Face and gait recognition problems are challenging due to largely varying appearances, highly complex pattern distributions, and insufficient training samples. This dissertation focuses on multilinear subspace learning for face and gait recognition, where low-dimensional representations are learned directly from tensorial face or gait objects. This research introduces a unifying multilinear subspace learning framework for systematic treatment of the multilinear subspace learning problem. Three multilinear projections are categorized according to the input-output space mapping as: vector-to-vector projection, tensor-to-tensor projection, and tensor-to-vector projection. Techniques for subspace learning from tensorial data are then proposed and analyzed. Multilinear principal component analysis (MPCA) seeks a tensor-to-tensor projection that maximizes the variation captured in the projected space, and it is further combined with linear discriminant analysis and boosting for better recognition performance. Uncorrelated MPCA (UMPCA) solves for a tensor-to-vector projection that maximizes the captured variation in the projected space while enforcing the zero-correlation constraint. Uncorrelated multilinear discriminant analysis (UMLDA) aims to produce uncorrelated features through a tensor-to-vector projection that maximizes a ratio of the between-class scatter over the within-class scatter defined in the projected space. Regularization and aggregation are incorporated in the UMLDA solution for enhanced performance. Experimental studies and comparative evaluations are presented and analyzed on the PIE and FERET face databases, and the USF gait database. The results indicate that the MPCA-based solution has achieved the best overall performance in various learning scenarios, the UMLDA-based solution has produced the most stable and competitive results with the same parameter setting, and the UMPCA algorithm is effective in unsupervised learning in low-dimensional subspace. Besides advancing the state-of-the-art of multilinear subspace learning for face and gait recognition, this dissertation also has potential impact in both the development of new multilinear subspace learning algorithms and other applications involving tensor objects.

Image descriptors play an important role in image representation and analysis. Multi-resolution image descriptors can effectively characterize complex images and extract their hidden information. Wavelets descriptors have been widely used in multi-resolution image analysis. However, making the wavelets transform shift and rotation invariant produces redundancy and requires complex matching processes. As to other multi-resolution descriptors, they usually depend on other theories or information, such as filtering function, prior-domain knowledge, etc.; that not only increases the computation complexity, but also generates errors. We propose a novel multi-resolution scheme that is capable of transforming any kind of image descriptor into its multi-resolution structure with high computation accuracy and efficiency. Our multi-resolution scheme is based on sub-sampling an image into an odd-even image tree. Through applying image descriptors to the odd-even image tree, we get the relative multi-resolution image descriptors. Multi-resolution analysis is based on downsampling expansion with maximum energy extraction followed by upsampling reconstruction. Since the maximum energy usually retained in the lowest frequency coefficients; we do maximum energy extraction through keeping the lowest coefficients from each resolution level. Our multi-resolution scheme can analyze images recursively and effectively without introducing artifacts or changes to the original images, produce multi-resolution representations, obtain higher resolution images only using information from lower resolutions, compress data, filter noise, extract effective image features and be implemented in parallel processing.

Brain-Computer Interface (BCI) systems aim to provide a non-muscular channel for the brain to control external devices using electrical activities of the brain. These BCI systems can be used in various applications, such as controlling a wheelchair, neuroprosthesis, or speech synthesizer for disabled individuals, navigation in virtual environment, and assisting healthy individuals in performing highly demanding tasks. Motor-imagery BCI systems in particular are based on decoding imagination of motor tasks, e.g., to control the movement of a wheelchair or a mouse curser on the computer screen and move it to the right or left directions by imagining right/left hand movement. During the past decade, there has been a growing interest in utilization of electroencephalogram (EEG) signals for non-invasive motor-imagery BCI systems, due to their low cost, ease of use, and widespread availability. During motor-imagery tasks, multichannel EEG signals exhibit task-specific features in both spatial domain and spectral (or frequency) domain. This thesis studies the statistical characteristics of the multichannel EEG signals in these two domains and proposes a new approach for spatio-spectral feature extraction in motor-imagery BCI systems. This approach is based on the fact that due to the multichannel structure of the EEG data, its spatio-spectral features have a matrix-variate structure. This structure, which has been overlooked in the literate, can be exploited to design more efficient feature extraction methods for motor-imagery BCIs. Towards this end, this research work adopts a matrix-variate Gaussian model for the spatio-spectral features, which assumes a separable Kronecker product structure for the covariance of these features. This separable structure, together with the general properties of the Gaussian model, enables us to design new feature extraction schemes which can operate on the data in its inherent matrix-variate structure to reduce the computational cost of the BCI system while improving its performance. Throughout this thesis, the proposed matrix-variate model and its implications will be studied in various different feature extraction scenarios.

