Point Cloud Registration Algorithm Research Articles

Point cloud registration algorithm based on ICP algorithm and 3D-NDT algorithm

Point cloud registration algorithm based on ICP algorithm and 3D-NDT algorithm

Point cloud registration algorithm based on ICP algorithm and 3D-NDT algorithm

Point cloud registration algorithm based on ICP algorithm and 3D-NDT algorithm

Optimization of 3-D Pose Measurement Method Based on Binocular Vision

To improve the accuracy and speed of existing binocular vision methods in practical environments, this article presents an optimization method for accurate and stable 3-D pose measurement based on binocular vision. First, in the stereo-image-matching stage, we introduce a guide-point definition and propose an optimal path-searching method for dynamic programming (DP). Second, a distance-based adaptive-filtering method is added to remove noise points and outliers around the target so that the environmental noise interference in practical environments is reduced before registration. Third, an adaptive-sampling method based on distance is proposed to extract the key points of the target point-cloud, and finally a target surface model based on a hash table is established as an offline matching data structure of point-cloud registration to improve the matching speed and accuracy of the 3-D pose calculation. A series of experiments showed the translation error of the proposed method was less than 1 cm relative to the measurement of 100 cm, and the rotation error was less than 2° relative to the measurement of 180°. Compared with traditional point-cloud registration algorithms, our method exhibited higher accuracy and stability, indicating great potential for use in unstructured environments.

TIF-Reg: Point Cloud Registration with Transform-Invariant Features in SE(3).

Three-dimensional point cloud registration (PCReg) has a wide range of applications in computer vision, 3D reconstruction and medical fields. Although numerous advances have been achieved in the field of point cloud registration in recent years, large-scale rigid transformation is a problem that most algorithms still cannot effectively handle. To solve this problem, we propose a point cloud registration method based on learning and transform-invariant features (TIF-Reg). Our algorithm includes four modules, which are the transform-invariant feature extraction module, deep feature embedding module, corresponding point generation module and decoupled singular value decomposition (SVD) module. In the transform-invariant feature extraction module, we design TIF in SE(3) (which means the 3D rigid transformation space) which contains a triangular feature and local density feature for points. It fully exploits the transformation invariance of point clouds, making the algorithm highly robust to rigid transformation. The deep feature embedding module embeds TIF into a high-dimension space using a deep neural network, further improving the expression ability of features. The corresponding point cloud is generated using an attention mechanism in the corresponding point generation module, and the final transformation for registration is calculated in the decoupled SVD module. In an experiment, we first train and evaluate the TIF-Reg method on the ModelNet40 dataset. The results show that our method keeps the root mean squared error (RMSE) of rotation within 0.5 and the RMSE of translation error close to 0 m, even when the rotation is up to [−180, 180] or the translation is up to [−20 m, 20 m]. We also test the generalization of our method on the TUM3D dataset using the model trained on Modelnet40. The results show that our method’s errors are close to the experimental results on Modelnet40, which verifies the good generalization ability of our method. All experiments prove that the proposed method is superior to state-of-the-art PCReg algorithms in terms of accuracy and complexity.

Determination of the shear failure areas of rock joints using a laser scanning technique and artificial intelligence algorithms

Two artificial intelligence models-backpropagation neural network (BPNN) and support vector machines (SVM)-were created to investigate the effects of mesostructure characteristics on the shear mechanical behaviors of rock joints. This was achieved through learning training samples for the evaluation of the five basic geometrical parameters sensitivity (i.e., slope angle, horizontal orientation, elevation difference, curvature, and aperture distribution), and for determination of the shear failure regions when rock joints were subjected to low-normal and shear loads. First, the digital elevation models (DEMs) of rock joints were produced through point clouds collected using a laser scanning system. Five geometrical parameters were specified as the inputs for the artificial intelligence models, and an approach was developed to calculate the 3D aperture distribution using a point cloud registration algorithm. Shear failure regions were considered as the outputs, which were extracted from images taken after direct shear testing via the global threshold algorithm. Secondly, BPNN and SVM models were employed in order to establish relationships between the geometrical parameters and shear failure areas by machine learning on training samples. Thirdly, the information value (IV) algorithm was used to verify the two trained models. Results showed that the BPNN and SVM models made acceptable determinations of the shear failure areas, which corresponded to the real situation. The predictions from the BPNN and SVM models were more accurate than those from the IV algorithm. Furthermore, shear failure regions depended primarily on the aperture distribution of rock joints during the shearing process, followed by horizontal orientation, elevation difference, and then the slope angle. The effect of parameter curvature on shear failure was less than the other four parameters.

