[Paper Review] A comprehensive review of 3D point cloud descriptors.

Xian-Feng Han, Jesse S. Jin+3 more|arXiv (Cornell University)|Feb 7, 2018

Robotics and Sensor-Based Localization55 references47 citations

TL;DR

This paper provides a comprehensive review of 3D point cloud descriptors, categorizing them into local-based, global-based, and hybrid-based approaches. It evaluates state-of-the-art descriptors for descriptiveness, robustness, and efficiency through extensive experiments, offering a critical benchmark for 3D computer vision applications.

ABSTRACT

The introduction of inexpensive 3D data acquisition devices has promisingly facilitated the wide availability and popularity of 3D point cloud, which attracts more attention on the effective extraction of novel 3D point cloud descriptors for accurate and efficient of 3D computer vision tasks. However, how to de- velop discriminative and robust feature descriptors from various point clouds remains a challenging task. This paper comprehensively investigates the exist- ing approaches for extracting 3D point cloud descriptors which are categorized into three major classes: local-based descriptor, global-based descriptor and hybrid-based descriptor. Furthermore, experiments are carried out to present a thorough evaluation of performance of several state-of-the-art 3D point cloud descriptors used widely in practice in terms of descriptiveness, robustness and efficiency.

Motivation & Objective

• To systematically categorize existing 3D point cloud descriptor approaches into local-based, global-based, and hybrid-based classes.
• To evaluate the performance of state-of-the-art descriptors in terms of descriptiveness, robustness, and efficiency.
• To provide a comparative analysis that supports informed selection of descriptors for 3D computer vision tasks.

Proposed method

• The paper classifies 3D point cloud descriptors into three main categories: local-based, global-based, and hybrid-based descriptors.
• It reviews representative methods within each category, highlighting their design principles and underlying assumptions.
• Experiments are conducted to evaluate descriptors using standardized metrics for descriptiveness, robustness, and computational efficiency.
• Performance is assessed across diverse point cloud datasets under varying conditions such as noise, downsampling, and viewpoint changes.
• The evaluation framework includes quantitative comparisons of descriptor performance across multiple benchmarking scenarios.

Experimental results

Research questions

• RQ1 How do local-based, global-based, and hybrid-based 3D point cloud descriptors differ in design and performance?
• RQ2 Which descriptor types demonstrate superior descriptiveness and robustness under challenging conditions like noise and downsampling?
• RQ3 What trade-offs exist between descriptor efficiency and performance across different 3D vision tasks?

Key findings

• The review identifies that hybrid-based descriptors often achieve a favorable balance between descriptiveness and robustness.
• Local-based descriptors tend to be more efficient but may lack discriminative power in complex scenes.
• Global-based descriptors show strong performance in unique shape representation but can be sensitive to noise and occlusion.
• Robustness varies significantly across descriptors, with some showing high resilience to noise and downsampling.
• Efficiency varies widely, with simpler descriptors offering faster computation at the cost of reduced discriminative capability.