Unorganized Point Clouds Research Articles

EGST: Enhanced Geometric Structure Transformer for Point Cloud Registration.

We explore the effect of geometric structure descriptors on extracting reliable correspondences and obtaining accurate registration for point cloud registration. The point cloud registration task involves the estimation of rigid transformation motion in unorganized point cloud, hence it is crucial to capture the contextual features of the geometric structure in point cloud. Recent coordinates-only methods ignore numerous geometric information in the point cloud which weaken ability to express the global context. We propose Enhanced Geometric Structure Transformer to learn enhanced contextual features of the geometric structure in point cloud and model the structure consistency between point clouds for extracting reliable correspondences, which encodes three explicit enhanced geometric structures and provides significant cues for point cloud registration. More importantly, we report empirical results that Enhanced Geometric Structure Transformer can learn meaningful geometric structure features using none of the following: (i) explicit positional embeddings, (ii) additional feature exchange module such as cross-attention, which can simplify network structure compared with plain Transformer. Extensive experiments on the synthetic dataset and real-world datasets illustrate that our method can achieve competitive results.

Prototype pipeline modelling using interval scanning point clouds

AbstractWith the aid of computer aided design (CAD) and building information modelling (BIM), as-built to as-designed comparison has seen many developments in improving the workflow of manufacturing and construction tasks. Recently, evolution has been centred around automation of scene interpretation from three-dimensional (3D) scan data. The scope of this paper is to assess assemblies as the installation process progresses and inferring if arising deviations are problematic (complex task). The adequacy of utilising unorganised point clouds to compliance check are trialled with a real life down-scaled prototype model in conjunction with a synthetic dataset. This work aims to highlight areas where large rework could be avoided, in part by the detection of potential clashes of components early in the pipeline installation process. With the help of an extracted model in the form of a point cloud generated from a scanned physical model and a 3D CAD model representing the nominal geometry, an operator can be made visually aware of potential deviations and component clashes during a pipeline assembly process.

Shape reconstruction of trapezoidal surfaces from unorganized point clouds

A smooth T-surface can be thought of as a generalization of a surface of revolution in such a way that the axis of rotation is not fixed at one point but rather traces a smooth path on the base plane. Furthermore, the action, by which the aforementioned surface is obtained does not need to be merely rotation but any “suitable” planar equiform transformation applied to the points of a certain smooth profile curve. In analogy to the smooth setting, if the axis footpoints sweep a polyline on the base plane and if the profile curve is discretely chosen then a T-hedra (discrete T-surface) with trapezoidal faces is obtained.The goal of this article is to reconstruct a T-hedron from an already given unorganized point cloud of a T-surface. In doing so, a kinematic approach is taken into account, where the algorithm at first tries to find the aforementioned axis direction associated with the point cloud. Then the algorithm finds a polygonal path through which the axis footpoint moves. Finally, by properly cutting the point cloud with the planes passing through the axis and its footpoints, it reconstructs the surface. The presented method is demonstrated using examples.From an applied point of view, the straightforwardness of the generation of these surfaces predestines them for building and design processes. In fact, one can find many built objects belonging to the sub-classes of T-surfaces such as surfaces of revolution and moulding surfaces. Furthermore, the planarity of the faces of the discrete version paves the way for steel/glass construction in industry. Finally, these surfaces are also suitable for transformable designs as they allow an isometric deformation within their class.

Bridging Formal Shape Models and Deep Learning: A Novel Fusion for Understanding 3D Objects

This article describes a novel fusion of a generative formal model for three-dimensional (3D) shapes with deep learning (DL) methods to understand the geometric structure of 3D objects and the relationships between their components, given a collection of unorganized point cloud measurements. Formal 3D shape models are implemented as shape grammar programs written in Procedural Shape Modeling Language (PSML). Users write PSML programs to describe complex objects, and DL networks estimate the configured free parameters of the program to generate 3D shapes. Users write PSML programs to enforce fundamental rules that define an object class and encode object attributes, including shapes, components, size, position, etc., into a parametric representation of objects. This fusion of the generative model with DL offers artificial intelligence (AI) models an opportunity to better understand the geometric organization of objects in terms of their components and their relationships to other objects. This approach allows human-in-the-loop control over DL estimates by specifying lists of candidate objects, the shape variations that each object can exhibit, and the level of detail or, equivalently, dimension of the latent representation of the shape. The results demonstrate the advantages of the proposed method over competing approaches.

