Desarrollo de un sistema de reconocimiento facial

El objetivo principal alrededor del cual se desenvuelve este proyecto es el desarrollo de un sistema de reconocimiento facial. Entre sus objetivos específicos se encuentran: realizar una primera aproximación sobre las técnicas de reconocimiento facial existentes en la actualidad, elegir una aplicación donde pueda ser útil el reconocimiento facial, diseñar y desarrollar un programa en MATLAB que lleve a cabo la función de reconocimiento facial, y evaluar el funcionamiento del sistema desarrollado. Este documento se encuentra dividido en cuatro partes: INTRODUCCIÓN, MARCO TEÓRICO, IMPLEMENTACIÓN, y RESULTADOS, CONCLUSIONES Y LÍNEAS FUTURAS. En la primera parte, se hace una introducción relativa a la actualidad del reconocimiento facial y se comenta brevemente sobre las técnicas existentes para desarrollar un sistema biométrico de este tipo. En ella se justifican también aquellas técnicas que acabaron formando parte de la implementación. En la segunda parte, el marco teórico, se explica la estructura general que tiene un sistema de reconocimiento biométrico, así como sus modos de funcionamiento, y las tasas de error utilizadas para evaluar y comparar su rendimiento. Así mismo, se lleva a cabo una descripción más profunda sobre los conceptos y métodos utilizados para efectuar la detección y reconocimiento facial en la tercera parte del proyecto. La tercera parte abarca una descripción detallada de la solución propuesta. En ella se explica el diseño, características y aplicación de la implementación; que trata de un programa elaborado en MATLAB con interfaz gráfica, y que utiliza cuatro sistemas de reconocimiento facial, basados cada uno en diferentes técnicas: Análisis por componentes principales, análisis lineal discriminante, wavelets de Gabor, y emparejamiento de grafos elásticos. El programa ofrece además la capacidad de crear y editar una propia base de datos con etiquetas, dándole aplicación directa sobre el tema que se trata. Se proponen además una serie de características con el objetivo de ampliar y mejorar las funcionalidades del programa diseñado. Dentro de dichas características destaca la propuesta de un modo de verificación híbrido aplicable a cualquier rama de la biometría y un programa de evaluación capaz de medir, graficar, y comparar las configuraciones de cada uno de los sistemas de reconocimiento implementados. Otra característica destacable es la herramienta programada para la creación de grafos personalizados y generación de modelos, aplicable a reconocimiento de objetos en general. En la cuarta y última parte, se presentan al principio los resultados obtenidos. En ellos se contemplan y analizan las comparaciones entre las distintas configuraciones de los sistemas de reconocimiento implementados para diferentes bases de datos (una de ellas formada con imágenes con condiciones de adquisición no controladas). También se miden las tasas de error del modo de verificación híbrido propuesto. Finalmente, se extraen conclusiones, y se proponen líneas futuras de investigación. ABSTRACT The main goal of this project is to develop a facial recognition system. To meet this end, it was necessary to accomplish a series of specific objectives, which were: researching on the existing face recognition technics nowadays, choosing an application where face recognition might be useful, design and develop a face recognition system using MATLAB, and measure the performance of the implemented system. This document is divided into four parts: INTRODUCTION, THEORTICAL FRAMEWORK, IMPLEMENTATION, and RESULTS, CONCLUSSIONS AND FUTURE RESEARCH STUDIES. In the first part, an introduction is made in relation to facial recognition nowadays, and the techniques used to develop a biometric system of this kind. Furthermore, the techniques chosen to be part of the implementation are justified. In the second part, the general structure and the two basic modes of a biometric system are explained. The error rates used to evaluate and compare the performance of a biometric system are explained as well. Moreover, a description of the concepts and methods used to detect and recognize faces in the third part is made. The design, characteristics, and applications of the systems put into practice are explained in the third part. The implementation consists in developing a program with graphical user interface made in MATLAB. This program uses four face recognition systems, each of them based on a different technique: Principal Component Analysis (PCA), Fisher’s Linear Discriminant (FLD), Gabor wavelets, and Elastic Graph Matching (EGM). In addition, with this implementation it is possible to create and edit one´s tagged database, giving it a direct application. Also, a group of characteristics are proposed to enhance the functionalities of the program designed. Among these characteristics, three of them should be emphasized in this summary: A proposal of an hybrid verification mode of a biometric system; and an evaluation program capable of measuring, plotting curves, and comparing different configurations of each implemented recognition system; and a tool programmed to create personalized graphs and models (tagged graph associated to an image of a person), which can be used generally in object recognition. In the fourth and last part of the project, the results of the comparisons between different configurations of the systems implemented are shown for three databases (One of them created with pictures taken under non-controlled environments). The error rates of the proposed hybrid verification mode are measured as well. Finally, conclusions are extracted and future research studies are proposed.

