No description available.

No description available.

In this dissertation, we focus on solving Nonlinear Least Squares problems using a supervised approach. In particular, we developed a Supervised Descent Method (SDM), performed thorough theoretical analysis, and demonstrated its effectiveness on optimizing analytic functions, and four other real-world applications: Inverse Kinematics, Rigid Tracking, Face Alignment (frontal and multi-view), and 3D Object Pose Estimation. In Rigid Tracking, SDM was able to take advantage of more robust features, such as, HoG and SIFT. Those non-differentiable image features were out of consideration of previous work because they relied on gradient-based methods for optimization. In Inverse Kinematics where we minimize a non-convex function, SDM achieved significantly better convergence than gradient-based approaches. In Face Alignment, SDM achieved state-of-the-arts results. Moreover, it was extremely computationally efficient, which makes it applicable for many mobile applications. In addition, we provided a unified view of several popular methods including SDM on sequential prediction, and reformulated them as a sequence of function compositions. Finally, we suggested some future research directions on SDM and sequential prediction.

The Darwin-Hatherton glacial system is an outlet glacial system in the Transantarctic Mountains, Antarctica, which drains ice from the East Antarctic Ice Sheet into the Ross Ice Shelf. This research provides remotely sensed data that can be used in modeling research for the Darwin-Hatherton glacial system, which in turn can be used in mass balance research for the West Antarctic Ice Sheet. Two improved digital elevation models (DEM) are produced to cover the lower Darwin Glacier and to cover the upper Darwin and Hatherton Glaciers. The new improved DEMs are generated from Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) satellite data, with a resolution of 45 m. To produce the two final DEMs, multiple DEMs are firstly adjusted to remove systematic errors and are then stacked and averaged to increase the accuracy and produce the final two DEMs. For the lower Darwin Glacier, 5 DEMs were averaged and in the upper Darwin and Hatherton Glaciers, 6 DEMs were averaged. The accuracy is quantified by a remaining error of + 9 m for the lower Darwin Glacier DEM and + 37 m for the upper Darwin and Hatherton Glaciers DEM. This is a significant improvement from the existing 200 m resolution Radarsat Antarctic mapping project (RAMPv2) DEM which has a remaining error of + 138 m over the lower Darwin Glacier and + 152 m over the upper Darwin and Hatherton Glaciers. The accuracy is assessed by comparing the ASTER and RAMPv2 DEMs to highly accurate ice, cloud and land elevation satellite (ICESat) laser altimetry data. A 15 m resolution, true colour, orthorectified image is provided for the entire Darwin-Hatherton glacial system from ASTER satellite imagery. The DEMs used to orthorectify the ASTER satellite imagery are the two new 45 m resolution ASTER DEMs. Lastly feature tracking was explored as a method for measuring surface ice velocity. This research shows that feature tracking is unsuitable for the Darwin-Hatherton glacial system if using 15 m resolution satellite imagery over a 1 to 4 year time period.

An algorithm for tracking single visual targets is developed in this study. Feature detection is the necessary and appropriate image processing technique for this algorithm. The main point of this approach is to use the data supplied by the feature detection as the observation from a group of targets having similar motion dynamics. Therefore a single visual target is regarded as a group of multiple targets. Accurate data association and state estimation under clutter are desired for this application similar to other multi-target tracking applications. The group tracking approach is used with the well-known probabilistic data association technique to cope with data association and estimation problems. The applicability of this method particularly for visual tracking and for other cases is also discussed.