Geometry-aware 3D pose transfer using transformer autoencoder

3D pose transfer over unorganized point clouds is a challenging generation task, which transfers a source’s pose to a target shape and keeps the target’s identity. Recent deep models have learned deformations and used the target’s identity as a style to modulate the combined features of two shapes or the aligned vertices of the source shape. However, all operations in these models are point-wise and independent and ignore the geometric information on the surface and structure of the input shapes. This disadvantage severely limits the generation and generalization capabilities. In this study, we propose a geometry-aware method based on a novel transformer autoencoder to solve this problem. An efficient self-attention mechanism, that is, cross-covariance attention, was utilized across our framework to perceive the correlations between points at different distances. Specifically, the transformer encoder extracts the target shape’s local geometry details for identity attributes and the source shape’s global geometry structure for pose information. Our transformer decoder efficiently learns deformations and recovers identity properties by fusing and decoding the extracted features in a geometry attentional manner, which does not require corresponding information or modulation steps. The experiments demonstrated that the geometry-aware method achieved state-of-the-art performance in a 3D pose transfer task. The implementation code and data are available at https://github.com/SEULSH/Geometry-Aware-3D-Pose-Transfer-Using-Transformer-Autoencoder.

Robust Surface Area Measurement of Unorganized Point Clouds Based on Multiscale Supervoxel Segmentation

Robust Surface Area Measurement of Unorganized Point Clouds Based on Multiscale Supervoxel Segmentation

Completing 3D Point Clouds of Thin Corn Leaves for Phenotyping Using 3D Gridding Convolutional Neural Networks

Estimating the complete 3D points of crop plants from incomplete points is vital for phenotyping and smart agriculture management. Compared with the completion of regular man-made objects such as airplanes, chairs, and desks, the completion of corn plant points is more difficult for thin, curled, and irregular corn leaves. This study focuses on MSGRNet+OA, which is based on GRNet, to complete a 3D point cloud of thin corn plants. The developed MSGRNet+OA was accompanied by gridding, multi-scale 3DCNN, gridding reverse, cubic feature sampling, and offset-attention. In this paper, we propose the introduction of a 3D grid as an intermediate representation to regularize the unorganized point cloud, use multi-scale predictive fusion to utilize global information at different scales, and model the geometric features by adding offset-attention to compute the point position offsets. These techniques enable the network to exhibit good adaptability and robustness in dealing with irregular and varying point cloud structures. The accuracy assessment results show that the accuracy of completion using MSGRNet+OA is superlative, with a CD (×10−4) of 1.258 and an F-Score@1% of 0.843. MSGRNet+OA is the most effective when compared with other networks (PCN, shape inversion, the original GRNet, SeedFormer, and PMP-Net++), and it improves the accuracy of the CD (×10−4)/F-Score@1% with −15.882/0.404, −15.96/0.450, −0.181/0.018, −1.852/0.274, and −1.471/0.203, respectively. These results reveal that the developed MSGRNet+OA can be used to complete a 3D point cloud of thin corn leaves for phenotyping.

Robust Reconstruction of Closed Parametric Curves by Topological Understanding with Persistent Homology

Curve reconstruction is a fundamental problem in reverse engineering, which has intrigued researchers for decades. In this paper, we propose a topological understanding based method for reconstructing parametric curves robustly from unorganized point clouds. Given a point cloud, we firstly understand the number of closed curves which need to be reconstructed using persistent homology. Then, by calculating the persistent 1-cycles of the point cloud, the initial shapes of the reconstructed parametric curves are generated. Finally, the closed parametric curves are reconstructed with the weighted least-squares progressive iterative approximation (W-LSPIA) method. Due to the topological understanding, the reconstructed parametric curves are faithful to the salient topological structure of the point cloud. Moreover, the developed reconstruction method is robust, and the reconstructed curve is much less affected by noise points and outliers, compared with the conventional parametric curve reconstruction algorithms. Experimental results demonstrated in this paper show the effectiveness of the developed curve reconstruction method.

Robust and Fast Normal Mollification via Consistent Neighborhood Reconstruction for Unorganized Point Clouds.

This paper introduces a robust normal estimation method for point cloud data that can handle both smooth and sharp features. Our method is based on the inclusion of neighborhood recognition into the normal mollification process in the neighborhood of the current point: First, the point cloud surfaces are assigned normals via a normal estimator of robust location (NERL), which guarantees the reliability of the smooth region normals, and then a robust feature point recognition method is proposed to identify points around sharp features accurately. Furthermore, Gaussian maps and clustering are adopted for feature points to seek a rough isotropic neighborhood for the first-stage normal mollification. In order to further deal with non-uniform sampling or various complex scenes efficiently, the second-stage normal mollification based on residual is proposed. The proposed method was experimentally validated on synthetic and real-world datasets and compared to state-of-the-art methods.

