Thesis (MScEng (Mathematical Sciences. Applied Mathematics))--University of Stellenbosch, 2007. / The face is one of the most characteristic parts of the human body and has been used by people for personal identification for centuries. In this thesis an automatic process for frontal face recognition from 2–dimensional images is presented based on principal component analysis. The goal is to use these concepts in eventual face–recognizing login software. The first step is detecting faces in images that are allowed a certain degree of clutter. This is achieved by skin colour detection in the HSV colourspace. This process indicates the area of the image most likely corresponding to the face. Extracting the face is achieved by morphological processing of this area of the image. The face is then normalized by a transformation that uses the eye coordinates as input. Automatic eye detection is implemented based on colour analysis of the facial images and a 91.1% success rate is achieved. Recognition of the normalized faces is achieved using eigenfaces. To calculate these, a large enough database of facial images is needed. The xm2vts database is used in this thesis as the images have very constant lighting conditions throughout – an important factor affecting the accuracy of the recognition stage. Distinction is also made between identification and verification of faces. For identification, up to 80.1% accuracy is achieved, while for verification, the equal error rate is approximately 3.5%.

With the growing installation of surveillance video cameras in both private and public areas, it is an immediate requirement to develop intelligent video analysis system for the large-scale camera network. As a prerequisite step of person tracking and person retrieval in intelligent video analysis, person re-identification, which targets in matching person images across camera views is an important topic in computer vision community and has been received increasing attention in the recent years. In the supervised learning methods, the person re-identification task is formulated as a classification problem to extract matched person images/videos (positives) from unmatched person images/videos (negatives). Although the state-of-the-art supervised classification models could achieve encouraging re-identification performance, the assumption that label information is available for all the cameras, is impractical in large-scale camera network. That is because collecting the label information of every training subject from every camera in the large-scale network can be extremely time-consuming and expensive. While the unsupervised learning methods are flexible, their performance is typically weaker than the supervised ones. Though sufficient labels of the training subjects are not available from all the camera views, it is still reasonable to collect sufficient labels from a pair of camera views in the camera network or a few labeled data from each camera pair. Along this direction, we address two scenarios of person re-identification in large-scale camera network in this thesis, i.e. unsupervised domain adaptation and semi-supervised learning and proposed three methods to learn discriminative model using all available label information and domain knowledge in person re-identification. In the unsupervised domain adaptation scenario, we consider data with sufficient labels as the source domain, while data from the camera pair missing label information as the target domain. A novel domain adaptive approach is proposed to estimate the target label information and incorporate the labeled data from source domain with the estimated target label information for discriminative learning. Since the discriminative constraint of Support Vector Machines (SVM) can be relaxed into a necessary condition, which only relies on the mean of positive pairs (positive mean), a suboptimal classification model learning without target positive data can be those using target positive mean. A reliable positive mean estimation is given by using both the labeled data from the source domain and potential positive data selected from the unlabeled data in the target domain. An Adaptive Ranking Support Vector Machines (AdaRSVM) method is also proposed to improve the discriminability of the suboptimal mean based SVM model using source labeled data. Experimental results demonstrate the effectiveness of the proposed method. Different from the AdaRSVM method that using source labeled data, we can also improve the above mean based method by adapting it onto target unlabeled data. In more general situation, we improve a pre-learned classifier by adapting it onto target unlabeled data, where the pre-learned classifier can be domain adaptive or learned from only source labeled data. Since it is difficult to estimate positives from the imbalanced target unlabeled data, we propose to alternatively estimate positive neighbors which refer to data close to any true target positive. An optimization problem for positive neighbor estimation from unlabeled data is derived and solved by aligning the cross-person score distributions together with optimizing for multiple graphs based label propagation. To utilize the positive neighbors to learn discriminative classification model, a reliable multiple region metric learning method is proposed to learn a target adaptive metric using regularized affine hulls of positive neighbors as positive regions. Experimental results demonstrate the effectiveness of the proposed method. In the semi-supervised learning scenario, we propose a discriminative feature learning using all available information from the surveillance videos. To enrich the labeled data from target camera pair, image sequences (videos) of the tagged persons are collected from the surveillance videos by human tracking. To extract the discriminative and adaptable video feature representation, we propose to model the intra-view variations by a video variation dictionary and a video level adaptable feature by multiple sources domain adaptation and an adaptability-discriminability fusion. First, a novel video variation dictionary learning is proposed to model the large intra-view variations and solved as a constrained sparse dictionary learning problem. Second, a frame level adaptable feature is generated by multiple sources domain adaptation using the variation modeling. By mining the discriminative information of the frames from the reconstruction error of the variation dictionary, an adaptability-discriminability (AD) fusion is proposed to generate the video level adaptable feature. Experimental results demonstrate the effectiveness of the proposed method.

