由超点引导的无监督 3D 场景流估计（英文版）

英文标题：Self-Supervised 3D Scene Flow Estimation Guided by Superpoints中文摘要：本文提出了一种基于可动态更新超级点的 3D 场景流估计框架，该框架结合超级点生成模块和超级点导向流细化模块，通过流提导超级点的生成和可适应聚合超级点级流来重建点的流，并将一致性编码和重建流输入 GRU 以细化点级流，实现了对复杂 3D 场景的准确估计。英文摘要：3D scene flow estimation aims to estimate point-wise motions between twoconsecutive frames of point clouds. Superpoints, i.e., points with similargeometric features, are usually employed to capture similar motions of localregions in 3D scenes for scene flow estimation. However, in existing method