Computer Aided Geometric Design最新文献

Transferable and undefendable point cloud attacks via medial axis transform 通过中轴变换转移和不可防御的点云攻击

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2026-03-01 Epub Date: 2026-02-03 DOI: 10.1016/j.cagd.2026.102515

Keke Tang , Yuze Gao , Weilong Peng , Xiaofei Wang , Meie Fang , Peican Zhu

{"title":"Transferable and undefendable point cloud attacks via medial axis transform","authors":"Keke Tang , Yuze Gao , Weilong Peng , Xiaofei Wang , Meie Fang , Peican Zhu","doi":"10.1016/j.cagd.2026.102515","DOIUrl":"10.1016/j.cagd.2026.102515","url":null,"abstract":"<div><div>Studying adversarial attacks on point clouds is essential for evaluating and improving the robustness of 3D deep learning models. However, most existing attack methods are developed under ideal white-box settings and often suffer from limited transferability to unseen models and insufficient robustness against common defense mechanisms. In this paper, we propose MAT-Adv, a novel adversarial attack framework that enhances both transferability and undefendability by explicitly perturbing the medial axis transform (MAT) representations, in order to induce inherent adversarialness in the resulting point clouds. Specifically, we employ an autoencoder to project input point clouds into compact MAT representations that capture the intrinsic geometric structure of point clouds. By perturbing these intrinsic representations, MAT-Adv introduces structural-level adversarial characteristics that remain effective across diverse models and defense strategies. To mitigate overfitting and prevent perturbation collapse, we incorporate a dropout strategy into the optimization of MAT perturbations, further improving transferability and undefendability. Extensive experiments demonstrate that MAT-Adv significantly outperforms existing state-of-the-art methods in both transferability and undefendability.</div></div>","PeriodicalId":55226,"journal":{"name":"Computer Aided Geometric Design","volume":"125 ","pages":"Article 102515"},"PeriodicalIF":1.7,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146135630","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Fast extremum graph computation for large-scale arbitrary grids 大规模任意网格的快速极值图计算

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2026-03-01 Epub Date: 2026-02-17 DOI: 10.1016/j.cagd.2026.102519

Quming Li , Zhengwen Liu , Zhibin Huang , Wantong Li , Zhiqiang Chu , Zhitao Dai , Min Yu , Fang Deng

{"title":"Fast extremum graph computation for large-scale arbitrary grids","authors":"Quming Li , Zhengwen Liu , Zhibin Huang , Wantong Li , Zhiqiang Chu , Zhitao Dai , Min Yu , Fang Deng","doi":"10.1016/j.cagd.2026.102519","DOIUrl":"10.1016/j.cagd.2026.102519","url":null,"abstract":"<div><div>The extremum graph is a compact topological representation designed to capture critical structures in scalar fields, and has been widely applied in flow field analysis, scientific visualization, and feature extraction. However, existing methods for constructing extremum graphs are typically restricted to structured grids and rely on exhaustive saddle point detection for every sample, which results in extremely low computational efficiency or significant precision degradation when large-scale unstructured data must be resampled onto structured grids. To overcome these limitations, we propose a fast extremum graph construction framework, denoted as <em>AnyEG</em>, which supports both structured and unstructured grids in arbitrary dimensions. The method integrates local path compression with gradient flow analysis, and introduces the concept of <em>boundary saddle points</em> to achieve efficient and accurate saddle point detection. Grounded in discrete Morse theory, our approach substantially reduces redundant computation and memory access overhead. Furthermore, to the best of our knowledge, <em>AnyEG</em> is the first GPU-based implementation of an extremum graph algorithm within the Topology ToolKit (TTK) framework. Comprehensive experiments on multiple structured and unstructured datasets demonstrate that our method achieves an average speedup of nearly five times over the state-of-the-art Tachyon method, while preserving complete topological correctness.</div></div>","PeriodicalId":55226,"journal":{"name":"Computer Aided Geometric Design","volume":"125 ","pages":"Article 102519"},"PeriodicalIF":1.7,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147403593","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A geometric perturbation-based method for tolerance analysis 基于几何微扰的公差分析方法

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2026-03-01 Epub Date: 2026-02-04 DOI: 10.1016/j.cagd.2026.102517

Hui Chen , Li-Yong Shen , Wenkai Hu , Zaisheng Lin , Shaoqiang Ma

引用次数: 0

Accurate matrix conversion between Bernstein and h-Bernstein bases Bernstein和h-Bernstein基之间精确的矩阵转换

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2026-03-01 Epub Date: 2026-02-07 DOI: 10.1016/j.cagd.2026.102518

