武汉大学 李由教授团队:基于彩色点云地图双稀疏约束与分层优化的视觉里程计方法 | SANA佳文速递

Robust and efficient visual odometry using colored point cloud maps via dual-sparsity and hierarchical optimization

Xuanxuan Zhang, You Li*, Tianxiang Zhang, Zhenqi Zheng, Xueli Guo, Zhichao Wen & Fuqiang Gu

Satellite Navigation (2026) 7: 9

引用文章:

Zhang, X., Li, Y., Zhang, T. et al. Robust and efficient visual odometry using colored point cloud maps via dual-sparsity and hierarchical optimization. Satell Navig 7, 9 (2026). https://doi.org/10.1186/s43020-026-00196-x

PDF文件下载链接：

https://link.springer.com/article/10.1186/s43020-026-00196-x

-长按识别二维码查看/下载全文-

Reliable localization is essential for navigation in GNSS-challenged environments, where monocular visual odometry often suffers from accumulated drift and multi-sensor fusion increases hardware complexity. This paper proposes a robust and efficient camera-only visual odometry system using prebuilt colored point cloud maps. The method directly exploits the global map as a constraint and leverages the shared color information between camera images and colored point clouds to improve localization accuracy and robustness. A dual-sparsity matching strategy is introduced to retain high-gradient points in both the map and image observations, reducing redundant computation and avoiding local optima. Furthermore, a hierarchical optimization-based iterative Kalman filtering framework combines geometric reprojection with photometric refinement in a coarse-to-fine manner. Experiments on public and self-collected datasets demonstrate significant improvements in accuracy, robustness, and efficiency for long-term localization in complex environments.