Scene Reconstruction And Understanding By RGB-D Sensors

This thesis investigates interactive scene reconstruction and understanding using RGB-D data only. Indeed, we believe that depth cameras will still be in the near future a cheap and low-power 3D sensing alternative suitable for mobile devices too. Therefore, our contributions build on top of state-of-the-art approaches to achieve advances in three main challenging scenarios, namely mobile mapping, large scale surface reconstruction and semantic modeling. First, we will describe an effective approach dealing with Simultaneous Localization And Mapping (SLAM) on platforms with limited resources, such as a tablet device. Unlike previous methods, dense reconstruction is achieved by reprojection of RGB-D frames, while local consistency is maintained by deploying relative bundle adjustment principles. We will show quantitative results comparing our technique to the state-of-the-art as well as detailed reconstruction of various environments ranging from rooms to small apartments. Then, we will address large scale surface modeling from depth maps exploiting parallel GPU computing. We will develop a real-time camera tracking method based on the popular KinectFusion system and an online surface alignment technique capable of counteracting drift errors and closing small loops. We will show very high quality meshes outperforming existing methods on publicly available datasets as well as on data recorded with our RGB-D camera even in complete darkness. Finally, we will move to our Semantic Bundle Adjustment framework to effectively combine object detection and SLAM in a unified system. Though the mathematical framework we will describe does not restrict to a particular sensing technology, in the experimental section we will refer, again, only to RGB-D sensing. We will discuss successful implementations of our algorithm showing the benefit of a joint object detection, camera tracking and environment mapping.

Combining visual features and Growing Neural Gas networks for robotic 3D SLAM

The use of 3D data in mobile robotics provides valuable information about the robot’s environment. Traditionally, stereo cameras have been used as a low-cost 3D sensor. However, the lack of precision and texture for some surfaces suggests that the use of other 3D sensors could be more suitable. In this work, we examine the use of two sensors: an infrared SR4000 and a Kinect camera. We use a combination of 3D data obtained by these cameras, along with features obtained from 2D images acquired from these cameras, using a Growing Neural Gas (GNG) network applied to the 3D data. The goal is to obtain a robust egomotion technique. The GNG network is used to reduce the camera error. To calculate the egomotion, we test two methods for 3D registration. One is based on an iterative closest points algorithm, and the other employs random sample consensus. Finally, a simultaneous localization and mapping method is applied to the complete sequence to reduce the global error. The error from each sensor and the mapping results from the proposed method are examined.

Towards Mapping of Rock Walls Using a UAV-Mounted 2D Laser Scanner in GPS Denied Environments

In geotechnical engineering, the stability of rock excavations and walls is estimated by using tools that include a map of the orientations of exposed rock faces. However, measuring these orientations by using conventional methods can be time consuming, sometimes dangerous, and is limited to regions of the exposed rock that are reachable by a human. This thesis introduces a 2D, simulated, quadcopter-based rock wall mapping algorithm for GPS denied environments such as underground mines or near high walls on surface. The proposed algorithm employs techniques from the field of robotics known as simultaneous localization and mapping (SLAM) and is a step towards 3D rock wall mapping. Not only are quadcopters agile, but they can hover. This is very useful for confined spaces such as underground or near rock walls. The quadcopter requires sensors to enable self localization and mapping in dark, confined and GPS denied environments. However, these sensors are limited by the quadcopter payload and power restrictions. Because of these restrictions, a light weight 2D laser scanner is proposed. As a first step towards a 3D mapping algorithm, this thesis proposes a simplified scenario in which a simulated 1D laser range finder and 2D IMU are mounted on a quadcopter that is moving on a plane. Because the 1D laser does not provide enough information to map the 2D world from a single measurement, many measurements are combined over the trajectory of the quadcopter. Least Squares Optimization (LSO) is used to optimize the estimated trajectory and rock face for all data collected over the length of a light. Simulation results show that the mapping algorithm developed is a good first step. It shows that by combining measurements over a trajectory, the scanned rock face can be estimated using a lower-dimensional range sensor. A swathing manoeuvre is introduced as a way to promote loop closures within a short time period, thus reducing accumulated error. Some suggestions on how to improve the algorithm are also provided.

Real-Time Visual SLAM Using Pre - Existing Floor Lines as Landmarks and a Single Camera

SANTANA, André M.; SANTIAGO, Gutemberg S.; MEDEIROS, Adelardo A. D. Real-Time Visual SLAM Using Pre-Existing Floor Lines as Landmarks and a Single Camera. In: CONGRESSO BRASILEIRO DE AUTOMÁTICA, 2008, Juiz de Fora, MG. Anais... Juiz de Fora: CBA, 2008.

