Andrew J. Davison

KinectFusion: Real-time dense surface mapping and tracking

We present a system for accurate real-time mapping of complex and arbitrary indoor scenes in variable lighting conditions, using only a moving low-cost depth camera and commodity graphics hardware. We fuse all of the depth data streamed from a Kinect sensor into a single global implicit surface model of the observed scene in real-time. The current sensor pose is simultaneously obtained by tracking the live depth frame relative to the global model using a coarse-to-fine iterative closest point (ICP) algorithm, which uses all of the observed depth data available. We demonstrate the advantages of tracking against the growing full surface model compared with frame-to-frame tracking, obtaining tracking and mapping results in constant time within room sized scenes with limited drift and high accuracy. We also show both qualitative and quantitative results relating to various aspects of our tracking and mapping system. Modelling of natural scenes, in real-time with only commodity sensor and GPU hardware, promises an exciting step forward in augmented reality (AR), in particular, it allows dense surfaces to be reconstructed in real-time, with a level of detail and robustness beyond any solution yet presented using passive computer vision.

KinectFusion: real-time 3D reconstruction and interaction using a moving depth camera

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. Uses of the core system for low-cost handheld scanning, and geometry-aware augmented reality and physics-based interactions are shown. 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.

DTAM: Dense tracking and mapping in real-time

DTAM is a system for real-time camera tracking and reconstruction which relies not on feature extraction but dense, every pixel methods. As a single hand-held RGB camera flies over a static scene, we estimate detailed textured depth maps at selected keyframes to produce a surface patchwork with millions of vertices. We use the hundreds of images available in a video stream to improve the quality of a simple photometric data term, and minimise a global spatially regularised energy functional in a novel non-convex optimisation framework. Interleaved, we track the cameras 6DOF motion precisely by frame-rate whole image alignment against the entire dense model. Our algorithms are highly parallelisable throughout and DTAM achieves real-time performance using current commodity GPU hardware. We demonstrate that a dense model permits superior tracking performance under rapid motion compared to a state of the art method using features; and also show the additional usefulness of the dense model for real-time scene interaction in a physics-enhanced augmented reality application.

Simultaneous localization and map-building using active vision

An active approach to sensing can provide the focused measurement capability over a wide field of view which allows correctly formulated simultaneous localization and map-building (SLAM) to be implemented with vision, permitting repeatable longterm localization using only naturally occurring, automatically-detected features. In this paper, we present the first example of a general system for autonomous localization using active vision, enabled here by a high-performance stereo head, addressing such issues as uncertainty-based measurement selection, automatic map-maintenance, and goal-directed steering. We present varied real-time experiments in a complex environment.

Live dense reconstruction with a single moving camera

We present a method which enables rapid and dense reconstruction of scenes browsed by a single live camera. We take point-based real-time structure from motion (SFM) as our starting point, generating accurate 3D camera pose estimates and a sparse point cloud. Our main novel contribution is to use an approximate but smooth base mesh generated from the SFM to predict the view at a bundle of poses around automatically selected reference frames spanning the scene, and then warp the base mesh into highly accurate depth maps based on view-predictive optical flow and a constrained scene flow update. The quality of the resulting depth maps means that a convincing global scene model can be obtained simply by placing them side by side and removing overlapping regions. We show that a cluttered indoor environment can be reconstructed from a live hand-held camera in a few seconds, with all processing performed by current desktop hardware. Real-time monocular dense reconstruction opens up many application areas, and we demonstrate both real-time novel view synthesis and advanced augmented reality where augmentations interact physically with the 3D scene and are correctly clipped by occlusions.

Unified Inverse Depth Parametrization for Monocular SLAM

Recent work has shown that the probabilistic SLAM approach of explicit uncertainty propagation can succeed in permitting repeatable 3D real-time localization and mapping even in the ‘pure vision’ domain of a single agile camera with no extra sensing. An issue which has caused difficulty in monocular SLAM however is the initialization of features, since information from multiple images acquired during motion must be combined to achieve accurate depth estimates. This has led algorithms to deviate from the desirable Gaussian uncertainty representation of the EKF and related probabilistic filters during special initialization steps. In this paper we present a new unified 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 concept is direct parametrization of inverse depth, where there is a high degree of linearity. Importantly, our parametrization can cope with features which are so far from the camera that they present little parallax during motion, maintaining sufficient representative uncertainty that these points retain the opportunity to ‘come in’ from infinity if the camera makes larger movements. We demonstrate the parametrization using real image sequences of large-scale indoor and outdoor scenes.

