abstract: In this thesis we consider the problem of facial expression recognition (FER) from video sequences. Our method is based on subspace representations and Grassmann manifold based learning. We use Local Binary Pattern (LBP) at the frame level for representing the facial features. Next we develop a model to represent the video sequence in a lower dimensional expression subspace and also as a linear dynamical system using Autoregressive Moving Average (ARMA) model. As these subspaces lie on Grassmann space, we use Grassmann manifold based learning techniques such as kernel Fisher Discriminant Analysis with Grassmann kernels for classification. We consider six expressions namely, Angry (AN), Disgust (Di), Fear (Fe), Happy (Ha), Sadness (Sa) and Surprise (Su) for classification. We perform experiments on extended Cohn-Kanade (CK+) facial expression database to evaluate the expression recognition performance. Our method demonstrates good expression recognition performance outperforming other state of the art FER algorithms. We achieve an average recognition accuracy of 97.41% using a method based on expression subspace, kernel-FDA and Support Vector Machines (SVM) classifier. By using a simpler classifier, 1-Nearest Neighbor (1-NN) along with kernel-FDA, we achieve a recognition accuracy of 97.09%. We find that to process a group of 19 frames in a video sequence, LBP feature extraction requires majority of computation time (97 %) which is about 1.662 seconds on the Intel Core i3, dual core platform. However when only 3 frames (onset, middle and peak) of a video sequence are used, the computational complexity is reduced by about 83.75 % to 260 milliseconds at the expense of drop in the recognition accuracy to 92.88 %. / Dissertation/Thesis / Masters Thesis Electrical Engineering 2014

The main theme of the thesis is a video sequence, its production and use in physics education at primary and secondary schools. Teaching enriched of short videos is dynamic and illustrative. Students see a situation that could otherwise only read in an austere entry in a textbook example. It is based on the assumption that the demonstration, in addition to teaching, is an appropriate means for obtaining and verification of theoretical knowledge. Moreover, if the children create it themselves, it will be suitable method to develop their skills and creativity, which is as important in the process of learning as knowledge itself.

The aim of this thesis is to certify appropriateness of using the video sequences during the Physics lessons. The school lessons are complemented by multimedia items which are not found in traditional education and that are why pupils can verify their theoretical knowledge during the work with the video sequences. The main point is the conditions that if the theoretical knowledge and competences given during the Psychics lessons are complemented by the concrete samples, pupils will easily understand and master it. It is a well chosen educational strategy which will lead to better fulfilment of the key competences in Physics.

The aim of this thesis is a study of advanced algorithms for removing interlacing in digital video sequences. The first part of the work is devoted to the theory, basic knowledge and processing characteristics using multimedia data. The second part is devoted to displaying the signal using interlaced and progressive spacing. The following part is focused on methods of conversion between the interlaced line spacing and progressive spacing. The last part deals with implementation of the proposed methods. The algorithm is implemented in C++ language, which provides sufficiently fast processing algorithms. The conclusion of work is focused in testing and verification of the implemented algorithms.

This thesis deals with design, implementation and testing of face recognition system processing video sequences captured by CCTV systems. The use of Local Binary Pattern Histograms (LPBH) and Nearest Neighbor (NN) classifier was suggested according to the survey of face recognition methods. Discrimination power of LBPH features was examined and individual informative features were searched based on Fisher discrimination ratio and mutual correlation. Cluster’s centorid method was utilized for pattern creation because of its best effect on system’s face recognition capability comparing several proposed methods. Software tool for effective face recognition system algorithms performance testing was developed. Video database IFaViD was assembled for training and performance testing of implemented face recognition system.

