur le critère LDM à noyau. Une nouvelle méthode de ''combinaison de clustering'' est présentée dans la thèse pour profiter de différents algorithmes de clustering. Nous développons un algorithme hiérarchique pour optimiser la fonction objective basée sur une matrice de co-association. Une deuxième méthode est proposée qui converge à une solution globale. Nous prouvons que le minimum global peut être trouvé en utilisant l'algorithme de type ''mean shift''. Les avantages de cette méthode sont une convergence rapide et une complexité linéaire. Dans la troisième partie de la thèse un protocole complet de la fouille d'images est proposé. Différents clusterings sont représentés via les relations sémantiques entre les concepts.