Traditional models for semantic segmentation in point clouds primarily focus on smaller scales.However,in real-world applications,point clouds often exhibit larger scales,leading to heavy computational and memory requ...Traditional models for semantic segmentation in point clouds primarily focus on smaller scales.However,in real-world applications,point clouds often exhibit larger scales,leading to heavy computational and memory requirements.The key to handling large-scale point clouds lies in leveraging random sampling,which offers higher computational efficiency and lower memory consumption compared to other sampling methods.Nevertheless,the use of random sampling can potentially result in the loss of crucial points during the encoding stage.To address these issues,this paper proposes cross-fusion self-attention network(CFSA-Net),a lightweight and efficient network architecture specifically designed for directly processing large-scale point clouds.At the core of this network is the incorporation of random sampling alongside a local feature extraction module based on cross-fusion self-attention(CFSA).This module effectively integrates long-range contextual dependencies between points by employing hierarchical position encoding(HPC).Furthermore,it enhances the interaction between each point’s coordinates and feature information through cross-fusion self-attention pooling,enabling the acquisition of more comprehensive geometric information.Finally,a residual optimization(RO)structure is introduced to extend the receptive field of individual points by stacking hierarchical position encoding and cross-fusion self-attention pooling,thereby reducing the impact of information loss caused by random sampling.Experimental results on the Stanford Large-Scale 3D Indoor Spaces(S3DIS),Semantic3D,and SemanticKITTI datasets demonstrate the superiority of this algorithm over advanced approaches such as RandLA-Net and KPConv.These findings underscore the excellent performance of CFSA-Net in large-scale 3D semantic segmentation.展开更多
3D laser scanning technology is widely used in underground openings for high-precision,rapid,and nondestructive structural evaluations.Segmenting large 3D point cloud datasets,particularly in coal mine roadways with m...3D laser scanning technology is widely used in underground openings for high-precision,rapid,and nondestructive structural evaluations.Segmenting large 3D point cloud datasets,particularly in coal mine roadways with multi-scale targets,remains challenging.This paper proposes an enhanced segmentation method integrating improved PointNet++with a coverage-voted strategy.The coverage-voted strategy reduces data while preserving multi-scale target topology.The segmentation is achieved using an enhanced PointNet++algorithm with a normalization preprocessing head,resulting in a 94%accuracy for common supporting components.Ablation experiments show that the preprocessing head and coverage strategies increase segmentation accuracy by 20%and 2%,respectively,and improve Intersection over Union(IoU)for bearing plate segmentation by 58%and 20%.The accuracy of the current pretraining segmentation model may be affected by variations in surface support components,but it can be readily enhanced through re-optimization with additional labeled point cloud data.This proposed method,combined with a previously developed machine learning model that links rock bolt load and the deformation field of its bearing plate,provides a robust technique for simultaneously measuring the load of multiple rock bolts in a single laser scan.展开更多
Evaluating rock mass quality using three-dimensional(3D)point clouds is crucial for discontinuity extraction and is widely applied in various industrial sectors.However,the utilization of this method in geological sur...Evaluating rock mass quality using three-dimensional(3D)point clouds is crucial for discontinuity extraction and is widely applied in various industrial sectors.However,the utilization of this method in geological surveys remains limited.Notable limitations of current research include the scarcity of validation using simple geometric shapes for discontinuity extraction methods,and the lack of studies that target both planar and linear discontinuity.To address these gaps,this study proposes a workflow for identifying discontinuity planes and traces in rock outcrops from photogrammetric 3D modeling,employing the Compass and Facets plugins in the open-source CloudCompare software.Prior to field application,the efficacy of the extraction methods was first evaluated using experimental datasets of a cube and an isosceles triangular prism generated under laboratory-controlled conditions.This validation demonstrated exceptional accuracy,with the dip and dip direction(DDD)of extracted structures consistently within±2°of the actual values.Following this rigorous laboratory validation,this methodology was applied to a more complex natural rock outcrop(Miocene–Pliocene deposits in Japan),demonstrating its applicability in realistic geological settings for identifying structures.The results showed that the dip and dip direction trends of the extracted bedding planes and faults were consistent with field measurements,achieving a time reduction of approximately 40%compared to traditional methods.In conclusion,through strictly controlled initial verification and subsequent successful application to a complex natural setting,this study confirmed that the proposed workflow can effectively and efficiently extract discontinuous geological structures from point clouds.展开更多
Recently,large-scale deep learning models have been increasingly adopted for point cloud classification.However,thesemethods typically require collecting extensive datasets frommultiple clients,which may lead to priva...Recently,large-scale deep learning models have been increasingly adopted for point cloud classification.However,thesemethods typically require collecting extensive datasets frommultiple clients,which may lead to privacy leaks.Federated learning provides an effective solution to data leakage by eliminating the need for data transmission,relying instead on the exchange of model parameters.However,the uneven distribution of client data can still affect the model’s ability to generalize effectively.To address these challenges,we propose a new framework for point cloud classification called Federated Dynamic Aggregation Selection Strategy-based Multi-Receptive Field Fusion Classification Framework(FDASS-MRFCF).Specifically,we tackle these challenges with two key innovations:(1)During the client local training phase,we propose a Multi-Receptive Field Fusion Classification Model(MRFCM),which captures local and global structures in point cloud data through dynamic convolution and multi-scale feature fusion,enhancing the robustness of point cloud classification.(2)In the server aggregation phase,we introduce a Federated Dynamic Aggregation Selection Strategy(FDASS),which employs a hybrid strategy to average client model parameters,skip aggregation,or reallocate local models to different clients,thereby balancing global consistency and local diversity.We evaluate our framework using the ModelNet40 and ShapeNetPart benchmarks,demonstrating its effectiveness.The proposed method is expected to significantly advance the field of point cloud classification in a secure environment.展开更多