Cette thèse s'intéresse à la problématique de la classification automatique des séquences vidéo. L'idée est de se démarquer de la méthodologie dominante qui se base sur l'utilisation de caractéristiques conçues manuellement, et de proposer des modèles qui soient les plus génériques possibles et indépendants du domaine. Ceci est fait en automatisant la phase d'extraction des caractéristiques, qui sont dans notre cas générées par apprentissage à partir d'exemples, sans aucune connaissance a priori. Nous nous appuyons pour ce faire sur des travaux existants sur les modèles neuronaux pour la reconnaissance d'objets dans les images fixes, et nous étudions leur extension au cas de la vidéo. Plus concrètement, nous proposons deux modèles d'apprentissage des caractéristiques spatio-temporelles pour la classification vidéo : (i) Un modèle d'apprentissage supervisé profond, qui peut être vu comme une extension des modèles ConvNets au cas de la vidéo, et (ii) Un modèle d'apprentissage non supervisé, qui se base sur un schéma d'auto-encodage, et sur une représentation parcimonieuse sur-complète des données. Outre les originalités liées à chacune de ces deux approches, une contribution supplémentaire de cette thèse est une étude comparative entre plusieurs modèles de classification de séquences parmi les plus populaires de l'état de l'art. Cette étude a été réalisée en se basant sur des caractéristiques manuelles adaptées à la problématique de la reconnaissance d'actions dans les vidéos de football. Ceci a permis d'identifier le modèle de classification le plus performant (un réseau de neurone récurrent bidirectionnel à longue mémoire à court-terme -BLSTM-), et de justifier son utilisation pour le reste des expérimentations. Enfin, afin de valider la généricité des deux modèles proposés, ceux-ci ont été évalués sur deux problématiques différentes, à savoir la reconnaissance d'actions humaines (sur la base KTH), et la reconnaissance d'expressions faciales (sur la base GEMEP-FERA). L'étude des résultats a permis de valider les approches, et de montrer qu'elles obtiennent des performances parmi les meilleures de l'état de l'art (avec 95,83% de bonne reconnaissance pour la base KTH, et 87,57% pour la base GEMEP-FERA). / This thesis focuses on the issue of automatic classification of video sequences. We aim, through this work, at standing out from the dominant methodology, which relies on so-called hand-crafted features, by proposing generic and problem-independent models. This can be done by automating the feature extraction process, which is performed in our case through a learning scheme from training examples, without any prior knowledge. To do so, we rely on existing neural-based methods, which are dedicated to object recognition in still images, and investigate their extension to the video case. More concretely, we introduce two learning-based models to extract spatio-temporal features for video classification: (i) A deep learning model, which is trained in a supervised way, and which can be considered as an extension of the popular ConvNets model to the video case, and (ii) An unsupervised learning model that relies on an auto-encoder scheme, and a sparse over-complete representation. Moreover, an additional contribution of this work lies in a comparative study between several sequence classification models. This study was performed using hand-crafted features especially designed to be optimal for the soccer action recognition problem. Obtained results have permitted to select the best classifier (a bidirectional long short-term memory recurrent neural network -BLSTM-) to be used for all experiments. In order to validate the genericity of the two proposed models, experiments were carried out on two different problems, namely human action recognition (using the KTH dataset) and facial expression recognition (using the GEMEP-FERA dataset). Obtained results show that our approaches achieve outstanding performances, among the best of the related works (with a recognition rate of 95,83% for the KTH dataset, and 87,57% for the GEMEP-FERA dataset).

This work deals with the metal surface wetting problems of molten lead-free solder and monitoring of ongoing processes at the inter-phase interface using the method of evaluation of the height of the molten solder deducted from the video sequences. The work is aimed at evaluating the metal surface wettability, wetting angle determined. Wettability of the metal surface is compared for different types of surface treatments and for different ages of the measured samples. Measurement is performed at the improved workplace, is used to evaluate the newly derived formula for calculating the wetting angle and created program for automatic evaluation of Picture is used to.

This thesis deals with digital image stabilization. It contains a brief overview of the problem and available methods for digital image stabilization. The aim was to design and implement image stabilization system in JAVA, which is designed for RapidMiner. Two new stabilization methods have been proposed. The first is based on the motion estimation and motion compensation using Full-search and Three-step search algorithms. The basis of the second method is the detection of object boundaries. The functionality of the proposed method was tested on video sequences with contain visible shake of the scene, which has beed created for this purpose. Testing results show that with the proper set of input parameters for the object border detection method, successful stabilization of the scene is achieved. The rate of error reduction between images is approximately about 65 to 85%. The output of the method is stabilized image sequence and a set of metadata collected during stabilization, which can be further processed in an environment of RapidMiner.

The aim of this work was to develop an automated methods for removing unwanted objects from video sequences. The proposed method is able to autonomously tackle the static and the moving object with no user intervention into the process. The user only determines the object to deleted.