Morphological Operations on Unorganized Point Clouds Using Octree Graphs

Point clouds resulting from digital scanning are increasingly being used in the heritage field to create knowledge-based models, such as building information models (BIM) . Nevertheless, the use of digital image processing techniques in point cloud unorganized data is not well explored yet because of the lack of information about point adjacency and corresponding difficulty in structuring and manipulating data. In this study we propose an approach to deal with image processing-like mathematical morphological operations in unorganized point cloud using the octree graph. Our method consists of three main steps, namely: (i) computing the best distance value between points in the cloud; (ii) constructing an octree from point cloud; and (iii) producing a graph using rectangular layout. We exemplify the use of the proposed methodology performing morphological operations in grayscale and color point clouds of historical buildings. Results prove the effectiveness of the octree representation, which can efficiently generate an adjacency map, demonstrated by the resulting images, and has the potential to be applied at different areas of 3D data analysis.

A Novel Type of Boundary Extraction Method and Its Statistical Improvement for Unorganized Point Clouds Based on Concurrent Delaunay Triangular Meshes.

Currently, three-dimensional (3D) laser-scanned point clouds have been broadly applied in many important fields, such as non-contact measurements and reverse engineering. However, it is a huge challenge to efficiently and precisely extract the boundary features of unorganized point cloud data with strong randomness and distinct uncertainty. Therefore, a novel type of boundary extraction method will be developed based on concurrent Delaunay triangular meshes (CDTMs), which adds the vertex-angles of all CDTMs around a common data point together as an evaluation index to judge whether this targeted point will appear at boundary regions. Based on the statistical analyses on the CDTM numbers of every data point, another new type of CDTM-based boundary extraction method will be further improved by filtering out most of potential non-edge points in advance. Then these two CDTM-based methods and popular α-shape method will be employed in conducting boundary extractions on several point cloud datasets for comparatively analyzing and discussing their extraction accuracies and time consumptions in detail. Finally, all obtained results can strongly demonstrate that both these two CDTM-based methods present superior accuracies and strong robustness in extracting the boundary features of various unorganized point clouds, but the statistically improved version can greatly reduce time consumption.

Hybrid simplification algorithm for unorganized point cloud based on two-level fuzzy decision making

Traditional point cloud simplification algorithms have specific application scenarios. If these well-known methods can be accurately applied to different feature regions of a point cloud, high-quality point cloud feature preservation can be realized. In this paper, a hybrid point cloud simplification method, based on two-level fuzzy decision making, is proposed. Specifically, the number of peak points and bin intervals of each dimension density histogram of the point cloud are counted, and the cluster number and initial centers of fuzzy c-means (FCM) clustering method are determined, hence the autonomous FCM clustering method of point clouds is realized. Based on the expert linguistic inference rules of type-1 fuzzy system, its output is used to accurately identify clusters as flat, transitional and drastic type. According to the attributes of the clusters, predefined thresholds or algorithms are applied in each one, achieving high quality point cloud simplification. The experimental results show that the algorithm can effectively determine the number of clusters, improve the speed of FCM clustering algorithm and achieve feature preserving point cloud simplification, after removing 50–90 % of the points.

A method for identifying and repairing holes on the surface of unorganized point cloud

During scanning, reflection and occlusion of objects, produce holes in the acquired point clouds, causing problems to functions such as map building and target detection. In this paper, a new method for identifying and repairing such holes is proposed. First, the density and the 2D projection points of the point cloud are calculated. Based on the Euclidean distance, the hole boundary points of the original point cloud are obtained. Following, a novel preprocessing segmentation algorithm is used to obtain the boundary points of holes. Next, even if the boundary points are not fully identified, the total 2D repairing points are obtained by the proposed algorithm. Finally, a new hole repairing model, based on fuzzy inference system, is proposed, integrating geometry and curvature information of boundary points, while it can fast and accurately estimate the height coordinates of the repairing points. The experimental results show that the proposed algorithm is efficient.