The virtual preassembly of super-high steel bridge towers faces a challenge in the efficient and precise extraction of complex cross-sectional features.Factors such as fabrication errors,gravity-induced deformations,a...The virtual preassembly of super-high steel bridge towers faces a challenge in the efficient and precise extraction of complex cross-sectional features.Factors such as fabrication errors,gravity-induced deformations,and temperature fluctuations can compromise the accuracy of contour extraction.To address these limitations,an improved Alpha-shape-based point cloud contour extraction method is proposed.The proposed approach uses a hierarchical strategy to process three-dimensional laser scanning point clouds.The processed data are then subjected to curvatureadaptive voxel filtering to reduce acquisition noise.In addition,an enhanced iterative closest point(ICP)variant with correspondence validation accurately aligns the discrete point cloud segments.The proposed curvature-responsive Alpha-shape framework enables multiscale contour delineation through topology-adaptive threshold modulation,which resolves boundary ambiguities in geometrically complex cross-sections.The method was experimentally validated using field-acquired measurement datasets from the Zhangjinggao Yangtze River Bridge tower segments,confirming its capability to reconstruct noncanonical cross-sectional geometries.Three contour extraction methods,including Poisson reconstruction,the conventional Alpha-shape algorithm,and random sample consensus with ICP(RANSAC-ICP),were compared to evaluate the performance of the proposed Alpha-shape algorithm.The results demonstrate that the proposed method achieves superior contour extraction accuracy and data reduction efficiency,highlighting its effectiveness in contour extraction tasks.展开更多
Automatic identificationof discontinuities is a key focus in rock slope research.Conventional methods typically target small areas,which limits efficiencyand applicability for complex discontinuities in large-scale ro...Automatic identificationof discontinuities is a key focus in rock slope research.Conventional methods typically target small areas,which limits efficiencyand applicability for complex discontinuities in large-scale rock slopes.This study uses multi-angle unmanned aerial vehicle(UAV)nap-of-the-object photogrammetry to construct a high-definitionthree-dimensional(3D)point cloud model of the slope.The edge-firstconnection algorithm identifiesall edge points of discontinuities in the point cloud and completes recognition through simple connection analysis.This method avoids the complex calculations required for sequentially identifying discontinuity edges in conventional methods and achieves significantacceleration through algorithm optimization and parallel computation support.Based on this algorithm,the RockDiscontinuity Identification(RD ID)software is developed and applied to identify numerous highly disordered discontinuities on the Xulong slope in the Jinsha River suture zone.Processing tens of millions of point clouds within approximately 2 h demonstrates exceptional computational efficiency.The automatic algorithm accurately identifiesnearly 80%of planar discontinuities,with orientations and trace lengths closely matching manual results,highlighting its potential for large-scale rock outcrop applications.Comparisons with region growing algorithms further emphasize its effectiveness and accuracy.However,the algorithm struggles to identify linear discontinuities,which are a major source of error.Additionally,high roughness and smooth edges of discontinuities affect recognition accuracy,indicating areas for further improvement.展开更多
To support the process of grasping objects on a tabletop for the blind or robotic arm,it is necessary to address fundamental computer vision tasks,such as detecting,recognizing,and locating objects in space,and determ...To support the process of grasping objects on a tabletop for the blind or robotic arm,it is necessary to address fundamental computer vision tasks,such as detecting,recognizing,and locating objects in space,and determining the position of the grasping information.These results can then be used to guide the visually impaired or to execute grasping tasks with a robotic arm.In this paper,we collected,annotated,and published the benchmark TQUGraspingObject dataset for testing,validation,and evaluation of deep learning(DL)models for detecting,recognizing,and localizing grasping objects in 2D and 3D space,especially 3D point cloud data.Our dataset is collected in a shared room,with common everyday objects placed on the tabletop in jumbled positions by Intel RealSense D435(IR-D435).This dataset includes more than 63k RGB-D pairs and related data such as normalized 3D object point cloud,3D object point cloud segmented,coordinate system normalizationmatrix,3D object point cloud normalized,and hand pose for grasping each object.At the same time,we also conducted experiments on fourDL networks with the best performance:SSD-MobileNetV3,ResNet50-Transformer,ResNet101-Transformer,and YOLOv12.The results present that YOLOv12 has the most suitable results in detecting and recognizing objects in images.All data,annotations,toolkit,source code,point cloud data,and results are publicly available on our project website:https://github.com/HuaTThanhIT2327Tqu/datasetv2.展开更多