Crack initiation and propagation in the bladed disks of aero-engines caused by high-cycle fatigue under cyclic loads could result in the breakdown of the engines if not detected at an early stage. Although a number of fault detection methods have been reported in the literature, it still remains very challenging to develop a reliable online technique to accurately diagnose defects in bladed disks. One of the main challenges is to characterize signals contaminated by noises. These noises caused by very dynamic engine operation environment. This work presents a new technique for engine bladed disk crack detection, which utilizes advanced analysis of clearance and time-of-arrival signals acquired from blade tip sensors. This technique involves two stages of signal processing: 1) signal pre-processing for noise elimination from predetermined causes; and 2) signal post-processing for characterizing crack initiation and location. Experimental results from the spin rig test were used to validate technique predictions.

Image descriptors play an important role in image representation and analysis. Multi-resolution image descriptors can effectively characterize complex images and extract their hidden information. Wavelet descriptors have been widely used in multi-resolution image analysis. However, making the wavelet transform shift and rotation invariant produces redundancy and requires complex matching processes. As to other multi-resolution descriptors, they usually depend on other methods, such as filtering function, prior-domain knowledge, etc.; that not only increases the computation complexity, but also generates errors. We propose a novel multi-resolution scheme that is capable of transforming any kind of image descriptor into its multi-resolution structure with high computation accuracy and efficiency. Our multi-resolution scheme is based on sub-sampling each image into an odd-even image tree. Through applying image descriptors to the odd-even image tree, we get the relative multi-resolution image descriptors. Multi-resolution analysis is based on downsampling expansion with maximum energy extraction followed by upsampling reconstruction. Since the maximum energy usually retained in the lowest frequency coefficients; we do maximum energy extraction through keeping the lowest coefficients from each resolution level. Our multi-resolution scheme can analyze images recursively and effectively without introducing artifacts or changes to the original images, produce multi-resolution representations, obtain higher resolution images only using information from lower resolutions, compress data, filter noise, extract effective image features and be implemented in parallel processing.

Face detection in video sequence is becoming popular in surveillance applications. The trade-off between obtaining discriminative features to achieve accurate detection versus computational overhead of extracting these features, which affects the classification speed, is a persistent problem. Two ideas are introduced to increase the features’ discriminative power. These ideas are used to implement two frontal face detectors examined on a 2D low-resolution surveillance sequence. First contribution is the parallel classifier. High discriminative power features are achieved by fusing the decision from two different features trained classifiers where each type of the features targets different image structure. Accurate and fast to train classifier is achieved. Co-occurrence of Local Binary Patterns (CoLBP) features is proposed, the pixels of the image are targeted. CoLBP features find the joint probability of multiple LBP features. These features have computationally efficient feature extraction and provide high discriminative features; hence, accurate detection is achieved.

This research focused on the development of a methodology for analyzing functional magnetic resonance imaging (fMRI) data collected from patients with epilepsy in order to map epileptic networks. Epilepsy, a chronic neurological disorder characterized by recurrent, unprovoked seizures, affects up to 1% of the world's population. Antiepileptic drug therapies either do not successfully control seizures or have unacceptable side effects in over 30% of patients. Approximately one-third of patients whose seizures cannot be controlled by medication are candidates for surgical removal of the affected area of the brain, potentially rendering them seizure free. Accurate localization of the epileptogenic focus, i.e., the area of seizure onset, is critical for the best surgical outcome. The main objective of the research was to develop a set of fMRI data features that could be used to distinguish between normal brain tissue and the epileptic focus. To determine the best combination of features from various domains for mapping the focus, genetic programming and several feature selection methods were employed. These composite features and feature sets were subsequently used to train a classifier capable of discriminating between the two classes of voxels. The classifier was then applied to a separate testing set in order to generate maps showing brain voxels labeled as either normal or epileptogenic based on the best feature or set of features. It should be noted that although this work focuses on the application of fMRI analysis to epilepsy data, similar techniques could be used when studying brain activations due to other sources. In addition to investigating in vivo data collected from temporal lobe epilepsy patients with uncertain epileptic foci, phantom (simulated) data were created and processed to provide quantitative measures of the efficacy of the techniques.