Real-Time Visual SLAM Using Pre - Existing Floor Lines as Landmarks and a Single Camera

SANTANA, André M.; SANTIAGO, Gutemberg S.; MEDEIROS, Adelardo A. D. Real-Time Visual SLAM Using Pre-Existing Floor Lines as Landmarks and a Single Camera. In: CONGRESSO BRASILEIRO DE AUTOMÁTICA, 2008, Juiz de Fora, MG. Anais... Juiz de Fora: CBA, 2008.

POSE ESTIMATION AND 3D RECONSTRUCTION USING SENSOR FUSION

A camera maps 3-dimensional (3D) world space to a 2-dimensional (2D) image space. In the process it loses the depth information, i.e., the distance from the camera focal point to the imaged objects. It is impossible to recover this information from a single image. However, by using two or more images from different viewing angles this information can be recovered, which in turn can be used to obtain the pose (position and orientation) of the camera. Using this pose, a 3D reconstruction of imaged objects in the world can be computed. Numerous algorithms have been proposed and implemented to solve the above problem; these algorithms are commonly called Structure from Motion (SfM). State-of-the-art SfM techniques have been shown to give promising results. However, unlike a Global Positioning System (GPS) or an Inertial Measurement Unit (IMU) which directly give the position and orientation respectively, the camera system estimates it after implementing SfM as mentioned above. This makes the pose obtained from a camera highly sensitive to the images captured and other effects, such as low lighting conditions, poor focus or improper viewing angles. In some applications, for example, an Unmanned Aerial Vehicle (UAV) inspecting a bridge or a robot mapping an environment using Simultaneous Localization and Mapping (SLAM), it is often difficult to capture images with ideal conditions. This report examines the use of SfM methods in such applications and the role of combining multiple sensors, viz., sensor fusion, to achieve more accurate and usable position and reconstruction information. This project investigates the role of sensor fusion in accurately estimating the pose of a camera for the application of 3D reconstruction of a scene. The first set of experiments is conducted in a motion capture room. These results are assumed as ground truth in order to evaluate the strengths and weaknesses of each sensor and to map their coordinate systems. Then a number of scenarios are targeted where SfM fails. The pose estimates obtained from SfM are replaced by those obtained from other sensors and the 3D reconstruction is completed. Quantitative and qualitative comparisons are made between the 3D reconstruction obtained by using only a camera versus that obtained by using the camera along with a LIDAR and/or an IMU. Additionally, the project also works towards the performance issue faced while handling large data sets of high-resolution images by implementing the system on the Superior high performance computing cluster at Michigan Technological University.

Sistema de modelização tridimensional de galerias subterrâneas

A presente dissertação apresenta uma solução para o problema de modelização tridimensional de galerias subterrâneas. O trabalho desenvolvido emprega técnicas provenientes da área da robótica móvel para obtenção um sistema autónomo móvel de modelização, capaz de operar em ambientes não estruturados sem acesso a sistemas de posicionamento global, designadamente GPS. Um sistema de modelização móvel e autónomo pode ser bastante vantajoso, pois constitui um método rápido e simples de monitorização das estruturas e criação de representações virtuais das galerias com um elevado nível de detalhe. O sistema de modelização desloca-se no interior dos túneis para recolher informações sensoriais sobre a geometria da estrutura. A tarefa de organização destes dados com vista _a construção de um modelo coerente, exige um conhecimento exacto do percurso praticado pelo sistema, logo o problema de localização da plataforma sensorial tem que ser resolvido. A formulação de um sistema de localização autónoma tem que superar obstáculos que se manifestam vincadamente nos ambientes underground, tais como a monotonia estrutural e a já referida ausência de sistemas de posicionamento global. Neste contexto, foi abordado o conceito de SLAM (Simultaneous Loacalization and Mapping) para determinação da localização da plataforma sensorial em seis graus de liberdade. Seguindo a abordagem tradicional, o núcleo do algoritmo de SLAM consiste no filtro de Kalman estendido (EKF { Extended Kalman Filter ). O sistema proposto incorpora métodos avançados do estado da arte, designadamente a parametrização em profundidade inversa (Inverse Depth Parametrization) e o método de rejeição de outliers 1-Point RANSAC. A contribuição mais importante do método por nós proposto para o avanço do estado da arte foi a fusão da informação visual com a informação inercial. O algoritmo de localização foi testado com base em dados reais, adquiridos no interior de um túnel rodoviário. Os resultados obtidos permitem concluir que, ao fundir medidas inerciais com informações visuais, conseguimos evitar o fenómeno de degeneração do factor de escala, comum nas aplicações de localização através de sistemas puramente monoculares. Provámos simultaneamente que a correcção de um sistema de localização inercial através da consideração de informações visuais é eficaz, pois permite suprimir os desvios de trajectória que caracterizam os sistemas de dead reckoning. O algoritmo de modelização, com base na localização estimada, organiza no espaço tridimensional os dados geométricos adquiridos, resultando deste processo um modelo em nuvem de pontos, que posteriormente _e convertido numa malha triangular, atingindo-se assim uma representação mais realista do cenário original.