Semantic segmentation in the context of 3D point clouds for the railway environment holds a significant economic value,but its development is severely hindered by the lack of suitable and specific datasets.Additionall...Semantic segmentation in the context of 3D point clouds for the railway environment holds a significant economic value,but its development is severely hindered by the lack of suitable and specific datasets.Additionally,the models trained on existing urban road point cloud datasets demonstrate poor generalisation on railway data due to a large domain gap caused by non-overlapping special/rare categories,for example,rail track,track bed etc.To harness the potential of supervised learning methods in the domain of 3D railway semantic segmentation,we introduce RailPC,a new point cloud benchmark.RailPC provides a large-scale dataset with rich annotations for semantic segmentation in the railway environment.Notably,RailPC contains twice the number of annotated points compared to the largest available mobile laser scanning(MLS)point cloud dataset and is the first railway-specific 3D dataset for semantic segmentation.It covers a total of nearly 25 km railway in two different scenes(urban and mountain),with 3 billion points that are finely labelled as 16 most typical classes with respect to railway,and the data acquisition process is completed in China by MLS systems.Through extensive experimentation,we evaluate the performance of advanced scene understanding methods on the annotated dataset and present a synthetic analysis of semantic segmentation results.Based on our findings,we establish some critical challenges towards railway-scale point cloud semantic segmentation.The dataset is available at https://github.com/NNU-GISA/GISA-RailPC,and we will continuously update it based on community feedback.展开更多
Automatic registration of unordered point clouds is the prerequisite forurban reconstruction. However, most of the existing technologies stillsuffer from some limitations. On one hand, most of them are sensitive tonoi...Automatic registration of unordered point clouds is the prerequisite forurban reconstruction. However, most of the existing technologies stillsuffer from some limitations. On one hand, most of them are sensitive tonoise and repetitive structures, which makes them infeasible for theregistration of large-scale point clouds. On the other hand, most of themdealing with point clouds with limited overlaps and unpredictablelocation. All these problems make it difficult for registration technology toobtain qualified results in outdoor point cloud. To overcome theselimitations, this paper presents a grid graph-based point cloud registration(GGR) algorithm to align pairwise scans. First, point cloud is divided into aset of 3D grids. We propose a voting strategy to measure the similaritybetween two grids based on feature descriptor, transforming thesuperficial correspondence into 3D grid expression. Next, a graphmatching is proposed to capture the spatial consistency from putativecorrespondences, and graph matching hierarchically refines thecorresponding grids until obtaining point-to-point correspondences.Comprehensive experiments demonstrated that the proposed algorithmobtains good performance in terms of successful registration rate, rotationerror, translation error, and outperformed the state-of-the-art approaches.展开更多
针对自动驾驶场景下,近处干扰点云误检率高、远处稀疏点云漏检率高的问题,提出了一种基于改进PointPillars的自动驾驶障碍物点云检测算法.首先,通过聚合模块和共享多层感知机(shared multi-layer perceptron,MLP)对柱体内点云进行特征编...针对自动驾驶场景下,近处干扰点云误检率高、远处稀疏点云漏检率高的问题,提出了一种基于改进PointPillars的自动驾驶障碍物点云检测算法.首先,通过聚合模块和共享多层感知机(shared multi-layer perceptron,MLP)对柱体内点云进行特征编码,采用最大池化与平均池化叠加的方法将点云的显著特征与细节特征映射为柱体特征;其次,针对算法对伪图特征关注与利用不充分的问题,引入坐标注意力(coordinate attention,CA)机制和残差连接的伪图特征提取模块(attention and residual second block,ARSB),将深层与浅层特征图进行融合,优化算法梯度,增强算法对有效目标的关注度.试验结果表明:改进算法对全局点云检测精度较高,平均精度优于PointPillars、稀疏到稠密3D目标检测器(STD)等点云目标检测算法,在汽车类别上的检测精度优势明显,检测速度较快,符合实时性要求.展开更多
Large-scale point cloud datasets form the basis for training various deep learning networks and achieving high-quality network processing tasks.Due to the diversity and robustness constraints of the data,data augmenta...Large-scale point cloud datasets form the basis for training various deep learning networks and achieving high-quality network processing tasks.Due to the diversity and robustness constraints of the data,data augmentation(DA)methods are utilised to expand dataset diversity and scale.However,due to the complex and distinct characteristics of LiDAR point cloud data from different platforms(such as missile-borne and vehicular LiDAR data),directly applying traditional 2D visual domain DA methods to 3D data can lead to networks trained using this approach not robustly achieving the corresponding tasks.To address this issue,the present study explores DA for missile-borne LiDAR point cloud using a Monte Carlo(MC)simulation method that closely resembles practical application.Firstly,the model of multi-sensor imaging system is established,taking into account the joint errors arising from the platform itself and the relative motion during the imaging process.A distortion simulation method based on MC simulation for augmenting missile-borne LiDAR point cloud data is proposed,underpinned by an analysis of combined errors between different modal sensors,achieving high-quality augmentation of point cloud data.The effectiveness of the proposed method in addressing imaging system errors and distortion simulation is validated using the imaging scene dataset constructed in this paper.Comparative experiments between the proposed point cloud DA algorithm and the current state-of-the-art algorithms in point cloud detection and single object tracking tasks demonstrate that the proposed method can improve the network performance obtained from unaugmented datasets by over 17.3%and 17.9%,surpassing SOTA performance of current point cloud DA algorithms.展开更多
Identity-based public cloud storage auditing schemes can check the integrity of cloud data, and reduce the complicated certificate management. In such a scheme, one Private Key Generator(PKG) is employed to authentica...Identity-based public cloud storage auditing schemes can check the integrity of cloud data, and reduce the complicated certificate management. In such a scheme, one Private Key Generator(PKG) is employed to authenticate the identity and generate private keys for all users, and one Third Party Auditor(TPA) is employed to by users to check the integrity of cloud data. This approach is undesirable for large-scale users since the PKG and the TPA might not be able to afford the heavy workload. To solve the problem, we give a hierarchical Private Key Generator structure for large-scale user groups, in which a root PKG delegates lower-level PKGs to generate private keys and authenticate identities. Based on the proposed structure, we propose an authorized identity-based public cloud storage auditing scheme, in which the lowest-level PKGs play the role of TPA, and only the authorized lowest-level PKGs can represent users in their domains to check cloud data's integrity. Furthermore, we give the formal security analysis and experimental results, which show that our proposed scheme is secure and efficient.展开更多