The Darwin-Hatherton glacial system is an outlet glacial system in the Transantarctic Mountains, Antarctica, which drains ice from the East Antarctic Ice Sheet into the Ross Ice Shelf. This research provides remotely sensed data that can be used in modeling research for the Darwin-Hatherton glacial system, which in turn can be used in mass balance research for the West Antarctic Ice Sheet. Two improved digital elevation models (DEM) are produced to cover the lower Darwin Glacier and to cover the upper Darwin and Hatherton Glaciers. The new improved DEMs are generated from Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) satellite data, with a resolution of 45 m. To produce the two final DEMs, multiple DEMs are firstly adjusted to remove systematic errors and are then stacked and averaged to increase the accuracy and produce the final two DEMs. For the lower Darwin Glacier, 5 DEMs were averaged and in the upper Darwin and Hatherton Glaciers, 6 DEMs were averaged. The accuracy is quantified by a remaining error of + 9 m for the lower Darwin Glacier DEM and + 37 m for the upper Darwin and Hatherton Glaciers DEM. This is a significant improvement from the existing 200 m resolution Radarsat Antarctic mapping project (RAMPv2) DEM which has a remaining error of + 138 m over the lower Darwin Glacier and + 152 m over the upper Darwin and Hatherton Glaciers. The accuracy is assessed by comparing the ASTER and RAMPv2 DEMs to highly accurate ice, cloud and land elevation satellite (ICESat) laser altimetry data. A 15 m resolution, true colour, orthorectified image is provided for the entire Darwin-Hatherton glacial system from ASTER satellite imagery. The DEMs used to orthorectify the ASTER satellite imagery are the two new 45 m resolution ASTER DEMs. Lastly feature tracking was explored as a method for measuring surface ice velocity. This research shows that feature tracking is unsuitable for the Darwin-Hatherton glacial system if using 15 m resolution satellite imagery over a 1 to 4 year time period.

In this work, we propose a depth estimation system based on image sequence and odometry information. The key idea is that depth estimation is decoupled from pose estimation. Such approach results in multipurpose system applicable on different robot platforms and for different depth estimation related problems. Our implementation uses various filtration techniques, operates real-time and provides appropriate results. Although the system was aimed at and tested on drone platform, it can be well used on any other type of autonomous vehicle that provides odometry information and video output.

Orientador: Siome Klein Goldenstein / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-06T08:17:58Z (GMT). No. of bitstreams: 1 Dorini_LeyzaElmeriBaldo_M.pdf: 1659888 bytes, checksum: 3ac1fe51a9d8159f6fa5df97cb4f52bc (MD5) Previous issue date: 2006 / Resumo: O correto estabelecimento de correspondências entre imagens tomadas de diferentes pontos de vista é um problema fundamental na área de visão computacional, sendo base para diversas tarefas de alto nível, tais como reconstrução 3D e análise de movimento. A grande maioria dos algoritmos de rastreamento de características não possui uma incerteza associada a posição estimada das características sendo rastreadas, informação esta de extrema importância, considerando sua vasta aplicabilidade. Exatamente este o foco principal deste trabalho, onde introduzimos um framework genérico que expande algoritmos de rastreamento de tal forma que eles possam propagar também informações de incerteza. Neste trabalho, por questão de simplicidade, utilizamos o algoritmo de rastreamento de características Kanade-Lucas-Tomasi (KLT) para demonstrar as vantagens do nosso método, denominado Unscented Feature Tracking (UFT). A abordagem consiste na introdução de Variáveis Aleatórias Gaussianas (GRVs) para a representação da localização dos pontos característicos, e utiliza a Transformada Unscented com Escala (SUT) para propagar e combinar GRVs. Mostramos uma aplicação do UFT em um procedimento de bundle adjustment, onde a função custo leva em conta a informação das GRVs, fornecendo melhores estimativas. O método é robusto, considerando que identifica e descarta anomalias, que podem comprometer de maneira expressiva o resultado de tarefas que utilizam as correspondências. Experimentos com seqüências de imagens reais e sintéticas comprovam os benefícios do método proposto / Abstract: To determine reliable correspondences between a pair of images is a fundamental problem in the computer vision community. It is the foundation of several high level tasks, such as 3D reconstruction and motion analysis. Although there are many feature tracking algorithms, most of them do not maintain information about the uncertainty of the feature locations' estimates. This information is very useful, since large errors can disturb the results of the correspondence-based tasks. This is the goal of this work, a new generic framework that augments feature tracking algorithms so that they also propagate uncertainty information. In this work, we use the well-known Kanade-Lucas-Tomasi (KLT) feature tracker to demonstrate the benefits of our method, called Unscented Feature Tracking (UFT). The approach consists on the introduction of Gaussian Random Variables (GRVs) for the representation of the features' locations, and on the use of the Scaled Unscented Transform (SUT) to propagate and combine GRVs. We also describe an improved bundle adjustment procedure as an application, where the cost function takes into account the information of the GRVs, and provides better estimates. Experiments with real and synthetic images confirm that UFT improves the quality of the feature tracking process and is a robust method for detect and reject outliers / Mestrado / Visão Computacional / Mestre em Ciência da Computação