An investigation of the performance of a new Mechanical thrombectomy device using Bond Graph modelling: application to the extraction of blood clots in the middle cerebral artery

A number of thrombectomy devices using a variety of methods have now been developed to facilitate clot removal. We present research involving one such experimental device recently developed in the UK, called a ‘GP’ Thrombus Aspiration Device (GPTAD). This device has the potential to bring about the extraction of a thrombus. Although the device is at a relatively early stage of development, the results look encouraging. In this work, we present an analysis and modeling of the GPTAD by means of the bond graph technique; it seems to be a highly effective method of simulating the device under a variety of conditions. Such modeling is useful in optimizing the GPTAD and predicting the result of clot extraction. The aim of this simulation model is to obtain the minimum pressure necessary to extract the clot and to verify that both the pressure and the time required to complete the clot extraction are realistic for use in clinical situations, and are consistent with any experimentally obtained data. We therefore consider aspects of rheology and mechanics in our modeling.

Uniformly Reweighted Belief Propagation: A Factor Graph Approach

Tree-reweighted belief propagation is a message passing method that has certain advantages compared to traditional belief propagation (BP). However, it fails to outperform BP in a consistent manner, does not lend itself well to distributed implementation, and has not been applied to distributions with higher-order interactions. We propose a method called uniformly-reweighted belief propagation that mitigates these drawbacks. After having shown in previous works that this method can substantially outperform BP in distributed inference with pairwise interaction models, in this paper we extend it to higher-order interactions and apply it to LDPC decoding, leading performance gains over BP.

A graph-based approach for latency modeling and optimization in multiview video encoding

We present a novel framework for encoding latency analysis of arbitrary multiview video coding prediction structures. This framework avoids the need to consider an specific encoder architecture for encoding latency analysis by assuming an unlimited processing capacity on the multiview encoder. Under this assumption, only the influence of the prediction structure and the processing times have to be considered, and the encoding latency is solved systematically by means of a graph model. The results obtained with this model are valid for a multiview encoder with sufficient processing capacity and serve as a lower bound otherwise. Furthermore, with the objective of low latency encoder design with low penalty on rate-distortion performance, the graph model allows us to identify the prediction relationships that add higher encoding latency to the encoder. Experimental results for JMVM prediction structures illustrate how low latency prediction structures with a low rate-distortion penalty can be derived in a systematic manner using the new model.

Probabilistic Atlas Based Segmentation Using Affine Moment Descriptors and Graph-Cuts

We show a procedure for constructing a probabilistic atlas based on affine moment descriptors. It uses a normalization procedure over the labeled atlas. The proposed linear registration is defined by closed-form expressions involving only geometric moments. This procedure applies both to atlas construction as atlas-based segmentation. We model the likelihood term for each voxel and each label using parametric or nonparametric distributions and the prior term is determined by applying the vote-rule. The probabilistic atlas is built with the variability of our linear registration. We have two segmentation strategy: a) it applies the proposed affine registration to bring the target image into the coordinate frame of the atlas or b) the probabilistic atlas is non-rigidly aligning with the target image, where the probabilistic atlas is previously aligned to the target image with our affine registration. Finally, we adopt a graph cut - Bayesian framework for implementing the atlas-based segmentation.

