LEADER：学习可靠局部到全局对应关系的激光雷达重定位
LEADER: Learning Reliable Local-to-Global Correspondences for LiDAR Relocalization
提出一种基于投影的几何编码器来提取多尺度几何特征，并设计截断相对可靠性损失来建模点级模糊性，可直接用于提升基于点云的回归式重定位在噪声和离群点下的鲁棒性。
原文依据 A Truncated Relative Reliability loss is then formulated to model point-wise ambiguity and mitigate the influence of unreliable predictions.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论激光雷达重定位因其能在复杂三维环境中提供精确的 6-DoF 姿态估计而受到越来越多的关注。近期基于学习的回归方法通过直接预测全局姿态而无需显式地图存储，提供了高效的解决方案。然而，这些方法通常对所有预测点同等对待，容易受到噪声和离群点的影响，因此在挑战性场景中往往表现不佳。本文提出 LEADER，一个由简单而有效的几何编码器增强的鲁棒激光雷达重定位框架。具体而言，首先提出一种基于投影的鲁棒几何编码器架构，该架构捕获多尺度几何特征以增强几何表示的描述性。然后，设计了一种截断相对可靠性损失来建模点级模糊性并减轻不可靠预测的影响。在 Oxford RobotCar 和 NCLT 数据集上的大量实验表明，LEADER 优于最优方法，相较于现有技术分别实现了 24.1% 和 73.9% 的位置误差相对降低。源代码已在 https://github.com/JiansW/LEADER 发布。
英文原文摘要 1197 字符
LiDAR relocalization has attracted increasing attention as it can deliver accurate 6-DoF pose estimation in complex 3D environments. Recent learning-based regression methods offer efficient solutions by directly predicting global poses without the need for explicit map storage. However, these methods often struggle in challenging scenes due to their equal treatment of all predicted points, which is vulnerable to noise and outliers. In this paper, we propose LEADER, a robust LiDAR-based relocalization framework enhanced by a simple, yet effective geometric encoder. Specifically, a Robust Projection-based Geometric Encoder architecture which captures multi-scale geometric features is first presented to enhance descriptiveness in geometric representation. A Truncated Relative Reliability loss is then formulated to model point-wise ambiguity and mitigate the influence of unreliable predictions. Extensive experiments on the Oxford RobotCar and NCLT datasets demonstrate that LEADER outperforms state-of-the-art methods, achieving 24.1% and 73.9% relative reductions in position error over existing techniques, respectively. The source code is released on https://github.com/JiansW/LEADER.