Biometric systems are being increasingly used, from daily entertainment to critical applications such as security access and identity management. It is known that biometric systems should meet the stringent requirement of low error rate. In addition, for critical applications, the security and privacy issues of biometric systems are required to be concerned. Otherwise, severe consequence such as the unauthorized access (security) or the exposure of identity-related information (privacy) can be caused. Therefore, it is imperative to study the vulnerability to potential attacks and identify the corresponding risks. Furthermore, the countermeasures should also be devised and patched on the systems. In this thesis, we study the security and privacy issues in biometric systems. We first make an attempt to reconstruct raw biometric data from biometric templates and demonstrate the security and privacy issues caused by the data reconstruction. Then, we make two attempts to protect biometric templates from being reconstructed and improve the state-of-the-art biometric template protection techniques.

Made available in DSpace on 2015-09-17T15:26:23Z (GMT). No. of bitstreams: 0 Previous issue date: 2015-05-15. Added 1 bitstream(s) on 2015-09-17T15:45:43Z : No. of bitstreams: 1 000846199.pdf: 4785482 bytes, checksum: d06441c7f33c2c6fc4bfe273884b0d5a (MD5) / O presente trabalho tem como objetivo realizar o reconhecimento de identidade por meio de um método baseado nos Sistemas Imunológicos Artificiais de Seleção Negativa. Para isso, foram explorados os tipos de recursos e alternativas adequadas para a análise de expressões faciais 3D, abordando a técnica de Padrão Binário que tem sido aplicada com sucesso para o problema 2D. Inicialmente, a geometria facial 3D foi convertida em duas representações em 2D, a Depth Map e a APDI, que foram implementadas com uma variedade de tipos de recursos, tais como o Local Phase Quantisers, Gabor Filters e Monogenic Filters, a fim de produzir alguns descritores para então fazer-se a análise de expressões faciais. Posteriormente, aplica-se o Algoritmo de Seleção Negativa onde são realizadas comparações e análises entre as imagens e os detectores previamente criados. Havendo afinidade entre as imagens previamente estabelecidas pelo operador, a imagem é classificada. Esta classificação é chamada de casamento. Por fim, para validar e avaliar o desempenho do método foram realizados testes com imagens diretamente da base de dados e posteriormente com dez descritores desenvolvidos a partir dos padrões binários. Esses tipos de testes foram realizados tendo em vista três objetivos: avaliar quais os melhores descritores e as melhores expressões para se realizar o reconhecimento de identidade e, por fim, validar o desempenho da nova solução de reconhecimento de identidades baseado nos Sistemas Imunológicos Artificiais. Os resultados obtidos pelo método apresentaram eficiência, robustez e precisão no reconhecimento de identidade facial / This work aims to perform the identity recognition by a method based on Artificial Immune Systems, the Negative Selection Algorithm. Thus, the resources and adequate alternatives for analyzing 3D facial expressions were explored, exploring the Binary Pattern technique that is successfully applied for the 2D problem. Firstly, the 3D facial geometry was converted in two 2D representations. The Depth Map and the Azimuthal Projection Distance Image were implemented with other resources such as the Local Phase Quantisers, Gabor Filters and Monogenic Filters to produce descriptors to perform the facial expression analysis. Afterwards, the Negative Selection Algorithm is applied, and comparisons and analysis with the images and the detectors previously created are done. If there is affinity with the images, than the image is classified. This classification is called matching. Finally, to validate and evaluate the performance of the method, tests were realized with images from the database and after with ten descriptors developed from the binary patterns. These tests aim to: evaluate which are the best descriptors and the best expressions to recognize the identities, and to validate the performance of the new solution of identity recognition based on Artificial Immune Systems. The results show efficiency, robustness and precision in recognizing facial identity

Past studies, that studied facial attractiveness, focus mainly on fluctuating asymmetry, averageness, masculinity, femininity, and skin coloration influences on attractiveness assessment. Their findings have been used by intercultural studies. They have revealed that people from diverse areas prefer differential development of sexually dimorphic cues, according to, e.g. local health and economic situation. We have done an online questionnaire survey. We included Czech (N = 100), Iranian (N = 87) and Turkish (N = 185) facial stimuli and raters of both sexes. We have studied whether members of all the populations would utilize the facial colouration and sexual dimorphism cues in a similar way. We have also tested whether there were differences in the facial width to height ratio (fWHR) between the sexes in all the populations. We have also wondered if the populations differed in the variance of facial skin colouration. Raters from all the populations saw feminine women as attractive. Turks and Czechs found masculine men as attractive, while Iranian women found more average Iranian men as attractive. Averageness positively predicted the attractiveness of the Czech and Turkish faces of both sexes. Older and more average Czech men were considered more masculine by Czech women. More average and younger Czech...