Pairwise Registration Algorithm for Large-Scale Planar Point Cloud Used in Flatness Measurement.

In this paper, an optimized three-dimensional (3D) pairwise point cloud registration algorithm is proposed, which is used for flatness measurement based on a laser profilometer. The objective is to achieve a fast and accurate six-degrees-of-freedom (6-DoF) pose estimation of a large-scale planar point cloud to ensure that the flatness measurement is precise. To that end, the proposed algorithm extracts the boundary of the point cloud to obtain more effective feature descriptors of the keypoints. Then, it eliminates the invalid keypoints by neighborhood evaluation to obtain the initial matching point pairs. Thereafter, clustering combined with the geometric consistency constraints of correspondences is conducted to realize coarse registration. Finally, the iterative closest point (ICP) algorithm is used to complete fine registration based on the boundary point cloud. The experimental results demonstrate that the proposed algorithm is superior to the current algorithms in terms of boundary extraction and registration performance.

V2I based environment perception for autonomous vehicles at intersections

In recent years, autonomous driving technology has made good progress, but the noncooperative intelligence of vehicle for autonomous driving still has many technical bottlenecks when facing urban road autonomous driving challenges. V2I (Vehicle-to-Infrastructure) communication is a potential solution to enable cooperative intelligence of vehicles and roads. In this paper, the RGB-PVRCNN, an environment perception framework, is proposed to improve the environmental awareness of autonomous vehicles at intersections by leveraging V2I communication technology. This framework integrates vision feature based on PVRCNN. The normal distributions transform(NDT) point cloud registration algorithm is deployed both on onboard and roadside to obtain the position of the autonomous vehicles and to build the local map objects detected by roadside multi-sensor system are sent back to autonomous vehicles to enhance the perception ability of autonomous vehicles for benefiting path planning and traffic efficiency at the intersection. The field-testing results show that our method can effectively extend the environmental perception ability and range of autonomous vehicles at the intersection and outperform the PointPillar algorithm and the VoxelRCNN algorithm in detection accuracy.

Initial alignment for point cloud registration by improved differential evolution algorithm

Aiming at the shortcomings of traditional point cloud registration algorithms, a point cloud initial alignment method based on the differential evolution algorithm is proposed. By designing an appropriate fitness function and selecting a proper evolution strategy, the probability and speed of finding the globally optimal solution are improved. Finally, the point to plane iterative closest point is used to complete the precise registration. Experimental results prove that the speed and robustness of the algorithm put forward here are improved.

Color Point Cloud Registration Algorithm Based on Hue

ICP is a well-known method for point cloud registration but it only uses geometric information to do this, which will result in bad results in some similar structures. Adding color information when registering will improve the performance. However, color information of point cloud, such as gray, varies differently under different lighting conditions. Using gray as the color information to register can cause large errors and even wrong results. To solve this problem, we propose a color point cloud registration algorithm based on hue, which has good robustness at different lighting conditions. We extract the hue component according to the color information of point clouds and make the hue distribution of the tangent plane continuous. The error function consists of color and geometric error of two point clouds under the current transformation. We optimize the error function using the Gauss–Newton method. If the value of the error function is less than the preset threshold or the maximum number of iterations is reached, the current transformation relationship is required. We use RGB-D Scenes V2 dataset to evaluate our algorithm and the results show that the average recall of our algorithms is 8.63% higher than that of some excellent algorithms, and its RMSE of 14.3% is lower than that of the other compared algorithms.

Towards a Fully Automated 3D Reconstruction System Based on LiDAR and GNSS in Challenging Scenarios

High-precision 3D maps play an important role in autonomous driving. The current mapping system performs well in most circumstances. However, it still encounters difficulties in the case of the Global Navigation Satellite System (GNSS) signal blockage, when surrounded by too many moving objects, or when mapping a featureless environment. In these challenging scenarios, either the global navigation approach or the local navigation approach will degenerate. With the aim of developing a degeneracy-aware robust mapping system, this paper analyzes the possible degeneration states for different navigation sources and proposes a new degeneration indicator for the point cloud registration algorithm. The proposed degeneracy indicator could then be seamlessly integrated into the factor graph-based mapping framework. Extensive experiments on real-world datasets demonstrate that the proposed 3D reconstruction system based on GNSS and Light Detection and Ranging (LiDAR) sensors can map challenging scenarios with high precision.