Made available in DSpace on 2014-06-11T19:30:31Z (GMT). No. of bitstreams: 0 Previous issue date: 2006-10-31Bitstream added on 2014-06-13T20:47:04Z : No. of bitstreams: 1 gallis_rba_dr_prud.pdf: 3261376 bytes, checksum: 6967d0b5771ef57a837696cfb04efa2f (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) / Este trabalho propõe uma nova metodologia para extração de rodovias utilizando imagens aéreas digitais. A inovação baseia-se no algoritmo de Programação dinâmica (PD), que nesta metodologia realiza o processo de otimização no espaço objeto, e não no espaço imagem como as metodologias tradicionais de extração de rodovias por PD. A feição rodovia é extraída no espaço objeto, o qual implica um rigoroso modelo matemático, que é necessário para estabelecer os pontos entre o espaço imagem e objeto. Necessita-se que o operador forneça alguns pontos sementes no espaço imagem para descrever grosseiramente a rodovia, e estes pontos devem ser transformados para o espaço objeto para inicialização do processo de otimização por PD. Esta metodologia pode operar em diferentes modos (modo mono e estéreo), e com diversos tipos de imagens, incluindo imagens multisensores. Este trabalho apresenta detalhes da metodologia mono e estéreo e também os experimentos realizados e os resultados obtidos. / This work proposes a novel road extraction methodology from digital images. The innovation is based on the dynamic programming (DP) algorithm to carry out the optimisation process in the object space, instead of doing it in the image space such as the DP traditional methodologies. Road features are traced in the object space, which implies that a rigorous mathematical model is necessary to be established between image and object space points. It is required that the operator measures a few seed points in the image space to describe sparsely and coarsely the roads, which must be transformed into the object space to make possible the initialisation of the DP optimisation process. Although the methodology can operate in different modes (mono-plotting or stereoplotting), and with several image types, including multisensor images, this work presents details of our single and stereo image methodology, along with the experimental results.

This work has aimed to implement an easy-to-use application which can be used to navigate through aerial imagery, assign sections of this image for different classes. Based on these category assignments the application can autonomously assign categories to so-far unknown fields, hence it helps the user in further classification. The output of the application is an index file, which can serve as underlying dataset for further analysis of a given area from geographic or economic point-of-view. To fulfil this task the program uses standard MPEG-7 descriptors to perform the feature extraction upon which the classification relies.

Thesis submitted in fulfillment of the requirements for the degree Master of Technology: Electrical Engineering in the Faculty of Engineering at the Cape Peninsula University of Technology / Small satellites, such as CubeSats are mainly utilized for space and earth imaging missions. Imaging CubeSats are equipped with high resolution cameras for the capturing of digital images, as well as mass storage devices for storing the images. The captured images are transmitted to the ground station and subsequently stored in a database. The main problem with stored images in a large image database, identified by researchers and developers within the last number of years, is the retrieval of precise, clear images and overcoming the semantic gap. The semantic gap relates to the lack of correlation between the semantic categories the user requires and the low level features that a content-based image retrieval system offers. Clear images are needed to be usable for applications such as mapping, disaster monitoring and town planning. The main objective of this thesis is the design and development of an image classification, storage and retrieval system for a CubeSat. This system enables efficient classification, storing and retrieval of images that are received on a daily basis from an in-orbit CubeSat. In order to propose such a system, a specific research methodology was chosen and adopted. This entails extensive literature reviews on image classification techniques and image feature extraction techniques, to extract content embedded within an image, and include studies on image database systems, data mining techniques and image retrieval techniques. The literature study led to a requirement analysis followed by the analyses of software development models in order to design the system. The proposed design entails classifying images using content embedded in the image and also extracting image metadata such as date and time. Specific features extraction techniques are needed to extract required content and metadata. In order to achieve extraction of information embedded in the image, colour feature (colour histogram), shape feature (Mathematical Morphology) and texture feature (GLCM) techniques were used. Other major contributions of this project include a graphical user interface which enables users to search for similar images against those stored in the database. An automatic image extractor algorithm was also designed to classify images according to date and time, and colour, texture and shape features extractor techniques were proposed. These ensured that when a user wishes to query the database, the shape objects, colour quantities and contrast contained in an image are extracted and compared to those stored in the database. Implementation and test results concluded that the designed system is able to categorize images automatically and at the same time provide efficient and accurate results. The features extracted for each image depend on colour, shape and texture methods. Optimal values were also incorporated in order to reduce retrieval times. The mathematical morphological technique was used to compute shape objects using erosion and dilation operators, and the co-occurrence matrix was used to compute the texture feature of the image.