Self-supervised rigid transformation equivariance for accurate 3D point cloud registration

作者：

Highlights：

• We build a dedicated RTE and design a Siamese structurefor 3D point cloud registration.

• We propose to learn the matching matrix from the LCV more effectively instead of the hand crafted matching strategy.

• Remarkable performance on several datasets topping the state of the art methods proves the effectiveness of our method.

摘要

•We build a dedicated RTE and design a Siamese structurefor 3D point cloud registration.•We propose to learn the matching matrix from the LCV more effectively instead of the hand crafted matching strategy.•Remarkable performance on several datasets topping the state of the art methods proves the effectiveness of our method.

论文关键词：Point cloud,Rigid transformation equivariance,Learned cost volume

论文评审过程：Received 23 August 2021, Revised 13 March 2022, Accepted 7 May 2022, Available online 10 May 2022, Version of Record 25 May 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108784