Cross-lingual ontology matching as a challenge for the Multilingual Semantic Web

Recently, the Semantic Web has experienced significant advancements in standards and techniques, as well as in the amount of semantic information available online. Nevertheless, mechanisms are still needed to automatically reconcile information when it is expressed in different natural languages on the Web of Data, in order to improve the access to semantic information across language barriers. In this context several challenges arise [1], such as: (i) ontology translation/localization, (ii) cross-lingual ontology mappings, (iii) representation of multilingual lexical information, and (iv) cross-lingual access and querying of linked data. In the following we will focus on the second challenge, which is the necessity of establishing, representing and storing cross-lingual links among semantic information on the Web. In fact, in a “truly” multilingual Semantic Web, semantic data with lexical representations in one natural language would be mapped to equivalent or related information in other languages, thus making navigation across multilingual information possible for software agents.

MultiFarm: A benchmark for multilingual ontology matching

In this paper we present the MultiFarm dataset, which has been designed as a benchmark for multilingual ontology matching. The MultiFarm dataset is composed of a set of ontologies translated in different languages and the corresponding alignments between these ontologies. It is based on the OntoFarm dataset, which has been used successfully for several years in the Ontology Alignment Evaluation Initiative (OAEI). By translating the ontologies of the OntoFarm dataset into eight different languages – Chinese, Czech, Dutch, French, German, Portuguese, Russian, and Spanish – we created a comprehensive set of realistic test cases. Based on these test cases, it is possible to evaluate and compare the performance of matching approaches with a special focus on multilingualism.

Graph based study of allergen cross-reactivity of plant lipid transfer proteins (LTPs) using microarray in a multicenter study.

The study of cross-reactivity in allergy is key to both understanding. the allergic response of many patients and providing them with a rational treatment In the present study, protein microarrays and a co-sensitization graph approach were used in conjunction with an allergen microarray immunoassay. This enabled us to include a wide number of proteins and a large number of patients, and to study sensitization profiles among members of the LTP family. Fourteen LTPs from the most frequent plant food-induced allergies in the geographical area studied were printed into a microarray specifically designed for this research. 212 patients with fruit allergy and 117 food-tolerant pollen allergic subjects were recruited from seven regions of Spain with different pollen profiles, and their sera were tested with allergen microarray. This approach has proven itself to be a good tool to study cross-reactivity between members of LTP family, and could become a useful strategy to analyze other families of allergens.

Time series irreversibility: a visibility graph approach

We propose a method to measure real-valued time series irreversibility which combines two different tools: the horizontal visibility algorithm and the Kullback-Leibler divergence. This method maps a time series to a directed network according to a geometric criterion. The degree of irreversibility of the series is then estimated by the Kullback-Leibler divergence (i.e. the distinguishability) between the in and out degree distributions of the associated graph. The method is computationally efficient and does not require any ad hoc symbolization process. We find that the method correctly distinguishes between reversible and irreversible stationary time series, including analytical and numerical studies of its performance for: (i) reversible stochastic processes (uncorrelated and Gaussian linearly correlated), (ii) irreversible stochastic processes (a discrete flashing ratchet in an asymmetric potential), (iii) reversible (conservative) and irreversible (dissipative) chaotic maps, and (iv) dissipative chaotic maps in the presence of noise. Two alternative graph functionals, the degree and the degree-degree distributions, can be used as the Kullback-Leibler divergence argument. The former is simpler and more intuitive and can be used as a benchmark, but in the case of an irreversible process with null net current, the degree-degree distribution has to be considered to identify the irreversible nature of the series

Segmentation of RV in 4D Cardiac MR Volumes using region-merging graph cuts

Non-invasive quantitative assessment of the right ventricular anatomical and functional parameters is a challenging task. We present a semi-automatic approach for right ventricle (RV) segmentation from 4D MR images in two variants, which differ in the amount of user interaction. The method consists of three main phases: First, foreground and background markers are generated from the user input. Next, an over-segmented region image is obtained applying a watershed transform. Finally, these regions are merged using 4D graph-cuts with an intensity based boundary term. For the first variant the user outlines the inside of the RV wall in a few end-diastole slices, for the second two marker pixels serve as starting point for a statistical atlas application. Results were obtained by blind evaluation on 16 testing 4D MR volumes. They prove our method to be robust against markers location and place it favourably in the ranks of existing approaches.

