Calibración geométrica de cámaras no métricas. Estudio de metodologías y modelos matemáticos de distorsión

La Fotogrametría, como ciencia y técnica de obtención de información tridimensional del espacio objeto a partir de imágenes bidimensionales, requiere de medidas de precisión y en ese contexto, la calibración geométrica de cámaras ocupa un lugar importante. El conocimiento de la geometría interna de la cámara es fundamental para lograr mayor precisión en las medidas realizadas. En Fotogrametría Aérea se utilizan cámaras métricas (fabricadas exclusivamente para aplicaciones cartográficas), que incluyen objetivos fotográficos con sistemas de lentes complejos y de alta calidad. Pero en Fotogrametría de Objeto Cercano se está trabajando cada vez con más asiduidad con cámaras no métricas, con ópticas de peor calidad que exigen una calibración geométrica antes o después de cada trabajo. El proceso de calibración encierra tres conceptos fundamentales: modelo de cámara, modelo de distorsión y método de calibración. El modelo de cámara es un modelo matemático que aproxima la transformación proyectiva original a la realidad física de las lentes. Ese modelo matemático incluye una serie de parámetros entre los que se encuentran los correspondientes al modelo de distorsión, que se encarga de corregir los errores sistemáticos de la imagen. Finalmente, el método de calibración propone el método de estimación de los parámetros del modelo matemático y la técnica de optimización a emplear. En esta Tesis se propone la utilización de un patrón de calibración bidimensional que se desplaza en la dirección del eje óptico de la cámara, ofreciendo así tridimensionalidad a la escena fotografiada. El patrón incluye un número elevado de marcas, lo que permite realizar ensayos con distintas configuraciones geométricas. Tomando el modelo de proyección perspectiva (o pinhole) como modelo de cámara, se realizan ensayos con tres modelos de distorsión diferentes, el clásico de distorsión radial y tangencial propuesto por D.C. Brown, una aproximación por polinomios de Legendre y una interpolación bicúbica. De la combinación de diferentes configuraciones geométricas y del modelo de distorsión más adecuado, se llega al establecimiento de una metodología de calibración óptima. Para ayudar a la elección se realiza un estudio de las precisiones obtenidas en los distintos ensayos y un control estereoscópico de un panel test construido al efecto. ABSTRACT Photogrammetry, as science and technique for obtaining three-dimensional information of the space object from two-dimensional images, requires measurements of precision and in that context, the geometric camera calibration occupies an important place. The knowledge of the internal geometry of the camera is fundamental to achieve greater precision in measurements made. Metric cameras (manufactured exclusively for cartographic applications), including photographic lenses with complex lenses and high quality systems are used in Aerial Photogrammetry. But in Close Range Photogrammetry is working increasingly more frequently with non-metric cameras, worst quality optical components which require a geometric calibration before or after each job. The calibration process contains three fundamental concepts: camera model, distortion model and method of calibration. The camera model is a mathematical model that approximates the original projective transformation to the physical reality of the lenses. The mathematical model includes a series of parameters which include the correspondents to the model of distortion, which is in charge of correcting the systematic errors of the image. Finally, the calibration method proposes the method of estimation of the parameters of the mathematical modeling and optimization technique to employ. This Thesis is proposing the use of a pattern of two dimensional calibration that moves in the direction of the optical axis of the camera, thus offering three-dimensionality to the photographed scene. The pattern includes a large number of marks, which allows testing with different geometric configurations. Taking the projection model perspective (or pinhole) as a model of camera, tests are performed with three different models of distortion, the classical of distortion radial and tangential proposed by D.C. Brown, an approximation by Legendre polynomials and bicubic interpolation. From the combination of different geometric configurations and the most suitable distortion model, brings the establishment of a methodology for optimal calibration. To help the election, a study of the information obtained in the various tests and a purpose built test panel stereoscopic control is performed.

Bernstein polynomials in element-free Galerkin method

In the recent decades, meshless methods (MMs), like the element-free Galerkin method (EFGM), have been widely studied and interesting results have been reached when solving partial differential equations. However, such solutions show a problem around boundary conditions, where the accuracy is not adequately achieved. This is caused by the use of moving least squares or residual kernel particle method methods to obtain the shape functions needed in MM, since such methods are good enough in the inner of the integration domains, but not so accurate in boundaries. This way, Bernstein curves, which are a partition of unity themselves,can solve this problem with the same accuracy in the inner area of the domain and at their boundaries.

Differential resultant formulas for linear od-polynomials.