Real-time monocular SLAM: Why filter?

While the most accurate solution to off-line structure from motion (SFM) problems is undoubtedly to extract as much correspondence information as possible and perform global optimisation, sequential methods suitable for live video streams must approximate this to fit within fixed computational bounds. Two quite different approaches to real-time SFM — also called monocular SLAM (Simultaneous Localisation and Mapping) — have proven successful, but they sparsify the problem in different ways. Filtering methods marginalise out past poses and summarise the information gained over time with a probability distribution. Keyframe methods retain the optimisation approach of global bundle adjustment, but computationally must select only a small number of past frames to process. In this paper we perform the first rigorous analysis of the relative advantages of filtering and sparse optimisation for sequential monocular SLAM. A series of experiments in simulation as well using a real image SLAM system were performed by means of covariance propagation and Monte Carlo methods, and comparisons made using a combined cost/accuracy measure. With some well-discussed reservations, we conclude that while filtering may have a niche in systems with low processing resources, in most modern applications keyframe optimisation gives the most accuracy per unit of computing time.

Scale Drift-Aware Large Scale Monocular SLAM

State of the art visual SLAM systems have recently been presented which are capable of accurate, large-scale and real-time performance, but most of these require stereo vision. Important application areas in robotics and beyond open up if similar performance can be demonstrated using monocular vision, since a single camera will always be cheaper, more compact and easier to calibrate than a multi-camera rig. With high quality estimation, a single camera moving through a static scene of course effectively provides its own stereo geometry via frames distributed over time. However, a classic issue with monocular visual SLAM is that due to the purely projective nature of a single camera, motion estimates and map structure can only be recovered up to scale. Without the known inter-camera distance of a stereo rig to serve as an anchor, the scale of locally constructed map portions and the corresponding motion estimates is therefore liable to drift over time. In this paper we describe a new near real-time visual SLAM system which adopts the continuous keyframe optimisation approach of the best current stereo systems, but accounts for the additional challenges presented by monocular input. In particular, we present a new pose-graph optimisation technique which allows for the efficient correction of rotation, translation and scale drift at loop closures. Especially, we describe the Lie group of similarity transformations and its relation to the corresponding Lie algebra. We also present in detail the system’s new image processing front-end which is able accurately to track hundreds of features per frame, and a filter-based approach for feature initialisation within keyframe-based SLAM. Our approach is proven via large-scale simulation and real-world experiments where a camera completes large looped trajectories.

KAZE features

In this paper, we introduce KAZE features, a novel multiscale 2D feature detection and description algorithm in nonlinear scale spaces. Previous approaches detect and describe features at different scale levels by building or approximating the Gaussian scale space of an image. However, Gaussian blurring does not respect the natural boundaries of objects and smoothes to the same degree both details and noise, reducing localization accuracy and distinctiveness. In contrast, we detect and describe 2D features in a nonlinear scale space by means of nonlinear diffusion filtering. In this way, we can make blurring locally adaptive to the image data, reducing noise but retaining object boundaries, obtaining superior localization accuracy and distinctiviness. The nonlinear scale space is built using efficient Additive Operator Splitting (AOS) techniques and variable conductance diffusion. We present an extensive evaluation on benchmark datasets and a practical matching application on deformable surfaces. Even though our features are somewhat more expensive to compute than SURF due to the construction of the nonlinear scale space, but comparable to SIFT, our results reveal a step forward in performance both in detection and description against previous state-of-the-art methods.

Mapping Large Loops with a Single Hand-Held Camera

This paper presents a method for Simultaneous Localization and Mapping (SLAM) relying on a monocular camera as the only sensor which is able to build outdoor, closedloop maps much larger than previously achieved with such input. Our system, based on the Hierarchical Map approach [1], builds independent local maps in real-time using the EKF-SLAM technique and the inverse depth representation proposed in [2]. The main novelty in the local mapping process is the use of a data association technique that greatly improves its robustness in dynamic and complex environments. A new visual map matching algorithm stitches these maps together and is able to detect large loops automatically, taking into account the unobservability of scale intrinsic to pure monocular SLAM. The loop closing constraint is applied at the upper level of the Hierarchical Map in near real-time. We present experimental results demonstrating monocular SLAM as a human carries a camera over long walked trajectories in outdoor areas with people and other clutter, even in the more difficult case of forward-looking camera, and show the closing of loops of several hundred meters.