Rapid and Robust Monocular Visual-Inertial Initialization with Gravity Estimation via Vertical Edges

Li, Jinyu; Bao, Hujun

doi:10.1109/iros40897.2019.8968456

Cited by 6 publications

(5 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To generate reliable correspondences for visual and monodepth constraints, our front-end uses gyro measurements as a prior for frame-toframe rotations following 2-pt RANSAC [58]. We first exhaustively evaluate VIO initialization performance on the whole trajectory by running our initialization routine in windows sampled throughout each trajectory in the dataset, which is commonly done in a variety initialization works [7,8,30]. Additionally, we also evaluate the effect of initialization on tracking performance by employing our method on a baseline similar to OpenVINS [47] in 10s time windows distributed uniformly across datasets.…”

Section: Methodsmentioning

confidence: 99%

“…Visual-inertial odometry [18,19] is a well-studied problem in both the computer vision and robotics communities and many works [20][21][22][23][24][25][26][27][28] have focused specifically on accurate initial estimation of states required by the inertial sensor. These works can be roughly classified into two categories -1) jointly solving a visual-inertial SFM problem directly in closed form or as a bundle adjustment problem [5,10,29] and 2) cascaded approaches which solve a pure visual SFM for up to scale pose followed by metric scale recovery using inertial observations [7][8][9]30]. Both approaches typically use a visual-inertial bundle adjustment (VI-BA) step to further refine their solution.…”

Section: Related Workmentioning

confidence: 99%

“…less data collection time) even in challenging low-motion scenarios. Some prior works have explored using higher order visual information such as lines [30] for increased system observability in monocular VIO. Additionally, RGB-D SLAM systems [12][13][14] have been tremendously successful in a number of domains (AR/VR, self driving cars, etc.)…”

Section: Related Workmentioning

confidence: 99%

“…For each metric, lower is better. Following prior related initialization works [7,8,30], we exhaustively create VIO initialization events across the whole trajectory to evaluate performance across different motion and visual scenarios. For a fair comparison, we split each dataset into segments evenly and attempt to initialize all methods on the same set of segments.…”

Section: Exhaustive Initialization Evaluationmentioning

confidence: 99%

See 3 more Smart Citations

Learned Monocular Depth Priors in Visual-Inertial Initialization

Zhou¹,

Kar²,

Turner³

et al. 2022

Preprint

View full text Add to dashboard Cite

Visual-inertial odometry (VIO) is the pose estimation backbone for most AR/VR and autonomous robotic systems today, in both academia and industry. However, these systems are highly sensitive to the initialization of key parameters such as sensor biases, gravity direction, and metric scale. In practical scenarios where high-parallax or variable acceleration assumptions are rarely met (e.g. hovering aerial robot, smartphone AR user not gesticulating with phone), classical visualinertial initialization formulations often become ill-conditioned and/or fail to meaningfully converge. In this paper we target visual-inertial initialization specifically for these low-excitation scenarios critical to in-the-wild usage. We propose to circumvent the limitations of classical visual-inertial structure-from-motion (SfM) initialization by incorporating a new learning-based measurement as a higher-level input. We leverage learned monocular depth images (mono-depth) to constrain the relative depth of features, and upgrade the mono-depth to metric scale by jointly optimizing for its scale and shift. Our experiments show a significant improvement in problem conditioning compared to a classical formulation for visual-inertial initialization, and demonstrate significant accuracy and robustness improvements relative to the state-of-the-art on public benchmarks, particularly under motion-restricted scenarios. We further extend this improvement to implementation within an existing odometry system to illustrate the impact of our improved initialization method on resulting tracking trajectories.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Exhaustive Initialization Evaluationmentioning

confidence: 99%

See 2 more Smart Citations

Learned Monocular Depth Priors in Visual-Inertial Initialization

Zhou¹,

Kar²,

Turner³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Disjoint visual-inertial initialization was pioneered by Mur-Artal et al in ORB-SLAM-VI [ 35 ] and later adopted by Qin et al in VINS-Mono [ 15 , 53 ]. All subsequent disjoint initialization methods are improvements to these two methods, such as [ 54 , 55 , 56 ]. Compared to joint initialization, disjoint initialization has faster solution speed and better robustness.…”

Section: Related Workmentioning

confidence: 99%

ESVIO: Event-Based Stereo Visual-Inertial Odometry

Liu

Shi

et al. 2023

Sensors

View full text Add to dashboard Cite

The emerging event cameras are bio-inspired sensors that can output pixel-level brightness changes at extremely high rates, and event-based visual-inertial odometry (VIO) is widely studied and used in autonomous robots. In this paper, we propose an event-based stereo VIO system, namely ESVIO. Firstly, we present a novel direct event-based VIO method, which fuses events’ depth, Time-Surface images, and pre-integrated inertial measurement to estimate the camera motion and inertial measurement unit (IMU) biases in a sliding window non-linear optimization framework, effectively improving the state estimation accuracy and robustness. Secondly, we design an event-inertia semi-joint initialization method, through two steps of event-only initialization and event-inertia initial optimization, to rapidly and accurately solve the initialization parameters of the VIO system, thereby further improving the state estimation accuracy. Based on these two methods, we implement the ESVIO system and evaluate the effectiveness and robustness of ESVIO on various public datasets. The experimental results show that ESVIO achieves good performance in both accuracy and robustness when compared with other state-of-the-art event-based VIO and stereo visual odometry (VO) systems, and, at the same time, with no compromise to real-time performance.

show abstract