Point Cloud Registration Research Articles

SVC: Sight view constraint for robust point cloud registration

Partial to Partial Point Cloud Registration (partial PCR) remains a challenging task, particularly when dealing with a low overlap rate. In comparison to the full-to-full registration task, we find that the objective of partial PCR is still not well-defined, indicating no metric can reliably identify the true transformation. We identify this as the most fundamental challenge in partial PCR tasks. In this paper, instead of directly seeking the optimal transformation, we propose a novel and general Sight View Constraint (SVC) to conclusively identify incorrect transformations, thereby enhancing the robustness of existing PCR methods. Extensive experiments validate the effectiveness of SVC on both indoor and outdoor scenes. On the challenging 3DLoMatch dataset, our approach increases the registration recall from 78% to 82%, achieving the state-of-the-art result. This research also highlights the significance of the decision version problem of partial PCR, which has the potential to provide novel insights into the partial PCR problem. Code will be available at: https://github.com/pppyj-m/SVC.

A Base-Map-Guided Global Localization Solution for Heterogeneous Robots Using a Co-View Context Descriptor

With the continuous advancement of autonomous driving technology, an increasing number of high-definition (HD) maps have been generated and stored in geospatial databases. These HD maps can provide strong localization support for mobile robots equipped with light detection and ranging (LiDAR) sensors. However, the global localization of heterogeneous robots under complex environments remains challenging. Most of the existing point cloud global localization methods perform poorly due to the different perspective views of heterogeneous robots. Leveraging existing HD maps, this paper proposes a base-map-guided heterogeneous robots localization solution. A novel co-view context descriptor with rotational invariance is developed to represent the characteristics of heterogeneous point clouds in a unified manner. The pre-set base map is divided into virtual scans, each of which generates a candidate co-view context descriptor. These descriptors are assigned to robots before operations. By matching the query co-view context descriptors of a working robot with the assigned candidate descriptors, the coarse localization is achieved. Finally, the refined localization is done through point cloud registration. The proposed solution can be applied to both single-robot and multi-robot global localization scenarios, especially when communication is impaired. The heterogeneous datasets used for the experiments cover both indoor and outdoor scenarios, utilizing various scanning modes. The average rotation and translation errors are within 1° and 0.30 m, indicating the proposed solution can provide reliable localization support despite communication failures, even across heterogeneous robots.

Geodesic-Based Maximal Cliques Search for Non-Rigid Human Point Cloud Registration.

Non-rigid point cloud registration holds significant importance for human body pose analysis in the fields of sports, medicine, gaming, etc. In this paper, we propose a non-rigid point cloud registration algorithm based on geodesic distance measurement, which can improve the accuracy of the registration for matching point pairs during non-rigid deformations. Firstly, a graph is constructed for two sets of point clouds using geodesic distance measurement considering that geodesic distance changes minimally during non-rigid deformation of the human body, which can preserve the point cloud matching information between corresponding points. Furthermore, a maximal clique search is employed to find combinations of matching pairs between point clouds. Finally, by driving the human body model parameters, sparse matching pairs are overlapped as much as possible to achieve non-rigid point cloud registration of the human body. The accuracy of the proposed algorithm is verified with FAUST and CAPE datasets.

Semi-automatic robotic puncture system based on deformable soft tissue point cloud registration.

Traditional surgical puncture robot systems based on computed tomography (CT) and infrared camera guidance have natural disadvantages for puncture of deformable soft tissues such as the liver. Liver movement and deformation caused by breathing are difficult to accurately assess and compensate by current technical solutions. We propose a semi-automatic robotic puncture system based on real-time ultrasound images to solve this problem. Real-time ultrasound images and their spatial position information can be obtained by robot in this system. By recognizing target tissue in these ultrasound images and using reconstruction algorithm, 3D real-time ultrasound tissue point cloud can be constructed. Point cloud of the target tissue in the CT image can be obtained by using developed software. Through the point cloud registration method based on feature points, two point clouds above are registered. The puncture target will be automatically positioned, then robot quickly carries the puncture guide mechanism to the puncture site and guides the puncture. It takes about just tens of seconds from the start of image acquisition to completion of needle insertion. Patient can be controlled by a ventilator to temporarily stop breathing, and patient's breathing state does not need to be the same as taking CT scan. The average operation time of 24 phantom experiments is 64.5 s, and the average error between the needle tip and the target point after puncture is 0.8mm. Two animal puncture surgeries were performed, and the results indicated that the puncture errors of these two experiments are 1.76mm and 1.81mm, respectively. Robot system can effectively carry out and implement liver tissue puncture surgery, and the success rate of phantom experiments and experiments is 100%. It also shows that the puncture robot system has high puncture accuracy, short operation time, and great clinical value.