This study presents a framework for the semi-automatic detection of rock discontinuities using a threedimensional(3D)point cloud(PC).The process begins by selecting an appropriate neighborhood size,a critical step for...This study presents a framework for the semi-automatic detection of rock discontinuities using a threedimensional(3D)point cloud(PC).The process begins by selecting an appropriate neighborhood size,a critical step for feature extraction from the PC.The effects of different neighborhood sizes(k=5,10,20,50,and 100)have been evaluated to assess their impact on classification performance.After that,17 geometric and spatial features were extracted from the PC.Next,ensemble methods,AdaBoost.M2,random forest,and decision tree,have been compared with Artificial Neural Networks to classify the main discontinuity sets.The McNemar test indicates that the classifiers are statistically significant.The random forest classifier consistently achieves the highest performance with an accuracy exceeding 95%when using a neighborhood size of k=100,while recall,F-score,and Cohen's Kappa also demonstrate high success.SHapley Additive exPlanations(SHAP),an Explainable AI technique,has been used to evaluate feature importance and improve the explainability of black-box machine learning models in the context of rock discontinuity classification.The analysis reveals that features such as normal vectors,verticality,and Z-values have the greatest influence on identifying main discontinuity sets,while linearity,planarity,and eigenvalues contribute less,making the model more transparent and easier to understand.After classification,individual discontinuity sets were detected using a revised DBSCAN from the main discontinuity sets.Finally,the orientation parameters of the plane fitted to each discontinuity were derived from the plane parameters obtained using the Random Sample Consensus(RANSAC).Two real-world datasets(obtained from SfM and LiDAR)and one synthetic dataset were used to validate the proposed method,which successfully identified rock discontinuities and their orientation parameters(dip angle/direction).展开更多
The spatial distribution of discontinuities and the size of rock blocks are the key indicators for rock mass quality evaluation and rockfall risk assessment.Traditional manual measurement is often dangerous or unreach...The spatial distribution of discontinuities and the size of rock blocks are the key indicators for rock mass quality evaluation and rockfall risk assessment.Traditional manual measurement is often dangerous or unreachable at some high-steep rock slopes.In contrast,unmanned aerial vehicle(UAV)photogrammetry is not limited by terrain conditions,and can efficiently collect high-precision three-dimensional(3D)point clouds of rock masses through all-round and multiangle photography for rock mass characterization.In this paper,a new method based on a 3D point cloud is proposed for discontinuity identification and refined rock block modeling.The method is based on four steps:(1)Establish a point cloud spatial topology,and calculate the point cloud normal vector and average point spacing based on several machine learning algorithms;(2)Extract discontinuities using the density-based spatial clustering of applications with noise(DBSCAN)algorithm and fit the discontinuity plane by combining principal component analysis(PCA)with the natural breaks(NB)method;(3)Propose a method of inserting points in the line segment to generate an embedded discontinuity point cloud;and(4)Adopt a Poisson reconstruction method for refined rock block modeling.The proposed method was applied to an outcrop of an ultrahigh steep rock slope and compared with the results of previous studies and manual surveys.The results show that the method can eliminate the influence of discontinuity undulations on the orientation measurement and describe the local concave-convex characteristics on the modeling of rock blocks.The calculation results are accurate and reliable,which can meet the practical requirements of engineering.展开更多
Airborne LiDAR(Light Detection and Ranging)is an evolving high-tech active remote sensing technology that has the capability to acquire large-area topographic data and can quickly generate DEM(Digital Elevation Model)...Airborne LiDAR(Light Detection and Ranging)is an evolving high-tech active remote sensing technology that has the capability to acquire large-area topographic data and can quickly generate DEM(Digital Elevation Model)products.Combined with image data,this technology can further enrich and extract spatial geographic information.However,practically,due to the limited operating range of airborne LiDAR and the large area of task,it would be necessary to perform registration and stitching process on point clouds of adjacent flight strips.By eliminating grow errors,the systematic errors in the data need to be effectively reduced.Thus,this paper conducts research on point cloud registration methods in urban building areas,aiming to improve the accuracy and processing efficiency of airborne LiDAR data.Meanwhile,an improved post-ICP(Iterative Closest Point)point cloud registration method was proposed in this study to determine the accurate registration and efficient stitching of point clouds,which capable to provide a potential technical support for applicants in related field.展开更多
Task-oriented point cloud sampling aims to select a representative subset from the input,tailored to specific application scenarios and task requirements.However,existing approaches rarely tackle the problem of redund...Task-oriented point cloud sampling aims to select a representative subset from the input,tailored to specific application scenarios and task requirements.However,existing approaches rarely tackle the problem of redundancy caused by local structural similarities in 3D objects,which limits the performance of sampling.To address this issue,this paper introduces a novel task-oriented point cloud masked autoencoder-based sampling network(Point-MASNet),inspired by the masked autoencoder mechanism.Point-MASNet employs a voxel-based random non-overlapping masking strategy,which allows the model to selectively learn and capture distinctive local structural features from the input data.This approach effectively mitigates redundancy and enhances the representativeness of the sampled subset.In addition,we propose a lightweight,symmetrically structured keypoint reconstruction network,designed as an autoencoder.This network is optimized to efficiently extract latent features while enabling refined reconstructions.Extensive experiments demonstrate that Point-MASNet achieves competitive sampling performance across classification,registration,and reconstruction tasks.展开更多
基金funded by the National Natural Science Foundation of China Youth Project(61603127).