Ponencia

Simplified Finite Element Assessment of Beams With and Without Shear Deformation

This paper presents a simplified finite element (FE) methodology for solving accurately beam models with (Timoshenko) and without (Bernoulli-Euler) shear deformation. Special emphasis is made on showing how it is possible to obtain the exact solution on the nodes and a good accuracy inside the element. The proposed simplifying concept, denominated as the equivalent distributed load (EDL) of any order, is based on the use of Legendre orthogonal polynomials to approximate the original or acting load for computing the results between the nodes. The 1-span beam examples show that this is a promising procedure that allows the aim of using either one FE and an EDL of slightly higher order or by using an slightly larger number of FEs leaving the EDL in the lowest possible order assumed by definition to be equal to 4 independently of how irregular the beam is loaded.

Differential resultants of super essential systems of linear OD-polynomials

The sparse differential resultant dres(P) of an overdetermined system P of generic nonhomogeneous ordinary differential polynomials, was formally defined recently by Li, Gao and Yuan (2011). In this note, a differential resultant formula dfres(P) is defined and proved to be nonzero for linear "super essential" systems. In the linear case, dres(P) is proved to be equal, up to a nonzero constant, to dfres(P*) for the supper essential subsystem P* of P.

Learning Bayesian networks from data by the incremental compilation of new network polynomials

Probabilistic graphical models are a huge research field in artificial intelligence nowadays. The scope of this work is the study of directed graphical models for the representation of discrete distributions. Two of the main research topics related to this area focus on performing inference over graphical models and on learning graphical models from data. Traditionally, the inference process and the learning process have been treated separately, but given that the learned models structure marks the inference complexity, this kind of strategies will sometimes produce very inefficient models. With the purpose of learning thinner models, in this master thesis we propose a new model for the representation of network polynomials, which we call polynomial trees. Polynomial trees are a complementary representation for Bayesian networks that allows an efficient evaluation of the inference complexity and provides a framework for exact inference. We also propose a set of methods for the incremental compilation of polynomial trees and an algorithm for learning polynomial trees from data using a greedy score+search method that includes the inference complexity as a penalization in the scoring function.

Conditional density approximations with mixtures of polynomials

Mixtures of polynomials (MoPs) are a non-parametric density estimation technique especially designed for hybrid Bayesian networks with continuous and discrete variables. Algorithms to learn one- and multi-dimensional (marginal) MoPs from data have recently been proposed. In this paper we introduce two methods for learning MoP approximations of conditional densities from data. Both approaches are based on learning MoP approximations of the joint density and the marginal density of the conditioning variables, but they differ as to how the MoP approximation of the quotient of the two densities is found. We illustrate and study the methods using data sampled from known parametric distributions, and we demonstrate their applicability by learning models based on real neuroscience data. Finally, we compare the performance of the proposed methods with an approach for learning mixtures of truncated basis functions (MoTBFs). The empirical results show that the proposed methods generally yield models that are comparable to or significantly better than those found using the MoTBF-based method.

A finite class of orthogonal functions generated by Routh-Romanovski polynomials

It is known that some orthogonal systems are mapped onto other orthogonal systems by the Fourier transform. In this article we introduce a finite class of orthogonal functions, which is the Fourier transform of Routh-Romanovski orthogonal polynomials, and obtain its orthogonality relation using Parseval identity.

Language recognition using phonotactic-based shifted delta coefficients and multiple phone recognizers

A new language recognition technique based on the application of the philosophy of the Shifted Delta Coefficients (SDC) to phone log-likelihood ratio features (PLLR) is described. The new methodology allows the incorporation of long-span phonetic information at a frame-by-frame level while dealing with the temporal length of each phone unit. The proposed features are used to train an i-vector based system and tested on the Albayzin LRE 2012 dataset. The results show a relative improvement of 33.3% in Cavg in comparison with different state-of-the-art acoustic i-vector based systems. On the other hand, the integration of parallel phone ASR systems where each one is used to generate multiple PLLR coefficients which are stacked together and then projected into a reduced dimension are also presented. Finally, the paper shows how the incorporation of state information from the phone ASR contributes to provide additional improvements and how the fusion with the other acoustic and phonotactic systems provides an important improvement of 25.8% over the system presented during the competition.

Bivariate Krawtchouk polynomials: Inversion and connection problems with the NAVIMA algorithm

In this paper we present a recurrent procedure to solve an inversion problem for monic bivariate Krawtchouk polynomials written in vector column form, giving its solution explicitly. As a by-product, a general connection problem between two vector column of monic bivariate Krawtchouk families is also explicitly solved. Moreover, in the non monic case and also for Krawtchouk families, several expansion formulas are given, but for polynomials written in scalar form.