This website requires JavaScript.
DOI: 10.1109/LRA.2022.3229224

GraffMatch: Global Matching of 3D Lines and Planes for Wide Baseline LiDAR Registration

Parker C. LuskDevarth ParikhJonathan P. How
Dec 2022
摘要
Using geometric landmarks like lines and planes can increase navigationaccuracy and decrease map storage requirements compared to commonly-used LiDARpoint cloud maps. However, landmark-based registration for applications likeloop closure detection is challenging because a reliable initial guess is notavailable. Global landmark matching has been investigated in the literature,but these methods typically use ad hoc representations of 3D line and planelandmarks that are not invariant to large viewpoint changes, resulting inincorrect matches and high registration error. To address this issue, we adoptthe affine Grassmannian manifold to represent 3D lines and planes and provethat the distance between two landmarks is invariant to rotation andtranslation if a shift operation is performed before applying the Grassmannianmetric. This invariance property enables the use of our graph-based dataassociation framework for identifying landmark matches that can subsequently beused for registration in the least-squares sense. Evaluated on a challenginglandmark matching and registration task using publicly-available LiDARdatasets, our approach yields a 1.7x and 3.5x improvement in successfulregistrations compared to methods that use viewpoint-dependent centroid and"closest point" representations, respectively.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答