文摘Traditional models for semantic segmentation in point clouds primarily focus on smaller scales.However,in real-world applications,point clouds often exhibit larger scales,leading to heavy computational and memory requirements.The key to handling large-scale point clouds lies in leveraging random sampling,which offers higher computational efficiency and lower memory consumption compared to other sampling methods.Nevertheless,the use of random sampling can potentially result in the loss of crucial points during the encoding stage.To address these issues,this paper proposes cross-fusion self-attention network(CFSA-Net),a lightweight and efficient network architecture specifically designed for directly processing large-scale point clouds.At the core of this network is the incorporation of random sampling alongside a local feature extraction module based on cross-fusion self-attention(CFSA).This module effectively integrates long-range contextual dependencies between points by employing hierarchical position encoding(HPC).Furthermore,it enhances the interaction between each point’s coordinates and feature information through cross-fusion self-attention pooling,enabling the acquisition of more comprehensive geometric information.Finally,a residual optimization(RO)structure is introduced to extend the receptive field of individual points by stacking hierarchical position encoding and cross-fusion self-attention pooling,thereby reducing the impact of information loss caused by random sampling.Experimental results on the Stanford Large-Scale 3D Indoor Spaces(S3DIS),Semantic3D,and SemanticKITTI datasets demonstrate the superiority of this algorithm over advanced approaches such as RandLA-Net and KPConv.These findings underscore the excellent performance of CFSA-Net in large-scale 3D semantic segmentation.
基金supported by the National Natural Science Foundation of China(Grant Nos.52304139,52325403)the CCTEG Coal Mining Research Institute funding(Grant No.KCYJY-2024-MS-10).
文摘3D laser scanning technology is widely used in underground openings for high-precision,rapid,and nondestructive structural evaluations.Segmenting large 3D point cloud datasets,particularly in coal mine roadways with multi-scale targets,remains challenging.This paper proposes an enhanced segmentation method integrating improved PointNet++with a coverage-voted strategy.The coverage-voted strategy reduces data while preserving multi-scale target topology.The segmentation is achieved using an enhanced PointNet++algorithm with a normalization preprocessing head,resulting in a 94%accuracy for common supporting components.Ablation experiments show that the preprocessing head and coverage strategies increase segmentation accuracy by 20%and 2%,respectively,and improve Intersection over Union(IoU)for bearing plate segmentation by 58%and 20%.The accuracy of the current pretraining segmentation model may be affected by variations in surface support components,but it can be readily enhanced through re-optimization with additional labeled point cloud data.This proposed method,combined with a previously developed machine learning model that links rock bolt load and the deformation field of its bearing plate,provides a robust technique for simultaneously measuring the load of multiple rock bolts in a single laser scan.
文摘Evaluating rock mass quality using three-dimensional(3D)point clouds is crucial for discontinuity extraction and is widely applied in various industrial sectors.However,the utilization of this method in geological surveys remains limited.Notable limitations of current research include the scarcity of validation using simple geometric shapes for discontinuity extraction methods,and the lack of studies that target both planar and linear discontinuity.To address these gaps,this study proposes a workflow for identifying discontinuity planes and traces in rock outcrops from photogrammetric 3D modeling,employing the Compass and Facets plugins in the open-source CloudCompare software.Prior to field application,the efficacy of the extraction methods was first evaluated using experimental datasets of a cube and an isosceles triangular prism generated under laboratory-controlled conditions.This validation demonstrated exceptional accuracy,with the dip and dip direction(DDD)of extracted structures consistently within±2°of the actual values.Following this rigorous laboratory validation,this methodology was applied to a more complex natural rock outcrop(Miocene–Pliocene deposits in Japan),demonstrating its applicability in realistic geological settings for identifying structures.The results showed that the dip and dip direction trends of the extracted bedding planes and faults were consistent with field measurements,achieving a time reduction of approximately 40%compared to traditional methods.In conclusion,through strictly controlled initial verification and subsequent successful application to a complex natural setting,this study confirmed that the proposed workflow can effectively and efficiently extract discontinuous geological structures from point clouds.
基金supported in part by the National Key Research and Development Program of Chinaunder(Grant 2021YFB3101100)in part by the National Natural Science Foundation of Chinaunder(Grant 42461057),(Grant 62272123),and(Grant 42371470)+1 种基金in part by the Fundamental Research Program of Shanxi Province under(Grant 202303021212164)in part by the Postgraduate Education Innovation Program of Shanxi Province under(Grant 2024KY474).
文摘Recently,large-scale deep learning models have been increasingly adopted for point cloud classification.However,thesemethods typically require collecting extensive datasets frommultiple clients,which may lead to privacy leaks.Federated learning provides an effective solution to data leakage by eliminating the need for data transmission,relying instead on the exchange of model parameters.However,the uneven distribution of client data can still affect the model’s ability to generalize effectively.To address these challenges,we propose a new framework for point cloud classification called Federated Dynamic Aggregation Selection Strategy-based Multi-Receptive Field Fusion Classification Framework(FDASS-MRFCF).Specifically,we tackle these challenges with two key innovations:(1)During the client local training phase,we propose a Multi-Receptive Field Fusion Classification Model(MRFCM),which captures local and global structures in point cloud data through dynamic convolution and multi-scale feature fusion,enhancing the robustness of point cloud classification.(2)In the server aggregation phase,we introduce a Federated Dynamic Aggregation Selection Strategy(FDASS),which employs a hybrid strategy to average client model parameters,skip aggregation,or reallocate local models to different clients,thereby balancing global consistency and local diversity.We evaluate our framework using the ModelNet40 and ShapeNetPart benchmarks,demonstrating its effectiveness.The proposed method is expected to significantly advance the field of point cloud classification in a secure environment.
