From Spherical to Gaussian: A Comparative Analysis of Point Cloud Cropping Strategies in Large-Scale 3D Environments

TL;DR

This paper compares spherical, exponential, Gaussian, and linear cropping strategies for dividing large-scale 3D point clouds into subclouds, demonstrating that alternative methods improve model performance, especially in outdoor scenes.

Contribution

It introduces and evaluates alternative cropping methods to spherical cropping, showing improved performance in large-scale 3D scene understanding.

Findings

Gaussian and exponential cropping outperform spherical cropping in outdoor scenes.

Altered cropping strategies lead to state-of-the-art results in 3D scene understanding.

Method improves geometric context retention in subclouds.

Abstract

Large-scale 3D point clouds can consist of hundreds of millions of points. Even after downsampling, these point clouds are too large for modern 3D neural networks. In order to develop a semantic understanding of the scene, the point clouds are divided into smaller subclouds that can be processed. Typically, this division is done using spherical crops, resulting in a loss of surrounding geometric context. To address this issue, we propose alternative methods that produce subclouds with larger crop sizes while maintaining a similar number of points. Specifically, we compare exponential, Gaussian, and linear cropping methods with the spherical method. We evaluated three 3D deep learning model architectures using multiple indoor and outdoor environment datasets. Our results demonstrate that altering the cropping strategy can enhance model performance, especially for large-scale outdoor scenes, yielding new state-of-the-art results. Code is available at https://github.com/mvg-inatech/point_cloud_cropping