PDConv: Rigid transformation invariant convolution for 3D point clouds

作者：

Highlights：

• A new representation for 3D point clouds based on distances.

• A rigid transformation invariant convolution operator for point clouds.

• Deep neural network for learning rigid transformation invariant features.

• Experiments on point clouds classification and parts and semantic segmentation.

摘要

•A new representation for 3D point clouds based on distances.•A rigid transformation invariant convolution operator for point clouds.•Deep neural network for learning rigid transformation invariant features.•Experiments on point clouds classification and parts and semantic segmentation.

论文关键词：Point clouds,Transformation invariance,Rotation,Translation,Classification,Parts segmentation

论文评审过程：Received 16 December 2021, Revised 28 July 2022, Accepted 1 August 2022, Available online 5 August 2022, Version of Record 12 August 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118356