基金The National Natural Science Foundation of China(No.52338011)the Start-up Research Fund of Southeast University(No.RF1028624058)+1 种基金the Southeast University Interdisciplinary Research Program for Young Scholarsthe National Key Research and Development Program of China(No.2024YFC3014103).
文摘The virtual preassembly of super-high steel bridge towers faces a challenge in the efficient and precise extraction of complex cross-sectional features.Factors such as fabrication errors,gravity-induced deformations,and temperature fluctuations can compromise the accuracy of contour extraction.To address these limitations,an improved Alpha-shape-based point cloud contour extraction method is proposed.The proposed approach uses a hierarchical strategy to process three-dimensional laser scanning point clouds.The processed data are then subjected to curvatureadaptive voxel filtering to reduce acquisition noise.In addition,an enhanced iterative closest point(ICP)variant with correspondence validation accurately aligns the discrete point cloud segments.The proposed curvature-responsive Alpha-shape framework enables multiscale contour delineation through topology-adaptive threshold modulation,which resolves boundary ambiguities in geometrically complex cross-sections.The method was experimentally validated using field-acquired measurement datasets from the Zhangjinggao Yangtze River Bridge tower segments,confirming its capability to reconstruct noncanonical cross-sectional geometries.Three contour extraction methods,including Poisson reconstruction,the conventional Alpha-shape algorithm,and random sample consensus with ICP(RANSAC-ICP),were compared to evaluate the performance of the proposed Alpha-shape algorithm.The results demonstrate that the proposed method achieves superior contour extraction accuracy and data reduction efficiency,highlighting its effectiveness in contour extraction tasks.
基金supported by the National Key R&D Program of China(Grant No.2022YFC3080200)the China Postdoctoral Science Foundation(Grant No.2023M731264)the Science and Technology Development Plan Project of Jilin Province,China(Grant No.20250602007RC).
文摘Automatic identificationof discontinuities is a key focus in rock slope research.Conventional methods typically target small areas,which limits efficiencyand applicability for complex discontinuities in large-scale rock slopes.This study uses multi-angle unmanned aerial vehicle(UAV)nap-of-the-object photogrammetry to construct a high-definitionthree-dimensional(3D)point cloud model of the slope.The edge-firstconnection algorithm identifiesall edge points of discontinuities in the point cloud and completes recognition through simple connection analysis.This method avoids the complex calculations required for sequentially identifying discontinuity edges in conventional methods and achieves significantacceleration through algorithm optimization and parallel computation support.Based on this algorithm,the RockDiscontinuity Identification(RD ID)software is developed and applied to identify numerous highly disordered discontinuities on the Xulong slope in the Jinsha River suture zone.Processing tens of millions of point clouds within approximately 2 h demonstrates exceptional computational efficiency.The automatic algorithm accurately identifiesnearly 80%of planar discontinuities,with orientations and trace lengths closely matching manual results,highlighting its potential for large-scale rock outcrop applications.Comparisons with region growing algorithms further emphasize its effectiveness and accuracy.However,the algorithm struggles to identify linear discontinuities,which are a major source of error.Additionally,high roughness and smooth edges of discontinuities affect recognition accuracy,indicating areas for further improvement.
文摘To support the process of grasping objects on a tabletop for the blind or robotic arm,it is necessary to address fundamental computer vision tasks,such as detecting,recognizing,and locating objects in space,and determining the position of the grasping information.These results can then be used to guide the visually impaired or to execute grasping tasks with a robotic arm.In this paper,we collected,annotated,and published the benchmark TQUGraspingObject dataset for testing,validation,and evaluation of deep learning(DL)models for detecting,recognizing,and localizing grasping objects in 2D and 3D space,especially 3D point cloud data.Our dataset is collected in a shared room,with common everyday objects placed on the tabletop in jumbled positions by Intel RealSense D435(IR-D435).This dataset includes more than 63k RGB-D pairs and related data such as normalized 3D object point cloud,3D object point cloud segmented,coordinate system normalizationmatrix,3D object point cloud normalized,and hand pose for grasping each object.At the same time,we also conducted experiments on fourDL networks with the best performance:SSD-MobileNetV3,ResNet50-Transformer,ResNet101-Transformer,and YOLOv12.The results present that YOLOv12 has the most suitable results in detecting and recognizing objects in images.All data,annotations,toolkit,source code,point cloud data,and results are publicly available on our project website:https://github.com/HuaTThanhIT2327Tqu/datasetv2.
基金Key Laboratory of Degraded and Unused Land Consolidation Engineering,Ministry of Natural Resources of China,Grant/Award Number:SXDJ2024-22Technology Innovation Centre for Integrated Applications in Remote Sensing and Navigation,Ministry of Natural Resources of China,Grant/Award Number:TICIARSN-2023-06+2 种基金National Natural Science Foundation of China,Grant/Award Numbers:42171446,62302246Zhejiang Provincial Natural Science Foundation of China,Grant/Award Number:LQ23F010008Science and Technology Program of Tianjin,China,Grant/Award Number:23ZGSSSS00010。
文摘Semantic segmentation in the context of 3D point clouds for the railway environment holds a significant economic value,but its development is severely hindered by the lack of suitable and specific datasets.Additionally,the models trained on existing urban road point cloud datasets demonstrate poor generalisation on railway data due to a large domain gap caused by non-overlapping special/rare categories,for example,rail track,track bed etc.To harness the potential of supervised learning methods in the domain of 3D railway semantic segmentation,we introduce RailPC,a new point cloud benchmark.RailPC provides a large-scale dataset with rich annotations for semantic segmentation in the railway environment.Notably,RailPC contains twice the number of annotated points compared to the largest available mobile laser scanning(MLS)point cloud dataset and is the first railway-specific 3D dataset for semantic segmentation.It covers a total of nearly 25 km railway in two different scenes(urban and mountain),with 3 billion points that are finely labelled as 16 most typical classes with respect to railway,and the data acquisition process is completed in China by MLS systems.Through extensive experimentation,we evaluate the performance of advanced scene understanding methods on the annotated dataset and present a synthetic analysis of semantic segmentation results.Based on our findings,we establish some critical challenges towards railway-scale point cloud semantic segmentation.The dataset is available at https://github.com/NNU-GISA/GISA-RailPC,and we will continuously update it based on community feedback.