Massive Point Cloud Processing for Efficient Construction Quality Inspection and Control.

The construction of large-scale civil infrastructures requires massive spatiotemporal data to support the management and control of scheduling, quality control, and safety monitoring. Existing artificial-intelligence-based data processing algorithms rely heavily on experienced engineers to adjust the parameters of data processing, which is inefficient and time-consuming when dealing with huge datasets. Limited studies have compared the performance of different algorithms on a unified dataset. This study proposes a framework and evaluation system for comparing different data processing policies for processing huge spatiotemporal data in construction quality control. The proposed method compares the combination of multiple types of algorithms involved in the processing of massive point cloud data. The performance of data processing strategies is evaluated through this framework, and the optimal point cloud processing strategies are explored based on registration accuracy and data fidelity. Results show that a reasonable choice of combinations of point cloud sampling, filtering, and registration algorithms can significantly improve the efficiency of point cloud data processing and satisfy engineering demands for data accuracy and completeness. The proposed method can be applied to the civil engineering problem of processing a large amount of point cloud data and selecting the optimal processing method.

Localization in MR-based Indoor Navigation System using point cloud registration

Abstract. This paper addresses the pervasive challenge of localization within indoor navigation systems. Many existing systems grapple with this issue, relying on marker-based methods, Bluetooth beacon localization, and image processing. However, these solutions often necessitate the placement of predetermined markers within the environment, presenting limitations such as marker deterioration over time, spatial constraints, and associated costs. In response, we propose a novel point cloud registration-based approach. This method involves the creation of a comprehensive global point cloud representing the entire indoor environment, effectively serving as a three-dimensional virtual map. Concurrently, a mixed-reality device is utilized to generate a local point cloud specific to the device’s current location within the building. By aligning and registering the local cloud with the global counterpart, the device’s precise location within the virtual map can be determined. This approach enables seamless navigation within indoor environments without the reliance on physical markers, offering a versatile and cost-effective solution to the challenge of localization in indoor navigation systems. Finally, an error analysis was also done to test localization accuracy for different illumination conditions, demonstrating the effectiveness of this method.

CRTF-MoeICP: A robust coarse-to-fine reflector-based LiDAR indoor positioning algorithm

The reflector-based Light Detection and Ranging (LiDAR) positioning method is susceptible to environmental interferences, resulting in instability. This instability not only reduces movement accuracy but also poses safety hazards. To solve the above problems in the application of LiDAR sensors in the field of indoor positioning, we propose a Coarse Registration algorithm based on the Triangular Feature (CRTF) and a fine registration algorithm based on Multi-level outlier elimination and Iterative Closest Point (MoeICP) for the reflector-based LiDAR positioning. The proposed coarse-to-fine positioning algorithm CRTF-MoeICP addresses the issue of reflector-based LiDAR positioning failure arising from the improper selection of the initial transformation matrix and outlier interference in indoor structured industrial environments. The experiment results show that the CRTF-MoeICP algorithm can ensure the stable registration of the LiDAR point cloud and the reflector map by completely removing all outliers, greatly improving the indoor positioning stability of LiDAR sensors. Besides, the proposed algorithm can be realized by LiDARs with different performance, and improve the static positioning repeatability to ±3 mm. The high precision and stable positioning results improve the motion accuracy, ensuring that the Automatic Guided Vehicle (AGV) can accurately and stably complete the handling task.

Geometry-Aware Enhanced Mutual-Supervised Point Elimination with Overlapping Mask Contrastive Learning for Partitial Point Cloud Registration

Point cloud registration is one of the fundamental tasks in computer vision, but faces challenges under low overlap conditions. Recent approaches use transformers and overlapping masks to improve perception, but mask learning only considers Euclidean distances between features, ignores mismatches caused by fuzzy geometric structures, and is often computationally inefficient. To address these issues, we introduce a novel matching framework. Firstly, we fuse adaptive graph convolution with PPF features to obtain rich feature perception. Subsequently, we construct a PGT framework that uses GeoTransformer and combines it with location information encoding to enhance the geometry perception between source and target clouds. In addition, we improve the visibility of overlapping regions through information exchange and the AIS module, aiming at subsequent keypoint extraction, preserving points with distinct geometrical structures while suppressing the influence of non-overlapping regions to improve computational efficiency. Finally, the mask is refined through contrast learning to preserve geometric and distance similarity, which helps to compute the transformation parameters more accurately. We have conducted comprehensive experiments on synthetic and real-world scene datasets, demonstrating superior registration performance compared to recent deep learning methods. Our approach shows remarkable improvements of 68.21% in RRMSE and 76.31% in tRMSE on synthetic data, while also excelling in real-world scenarios with enhancements of 76.46% in RRMSE and 45.16% in tRMSE.

