Publications
166 results found
Civera J, Davison AJ, Martínez Montiel JM, 2012, Conclusions, Springer Tracts in Advanced Robotics, Pages: 123-125
This chapter presents the main conclusions and summarizes the content of the book. The algorithms, models and methods presented in the previous chapters cover the main topics in sequential SfM or monocular SLAM: a projective point model, an efficient and robust search for correspondences and altorithms for model selection and internal self-calibration. Together, the contributions presented in the different chapters of the book form a robust system for sequential scene and camera motion estimation potentially able to deal with any image sequence in real-time at 30 frames per second.
Civera J, Davison AJ, Martínez Montiel JM, 2012, Self-calibration, Springer Tracts in Advanced Robotics, Pages: 111-122
Computer vision researchers have proved the feasibility of camera self-calibration –the estimation of a camera’s internal parameters from an image sequence without any known scene structure. Nevertheless, all of the recent sequential approaches to 3D structure and motion estimation from image sequences which have arisen in robotics and aim at real-time operation (often classed as visual SLAM or visual odometry) have relied on pre-calibrated cameras and have not attempted online calibration. In this chapter, we present a sequential filtering algorithm for simultaneous estimation of 3D scene estimation, camera trajectory and full camera calibration from a sequence of fixed but unknown calibration. This calibration comprises the standard projective parameters of focal length and principal point along with two radial distortion coefficients.
Civera J, Davison AJ, Martínez Montiel JM, 2012, 1-point RANSAC, Springer Tracts in Advanced Robotics, Pages: 65-97
Random Sample Consensus (RANSAC) has become one of the most successful techniques for robust estimation from a data set that may contain outliers. It works by constructing model hypotheses from random minimal data subsets and evaluating their validity from the support of the whole data. In this chapter we present an efficient RANSAC algorithm for an Extended Kalman Filter (EKF) framework that uses the available prior probabilistic information from the EKF in the RANSAC model hypothesize stage. This allows the minimal sample size to be reduced to one, resulting in large computational savings without performance degradation. 1-Point RANSAC is also shown to outperform both in accuracy and computational cost the Joint Compatibility Branch and Bound (JCBB) algorithm, a gold-standard technique for spurious rejection within the EKF framework. The combination of this 1-point RANSAC and the robocentric formulation of the EKF SLAM allows a qualitative jump on the general performance of the algorithms presented in this book: In this chapter, sequences covering trajectories of several hundreds of metres are processed showing highly accurate camera motion estimation results.
Civera J, Davison AJ, Martínez Montiel JM, 2012, Inverse depth parametrization, Springer Tracts in Advanced Robotics, Pages: 33-63
This chapter presents a parametrization for point features within monocular SLAM which permits efficient and accurate representation of uncertainty during undelayed initialisation and beyond, all within the standard EKF (Extended Kalman Filter). The key concepts are direct parametrization of the inverse depth of features relative to the camera locations from which they were first viewed and the addition of this camera position at initialization to the state vector. The parametrization fits the projective nature of the camera in the sense that is able to code the infinite depth case. Also, the projection equation holds the high degree of linearity required by the Extended Kalman Filter. The chapter also shows that once the depth estimate of a feature is sufficiently accurate, its representation can safely be converted to the Euclidean XYZ form, and proposes a linearity index which allows automatic detection and conversion to maintain maximum efficiency –only low parallax features need be maintained in inverse depth form for long periods.
Civera J, Davison AJ, Martínez Montiel JM, 2012, Degenerate camera motions and model selection, Springer Tracts in Advanced Robotics, Pages: 99-110
The assumption of a general camera motion –translation and rotation– between frames in an image sequence leads to inconsistent estimations when the camera performs more restricted motions, like pure rotation, or even no motion. In this situation, the noise present in the data fits the extra terms in the overparametrized model and artificially estimates dimensions for which we have no information. This chapter presents an Interacting Multiple Models (IMM) framework which can switch automatically between parameter sets in monocular SLAM; selecting the most appropriate motion model at each step. Remarkably, this approach of full sequential probability propagation means that there is no need for penalty terms to achieve the Occam property of favouring simpler models –this arises automatically. We demonstrate our method with results on a complex real image sequence with varied motion.