Point cloud registration algorithm based on curvature feature similarity

In this paper, an improved iterative closest point (ICP) algorithm based on the curvature feature similarity of the point cloud is proposed to improve the performance of classic ICP algorithm in an unstructured environment, such as alignment accuracy, robustness and stability. A K-D tree is introduced to segment the 3D point cloud for speeding up the search of nearest neighbor points using principal component analysis for coarse alignment based on the classical ICP algorithm. In the curvature calculation process, discrete index mapping and templates are taken for sampling to simplify the point cloud data and improve the registration efficiency. The nearest point of Euclidean distance is searched based on curvature feature similarity to improve alignment accuracy. The results of the point cloud alignment experiments show that the response time of the algorithm is less than 5s in the case where the curvature features of point cloud are obvious. And the alignment error can be reduced to 1% of the pre-alignment error. Therefore, the algorithm is an efficient, accurate and practical option for unstructured environment applications.

Large-size sprocket repairing based on robotic GMAW additive manufacturing

The variable and irregular surface damage of large-size sprockets has led to many challenges in the robotic GMAW-based automatic repair, which is a problem that has not been solved well. In this paper, a highly adaptive method based on robotic GMAW additive manufacturing was developed to repair the large-size sprocket with high quality. Key technologies including 3d rebuilding of damage surface and automatic robot deposition path planning have been researched. The model of damaged sprocket surface was obtained by point cloud registration algorithm including improved Iterative Closest Point (ICP) and Boolean subtraction operations. A highly adaptive slicing and path planning method for repairing sprockets with variable and irregular surface damage was proposed. Deposition parameters were obtained through experiments and a database with neural networks. The robot repairing process was simulated by 3D animation before executing the robot codes. The developed robot GMAW repairing system has integrated 3D scan modeling, slicing, path planning, parameters planning, and 3D animation simulation. The validation experiment results of repairing the damaged sprockets showed that the developed system has strong adaptability and high efficiency, and the repair accuracy met the actual requirements.

SVS-NLMS point cloud registration algorithm based on geometric algebra

SVS-NLMS point cloud registration algorithm based on geometric algebra

Point Cloud Registration Algorithm Based on Laplace Mixture Model

Registering point clouds quickly and accurately has always been a challenging task. A lot of research based on Gaussian mixture model is widely used in recent years. However, few people use other models for point cloud matching. Therefore, this paper proposes a point cloud registration algorithm based on the Laplace mixture model. In this paper, sampling variance is used to replace the variance of the likelihood estimation to successfully overcome the nonlinear problem. In addition, the Laplace model has strong robustness, which is very suitable for point cloud matching of 3D laser scanning. In the experiment, compared with several other algorithms, proposed method quickly and accurately registers point clouds.

一种改进的基于快速点特征直方图的ICP点云配准算法

一种改进的基于快速点特征直方图的ICP点云配准算法

Point Cloud Registration Algorithm Based on Cauchy Mixture Model

In this paper, the Cauchy mixture distribution is used for rigid and affine registration. The Cauchy function has a wider effective range than the general window function, so it has a greater information utilization ratio than most registration algorithms. This means that the method can quickly match point clouds without initial registration and corresponding points. On the other hand, the computational complexity of Cauchy mixture model is lower than that of widely used Gaussian mixture model. This means that the introduction of Cauchy mixture model will save computing resources. However, the log-likelihood function of the Cauchy mixture distribution is a nonlinear function that is difficult to solve. This paper proposes an optimization method for the Cauchy mixture model, and discusses the advantages of the Cauchy mixture model in rigid and affine registration. In the experiment, compared with several other algorithms, our method quickly and accurately registers point clouds.