A precise registration method for large-scale urban point clouds based on phased and spatial geometric features

Abstract The dense high-rise buildings and multipath effects in urban areas significantly reduce the positioning signal accuracy of laser scanning systems, leading to layering and offset issues in the collected point cloud data on the same road. In order to acquire comprehensive and consistent three-dimensional information on the objects, thereby providing field inspection data for large-scale road traffic network scenarios, in this paper, an improved point cloud registration method is proposed to divide the registration process into two stages: elevation registration and plane registration. Elevation registration takes the ground point cloud as the registration primitive, reduces the number of point clouds through curvature down-sampling, and constrains the feature point sequence with a fixed range to provide a good initial pose for fine registration. The plane registration first inherits the elevation registration parameters, combining the dynamic distance parameters of spherical region step based on the median, using robust multi-scale loss functions to address residual points, effective adjacent point pairs are selected to obtain the spatial transformation matrix, and realizes accurate registration. Experimental results with multiple sets of urban point cloud data show that the root mean square error of point cloud registration can be controlled within 0.06 m, achieving a relatively superior registration accuracy, it can provide detailed prior data for measurement information analysis.

Robust and correspondence-free point cloud registration: an extended approach with multiple hypotheses evaluation

AbstractPoint cloud registration is a fundamental problem in robotics, critical for tasks such as localization and mapping. Most approaches to this problem use feature-based techniques. However, these approaches have issues when dealing with unstructured environments where meaningful features are difficult to extract. Recently, an innovative global point cloud registration algorithm, PHASER, which does not rely on geometric features or point correspondences, has been introduced. It leverages Fourier transforms to identify the optimal rigid transform that maximizes cross-correlation between source and target point clouds. PHASER can also incorporate additional data channels, like LiDAR intensity, to enhance registration results. Because it does not rely on local features and because of its ability to exploit additional data, PHASER is particularly useful when dealing with very noisy point clouds or with many outliers. For this reason, we propose an extension to PHASER that considers multiple plausible rototranslation hypotheses. Our extended approach outperforms the original PHASER algorithm, especially in challenging scenarios where point clouds are widely separated. We validate its effectiveness on the DARPA SubT, and the Newer College datasets, showcasing its potential for improving registration accuracy in complex environments.

Dynamic indoor mapping for AVP: Crowdsourcing mapping without prior maps

AbstractHigh‐definition maps are essential for autonomous vehicle navigation, but indoor parking lots remain poorly mapped due to high costs. To address this, a crowdsourcing model gathers data from consumer‐grade sensors in mass‐produced vehicles to create semantic maps. Indoor parking lots lack GNSS signals, and most of them do not have high‐definition maps or navigation maps as references, making it difficult to ensure the accuracy of the final mapping results. Additionally, the semantic information of indoor parking lots is relatively limited, and the geometric features are overly similar, which significantly impacts the accuracy of point cloud registration. Therefore, this article proposes a crowdsourcing‐based approach, where vehicles generate local semantic maps at the client end and upload them to the cloud. Leveraging the scene characteristics of indoor parking lots, the cloud optimizes and fits a large amount of crowdsourced data to obtain a high‐precision base map without prior information. Enhanced ICP point cloud registration merges subsequent maps with the base. Additionally, parking space occupancy information is provided. This map can furnish the necessary information for Autonomous Valet Parking (AVP) tasks. Evaluation on the BEVIS dataset shows a root mean square error of 0.482446 m for vehicle localization on the cloud‐based map.

DGRO: Doppler Velocity and Gyroscope-Aided Radar Odometry

A stable and robust odometry system is essential for autonomous robot navigation. The 4D millimeter-wave radar, known for its resilience in harsh weather conditions, has attracted considerable attention. As the latest generation of FMCW radar, 4D millimeter-wave radar provides point clouds with both position and Doppler velocity information. However, the increased uncertainty and noise in 4D radar point clouds pose challenges that prevent the direct application of LiDAR-based SLAM algorithms. To address this, we propose a SLAM framework that fuses 4D radar data with gyroscope readings using graph optimization techniques. Initially, Doppler velocity is employed to estimate the radar’s ego velocity, with dynamic points being removed accordingly. Building on this, we introduce a pre-integration factor that combines ego-velocity and gyroscope data. Additionally, leveraging the stable RCS characteristics of radar, we design a corresponding point selection method based on normal direction and propose a scan-to-submap point cloud registration technique weighted by RCS intensity. Finally, we validate the reliability and localization accuracy of our framework using both our own dataset and the NTU dataset. Experimental results show that the proposed DGRO system outperforms traditional 4D radar odometry methods, especially in environments with slow speeds and fewer dynamic objects.