Y. Khiar , E. Mainar , J.M. Peña , E. Royo-Amondarain

{"title":"Accurate matrix conversion between Bernstein and h-Bernstein bases","authors":"Y. Khiar , E. Mainar , J.M. Peña , E. Royo-Amondarain","doi":"10.1016/j.cagd.2026.102518","DOIUrl":"10.1016/j.cagd.2026.102518","url":null,"abstract":"<div><div>This paper investigates the matrix conversion between the classical Bernstein basis and its one-parameter generalization, the <span><math><mi>h</mi></math></span>-Bernstein basis. New <span><math><mi>h</mi></math></span>-analogues of the binomial coefficients are introduced, providing explicit and compact expressions for the entries of the corresponding change-of-basis matrices. Structural properties such as symmetry and recurrence relations are derived, offering both theoretical insight and practical computational advantages. The proposed recurrence formulations enable the generation of the conversion matrices with high relative accuracy, avoiding subtractive cancellations and the numerical instabilities associated with direct collocation-based approaches. These results ensure reliable computations even for very large degrees and establish a foundation for the development of accurate and efficient algorithms in geometric modeling and related numerical applications involving <span><math><mi>h</mi></math></span>-Bernstein polynomials. Numerical experiments confirm the theoretical findings and highlight the advantages of the proposed approach.</div></div>","PeriodicalId":55226,"journal":{"name":"Computer Aided Geometric Design","volume":"125 ","pages":"Article 102518"},"PeriodicalIF":1.7,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147403592","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

UVConv: A UV grid-based convolutional neural network for classification and segmentation of 3D CAD models UVConv：一种基于UV网格的卷积神经网络，用于三维CAD模型的分类和分割

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2026-03-01 Epub Date: 2026-02-03 DOI: 10.1016/j.cagd.2026.102516

Zenghui Su , Jing Bai , Xiangyu Ma , Feiwei Qin

{"title":"UVConv: A UV grid-based convolutional neural network for classification and segmentation of 3D CAD models","authors":"Zenghui Su , Jing Bai , Xiangyu Ma , Feiwei Qin","doi":"10.1016/j.cagd.2026.102516","DOIUrl":"10.1016/j.cagd.2026.102516","url":null,"abstract":"<div><div>The widespread adoption of 3D CAD models in intelligent manufacturing and digital twin systems demands efficient and accurate classification and segmentation. While effective approaches typically model geometry and topology jointly using B-rep face-edge graphs, prevailing methods rely heavily on auxiliary information (e.g., supplemental geometry/semantics) to compensate for perceived shortcomings in raw B-rep data. This dependency incurs high preprocessing overhead, limits cross-platform generalization, and hinders industrial deployment. Furthermore, inherent redundancies within B-rep graphs can obscure essential semantics during encoding. To overcome these limitations, we propose UVConv, an end-to-end dual-path framework that operates exclusively on the original UV grid and face adjacency graph, eliminating auxiliary inputs. UVConv employs a serial geometry-topology processing: the geometric encoder utilizes bottleneck compression to amplify discriminative features, while the topological encoder integrates a Bidirectional Interaction Module (BIM) and an Attention Filtering Module (AFM) to extract critical structural semantics and suppress noise. Comprehensive experiments on multiple benchmark CAD datasets demonstrate that UVConv matches or surpasses state-of-the-art performance without auxiliary data, exhibiting superior representational capacity, robust generalization, and enhanced industrial deployability.</div></div>","PeriodicalId":55226,"journal":{"name":"Computer Aided Geometric Design","volume":"125 ","pages":"Article 102516"},"PeriodicalIF":1.7,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146135629","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Weighted transfinite interpolation for slightly unstructured curvilinear patches 略非结构化曲线斑块的加权超限插值

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2026-01-01 Epub Date: 2025-12-01 DOI: 10.1016/j.cagd.2025.102497

Christopher G. Provatidis

{"title":"Weighted transfinite interpolation for slightly unstructured curvilinear patches","authors":"Christopher G. Provatidis","doi":"10.1016/j.cagd.2025.102497","DOIUrl":"10.1016/j.cagd.2025.102497","url":null,"abstract":"<div><div>This paper presents an extension of the classical transfinite interpolation formula originally developed by Gordon and Hall, aimed at improving geometric modeling over slightly unstructured curvilinear patches. While traditional transfinite schemes rely on structured boundary data and cardinal blending functions, the proposed method introduces a weighted framework that accommodates irregularities in the layout of interpolation stations. By assigning suitable weights to both the blending and trial functions—such as Bernstein polynomials, B-splines, or NURBS—the interpolation scheme gains flexibility while preserving smoothness and continuity across the patch. The approach maintains the intuitive geometric structure of the original formulation, but significantly expands its applicability to more general domains encountered in computer-aided geometric design (CAGD), mesh generation, and isogeometric analysis. With the cylindrical patch serving as the primary benchmark and the spherical cap as the secondary, various combinations of blending and trial functions are investigated to illustrate the method’s capability to accurately represent mild irregularities in data distribution.</div></div>","PeriodicalId":55226,"journal":{"name":"Computer Aided Geometric Design","volume":"124 ","pages":"Article 102497"},"PeriodicalIF":1.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145750229","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A quaternion approach to G2 Hermite interpolation by quintic spatial Pythagorean-hodograph curves 用五次空间毕达哥拉斯曲线求G2 Hermite插值的四元数方法