Civera J, Davison AJ, Martínez Montiel JM, 2012, Points at infinity: Mosaics using the extended kalman filter, Springer Tracts in Advanced Robotics, Pages: 13-32
This chapter introduces the use of zero-parallax points on filtering-based Structure from Motion (SfM) or monocular SLAM. A geometric model is proposed for the estimation of an accurate camera rotation and the directions of a set of tracked features using an Extended Kalman Filter. On top of that, it is proposed a sequential mosaicing algorithm able to build drift-free, consistent spherical mosaics in real- time, automatically and seamlessly even when previously viewed parts of the scene are re-visited. This method represents a significant advance on previous mosaicing techniques which either require an expensive global optimization or which run sequentially in real-time but use local alignment of nearby images and ultimately drift.
Jachnik J, Newcombe RA, Davison AJ, 2012, Real-Time Surface Light-field Capture for Augmentation of Planar Specular Surfaces, 11th IEEE and ACM International Symposium on Mixed and Augmented Reality (ISMAR), Publisher: IEEE, Pages: 91-97, ISSN: 1554-7868
- Author Web Link
- Cite
- Citations: 41
Handa A, Newcombe RA, Angeli A, et al., 2012, Real-Time Camera Tracking: When is High Frame-Rate Best?, 12th European Conference on Computer Vision (ECCV), Publisher: SPRINGER-VERLAG BERLIN, Pages: 222-235, ISSN: 0302-9743
- Author Web Link
- Cite
- Citations: 64
Alcantarilla PF, Bartoli A, Davison AJ, 2012, KAZE Features, 12th European Conference on Computer Vision (ECCV), Publisher: SPRINGER-VERLAG BERLIN, Pages: 214-227, ISSN: 0302-9743
- Author Web Link
- Cite
- Citations: 757
Civera J, Davison AJ, Martinez Montiel JM, 2012, Structure from Motion Using the Extended Kalman Filter, Publisher: SPRINGER-VERLAG BERLIN, ISBN: 978-3-642-24833-7
- Author Web Link
- Cite
- Citations: 14
Izadi S, Kim D, Hilliges O, et al., 2011, KinectFusion: Real-time 3D reconstruction and interaction using a moving depth camera, Pages: 559-568
KinectFusion enables a user holding and moving a standard Kinect camera to rapidly create detailed 3D reconstructions of an indoor scene. Only the depth data from Kinect is used to track the 3D pose of the sensor and reconstruct, geometrically precise, 3D models of the physical scene in real-time. The capabilities of KinectFusion, as well as the novel GPU-based pipeline are described in full. We show uses of the core system for low-cost handheld scanning, and geometry-aware augmented reality and physics-based interactions. Novel extensions to the core GPU pipeline demonstrate object segmentation and user interaction directly in front of the sensor, without degrading camera tracking or reconstruction. These extensions are used to enable real-time multi-touch interactions anywhere, allowing any planar or non-planar reconstructed physical surface to be appropriated for touch. © 2011 ACM.
Izadi S, Newcombe RA, Kim D, et al., 2011, KinectFusion: Real-time dynamic 3D surface reconstruction and interaction
We present KinectFusion, a system that takes live depth data from a moving Kinect camera and in real-time creates high-quality, geometrically accurate, 3D models. Our system allows a user holding a Kinect camera to move quickly within any indoor space, and rapidly scan and create a fused 3D model of the whole room and its contents within seconds. Even small motions, caused for example by camera shake, lead to new viewpoints of the scene and thus refinements of the 3D model, similar to the effect of image super-resolution. As the camera is moved closer to objects in the scene more detail can be added to the acquired 3D model. © 2011 ACM.
Handa A, Newcombe RA, Angeli A, et al., 2011, Applications of Legendre-Fenchel transformation to computer vision problems, Departmental Technical Report: 11/7, Publisher: Department of Computing, Imperial College London, 11/7
We aim to provide a small background on Lengenre-Fencheltransformation, the applications of which have been increasingly gettingpopular in computer vision. A general motivation follows up with standardexamples. Then we take a good view on their applications in solvingvarious standard computer vision problems e.g. image denoising, opticalflow, image deconvolution etc.