文摘Automatic registration of unordered point clouds is the prerequisite forurban reconstruction. However, most of the existing technologies stillsuffer from some limitations. On one hand, most of them are sensitive tonoise and repetitive structures, which makes them infeasible for theregistration of large-scale point clouds. On the other hand, most of themdealing with point clouds with limited overlaps and unpredictablelocation. All these problems make it difficult for registration technology toobtain qualified results in outdoor point cloud. To overcome theselimitations, this paper presents a grid graph-based point cloud registration(GGR) algorithm to align pairwise scans. First, point cloud is divided into aset of 3D grids. We propose a voting strategy to measure the similaritybetween two grids based on feature descriptor, transforming thesuperficial correspondence into 3D grid expression. Next, a graphmatching is proposed to capture the spatial consistency from putativecorrespondences, and graph matching hierarchically refines thecorresponding grids until obtaining point-to-point correspondences.Comprehensive experiments demonstrated that the proposed algorithmobtains good performance in terms of successful registration rate, rotationerror, translation error, and outperformed the state-of-the-art approaches.
文摘针对自动驾驶场景下,近处干扰点云误检率高、远处稀疏点云漏检率高的问题,提出了一种基于改进PointPillars的自动驾驶障碍物点云检测算法.首先,通过聚合模块和共享多层感知机(shared multi-layer perceptron,MLP)对柱体内点云进行特征编码,采用最大池化与平均池化叠加的方法将点云的显著特征与细节特征映射为柱体特征;其次,针对算法对伪图特征关注与利用不充分的问题,引入坐标注意力(coordinate attention,CA)机制和残差连接的伪图特征提取模块(attention and residual second block,ARSB),将深层与浅层特征图进行融合,优化算法梯度,增强算法对有效目标的关注度.试验结果表明:改进算法对全局点云检测精度较高,平均精度优于PointPillars、稀疏到稠密3D目标检测器(STD)等点云目标检测算法,在汽车类别上的检测精度优势明显,检测速度较快,符合实时性要求.
基金Postgraduate Innovation Top notch Talent Training Project of Hunan Province,Grant/Award Number:CX20220045Scientific Research Project of National University of Defense Technology,Grant/Award Number:22-ZZCX-07+2 种基金New Era Education Quality Project of Anhui Province,Grant/Award Number:2023cxcysj194National Natural Science Foundation of China,Grant/Award Numbers:62201597,62205372,1210456foundation of Hefei Comprehensive National Science Center,Grant/Award Number:KY23C502。
文摘Large-scale point cloud datasets form the basis for training various deep learning networks and achieving high-quality network processing tasks.Due to the diversity and robustness constraints of the data,data augmentation(DA)methods are utilised to expand dataset diversity and scale.However,due to the complex and distinct characteristics of LiDAR point cloud data from different platforms(such as missile-borne and vehicular LiDAR data),directly applying traditional 2D visual domain DA methods to 3D data can lead to networks trained using this approach not robustly achieving the corresponding tasks.To address this issue,the present study explores DA for missile-borne LiDAR point cloud using a Monte Carlo(MC)simulation method that closely resembles practical application.Firstly,the model of multi-sensor imaging system is established,taking into account the joint errors arising from the platform itself and the relative motion during the imaging process.A distortion simulation method based on MC simulation for augmenting missile-borne LiDAR point cloud data is proposed,underpinned by an analysis of combined errors between different modal sensors,achieving high-quality augmentation of point cloud data.The effectiveness of the proposed method in addressing imaging system errors and distortion simulation is validated using the imaging scene dataset constructed in this paper.Comparative experiments between the proposed point cloud DA algorithm and the current state-of-the-art algorithms in point cloud detection and single object tracking tasks demonstrate that the proposed method can improve the network performance obtained from unaugmented datasets by over 17.3%and 17.9%,surpassing SOTA performance of current point cloud DA algorithms.
基金supported by National Natural Science Foundation of China (No. 61572267, No. 61272425, No. 61402245)the Open Project of Co-Innovation Center for Information Supply & Assurance Technology, Anhui University+1 种基金the Open Project of the State Key Laboratory of Information Security,Institute of Information Engineering,Chinese Academy of Sciences(No.2017-MS-21, No.2016-MS-23)National Cryptography Development Fund of China (MMJJ20170118)
文摘Identity-based public cloud storage auditing schemes can check the integrity of cloud data, and reduce the complicated certificate management. In such a scheme, one Private Key Generator(PKG) is employed to authenticate the identity and generate private keys for all users, and one Third Party Auditor(TPA) is employed to by users to check the integrity of cloud data. This approach is undesirable for large-scale users since the PKG and the TPA might not be able to afford the heavy workload. To solve the problem, we give a hierarchical Private Key Generator structure for large-scale user groups, in which a root PKG delegates lower-level PKGs to generate private keys and authenticate identities. Based on the proposed structure, we propose an authorized identity-based public cloud storage auditing scheme, in which the lowest-level PKGs play the role of TPA, and only the authorized lowest-level PKGs can represent users in their domains to check cloud data's integrity. Furthermore, we give the formal security analysis and experimental results, which show that our proposed scheme is secure and efficient.