Simulation, construction and characterization of a piezoelectric transducer using rexolite as acoustic matching

This article describes the simulation and characterization of an ultrasonic transducer using a new material called Rexolite to be used as a matching element. This transducer was simulated using a commercial piezoelectric ceramic PIC255 at 8 MHz. Rexolite, the new material, presents an excellent acoustic matching, specially in terms of the acoustic impedance of water. Finite elements simulations were used in this work. Rexolite was considered as a suitable material in the construction of the transducer due to its malleability and acoustic properties, to validate the simulations a prototype transducer was constructed. Experimental measurements were used to determine the resonance frequency of the prototype transducer. Simulated and experimental results were very similar showing that Rexolite may be an excellent matching, particularly for medical applications.

Aerial video geo-registration using terrain models from dense and coherent stereo matching

In the context of aerial imagery, one of the first steps toward a coherent processing of the information contained in multiple images is geo-registration, which consists in assigning geographic 3D coordinates to the pixels of the image. This enables accurate alignment and geo-positioning of multiple images, detection of moving objects and fusion of data acquired from multiple sensors. To solve this problem there are different approaches that require, in addition to a precise characterization of the camera sensor, high resolution referenced images or terrain elevation models, which are usually not publicly available or out of date. Building upon the idea of developing technology that does not need a reference terrain elevation model, we propose a geo-registration technique that applies variational methods to obtain a dense and coherent surface elevation model that is used to replace the reference model. The surface elevation model is built by interpolation of scattered 3D points, which are obtained in a two-step process following a classical stereo pipeline: first, coherent disparity maps between image pairs of a video sequence are estimated and then image point correspondences are back-projected. The proposed variational method enforces continuity of the disparity map not only along epipolar lines (as done by previous geo-registration techniques) but also across them, in the full 2D image domain. In the experiments, aerial images from synthetic video sequences have been used to validate the proposed technique.

Multiple Hashing Integration for Real-Time Large Scale Part-to-Part Video Matching

A real-time large scale part-to-part video matching algorithm, based on the cross correlation of the intensity of motion curves, is proposed with a view to originality recognition, video database cleansing, copyright enforcement, video tagging or video result re-ranking. Moreover, it is suggested how the most representative hashes and distance functions - strada, discrete cosine transformation, Marr-Hildreth and radial - should be integrated in order for the matching algorithm to be invariant against blur, compression and rotation distortions: (R; _) 2 [1; 20]_[1; 8], from 512_512 to 32_32pixels2 and from 10 to 180_. The DCT hash is invariant against blur and compression up to 64x64 pixels2. Nevertheless, although its performance against rotation is the best, with a success up to 70%, it should be combined with the Marr-Hildreth distance function. With the latter, the image selected by the DCT hash should be at a distance lower than 1.15 times the Marr-Hildreth minimum distance.

Multi-Camera very wide baseline feature matching based on view-adaptive junction detection

This paper presents a strategy for solving the feature matching problem in calibrated very wide-baseline camera settings. In this kind of settings, perspective distortion, depth discontinuities and occlusion represent enormous challenges. The proposed strategy addresses them by using geometrical information, specifically by exploiting epipolar-constraints. As a result it provides a sparse number of reliable feature points for which 3D position is accurately recovered. Special features known as junctions are used for robust matching. In particular, a strategy for refinement of junction end-point matching is proposed which enhances usual junction-based approaches. This allows to compute cross-correlation between perfectly aligned plane patches in both images, thus yielding better matching results. Evaluation of experimental results proves the effectiveness of the proposed algorithm in very wide-baseline environments.

Hierarchical graph maps for visualization of collaborative recommender systems

In this paper we provide a method that allows the visualization of similarity relationships present between items of collaborative filtering recommender systems, as well as the relative importance of each of these. The objective is to offer visual representations of the recommender system?s set of items and of their relationships; these graphs show us where the most representative information can be found and which items are rated in a more similar way by the recommender system?s community of users. The visual representations achieved take the shape of phylogenetic trees, displaying the numerical similarity and the reliability between each pair of items considered to be similar. As a case study we provide the results obtained using the public database Movielens 1M, which contains 3900 movies.