Instant Visual Odometry Initialization for Mobile AR

Concha, Alejo; Burri, Michael; Briales, Jesús; Förster, Christian; Oth, Luc

doi:10.1109/tvcg.2021.3106505

Cited by 10 publications

(4 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [17], the continuous scanning pose of the robot was solved by using a two-step Levenberg-Marquardt optimization approach, and the higher performance in trajectory tracking and map accuracy could be got. In [18], Alejo et al implemented the six-degrees-of-freedom pose estimation and environmental mapping by using the optimized LM approach, which has been shown to have good results.…”

Section: A Related Work Of Slammentioning

confidence: 99%

An Autonomous Navigation Strategy Based on Improved Hector SLAM With Dynamic Weighted A* Algorithm

Zhang

Wang

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Aiming at improving the mapping accuracy and autonomous navigation efficiency of rescue robot in unknown environment, an improved Hector SLAM based autonomous navigation strategy is proposed, which is implemented on the Levenberg-Marquardt optimization and Bezier smooth dynamic weighted A * algorithm. Firstly, the scan match process of Hector SLAM is performed by an improved Levenberg-Marquart (LM) method to solve the problems of non-convergence of functions and inaccurate local approximation caused by the non-singularity for solving the Hessian matrix. Secondly, the Bezier smooth dynamic weighted A * algorithm is utilized to perform autonomous navigation based on the SLAM map. In which, the navigation target points selection is employed by the frontier exploration strategy in order to solve the search efficiency as for the increasing number of nodes in the A * algorithm, and the accuracy of SLAM mapping declines owing to the large angle amplitude. Finally, the experiments are carried out in ROS, the results show that there is a better performance for proposed method to implement the high mapping accuracy and efficient autonomous navigation.INDEX TERMS SLAM, Levenberg-Marquart method, autonomous navigation, dynamic weighted A * algorithm.

show abstract

Section: A Related Work Of Slammentioning

confidence: 99%

An Autonomous Navigation Strategy Based on Improved Hector SLAM With Dynamic Weighted A* Algorithm

Zhang

Wang

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…A more recent work by Zuñiga-Noël et al [28] showed that up-to-scale SfM results could be leveraged in a quadratically-constrained least-squares problem, similar to closed-form solutions, which constrains the known magnitude of gravity to improve the accuracy. Another work by Concha et al [29] proposed a method that quickly initializes the 6 degrees of freedom (DoF) pose without motion parallax by decoupling the problem into the rotation, translation direction (5DoF) and magnitude of the translation (1DoF). While promising due to their robustification with RANSAC to handle outliers, they do not directly leverage inertial information in these low parallax scenarios.…”

Section: A Loosely-coupled Algorithmsmentioning

confidence: 99%

Fast Monocular Visual-Inertial Initialization Leveraging Learned Single-View Depth

Merrill,

Geneva,

Katragadda

et al. 2023

Robotics: Science and Systems XIX

View full text Add to dashboard Cite

In monocular visual-inertial navigation systems, it is ideal to initialize as quickly and robustly as possible. State-of-theart initialization methods typically make linear approximations using the image features and inertial information in order to initialize in closed-form, and then refine the states with a nonlinear optimization. While the standard methods typically wait for a 2sec data window, a recent work has shown that it is possible to initialize faster (0.5sec) by adding constraints from a robust but only up-to-scale monocular depth network in the nonlinear optimization. To further expedite the initialization, in this work, we leverage the scale-less depth measurements instead in the linear initialization step that is performed prior to the nonlinear one, which only requires a single depth image for the first frame. We show that the typical estimation of each feature state independently in the closed-form solution can be replaced by just estimating the scale and offset parameters of the learned depth map. Interestingly, our formulation makes it possible to construct small minimal problems in a RANSAC loop, whereas the typical linear system's minimal problem is quite large and includes every feature state. Experiments show that our method can improve the overall initialization performance on popular public datasets (EuRoC MAV and TUM-VI) over stateof-the-art methods. For the TUM-VI dataset, we show superior initialization performance with only a 0.3sec window of data, which is the smallest ever reported, and show that our method can initialize more often, robustly, and accurately in different challenging scenarios.

show abstract

“…For example, in a Moving through Glass AR application ( Tunur et al, 2020 ), users are allowed to evoke virtual trainers that, however, cannot recognize the user’s body movements and respond accordingly. ARcore and ARkit exemplify Google and Apple’s respective platforms for creating AR experiences ( Concha et al, 2021 ), which enable more complex AR software solutions in conjunction with devices such as smartphones, tablets, smart glasses, and headsets. Unlike other XR technologies, AR can be experienced in a fully- or semi-immersive viewing environment.…”

Section: Definitionsmentioning

confidence: 99%

Combining physical and virtual worlds for motor-cognitive training interventions: Position paper with guidelines on technology classification in movement-related research

et al. 2022

View full text Add to dashboard Cite

Efficient movements require intact motor and cognitive function. There is a growing literature on motor-cognitive interventions to improve the overall quality of life of healthy or diseased older people. For such interventions, novel technological advances are crucial not only in terms of motivation but also to improve the user experience in a multi-stimuli world, usually offered as a mixture of real and virtual environments. This article provides a classification system for movement-related research dealing with motor-cognitive interventions performed in different extents of a virtual environment. The classification is divided into three categories: (a) type of digital device with the associated degree of immersiveness provided; (b) presence or absence of a human-computer interaction; and (c) activity engagement during training, defined by activity >1.5 Metabolic Equivalent of task. Since virtual reality (VR) often categorizes different technologies under the same term, we propose a taxonomy of digital devices ranging from computer monitors and projectors to head-mounted VR technology. All immersive technologies that have developed rapidly in recent years are grouped under the umbrella term Extended Reality (XR). These include augmented reality (AR), mixed reality (MR), and VR, as well as all technologies that have yet to be developed. This technology has potential not only for gaming and entertainment, but also for research, motor-cognitive training programs, rehabilitation, telemedicine, etc. This position paper provides definitions, recommendations, and guidelines for future movement-related interventions based on digital devices, human-computer interactions, and physical engagement to use terms more consistently and contribute to a clearer understanding of their implications.

show abstract

Instant Visual Odometry Initialization for Mobile AR

Cited by 10 publications

References 24 publications

An Autonomous Navigation Strategy Based on Improved Hector SLAM With Dynamic Weighted A* Algorithm

An Autonomous Navigation Strategy Based on Improved Hector SLAM With Dynamic Weighted A* Algorithm

Fast Monocular Visual-Inertial Initialization Leveraging Learned Single-View Depth

Combining physical and virtual worlds for motor-cognitive training interventions: Position paper with guidelines on technology classification in movement-related research

Contact Info

Product

Resources

About