文摘This study presents a framework for the semi-automatic detection of rock discontinuities using a threedimensional(3D)point cloud(PC).The process begins by selecting an appropriate neighborhood size,a critical step for feature extraction from the PC.The effects of different neighborhood sizes(k=5,10,20,50,and 100)have been evaluated to assess their impact on classification performance.After that,17 geometric and spatial features were extracted from the PC.Next,ensemble methods,AdaBoost.M2,random forest,and decision tree,have been compared with Artificial Neural Networks to classify the main discontinuity sets.The McNemar test indicates that the classifiers are statistically significant.The random forest classifier consistently achieves the highest performance with an accuracy exceeding 95%when using a neighborhood size of k=100,while recall,F-score,and Cohen's Kappa also demonstrate high success.SHapley Additive exPlanations(SHAP),an Explainable AI technique,has been used to evaluate feature importance and improve the explainability of black-box machine learning models in the context of rock discontinuity classification.The analysis reveals that features such as normal vectors,verticality,and Z-values have the greatest influence on identifying main discontinuity sets,while linearity,planarity,and eigenvalues contribute less,making the model more transparent and easier to understand.After classification,individual discontinuity sets were detected using a revised DBSCAN from the main discontinuity sets.Finally,the orientation parameters of the plane fitted to each discontinuity were derived from the plane parameters obtained using the Random Sample Consensus(RANSAC).Two real-world datasets(obtained from SfM and LiDAR)and one synthetic dataset were used to validate the proposed method,which successfully identified rock discontinuities and their orientation parameters(dip angle/direction).
基金supported by the National Natural Science Foundation of China(Grant Nos.41941017 and 42177139)Graduate Innovation Fund of Jilin University(Grant No.2024CX099)。
文摘The spatial distribution of discontinuities and the size of rock blocks are the key indicators for rock mass quality evaluation and rockfall risk assessment.Traditional manual measurement is often dangerous or unreachable at some high-steep rock slopes.In contrast,unmanned aerial vehicle(UAV)photogrammetry is not limited by terrain conditions,and can efficiently collect high-precision three-dimensional(3D)point clouds of rock masses through all-round and multiangle photography for rock mass characterization.In this paper,a new method based on a 3D point cloud is proposed for discontinuity identification and refined rock block modeling.The method is based on four steps:(1)Establish a point cloud spatial topology,and calculate the point cloud normal vector and average point spacing based on several machine learning algorithms;(2)Extract discontinuities using the density-based spatial clustering of applications with noise(DBSCAN)algorithm and fit the discontinuity plane by combining principal component analysis(PCA)with the natural breaks(NB)method;(3)Propose a method of inserting points in the line segment to generate an embedded discontinuity point cloud;and(4)Adopt a Poisson reconstruction method for refined rock block modeling.The proposed method was applied to an outcrop of an ultrahigh steep rock slope and compared with the results of previous studies and manual surveys.The results show that the method can eliminate the influence of discontinuity undulations on the orientation measurement and describe the local concave-convex characteristics on the modeling of rock blocks.The calculation results are accurate and reliable,which can meet the practical requirements of engineering.
基金Guangxi Key Laboratory of Spatial Information and Geomatics(21-238-21-12)Guangxi Young and Middle-aged Teachers’Research Fundamental Ability Enhancement Project(2023KY1196).
文摘Airborne LiDAR(Light Detection and Ranging)is an evolving high-tech active remote sensing technology that has the capability to acquire large-area topographic data and can quickly generate DEM(Digital Elevation Model)products.Combined with image data,this technology can further enrich and extract spatial geographic information.However,practically,due to the limited operating range of airborne LiDAR and the large area of task,it would be necessary to perform registration and stitching process on point clouds of adjacent flight strips.By eliminating grow errors,the systematic errors in the data need to be effectively reduced.Thus,this paper conducts research on point cloud registration methods in urban building areas,aiming to improve the accuracy and processing efficiency of airborne LiDAR data.Meanwhile,an improved post-ICP(Iterative Closest Point)point cloud registration method was proposed in this study to determine the accurate registration and efficient stitching of point clouds,which capable to provide a potential technical support for applicants in related field.
基金supported by the National Key Research and Development Program of China(2022YFB3103500)the National Natural Science Foundation of China(62473033,62571027)+1 种基金in part by the Beijing Natural Science Foundation(L231012)the State Scholarship Fund from the China Scholarship Council.
文摘Task-oriented point cloud sampling aims to select a representative subset from the input,tailored to specific application scenarios and task requirements.However,existing approaches rarely tackle the problem of redundancy caused by local structural similarities in 3D objects,which limits the performance of sampling.To address this issue,this paper introduces a novel task-oriented point cloud masked autoencoder-based sampling network(Point-MASNet),inspired by the masked autoencoder mechanism.Point-MASNet employs a voxel-based random non-overlapping masking strategy,which allows the model to selectively learn and capture distinctive local structural features from the input data.This approach effectively mitigates redundancy and enhances the representativeness of the sampled subset.In addition,we propose a lightweight,symmetrically structured keypoint reconstruction network,designed as an autoencoder.This network is optimized to efficiently extract latent features while enabling refined reconstructions.Extensive experiments demonstrate that Point-MASNet achieves competitive sampling performance across classification,registration,and reconstruction tasks.