Newcombe RA, Izadi S, Hilliges O, et al., 2011, KinectFusion: Real-Time Dense Surface Mapping and Tracking, ISMAR 2011, Pages: 127-136
Newcombe RA, Lovegrove S, Davison AJ, 2011, DTAM: Dense Tracking and Mapping in Real-Time, ICCV 2011, Pages: 2320-2327, ISSN: 1550-5499
Strasdat H, Davison AJ, Montiel JMM, et al., 2011, Double Window Optimisation for Constant Time Visual SLAM
Izadi S, Kim D, Hilliges O, et al., 2011, KinectFusion: Real-Time 3D Reconstruction and Interaction Using a Moving Depth Camera, Pages: 127-136
Carrera G, Angeli A, Davison AJ, 2011, Lightweight SLAM and Navigation with a Multi-Camera Rig
Carrera G, Angeli A, Davison AJ, 2011, SLAM-Based Automatic Extrinsic Calibration of a Multi-Camera Rig, International Conference on Robotics and Automation (ICRA)
Lovegrove SJ, Davison AJ, Ibanez-Guzmán J, 2011, Accurate Visual Odometry from a Rear Parking Camera
Strasdat H, Montiel JMM, Davison AJ, 2010, Real-Time Monocular SLAM: Why Filter?, IEEE International Conference on Robotics and Automation (ICRA)
Angeli A, Davison AJ, 2010, Live Feature Clustering in Video Using Appearance and 3D Geometry, British Machine Vision Conference
Civera J, Grasa O, Davison AJ, et al., 2010, 1-Point RANSAC for EKF Filtering. Application to Real-Time Structure from Motion and Visual Odometry, Journal of Field Robotics, Vol: 27, Pages: 609-631
Strasdat H, Montiel JMM, Davison AJ, 2010, Scale Drift-Aware Large Scale Monocular SLAM
Lovegrove SJ, Davison AJ, 2010, Real-Time Spherical Mosaicing using Whole Image Alignment
Newcombe RA, Davison AJ, 2010, Live Dense Reconstruction with a Single Moving Camera
Civera J, Davison AJ, Magallon JA, et al., 2009, Drift-Free Real-Time Sequential Mosaicing, INTERNATIONAL JOURNAL OF COMPUTER VISION, Vol: 81, Pages: 128-137, ISSN: 0920-5691
- Author Web Link
- Cite
- Citations: 27
Chli M, Davison AJ, 2009, Automatically and efficiently inferring the hierarchical structure of visual maps, Pages: 387-394, ISSN: 1050-4729
In Simultaneous Localisation and Mapping (SLAM), it is well known that probabilistic filtering approaches which aim to estimate the robot and map state sequentially suffer from poor computational scaling to large map sizes. Various authors have demonstrated that this problem can be mitigated by approximations which treat estimates of features in different parts of a map as conditionally independent, allowing them to be processed separately. When it comes to the choice of how to divide a large map into such 'submaps', straightforward heuristics may be sufficient in maps built using sensors such as laser range-finders with limited range, where a regular grid of submap boundaries performs well. With visual sensing, however, the ideal division of submaps is less clear, since a camera has potentially unlimited range and will often observe spatially distant parts of a scene simultaneously. In this paper we present an efficient and generic method for automatically determining a suitable submap division for SLAM maps, and apply this to visual maps built with a single agile camera. We use the mutual information between predicted measurements of features as an absolute measure of correlation, and cluster highly correlated features into groups. Via tree factorisation, we are able to determine not just a single level submap division but a powerful fully hierarchical correlation and clustering structure. Our analysis and experiments reveal particularly interesting structure in visual maps and give pointers to more efficient approximate visual SLAM algorithms.
Chli M, Davison AJ, 2009, Active Matching for visual tracking, Robotics and Autonomous Systems, Vol: 57, Pages: 1173-1187
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.