Framework for Geometric Information Extraction and Digital Modeling from LiDAR Data of Road Scenarios

Road geometric information and a digital model based on light detection and ranging (LiDAR) can perform accurate geometric inventories and three-dimensional (3D) descriptions for as-built roads and infrastructures. However, unorganized point clouds and complex road scenarios would reduce the accuracy of geometric information extraction and digital modeling. There is a standardization need for information extraction and 3D model construction that integrates point cloud processing and digital modeling. This paper develops a framework from semantic segmentation to geometric information extraction and digital modeling based on LiDAR data. A semantic segmentation network is improved for the purpose of dividing the road surface and infrastructure. The road boundary and centerline are extracted by the alpha-shape and Voronoi diagram methods based on the semantic segmentation results. The road geometric information is obtained by a coordinate transformation matrix and the least square method. Subsequently, adaptive road components are constructed using Revit software. Thereafter, the road route, road entity model, and various infrastructure components are generated by the extracted geometric information through Dynamo and Revit software. Finally, a detailed digital model of the road scenario is developed. The Toronto-3D and Semantic3D datasets are utilized for analysis through training and testing. The overall accuracy (OA) of the proposed net for the two datasets is 95.3 and 95.0%, whereas the IoU of segmented road surfaces is 95.7 and 97.9%. This indicates that the proposed net could accomplish superior performance for semantic segmentation of point clouds. The mean absolute errors between the extracted and manually measured geometric information are marginal. This demonstrates the effectiveness and accuracy of the proposed extraction methods. Thus, the proposed framework could provide a reference for accurate extraction and modeling from LiDAR data.

Rigorous Density Correction Model for Single-Scan TLS Point Clouds

Terrestrial laser scanning (TLS) can acquire high-precision and high-resolution 3D point clouds of the scanned targets, and the common technical challenge is how to accurately and efficiently process the massive unorganized point clouds. Density is one of the most fundamental and significant quantities that can be derived for each point by counting the number of neighbors in a given 3D space. It contains key information of the target geometric features and thus plays an important role in TLS data processing (e.g., classification and feature extraction). However, density usually varies remarkably for different scenes with varied scan geometry (e.g., distance and incidence angle), scan resolution (e.g., horizontal and vertical angular resolutions), and object geometry (e.g., slope, size, and spatial attitude), thereby greatly hindering its full capability in reflecting the geometric discrepancies of different targets. The density must be corrected before reliably used as a proxy for target geometric features. In this study, a generalized rigorous model is proposed to correct the density of single-scan TLS point clouds based on the mathematically deduced relation between density and related influencing factors. The superiority of the proposed model over existing models is that it is rigorously developed based on the instrumental scanning principles and target spatial geometry and can be used for different instruments, scanning scenes, targets, and scanning parameters. The proposed model is verified by a series of indoor quantitative control experiments and three natural scenes with totally different instruments and scanning conditions. The results show that the proposed model can accurately simulate the density variation at different scanning geometries with an average correlation coefficient of 0.9957 between the calculated and actual densities. The effects of all influencing factors on density are significantly removed in complex natural scenes by the proposed model where the coefficient of variation of the density from a homogeneous surface after correction is reduced by 72.31%, on the average. The proposed model exhibits good performance in terms of the feasibility, effectiveness, and generality and can derive a corrected density value that is a proxy of target geometric size. Additionally, the proposed model can be conducted on each scan individually before the co-registration in the actual TLS campaign with a number of single scans, which has tremendous application value in facilitating TLS data interpretation.

Evaluation Method for Shape Similarities between Archaeological Unorganized Point Cloud

Researchers assume that the bodies of several Tara statues may have been molded from the same template. Therefore, researchers are eager to study the shape similarities between the Tara statues. Motivated by the archaeologist's requirements, this study presents an efficient approach to evaluate shape similarities between point clouds of two different Tara statues right-arm models. The evaluation approach consists of two main stages. The first is the reconstruction of individual body parts without decoration using a B-spline surface approximation technique. In previous reconstruction process, the surrounding surface selection parameter was manually specified. If the manually applied parameter is inadequate, it reduces the accuracy of the surface approximation and hole-filling, resulting in reduced accuracy of the evaluation result. Therefore, in this study, the surrounding surface selection parameter is automatically determined based on the Golden-section-search algorithm. The second stage evaluates the shape similarities between two different point clouds. The evaluation method considers a simple point-to-surface distance metric method with overlapping-surface detection and when compared with generic point-to-point distance metric method, the evaluation results proved that the proposed methodology was reliable and effective for this application. The approximation accuracy affects the evaluation result when the reconstruction method fills the gap by adding new points through the approximated surface. Therefore, in this study, the shape similarities were investigated between the two arm models with holes, immediately after separating the decoration parts.

