Spherical-Model-Based SLAM on Full-View Images for Indoor Environments

Li, Jianfeng; Wang, Xiaowei; Li, Shigang

doi:10.3390/app8112268

Cited by 17 publications

(9 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Except these indirect methods, large-scale direct monocular SLAM (LSD-SLAM) [71] and direct sparse odometry (DSO) [72] are the representative direct works, and there is also semi-direct monocular visual odometry (SVO) [73]. Furthermore, in order to deal with the wide FOV spherical images, [74] proposed a panoramic image pose calculation method, after that, many works turn to research the panorama SLAM, like the early phase [75][76][77], and the recent works [78][79][80]focus on panoramas or omnidirectional visual SLAM. But drift or accumulate error in the optimization are still the challenge.…”

Section: B Local Featurementioning

confidence: 99%

The Relocalization of SLAM Tracking Based on Spherical Cameras

Chang¹,

Liu

Xin

et al. 2021

IEEE Access

View full text Add to dashboard Cite

This work proposes a novel solution to relocalize the SLAM tracking based on spherical cameras. It focuses on the imaging method of spherical camera, the feature extracting algorithm and the relocalization of SLAM tracking based on the 3D reconstruction. In the imaging method, we design a new camera containing eight fish-eye lenses, and then we propose a calibration method to calibrate the eight fish-eye lenses spherical camera; To get the high-performance feature points of panoramic image, we propose a network based on a separate network to extract local feature accurately and quickly. With the correct key points obtained by the feature extracting method, we reoptimize the SLAM tracking after the maximum posteriori estimation usually applied in common back-end SLAM to relocalize the SLAM tracking. The experiment results show that the calibration method achieved 0.973 reprojection error, lower than the common methods like Zhang's or DLT. The inlier rate and matching time of proposed SimpGeoDesc are all better than the reference models ContextDesc and GeoDesc. With the correct feature points, SLAM tracking is clearer and more steadily with our relocalization method. That is the solution of relocalization of SLAM tracking proposed in this work is effective. The AR application of the relocalization proves the feasibility of our propose relocalization method.

show abstract

Section: B Local Featurementioning

confidence: 99%

The Relocalization of SLAM Tracking Based on Spherical Cameras

Chang¹,

Liu

Xin

et al. 2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…To expand the versatility of the camera-based motion estimation, omnidirectional [24], fisheye [25], or spherical cameras [26] have been adopted in recent years. They have proven to be more beneficial compared to perspective cameras, owing to their wider field-of-view.…”

Section: Related Workmentioning

confidence: 99%

SelfSphNet: Motion Estimation of a Spherical Camera via Self-Supervised Learning

et al. 2020

View full text Add to dashboard Cite

In this paper, we propose SelfSphNet, that is, a self-supervised learning network to estimate the motion of an arbitrarily moving spherical camera without the need for any labeled training data. Recently, numerous learning-based methods for camera motion estimation have been proposed. However, most of these methods require an enormous amount of labeled training data, which is difficult to acquire experimentally. To solve this problem, our SelfSphNet employs two loss functions to estimate the frame-to-frame camera motion, thus giving two supervision signals to the network with the usage of unlabeled training data. First, a 5 DoF epipolar angular loss, which is composed of a dense optical flow of spherical images, estimates the 5 DoF motion between two image frames. This loss function utilizes a unique property of the spherical optical flow, which allows the rotational and translational components to be decoupled by using a derotation operation. This operation is derived from the fact that spherical images can be rotated to any orientation without any loss of information, hence making it possible to ''decouple'' the dense optical flow between pairs of spherical images to a pure translational state. Next, a photometric reprojection loss estimates the full 6 DoF motion using a depth map generated from the decoupled optical flow. This minimization strategy enables our network to be optimized without using any labeled training data. To confirm the effectiveness of our proposed approach (SelfSphNet), several experiments to estimate the camera trajectory, as well as the camera motion, were conducted in comparison to a previous self-supervised learning approach, SfMLearner, and a fully supervised learning approach whose baseline network is the same as SelfSphNet. Moreover, transfer learning in a new scene was also conducted to verify that our proposed method can optimize the network with newly collected unlabeled data. INDEX TERMS Motion estimation, computer vision, image processing, deep learning, convolutional neural networks.

show abstract

“…Second, the geometry and camera motion are estimated through the coordinates of corresponding points. J. Li et al [22] presented a SLAM system based on spherical model for full-view images in indoor environments, but the system was not real-time. S. Wang et al [23] proposed to extend ORB-SLAM framework by the unified camera model and the semi-dense depth map, whose map initialization method follows the idea from [18], but there was no evaluation on localization accuracy and robustness of the system.…”

Section: Related Workmentioning

confidence: 99%

Accurate and Robust Monocular SLAM with Omnidirectional Cameras

Liu

Guo

Feng

et al. 2019

Sensors

View full text Add to dashboard Cite

Simultaneous localization and mapping (SLAM) are fundamental elements for many emerging technologies, such as autonomous driving and augmented reality. For this paper, to get more information, we developed an improved monocular visual SLAM system by using omnidirectional cameras. Our method extends the ORB-SLAM framework with the enhanced unified camera model as a projection function, which can be applied to catadioptric systems and wide-angle fisheye cameras with 195 degrees field-of-view. The proposed system can use the full area of the images even with strong distortion. For omnidirectional cameras, a map initialization method is proposed. We analytically derive the Jacobian matrices of the reprojection errors with respect to the camera pose and 3D position of points. The proposed SLAM has been extensively tested in real-world datasets. The results show positioning error is less than 0.1% in a small indoor environment and is less than 1.5% in a large environment. The results demonstrate that our method is real-time, and increases its accuracy and robustness over the normal systems based on the pinhole model.

show abstract

Spherical-Model-Based SLAM on Full-View Images for Indoor Environments

Cited by 17 publications

References 24 publications

The Relocalization of SLAM Tracking Based on Spherical Cameras

The Relocalization of SLAM Tracking Based on Spherical Cameras

SelfSphNet: Motion Estimation of a Spherical Camera via Self-Supervised Learning

Accurate and Robust Monocular SLAM with Omnidirectional Cameras

Contact Info

Product

Resources

About