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2026-01-01 Epub Date: 2025-12-05 DOI: 10.1016/j.cagd.2025.102505

Rida T. Farouki

引用次数: 0

SceneFlow: Synthesizing indoor scenes via geometry-enhanced flow matching SceneFlow：通过几何增强流匹配合成室内场景

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2025-12-01 Epub Date: 2025-11-07 DOI: 10.1016/j.cagd.2025.102489

Wenming Wu , Akang Shen , Yanzhe Yin , Zixiang Chen , Gaofeng Zhang , Liping Zheng

{"title":"SceneFlow: Synthesizing indoor scenes via geometry-enhanced flow matching","authors":"Wenming Wu , Akang Shen , Yanzhe Yin , Zixiang Chen , Gaofeng Zhang , Liping Zheng","doi":"10.1016/j.cagd.2025.102489","DOIUrl":"10.1016/j.cagd.2025.102489","url":null,"abstract":"<div><div>Recent autoregressive and diffusion models are used to generate indoor scenes. However, they face challenges in maintaining geometric consistency and achieving real-time performance. Autoregressive models struggle with global dependencies, leading to functional inconsistencies, while diffusion-based methods suffer from slow sampling speeds. Motivated by simpler and smoother sampling trajectories, we propose SceneFlow, a novel indoor scene synthesis framework based on Flow Matching, which learns a flow to gradually move objects from an initial chaotic state to a structured layout. To further enhance geometric consistency, we introduce a geometry enhancement strategy comprising a non-overlap constraint to reduce object overlap and a geometry refinement to align objects with room boundaries. We evaluate our approach using the 3D-FRONT dataset. Extensive evaluations demonstrate that our method outperforms state-of-the-art methods in visual quality, geometric accuracy, and generation efficiency. Our method provides a practical solution for applications requiring high-quality indoor scene generation with real-time responsiveness.</div></div>","PeriodicalId":55226,"journal":{"name":"Computer Aided Geometric Design","volume":"123 ","pages":"Article 102489"},"PeriodicalIF":1.7,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145520919","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Surface reconstruction of narrow-band guided Gray-Scott model 窄带制导Gray-Scott模型的表面重构

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2025-12-01 Epub Date: 2025-11-06 DOI: 10.1016/j.cagd.2025.102488

Jian Wang , Xuejuan Chen , Bowen Geng , Mengyu Luo , Shanshan Ge , Haozhe Wang , Jialin Chen , Junseok Kim

引用次数: 0

Tooth mesh repair method based on side length-area constraint and dihedral angle driven subdivision 基于边长-面积约束和二面角驱动细分的齿网修复方法

IF 1.7 4区计算机科学

Computer Aided Geometric Design Pub Date : 2025-11-01 Epub Date: 2025-10-30 DOI: 10.1016/j.cagd.2025.102487

Hongan Li , Kun Yang , Long Peng

{"title":"Tooth mesh repair method based on side length-area constraint and dihedral angle driven subdivision","authors":"Hongan Li , Kun Yang , Long Peng","doi":"10.1016/j.cagd.2025.102487","DOIUrl":"10.1016/j.cagd.2025.102487","url":null,"abstract":"<div><div>Aiming at the problems of poor mesh quality and lack of details in hole repair in 3D tooth scanning model, this paper proposes a high-quality mesh repair algorithm combining multiple geometric constraints and local feature enhancement mechanism. This method first introduces the edge length-area joint constraint strategy, and standardizes the new triangle shape in the initial repair stage to avoid the generation of narrow or deformed grids. Then, according to the dihedral angle change between adjacent triangles, the high curvature region is adaptively selected for Sqrt3 subdivision operation, so as to enhance the detail expression ability of the repaired region. Finally, Laplacian smoothing is used to optimize the spatial distribution of new vertices, which improves the geometric consistency and visual continuity between the patched region and the original model. The experimental results show that the proposed algorithm shows high repair quality and detail fidelity on multiple tooth models. The shape regularity of the new triangular surface is better than that of the comparison method. The proportion of high-quality patches with quality factor q is more than 50%. Multi-metric evaluations demonstrate that this method outperforms the comparison method in terms of geometric accuracy and surface fidelity, making it suitable for high-precision dental modeling and digital restoration tasks.</div></div>","PeriodicalId":55226,"journal":{"name":"Computer Aided Geometric Design","volume":"122 ","pages":"Article 102487"},"PeriodicalIF":1.7,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145466129","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0