Presented here is a novel tool to extract and track believable vortex core lines from unsteady Computational Fluid Dynamics data sets using multiple feature extraction algorithms. Existing work explored the possibility of extracting features concurrent with a running simulation using intelligent software agents, combining multiple algorithms' capabilities using subjective logic. This work modifies the steady-state approach to work with unsteady fluid dynamics and is designed to work within the Concurrent Agent-enabled Feature Extraction concept. Each agent's belief tuple is quantified using a predefined set of information. The information and functions necessary to set each component in each agent's belief tuple is given along with an explanation of the methods for setting the components. This method is applied to the analyses of flow in a lid-driven cavity and flow around a cylinder, which highlight strengths and weaknesses of the chosen algorithms and the potential for subjective logic to aid in understanding the resulting features. Feature tracking is successfully applied and is observed to have a significant impact on the opinion of the vortex core lines. In the lid-driven cavity data set, unsteady feature extraction modifications are shown to impact feature extraction results with moving vortex core lines. The Sujudi-Haimes algorithm is shown to be more believable when extracting the main vortex core lines of the cavity simulation while the Roth-Peikert algorithm succeeding in extracting the weaker vortex cores in the same simulation. Mesh type and time step is shown to have a significant effect on the method. In the curved wake of the cylinder data set, the Roth-Peikert algorithm more reliably detects vortex core lines which exist for a significant amount of time. the method was finally applied to a massive wind turbine simulation, where the importance of performing feature extraction in parallel is shown. The use of multiple extraction algorithms with subjective logic and feature tracking helps determine the expected probability that an extracted vortex core is believable. This approach may be applied to massive data sets which will greatly reduce analysis time and data size and will aid in a greater understanding of complex fluid flows.

This thesis presents an autonomous vision-based navigation strategy applicable to the approach phase of a small body mission, developed within the Robotics Section at NASA Jet Propulsion Laboratory. Today, the operations performed to approach small planetary bodies are largely dependent on ground support and human decision-making, which demand operational complexity and restrict the spectrum of achievable activities throughout the mission. In contrast, the autonomous pipeline presented here could be run onboard, without ground intervention. Using optical data only, the pipeline estimates the target body's rotation, pole, shape, and performs identification and tracking of surface landmarks, for terrain relative navigation. An end-to-end simulation is performed to validate the pipeline, starting from input synthetic images and ending with an orbit determination solution. As a case study, the approach phase of the Rosetta mission is reproduced, and it is concluded that navigation performance is in line with the ground-based state-of-the-art. Such results are presented in detail in the paper attached in the appendix, which presents the pipeline architecture and navigation analysis. This thesis manuscript aims to provide additional context to the appended paper, further describing some implementation details used for the approach simulations. / Detta examensarbete presenterar en strategi för ett autonomt visionsbaserat navigationssystem för att närma sig en liten himlakropp. Strategin har utvecklats av robotikavdelningen vid NASA Jet Propulsion Laboratory i USA. Nuvarande system som används för att närma sig en liten himlakropp bygger till största delen på markstationer och mänskligt beslutsfattande, vilka utgör komplexa rutiner och begränsar spektrumet av möjliga aktiviteter under rymduppdraget. I jämförelse, det autonoma system presenterat i denna rapport är utformat för att köras helt från rymdfarkosten och utan krav på kontakt med markstationer. Genom att använda enbart optisk information uppskattar systemet himlakroppens rotation, poler och form samt genomför en identifiering och spårning av landmärken på himlakroppens yta för relativ terrängnavigering. En simulering har genomförts för att validera det autonoma navigationssystemet. Simuleringen utgick ifrån bilder av himlakroppen och avslutades med en lösning på banbestämningsproblemet. Fasen då rymdfarkosten i ESA:s Rosetta-rymduppdrag närmar sig kometen valdes som fallstudie för simuleringen och slutsatsen från denna fallstudie var att systemets autonoma navigationsprestanda var i linje med toppmoderna system. Den detaljerade beskrivningen av det autonoma systemet och resultaten från studien har presenterats i ett konferensbidrag, som ingår som bilaga till rapporten. Inledningen av rapporten syftar till att förtydliga bakgrunden och implementering som komplement till innehållet i bilagan.