DART: dense articulated real-time tracking with consumer depth cameras

Abstract

This paper introduces DART, a general framework for tracking articulated objects composed of rigid bodies connected through a kinematic tree. DART covers a broad set of objects encountered in indoor environments, including furniture and tools, and human and robot bodies, hands and manipulators. To achieve efficient and robust tracking, DART extends the signed distance function representation to articulated objects and takes full advantage of highly parallel GPU algorithms for data association and pose optimization. We demonstrate the capabilities of DART on different types of objects that have each required dedicated tracking techniques in the past.

Robust Human Pose Tracking For Realistic Service Robot Applications

“…A generalization of Signed Distance Functions (SDF) [12] for articulated objects is introduced by Schmidt et al [32]. Objects are represented by a symmetric version of the articulated SDF in 3D space, which can be precomputed, allowing for a dense representation of the data and gradient based optimization is employed to estimate the optimal pose, taking into consideration free space constraints similar to [16].…”

“…The framework's human pose tracker employs the articulated SDF model [32] and uses an articulated skinned human template to track the human pose in sequences of depth images, extracting the 3D positions of the skeletal joints.…”

“…The current work aspires to improve the robustness of human motion analysis in realistic settings, by introducing a real-time human pose estimation and tracking framework, which builds upon the articulated SDF-based model presented in [32] and is extended through a series of complementary features and mechanisms. The features proposed herein target specific problems encountered under real-life monitoring conditions, eventually leading to the development of a complete standalone framework suitable for deployment in real life, assistive robot applications.…”

New trends on moving object detection in video images captured by a moving camera: A survey

“…In another method [101], a complete model of articulated objects composed of rigid bodies was proposed and applied for tracking and pose estimation of manipulated robots. The success of this approach was mostly due to use of depth sensors in order to accurately produce a depth image representing different poses of a moving object.…”

TSDF-based change detection for consistent long-term dense reconstruction and dynamic object discovery

“…Our work as well as [5], [7], and [9] focus solely on change that happens in between observations. In contrast, Schmidt et al [12] make use of Signed Distance Function (SDF) submaps to track articulated objects that move during a single observation.…”