Learning a 3D descriptor for cross-source point cloud registration from synthetic data

TL;DR

This paper introduces a deep learning-based 3D descriptor trained on synthetic data to improve cross-source point cloud registration, effectively handling sensor differences, noise, and transformations.

Contribution

It presents a novel 3D descriptor learned from synthetic data that generalizes well to real cross-source point clouds, enhancing registration accuracy.

Findings

Outperforms state-of-the-art methods in cross-source registration

Generalizes to new scenes and different sensors

Robust to noise, missing data, and geometric transformations

Abstract

As the development of 3D sensors, registration of 3D data (e.g. point cloud) coming from different kind of sensor is dispensable and shows great demanding. However, point cloud registration between different sensors is challenging because of the variant of density, missing data, different viewpoint, noise and outliers, and geometric transformation. In this paper, we propose a method to learn a 3D descriptor for finding the correspondent relations between these challenging point clouds. To train the deep learning framework, we use synthetic 3D point cloud as input. Starting from synthetic dataset, we use region-based sampling method to select reasonable, large and diverse training samples from synthetic samples. Then, we use data augmentation to extend our network be robust to rotation transformation. We focus our work on more general cases that point clouds coming from different sensors, named cross-source point cloud. The experiments show that our descriptor is not only able to generalize to new scenes, but also generalize to different sensors. The results demonstrate that the proposed method successfully aligns two 3D cross-source point clouds which outperforms state-of-the-art method.