Accurate reconstruction of turbine blade point cloud and multiple point cloud registration based on structured light

In recent years, the industry has increasingly adopted vision-guided laser processing techniques to machine cooling holes in turbine blades, placing higher demands on the accuracy of visual measurement. Therefore, to meet the need for high-precision measurement of the three-dimensional shape of turbine blades in the context of laser processing of film cooling holes, this paper proposes a structured light-based method for precise reconstruction of turbine blade point clouds and multi-point cloud registration. Specifically, to improve the accuracy of point cloud reconstruction, an improved phase matching and reconstruction method based on epipolar constraints is proposed. To enhance the accuracy and stability of multi-point cloud registration, a weighted iterative closest point (ICP) method combined with a mark point registration strategy is proposed, optimizing the stability of the ICP algorithm and achieving high-quality multiple point cloud stitching. Experimental results validate the effectiveness and accuracy of the proposed method.

Full Transformer Framework for Robust Point Cloud Registration With Deep Information Interaction.

Point cloud registration is an essential technology in computer vision and robotics. Recently, transformer-based methods have achieved advanced performance in point cloud registration by utilizing the advantages of the transformer in order-invariance and modeling dependencies to aggregate information. However, they still suffer from indistinct feature extraction, sensitivity to noise, and outliers, owing to three major limitations: 1) the adoption of CNNs fails to model global relations due to their local receptive fields, resulting in extracted features susceptible to noise; 2) the shallow-wide architecture of transformers and the lack of positional information lead to indistinct feature extraction due to inefficient information interaction; and 3) the insufficient consideration of geometrical compatibility leads to the ambiguous identification of incorrect correspondences. To address the above-mentioned limitations, a novel full transformer network for point cloud registration is proposed, named the deep interaction transformer (DIT), which incorporates: 1) a point cloud structure extractor (PSE) to retrieve structural information and model global relations with the local feature integrator (LFI) and transformer encoders; 2) a deep-narrow point feature transformer (PFT) to facilitate deep information interaction across a pair of point clouds with positional information, such that transformers establish comprehensive associations and directly learn the relative position between points; and 3) a geometric matching-based correspondence confidence evaluation (GMCCE) method to measure spatial consistency and estimate correspondence confidence by the designed triangulated descriptor. Extensive experiments on the ModelNet40, ScanObjectNN, and 3DMatch datasets demonstrate that our method is capable of precisely aligning point clouds, consequently, achieving superior performance compared with state-of-the-art methods. The code is publicly available at https://github.com/CGuangyan-BIT/DIT.

Efficient and Robust Point Cloud Registration via Heuristics-Guided Parameter Search.

Estimating the rigid transformation with 6 degrees of freedom based on a putative 3D correspondence set is a crucial procedure in point cloud registration. Existing correspondence identification methods usually lead to large outlier ratios (>95% is common), underscoring the significance of robust registration methods. Many researchers turn to parameter search-based strategies (e.g., Branch-and-Bround) for robust registration. Although related methods show high robustness, their efficiency is limited to the high-dimensional search space. This paper proposes a heuristics-guided parameter search strategy to accelerate the search while maintaining high robustness. We first sample some correspondences (i.e., heuristics) and then just need to sequentially search the feasible regions that make each sample an inlier. Our strategy largely reduces the search space and can guarantee accuracy with only a few inlier samples, therefore enjoying an excellent trade-off between efficiency and robustness. Since directly parameterizing the 6-dimensional nonlinear feasible region for efficient search is intractable, we construct a three-stage decomposition pipeline to reparameterize the feasible region, resulting in three lower-dimensional sub-problems that are easily solvable via our strategy. Besides reducing the searching dimension, our decomposition enables the leverage of 1-dimensional interval stabbing at all three stages for searching acceleration. Moreover, we propose a valid sampling strategy to guarantee our sampling effectiveness, and a compatibility verification setup to further accelerate our search. Extensive experiments on both simulated and real-world datasets demonstrate that our approach exhibits comparable robustness with state-of-the-art methods while achieving a significant efficiency boost.

Robust affine point matching via quadratic assignment on Grassmannians

Robust Affine Matching with Grassmannians (RoAM) is a new algorithm to perform affine registration of point clouds. The algorithm is based on minimizing the Frobenius distance between two elements of the Grassmannian. For this purpose, an indefinite relaxation of the Quadratic Assignment Problem (QAP) is used, and several approaches to affine feature matching are studied and compared. Experiments demonstrate that RoAM is more robust to noise and point discrepancy than previous methods.