M.Ing. / Virtual classrooms and online-learning are growing in popularity, but there are still some factors limiting the potential. Limited bandwidth for audio and video, the resultant transmission quality and limited feedback during virtual classroom sessions are some of the problems that need to be addressed. This thesis presents information on the design and implementation of various components of a virtual classroom system for researching methods of student feedback with a focus on bandwidth conservation. A facial feature technique is implemented and used within the system to determine the viability of using facial feature extraction to provide and prioritise feedback from students to teacher while conserving bandwidth. This allows a teacher to estimate the comprehension level of the class and individual students based on student images. A server determines which student terminal transmits its images to the teacher using data obtained from the facial feature extraction process. Feedback is improved as teachers adapt to class circumstances using experience gained in traditional classrooms. Feedback is also less reliant on intentional student participation. New page-turner, page suggestion and class activity components are presented as possible methods for improving student feedback. In particular, the effect of virtual classroom system parameters on feedback delays and bandwidth usage is investigated. In general, delays are increased as bandwidth requirements decrease. The system shows promise for future use in research on facial feature extraction, student feedback and bandwidth conservation in virtual classrooms.

D. Ing. / This thesis describes a real world application in the field of pattern recognition. License plate recognition and face recognition algorithms are combined to implement automated access control at the gates of RAU campus. One image of the license plate and three images of the driver’s face are enough to check if the person driving a particular car into campus is the same as the person driving this car out. The license plate recognition module is based on learning vector quantization and performs well enough to be used in a realistic environment. The face recognition module is based on the Bayes rule and while performing satisfactory, extensive research is still necessary before this system can be implemented in real life. The main reasons for failure of the system were identified as the variable lighting and insufficient landmarks for effective warping.

We report on the design, implementation and performance issues of the DRUBIS (Distributed Rhodes University Biometric Identification System) experimentation framework. The Principal Component Analysis (PCA) face-recognition approach is used as a case study. DRUBIS is a flexible experimentation framework, distributed over a number of modules that are easily pluggable and swappable, allowing for the easy construction of prototype systems. Web services are the logical means of distributing DRUBIS components and a number of prototype applications have been implemented from this framework. Different popular PCA face-recognition related experiments were used to evaluate our experimentation framework. We extract recognition performance measures from these experiments. In particular, we use the framework for a more indepth study of the suitability of the DFFS (Difference From Face Space) metric as a means for image classification in the area of race and gender determination.

Recently, automatic age progression has gained popularity due to its numerous applications. Among these is the frequent search for missing people, in the UK alone up to 300,000 people are reported missing every year. Although many algorithms have been proposed, most of the methods are affected by image noise, illumination variations, and facial expressions. Furthermore, most of the algorithms use a pattern caricaturing approach which infers ages by manipulating the target image and a template face formed by averaging faces at the intended age. To this end, this thesis investigates the problem with a view to tackling the most prominent issues associated with the existing algorithms. Initially using active appearance models (AAM), facial features are extracted and mapped to people’s ages, afterward a formula is derived which allows the convenient generation of age progressed images irrespective of whether the intended age exists in the training database or not. In order to handle image noise as well as varying facial expressions, a nonlinear appearance model called kernel appearance model (KAM) is derived. To illustrate the real application of automatic age progression, both AAM and KAM based algorithms are then used to synthesise faces of two popular long missing British and Irish kids; Ben Needham and Mary Boyle. However, both statistical techniques exhibit image rendering artefacts such as low-resolution output and the generation of inconsistent skin tone. To circumvent this problem, a hybrid texture enhancement pipeline is developed. To further ensure that the progressed images preserve people’s identities while at the same time attaining the intended age, rigorous human and machine based tests are conducted; part of this tests resulted to the development of a robust age estimation algorithm. Eventually, the results of the rigorous assessment reveal that the hybrid technique is able to handle all existing problems of age progression with minimal error. / National Information Technology Development Agency of Nigeria (NITDA)

No description available.