Abstract:3D point cloud registration is a fundamental problem in computer vision, computer graphics, robotics, remote sensing, and etc. Over the last thirty years, we have witnessed the amazing advancement in this area with numerous kinds of solutions. Although a handful of relevant surveys have been conducted, their coverage is still limited. In this work, we present a comprehensive survey on 3D point cloud registration, covering a set of sub-areas such as pairwise coarse registration, pairwise fine registration, multi-view registration, cross-scale registration, and multi-instance registration. The datasets, evaluation metrics, method taxonomy, discussions of the merits and demerits, insightful thoughts of future directions are comprehensively presented in this survey. The regularly updated project page of the survey is available at <a class="link-external link-https" href="https://github.com/Amyyyy11/3D-Registration-in-30-Years-A-Survey" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

The problems that this paper attempts to solve are multiple challenges in 3D point cloud registration. 3D point cloud registration is a fundamental problem and has wide applications in fields such as computer vision, computer graphics, robotics, and remote sensing. Specifically, the paper aims to: 1. **Comprehensive Review and New Taxonomy**: Provide a comprehensive overview of 3D point cloud registration methods in the past three decades, covering multiple sub - fields including pairwise coarse registration, pairwise fine registration, multi - view registration, cross - scale registration, and multi - instance registration. It provides a systematic taxonomy and extensive literature coverage. 2. **Benchmark Overview and Performance Comparison**: Systematically summarize popular benchmark datasets and performance evaluation metrics, and report the comparison results of representative up - to - date methods on standard benchmarks. 3. **Outlook on Future Research Directions**: Highlight the features, advantages, and disadvantages of existing methods, and deeply discuss current challenges and several future research directions to inspire follow - up work in this field. Through these efforts, the paper fills the gaps in existing reviews that focus on the point cloud registration task in different parts or within a limited scope, providing a more comprehensive perspective covering relevant literature in the past three decades. ### Key Issues - **Fundamental Problem of 3D Point Cloud Registration**: Align multiple 3D point clouds into a unified coordinate system, generate a more complete point cloud, and obtain 6 - DoF (six - degrees - of - freedom) pose parameters. - **Different Sub - problems**: - **Pairwise Coarse Registration**: Handle the initial alignment between two point clouds, usually involving large pose changes. - **Pairwise Fine Registration**: Focus on minimizing small residual errors. - **Multi - view Registration**: Align multiple sequential or unordered point clouds. - **Cross - scale Registration**: Handle point clouds of different resolutions or scales. - **Multi - instance Registration**: Handle point clouds of multiple instances from the same object or scene. ### Method Classification The paper classifies 3D point cloud registration according to geometric methods and deep - learning - based methods, and details the advantages and limitations of each method. In addition, it also discusses the registration problem from different perspectives such as feature learning, correspondence learning, and robust 6 - DoF pose estimation. ### Datasets and Evaluation Metrics The paper also systematically summarizes various datasets and evaluation metrics for evaluating 3D point cloud registration performance, such as Registration Recall (RR), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Square Error (MSE), Isotropic Error (MIE), Rotation Error (RE), and Translation Error (TE), etc. Through the above content, the paper provides readers with a comprehensive and systematic overview of the 3D point cloud registration field, points out the advantages and disadvantages of existing methods, and provides directions for future research.

3D Registration in 30 Years: A Survey

A comprehensive survey on point cloud registration

Rigid pairwise 3D point cloud registration: A survey

A Survey of Non-Rigid 3D Registration

Point cloud registration for LiDAR and photogrammetric data: A critical synthesis and performance analysis on classic and deep learning algorithms

Deep learning-based low overlap point cloud registration for complex scenario: The review

A Systematic Approach for Cross-source Point Cloud Registration by Preserving Macro and Micro Structures

Learning-based Point Cloud Registration: A Short Review and Evaluation.

Registration of Laser Scanning Point Clouds: A Review

A review of non-rigid transformations and learning-based 3D point cloud registration methods

A Comprehensive Survey and Taxonomy on Point Cloud Registration Based on Deep Learning

A Tutorial Review on Point Cloud Registrations: Principle, Classification, Comparison, and Technology Challenges

Deep learning-based point cloud registration: a comprehensive investigation

A fast point cloud registration method based on spatial relations and features

Efficient Single Correspondence Voting for Point Cloud Registration

Deep Learning Based 3D Segmentation: A Survey

Cross-source Point Cloud Registration: Challenges, Progress and Prospects

A Comprehensive Performance Evaluation of 3-D Transformation Estimation Techniques in Point Cloud Registration

Efficient Pairwise 3-D Registration of Urban Scenes via Hybrid Structural Descriptors

A new method for two-stage partial-to-partial 3D point cloud registration: multi-level interaction perception

3D Scanning Data Processing and Registration for the Construction of Urban Scenes Via Image Methods