Geotri-VIO: 全景视觉惯性里程计
Published:
Overview / 项目概览
Geotri-VIO is a real-time Visual-Inertial Odometry (VIO) system designed for multi-prism panoramic (catadioptric) cameras. Traditional fisheye or pinhole models struggle to accurately describe these complex refractive cameras. Our work proposes a geometrically consistent multi-prism projection model that can accurately process visual information from the panoramic camera and tightly couple it with IMU data.
Geotri-VIO 是一个为多棱镜全景相机（panoramic camera）设计的实时视觉惯性里程计（VIO）系统。传统的鱼眼或针孔相机模型难以精确描述这类具有复杂折射的（catadioptric）全景相机。我们的工作提出了一种几何一致性的多棱镜投影模型，能够准确地处理来自全景相机的视觉信息，并将其与 IMU 数据紧密耦合。
Core Contributions / 核心贡献
- Proposed a novel, geometrically consistent multi-prism projection model.
- Integrated this model into an optimization-based, tightly-coupled VIO framework.
Experiments demonstrate that Geotri-VIO surpasses existing methods in both accuracy and robustness.
- 提出了一种新颖的、几何一致的多棱镜投影模型。
- 将该模型集成到一个基于优化的紧耦合 VIO 框架中。
- 实验证明，Geotri-VIO 在精度和鲁棒性上均优于现有的方法。