Exploração da Point Cloud Library aplicada à percepção em sistemas autónomos

A navegação e a interpretação do meio envolvente por veículos autónomos em ambientes não estruturados continua a ser um grande desafio na actualidade. Sebastian Thrun, descreve em [Thr02], que o problema do mapeamento em sistemas robóticos é o da aquisição de um modelo espacial do meio envolvente do robô. Neste contexto, a integração de sistemas sensoriais em plataformas robóticas, que permitam a construção de mapas do mundo que as rodeia é de extrema importância. A informação recolhida desses dados pode ser interpretada, tendo aplicabilidade em tarefas de localização, navegação e manipulação de objectos. Até à bem pouco tempo, a generalidade dos sistemas robóticos que realizavam tarefas de mapeamento ou Simultaneous Localization And Mapping (SLAM), utilizavam dispositivos do tipo laser rangefinders e câmaras stereo. Estes equipamentos, para além de serem dispendiosos, fornecem apenas informação bidimensional, recolhidas através de cortes transversais 2D, no caso dos rangefinders. O paradigma deste tipo de tecnologia mudou consideravelmente, com o lançamento no mercado de câmaras RGB-D, como a desenvolvida pela PrimeSense TM e o subsequente lançamento da Kinect, pela Microsoft R para a Xbox 360 no final de 2010. A qualidade do sensor de profundidade, dada a natureza de baixo custo e a sua capacidade de aquisição de dados em tempo real, é incontornável, fazendo com que o sensor se tornasse instantaneamente popular entre pesquisadores e entusiastas. Este avanço tecnológico deu origem a várias ferramentas de desenvolvimento e interacção humana com este tipo de sensor, como por exemplo a Point Cloud Library [RC11] (PCL). Esta ferramenta tem como objectivo fornecer suporte para todos os blocos de construção comuns que uma aplicação 3D necessita, dando especial ênfase ao processamento de nuvens de pontos de n dimensões adquiridas a partir de câmaras RGB-D, bem como scanners laser, câmaras Time-of-Flight ou câmaras stereo. Neste contexto, é realizada nesta dissertação, a avaliação e comparação de alguns dos módulos e métodos constituintes da biblioteca PCL, para a resolução de problemas inerentes à construção e interpretação de mapas, em ambientes indoor não estruturados, utilizando os dados provenientes da Kinect. A partir desta avaliação, é proposta uma arquitectura de sistema que sistematiza o registo de nuvens de pontos, correspondentes a vistas parciais do mundo, num modelo global consistente. Os resultados da avaliação realizada à biblioteca PCL atestam a sua viabilidade, para a resolução dos problemas propostos. Prova da sua viabilidade, são os resultados práticos obtidos, da implementação da arquitectura de sistema proposta, que apresenta resultados de desempenho interessantes, como também boas perspectivas de integração deste tipo de conceitos e tecnologia em plataformas robóticas desenvolvidas no âmbito de projectos do Laboratório de Sistemas Autónomos (LSA).

Towards a Mobile Three-dimensional Modelling System for Underground Structures

Robotica 2012: 12th International Conference on Autonomous Robot Systems and Competitions April 11, 2012, Guimarães, Portugal

Sistema de visão robótico baseado no sensor kinect para orientação no espaço e contorno de obstáculos

Dissertação de mestrado em Engenharia Eletrónica Industrial e Computadores (área de especialização em Robótica)

Depth sensors in augmented reality solutions. Literature review

The emergence of depth sensors has made it possible to track – not only monocular cues – but also the actual depth values of the environment. This is especially useful in augmented reality solutions, where the position and orientation (pose) of the observer need to be accurately determined. This allows virtual objects to be installed to the view of the user through, for example, a screen of a tablet or augmented reality glasses (e.g. Google glass, etc.). Although the early 3D sensors have been physically quite large, the size of these sensors is decreasing, and possibly – eventually – a 3D sensor could be embedded – for example – to augmented reality glasses. The wider subject area considered in this review is 3D SLAM methods, which take advantage of the 3D information available by modern RGB-D sensors, such as Microsoft Kinect. Thus the review for SLAM (Simultaneous Localization and Mapping) and 3D tracking in augmented reality is a timely subject. We also try to find out the limitations and possibilities of different tracking methods, and how they should be improved, in order to allow efficient integration of the methods to the augmented reality solutions of the future.

