Pose Estimation for Omni-directional Cameras using Sinusoid Fitting

et al. 2021

J Intell Robot Syst

Self Cite

Rotation estimation is important for localization, mapping and navigation in robotic applications. A novel rotation estimation method for geometric vision of omni-directional cameras is proposed in this paper. Firstly, we formulate the rotation estimation as a sinusoid fitting problem on the basis of column-wise and row-wise shifts in omni-directional images. Then the method is implemented in two steps: motion vector extraction and sinusoid fitting. We are extracting motion vectors by finding pixel deviations in row-and column-directions as well as image rotation for each sub-image by Fourier-Mellin transform (FMT) or by optical flow. Then we fit these column-wise motion vectors to two sinusoid functions. Due to sharing the optimization variables, column-wise shifts and image rotation of each motion vector can be optimized jointly. Afterwards, the 3D rotation between two frames is obtained from the offsets, amplitudes and phase-shifts of the two sinusoid functions. Finally, we perform experiments for 3D rotation, which show that our algorithm outperforms the geometry-based methods in accuracy, robustness and speed. Compared to our early version, this paper makes several improvements: 1) modeling both pixel shifts and sub-regions' rotation to the sinusoid fitting problem, instead of only pixel shifts; 2) exploiting optical flow to extract motion vectors in addition to FMT; 3) comparing with a geometry-based method for pure rotation as well as five-point algorithms.

Section: Methodsmentioning

confidence: 87%

Section: Methodsmentioning

confidence: 99%

Section: Motivationmentioning

confidence: 99%

See 1 more Smart Citation

Rotation Estimation for Omni-directional Cameras Using Sinusoid Fitting

Long

et al. 2021

J Intell Robot Syst

Self Cite

“…In [34], the oversampling technology and Dirichlet-based phase filter were used to make FMT robust to some image skew. Moreover, the sub-image extraction strategy [21,[35][36][37] is popular in addressing the 3D motion problem. In this paper, we mainly focus on the second case, i.e., to relax the constraints of equidistance and planar environments.…”

Section: Introductionmentioning

confidence: 99%

Rethinking the Fourier-Mellin Transform: Multiple Depths in the Camera’s View

Kneip

et al. 2021

Remote Sensing

Self Cite

Remote sensing and robotics often rely on visual odometry (VO) for localization. Many standard approaches for VO use feature detection. However, these methods will meet challenges if the environments are feature-deprived or highly repetitive. Fourier-Mellin Transform (FMT) is an alternative VO approach that has been shown to show superior performance in these scenarios and is often used in remote sensing. One limitation of FMT is that it requires an environment that is equidistant to the camera, i.e., single-depth. To extend the applications of FMT to multi-depth environments, this paper presents the extended Fourier-Mellin Transform (eFMT), which maintains the advantages of FMT with respect to feature-deprived scenarios. To show the robustness and accuracy of eFMT, we implement an eFMT-based visual odometry framework and test it in toy examples and a large-scale drone dataset. All these experiments are performed on data collected in challenging scenarios, such as, trees, wooden boards and featureless roofs. The results show that eFMT performs better than FMT in the multi-depth settings. Moreover, eFMT also outperforms state-of-the-art VO algorithms, such as ORB-SLAM3, SVO and DSO, in our experiments.

“…Apart from normal pin-hole cameras, omnidirectional cameras are becoming popular, due to their large field of view (FOV). ey have been widely used on ground robots [12][13][14][15][16]. Some research groups also studied omnidirectional cameras for underwater use since they provide more information than perspective ones on object detection, localization, and mapping.…”

Section: Introductionmentioning

confidence: 99%

Underwater Depth Estimation for Spherical Images

Cui

Jin

et al. 2021

Journal of Robotics

Self Cite

This paper proposes a method for monocular underwater depth estimation, which is an open problem in robotics and computer vision. To this end, we leverage publicly available in-air RGB-D image pairs for underwater depth estimation in the spherical domain with an unsupervised approach. For this, the in-air images are style-transferred to the underwater style as the first step. Given those synthetic underwater images and their ground truth depth, we then train a network to estimate the depth. This way, our learning model is designed to obtain the depth up to scale, without the need of corresponding ground truth underwater depth data, which is typically not available. We test our approach on style-transferred in-air images as well as on our own real underwater dataset, for which we computed sparse ground truth depths data via stereopsis. This dataset is provided for download. Experiments with this data against a state-of-the-art in-air network as well as different artificial inputs show that the style transfer as well as the depth estimation exhibit promising performance.