Probability driven approach for point cloud registration of indoor scene

Point cloud registration is a crucial step of localization and mapping with mobile robots or in object modeling pipelines. In this paper, we present a novel probability driven algorithm for point cloud registration of the indoor scene based on RGB-D images. Firstly, we extract the key points in RGB-D images and map the key points to 3D space as preprocessing. Then, we build the distance matrix and the difference matrix for each point cloud, respectively in scalarization and vectorization, to encode the structural proximity. And establish the corresponding point set by computing the matching probabilities. At last, we solve the transform matrix that aligns the source point cloud to the target point cloud. The entire registration framework consists of two phases: coarse registration based on the distance matrix (in scalarization) and fine registration based on the difference matrix (in vectorization). The two-phase registration strategy is able to greatly reduce the influence of inherent noise. Experiments demonstrate that our method outperforms in registration accuracy than the state-of-the-art methods. Furthermore, our method is more efficient than existing methods in computing speed because we utilize the location relationship between key points instead of point features. The source code is provided at our project website https://github.com/BeCoolGuy/Probability-Driven-Approach-for-Point-Cloud-Registration-of-Indoor-Scene .

Automatic 3D point cloud registration algorithm based on triangle similarity ratio consistency

Three-dimensional (3D) point cloud registration is a fundamental key issue in 3D reconstruction, 3D object recognition and augmented reality. In this study, the authors propose a novel local feature descriptor called local angle statistics histogram (LASH) for efficient 3D point cloud registration. LASH forms a description of local shape geometries by encoding their properties on angles between the normal vector of the point and the vector formed by the point and other points in its local neighbourhood. In addition, they propose a 3D point cloud registration algorithm based on LASH. The registration algorithm firstly detects triangle matching points with consistent similarity ratios, and then aggregates each pair of triangular matching points into a set of matching points. They can use these matching sets to calculate multiple transformations between two point clouds. Finally, they use the error function to identify the best transformation and to achieve coarse alignment of the two point clouds. Experiments and comparisons with other global algorithms demonstrate that the proposed approach can be applied to register point clouds with considerable or limited overlaps and is robust to noise.

Pose estimation and adaptable grasp configuration with point cloud registration and geometry understanding for fruit grasp planning

Picking and placing fruits is a fundamental capability and common operation for agricultural robots in harvesting, sorting and packing tasks. Secure grasping fruits without potential slip and damage is challenging due to the shape varying, bruise sensitizing, as well as hardness changing during fruit maturation. In this study, a state-of-art point cloud processing based pre-grasp planning method was proposed to complete accurate 6D pose estimation and adaptable grasp configuration with point cloud registration and geometry understanding algorithm for grasp planning. Fruits 6D pose was estimated with SAC-IA ICP algorithm by matching Kinect-based online target point cloud obtained from 3D reconstruction with laser scanner based offline template. Simultaneously, fruits geometry understanding is completed with RANSAC algorithm, and the shape of typical fruits were classified into sphere, cylinder and cone. To pre-define grasp configuration for all kinds of shapes, fruits mesh models are also created with greedy projection algorithm for rapid collision detection and contact determination in GraspIt. Finally, the grasp information can be retrieved and transformed to match the position and orientation of the object by combining the results of 6D pose and geometry understanding. Compared with deep learning based pose estimation algorithm, large training data set for each novel, unknown object is not required in our method. The performance of the proposed system and algorithm has been evaluated on orange, cucumber, pineapple, kiwifruit and other agricultural products. The result indicates that our method is reliable with high efficiency and accuracy, and the 6D pose estimation along with adaptable grasp configuration results can provide a theoretical basis for grasp planning.

Automatic point cloud registration algorithm based on the feature histogram of local surface.

In this paper, we present an efficient algorithm for point cloud registration in presence of low overlap rate and high noise. The proposed registration method mainly includes four parts: the loop voxel filtering, the curvature-based key point selection, the robust geometric descriptor, and the determining and optimization of correspondences based on key point spatial relationship. The loop voxel filtering filters point clouds to a specified resolution. We propose a key point selection algorithm which has a better anti-noise and fast ability. The feature descriptor of key points is highly exclusive which is based on the geometric relationship between the neighborhood points and the center of gravity of the neighborhood. The correspondences in the pair of two point clouds are determined according to the combined features of key points. Finally, the singular value decomposition and ICP algorithm are applied to align two point clouds. The proposed registration method can accurately and quickly register point clouds of different resolutions in noisy situations. We validate our proposal by presenting a quantitative experimental comparison with state-of-the-art methods. Experimental results show that the proposed point cloud registration algorithm has faster calculation speed, higher registration accuracy, and better anti-noise performance.