Non-contact measurement of conveyor belt speed based on fast point cloud registration of feature block

Abstract Non-contact, real-time measurement of conveyor belt speed is critical for energy-saving speed regulation and efficient development of coal mine conveyor systems. Existing speed measurement technologies for conveyor systems are often limited by the slippage and wear in contact measurement and complex environmental disturbance. This study introduces three-dimensional point cloud technology into coal flow information detection and innovatively presents a non-contact measurement method of conveyor belt speed based on fast point cloud registration of feature blocks. In the proposed method, a three-dimensional camera is used to capture point cloud data of the dynamically running conveyor belt, and the raw point cloud is preprocessed and segmented into blocks. Then, the C-MANV (curvature and mean angle of normal vectors) features of block point clouds are constructed based on the point cloud neighborhood curvature and the mean angle of neighborhood normal vectors. Finally, an improved block point cloud registration method based on C-MANV features is adopted to achieve the accurate measurement of the dynamic running speed of conveyor belt. Experimental results demonstrate that the proposed method achieves an average relative error of less than 1.8% in high-speed conveyor belt operation with an average processing time of less than 35 ms, which fulfils the accuracy and real-time requirements for conveyor belt speed detection in coal mines. This study provides an effective technical solution for the speed monitoring of coal mine conveyor systems.

MFINet: a multi-scale feature interaction network for point cloud registration

MFINet: a multi-scale feature interaction network for point cloud registration

A robust correspondence-based registration method for large-scale outdoor point cloud

ABSTRACT Large-scale outdoor point cloud registration is essential for the 3D reconstruction of outdoor scenes. Its central objective is to achieve accurate point cloud registration by determining accurate spatial transformation parameters. While feature-based methods eliminate the need for initial position estimation, they encounter challenges in handling high outlier rates. Therefore, a method capable of effectively managing outliers is crucial for enhancing the efficiency and accuracy of large-scale outdoor point cloud registration. This paper introduces the maximal clique with adaptive voting (MCAV) method, which leverages graph-based inlier compatibility to optimize potential matches. MCAV employs adaptive parameter voting (APV) to enhance computational efficiency, demonstrating significant speedup characteristics in datasets with a significant number of inliers. To further reduce outliers in potential matches, we integrate Black-Rangarajan Duality (BRD) and graduated non-convexity (GNC) into the truncated least squares (TLS) framework (BG-TLS). Accordingly, we propose the efficient BG-TLS (EBG-TLS) method for computing the registration model. Comparative analyses with traditional and deep learning-based methods across various real-world environments demonstrate that the proposed method outperforms existing algorithms in terms of rotation error, translation error, and efficiency, particularly in complex, high-noise settings. This method finds broad applications in geospatial mapping and surveying, autonomous navigation, and environmental monitoring.

MAFNet: a two-stage multiple attention fusion network for partial-to-partial point cloud registration

Abstract 3D point cloud registration is a critical technology in the fields of visual measurement and robot automation processing. In actual large-scale industrial production, the accuracy of point cloud registration directly affects the quality of automated welding processes. However, most existing methods are confronted with serious challenges such as the failure of partial-to-partial point cloud model registration when facing robot automatic processing guidance and error analysis work. Therefore, this paper proposes a novel two-stage network architecture for point cloud registration, which aims at robot pose adjustment and visual guidance in the field of automated welding by using 3D point cloud data. Specifically, we propose a neighborhood-based multi-head attention module in the coarse registration stage. The neighborhood information of each point can be aggregated through focusing on different weight coefficients of multi-head inputs. Then the spatial structure features which is used to establish the overlapping constraint of point clouds are obtained based on above neighborhood information. In the fine registration stage, we propose the similarity matching removal module based on multiple attention fusion features to explore deeper features from different aspects. By using deep fusion features to guide the similarity calculation, the interference of non-overlapping points is removed to achieve the finer registration. Eventually, we compare and analyze the proposed method with the SOTA ones through several error metrics and overlap estimation experiments based on the ModelNet40 dataset. The test results indicate that our method, relative to other mainstream techniques, achieves lower error rates and the most superior accuracy of 98.61% and recall of 98.37%. To demonstrate the generalization performance of proposed algorithm, extensive experiments on the Stanford 3D Scanning Repository, 7-Scenes and our own scanning dataset using partially overlapping point clouds individually under clean and noisy conditions show the validity and reliability of our proposed registration network.