Scale-Consistent Monocular Fisheye Visual Odometry via Spherical Bundle Adjustment-Guided Unsupervised Depth Learning
Scale-Consistent Monocular Fisheye Visual Odometry via Spherical Bundle Adjustment-Guided Unsupervised Depth Learning
This research introduces S$^2$D-FVO, a monocular fisheye visual odometry framework that combines spherical bundle adjustment with unsupervised depth learning to improve scale consistency and depth estimation accuracy. The method uses tangent-space angular reprojection residuals derived from the spherical logarithmic map to handle fisheye distortion and integrates a median depth-based scale regularization term. Experiments on multiple public datasets and two newly collected large-scale datasets show that S$^2$D-FVO outperforms existing approaches in depth and trajectory estimation across different camera models.
The paper proposes S$^2$D-FVO, a monocular fisheye visual odometry framework that integrates spherical bundle adjustment with unsupervised depth learning to address scale drift and distortion challenges; results are reported on public datasets and two newly collected large-scale datasets.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
Source：arXiv Robotics — research abstracts · arxiv.org