Selective submap joining SLAM for autonomous vehicles

Simultaneous Localization and Mapping (SLAM) do not result in consistent maps of large areas because of gradual increase of the uncertainty for long term missions. In addition, as the size of the map grows the computational cost increases, making SLAM solutions unsuitable for on-line applications. This thesis surveys SLAM approaches paying special attention to those approaches aimed to work on large scenarios. Special focus is given to existing underwater SLAM applications. A technique based on using independent local maps together with a global stochastic map is presented. This technique is called Selective Submap Joining SLAM (SSJS). A global map contains relative transformations between local maps, which are updated once a new loop is detected. Maps sharing several features are fused, maintaining the correlation between landmarks and vehicle's pose. The use of local maps reduces computational costs and improves map consistency as compared to state of the art techniques.

Localização e mapeamento simultâneos de ambientes planos usando visão monocular e representação híbrida do ambiente

The goal of this work is to propose a SLAM (Simultaneous Localization and Mapping) solution based on Extended Kalman Filter (EKF) in order to make possible a robot navigates along the environment using information from odometry and pre-existing lines on the floor. Initially, a segmentation step is necessary to classify parts of the image in floor or non floor . Then the image processing identifies floor lines and the parameters of these lines are mapped to world using a homography matrix. Finally, the identified lines are used in SLAM as landmarks in order to build a feature map. In parallel, using the corrected robot pose, the uncertainty about the pose and also the part non floor of the image, it is possible to build an occupancy grid map and generate a metric map with the obstacle s description. A greater autonomy for the robot is attained by using the two types of obtained map (the metric map and the features map). Thus, it is possible to run path planning tasks in parallel with localization and mapping. Practical results are presented to validate the proposal

Localização de robôs móveis autônomos utilizando fusão sensorial de odometria e visão monocular

The development and refinement of techniques that make simultaneous localization and mapping (SLAM) for an autonomous mobile robot and the building of local 3-D maps from a sequence of images, is widely studied in scientific circles. This work presents a monocular visual SLAM technique based on extended Kalman filter, which uses features found in a sequence of images using the SURF descriptor (Speeded Up Robust Features) and determines which features can be used as marks by a technique based on delayed initialization from 3-D straight lines. For this, only the coordinates of the features found in the image and the intrinsic and extrinsic camera parameters are avaliable. Its possible to determine the position of the marks only on the availability of information of depth. Tests have shown that during the route, the mobile robot detects the presence of characteristics in the images and through a proposed technique for delayed initialization of marks, adds new marks to the state vector of the extended Kalman filter (EKF), after estimating the depth of features. With the estimated position of the marks, it was possible to estimate the updated position of the robot at each step, obtaining good results that demonstrate the effectiveness of monocular visual SLAM system proposed in this paper

Técnicas visuais de localização e mapeamento simultâneos sem extração de primitivas geométricas da imagem

In Simultaneous Localization and Mapping (SLAM - Simultaneous Localization and Mapping), a robot placed in an unknown location in any environment must be able to create a perspective of this environment (a map) and is situated in the same simultaneously, using only information captured by the robot s sensors and control signals known. Recently, driven by the advance of computing power, work in this area have proposed to use video camera as a sensor and it came so Visual SLAM. This has several approaches and the vast majority of them work basically extracting features of the environment, calculating the necessary correspondence and through these estimate the required parameters. This work presented a monocular visual SLAM system that uses direct image registration to calculate the image reprojection error and optimization methods that minimize this error and thus obtain the parameters for the robot pose and map of the environment directly from the pixels of the images. Thus the steps of extracting and matching features are not needed, enabling our system works well in environments where traditional approaches have difficulty. Moreover, when addressing the problem of SLAM as proposed in this work we avoid a very common problem in traditional approaches, known as error propagation. Worrying about the high computational cost of this approach have been tested several types of optimization methods in order to find a good balance between good estimates and processing time. The results presented in this work show the success of this system in different environments