Surface reconstruction algorithm using a modified Allen–Cahn equation

In this paper, we propose a novel efficient surface reconstruction method from unorganized point cloud data in three-dimensional Euclidean space. The proposed method is based on the Allen–Cahn partial differential equation, with an edge indicating function to restrict the evolution. We applied the explicit Euler’s method to solve the discrete equation, and use the operator splitting technique to split the governing equation. Furthermore, we also modify the double well form to a periodic potential. Then we find that the proposed model can reconstruct the surface well even in the case of insufficient data. After selecting the appropriate parameters, we carried out various numerical experiments to demonstrate the robustness and accuracy of the proposed method. We adopt the proposed method to reconstruct the surfaces on simple, irregular and complex models, respectively, and can obtain smooth three-dimensional surfaces and visual effects. In addition, we also perform comparison tests to show the superiority of the proposed model. Statistic metrics such as the [Formula: see text], [Formula: see text], [Formula: see text], CPU time, and vertex numbers are evaluated. Results show that our model performs better than the other methods in statistical metrics even use far less point cloud data, and with the faster CPU computing speed.

AN EXAMPLE OF 3D RECONSTRUCTION ENVIRONMENT FROM RGB-D CAMERA

3D environment reconstruction is a very important research direction in robotics and computer vision. This helps the robot to locate and find directions in a real environment or to help build support systems for the blind and visually impaired people. In this paper, we introduce a simple and real-time approach for 3D environment reconstruction from data obtained from cheap cameras. The implementation is detailed step by step and illustrated with source code. Simultaneously, cameras that support reconstructing 3D environments in this approach are also presented and introduced. The unorganized point cloud datais also presented and visualized in the available figures .

A Dual Attention Neural Network for Airborne LiDAR Point Cloud Semantic Segmentation

With the development of airborne light detection and ranging (LiDAR) technology, it has become a common and efficient way to collect large-scale 3D spatial information. However, efficient and automatic semantic segmentation of LiDAR data, in the form of 3D point clouds, remains a persistent challenge. To address this, a dual attention neural network (DA-Net) is proposed, consisting of two different blocks, namely augmented edge representation (AER) and elevation attentive pooling (EAP). First, the AER can adaptively represent local orientation and position, thereby effectively enhancing geometric information. Second, the captured local features of centroid points are utilized to further encode discriminative features using the EAP with the learned attention scores. Finally, a location homogeneity (LH) module is devised to explore the long-range relationship in an encoder-decoder network. Benefiting from the dual attention module, geometric information hidden in unorganized point clouds can be effectively propagated. Besides, the LH forces the network to pay attention to the semantic consistency of elevated objects, which facilitates both point- and object-level point cloud semantic segmentation for scene understanding. A benchmark dataset is used to assess the proposed method, which achieves an overall accuracy of 85.98% and an average F1 score of 72.31%. In addition, comparisons with other latest deep learning methods on the 2019 Data Fusion Contest dataset further demonstrate the robustness and generalization ability of the proposed method.

Adaptive neighborhood size and effective geometric features selection for 3D scattered point cloud classification

Classification of 3D scatter and unorganized point cloud (PC) is an ongoing hard problem due to high redundancy, unbalanced sampling density, and large data structure of PC. Geometric and spectral features derived from the PC are generally used for classification. In this paper, an Omnivariance based adaptive neighborhood size selection method and a new feature set composed of 14 features are proposed for extraction of geometric features for each individual point within the local neighborhood. Performance of 8 modern classifiers with different strategies (i.e., boosting, ensemble, and deep learning etc.) were evaluated on the Oakland, Vaihingen, and ISPRS datasets. These 3 datasets are identified by 5, 9, and 2 distinct object classes, respectively. The results were compared with different neighborhood size selection methods (i.e., eigenentropy based, fixed number of the k-nearest neighbors) and feature set (i.e., 21 features). Only 3D local features were employed to classify datasets with varying characteristics and properties. The proposed optimum neighbor selection method and feature set provided the best statistical results with Auto-Encoder classifier (the overall accuracies are over 85%, 60% and, 90% the Oakland, Vaihingen, and ISPRS datasets, respectively). Especially for the ISPRS dataset, the Auto-Encoder obtained over 94%, 90%, and 93% precision, recall, and f-score, respectively.