Mostrar el registro sencillo del ítem

dc.contributor.advisorGonzález-Jiménez, Antonio Javier 
dc.contributor.authorBriales Garcia, Jesus
dc.contributor.otherIngeniería de Sistemas y Automáticaes_ES
dc.date.accessioned2021-09-30T17:16:39Z
dc.date.available2021-09-30T17:16:39Z
dc.date.issued2021-09
dc.date.submitted2021-06-07
dc.identifier.urihttps://hdl.handle.net/10630/22918
dc.descriptionIn this thesis, we address a set of fundamental problems whose core difficulty boils down to optimizing over 3D poses. This includes many geometric 3D registration problems, covering well-known problems with a long research history such as the Perspective-n-Point (PnP) problem and generalizations, extrinsic sensor calibration, or even the gold standard for Structure from Motion (SfM) pipelines: The relative pose problem from corresponding features. Likewise, this is also the case for a close relative of SLAM, Pose Graph Optimization (also commonly known as Motion Averaging in SfM). The crux of this thesis contribution revolves around the successful characterization and development of empirically tight (convex) semidefinite relaxations for many of the aforementioned core problems of 3D Computer Vision. Building upon these empirically tight relaxations, we are able to find and certify the globally optimal solution to these problems with algorithms whose performance ranges as of today from efficient, scalable approaches comparable to fast second-order local search techniques to polynomial time (worst case). So, to conclude, our research reveals that an important subset of core problems that has been historically regarded as hard and thus dealt with mostly in empirical ways, are indeed tractable with optimality guarantees.es_ES
dc.description.abstractArtificial Intelligence (AI) drives a lot of services and products we use everyday. But for AI to bring its full potential into daily tasks, with technologies such as autonomous driving, augmented reality or mobile robots, AI needs to be not only intelligent but also perceptive. In particular, the ability to see and to construct an accurate model of the environment is an essential capability to build intelligent perceptive systems. The ideas developed in Computer Vision for the last decades in areas such as Multiple View Geometry or Optimization, put together to work into 3D reconstruction algorithms seem to be mature enough to nurture a range of emerging applications that already employ as of today 3D Computer Vision in the background. However, while there is a positive trend in the use of 3D reconstruction tools in real applications, there are also some fundamental limitations regarding reliability and performance guarantees that may hinder a wider adoption, e.g. in more critical applications involving people's safety such as autonomous navigation. State-of-the-art 3D reconstruction algorithms typically formulate the reconstruction problem as a Maximum Likelihood Estimation (MLE) instance, which entails solving a high-dimensional non-convex non-linear optimization problem. In practice, this is done via fast local optimization methods, that have enabled fast and scalable reconstruction pipelines, yet lack of guarantees on most of the building blocks leaving us with fundamentally brittle pipelines where no guarantees exist.es_ES
dc.language.isoenges_ES
dc.publisherUMA Editoriales_ES
dc.rightsinfo:eu-repo/semantics/openAccesses_ES
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/4.0/*
dc.subjectRobótica - Tesis doctoraleses_ES
dc.subject.otherRobóticaes_ES
dc.subject.otherVisión artificiales_ES
dc.titleGlobal Optimality via Tight Convex Relaxations for Pose Estimation in Geometric 3D Computer Visiones_ES
dc.typeinfo:eu-repo/semantics/doctoralThesises_ES
dc.centroEscuela de Ingenierías Industrialeses_ES
dc.rights.ccAttribution-NonCommercial-NoDerivatives 4.0 Internacional*


Ficheros en el ítem

Este ítem aparece en la(s) siguiente(s) colección(ones)

Mostrar el registro sencillo del ítem

Attribution-NonCommercial-NoDerivatives 4.0 Internacional
Excepto si se señala otra cosa, la licencia del ítem se describe como Attribution-NonCommercial-NoDerivatives 4.0 Internacional