arXiv subjects

Hongwei Lin

Publications and source records attributed to Hongwei Lin.

At least 19 recordsLinked to original sources

arXiv2026

Progressive-Iterative Fairing of Curves and Surfaces with Localized Control Point Adjustment

Curve and surface fairing is crucial in computer-aided geometric design, influencing product quality, physical performance, and aesthetics. Traditional methods often apply global modifications, lacking fine-grained control. This paper introduces a novel progressive-iterative fairing method based on control point adjustment. By assigning independent weights to each control point, our approach enables precise, localized shape adjustments. The method functions both globally and locally, allowing for comprehensive shape fairing and fine control over the fairing effect. Furthermore, this paper provides an automatic control point selection method to adjust shapes, thereby eliminating the reliance on manual interaction. Numerical experiments demonstrate the efficiency and effectiveness of our approach.

cs.GR↗

arXiv2026

Topology Understanding of B-Spline Surface/Surface Intersection with Mapper

In the realm of computer-aided design (CAD) software, the intersection of B-spline surfaces stands as a fundamental operation. Despite the extensive history of surface intersection algorithms, the challenge of handling complex intersection topologies persists. While subdivision algorithms have demonstrated strong robustness in computing surface/surface intersection and are capable of addressing singular cases, determining the topology of the intersection obtained through these methods is a key factor for calculating correct intersection, and remains a difficult issue. To address this challenge, we propose a Mapper-based method for determining the topology of the intersection between two B-spline surfaces. Our algorithm is designed to efficiently handle various common and complex intersection topologies. Experimental results verify the robustness and topological correctness of this method.

cs.CG↗

arXiv2026

AW-MoE: All-Weather Mixture of Experts for Robust Multi-Modal 3D Object Detection

Robust 3D object detection under adverse weather conditions is crucial for autonomous driving. However, most existing methods simply combine all weather samples for training while overlooking data distribution discrepancies across different weather scenarios, leading to performance conflicts. To address this issue, we introduce AW-MoE, the framework that innovatively integrates Mixture of Experts (MoE) into weather-robust multi-modal 3D object detection approaches. AW-MoE incorporates Image-guided Weather-aware Routing (IWR), which leverages the superior discriminability of image features across weather conditions and their invariance to scene variations for precise weather classification. Based on this accurate classification, IWR selects the top-K most relevant Weather-Specific Experts (WSE) that handle data discrepancies, ensuring optimal detection under all weather conditions. Additionally, we propose a Unified Dual-Modal Augmentation (UDMA) for synchronous LiDAR and 4D Radar dual-modal data augmentation while preserving the realism of scenes. Extensive experiments on the real-world dataset demonstrate that AW-MoE achieves ~ 15% improvement in adverse-weather performance over state-of-the-art methods, while incurring negligible inference overhead. Moreover, integrating AW-MoE into established baseline detectors yields performance improvements surpassing current state-of-the-art methods. These results show the effectiveness and strong scalability of our AW-MoE. We will release the code publicly at https://github.com/windlinsherlock/AW-MoE.

cs.CV↗

arXiv2026

Persistent homology-based explicit topological control for 2D topology optimization with MMA

Controlling structural complexity, particularly the number of holes, remains a fundamental challenge in topology optimization, with significant implications for both theoretical analysis and manufacturability. Most existing approaches rely on indirect strategies, such as filtering techniques, minimum length-scale control, or specific level-set initializations, which influence topology only implicitly and do not allow precise regulation of topological features. In this work, we propose an explicit and differentiable topology-control framework by integrating persistent homology into the classical minimum-compliance topology optimization problem. The design domain and density field are represented using non-uniform rational B-splines (NURBS), while persistence diagrams are employed to rigorously and quantitatively characterize topological features. Given a prescribed number of holes, a differentiable topology-aware objective is constructed from the persistence pairs and incorporated into the compliance objective, leading to a unified optimization formulation. The resulting problem is efficiently solved using the method of moving asymptotes (MMA).Numerical experiments demonstrate that the proposed approach enables explicit control over structural connectivity and the number of holes, thereby providing a systematic and mathematically grounded strategy for topology regulation.

math.OC↗

arXiv2026

Application of Ideal Observer for Thresholded Data in Search Task

This study advances task-based image quality assessment by developing an anthropomorphic thresholded visual-search model observer. The model is an ideal observer for thresholded data inspired by the human visual system, allowing selective processing of high-salience features to improve discrimination performance. By filtering out irrelevant variability, the model enhances diagnostic accuracy and computational efficiency. The observer employs a two-stage framework: candidate selection and decision-making. Using thresholded data during candidate selection refines regions of interest, while stage-specific feature processing optimizes performance. Simulations were conducted to evaluate the effects of thresholding on feature maps, candidate localization, and multi-feature scenarios. Results demonstrate that thresholding improves observer performance by excluding low-salience features, particularly in noisy environments. Intermediate thresholds often outperform no thresholding, indicating that retaining only relevant features is more effective than keeping all features. Additionally, the model demonstrates effective training with fewer images while maintaining alignment with human performance. These findings suggest that the proposed novel framework can predict human visual search performance in clinically realistic tasks and provide solutions for model observer training with limited resources. Our novel approach has applications in other areas where human visual search and detection tasks are modeled such as in computer vision, machine learning, defense and security image analysis.

eess.IV↗

arXiv2026

Predicting Region of Interest in Human Visual Search Based on Statistical Texture and Gabor Features

Understanding human visual search behavior is a fundamental problem in vision science and computer vision, with direct implications for modeling how observers allocate attention in location-unknown search tasks. In this study, we investigate the relationship between Gabor-based features and gray-level co-occurrence matrix (GLCM) based texture features in modeling early-stage visual search behavior. Two feature-combination pipelines are proposed to integrate Gabor and GLCM features for narrowing the region of possible human fixations. The pipelines are evaluated using simulated digital breast tomosynthesis images. Results show qualitative agreement among fixation candidates predicted by the proposed pipelines and a threshold-based model observer. A strong correlation is observed between GLCM mean and Gabor feature responses, indicating that these features encode related image information despite their different formulations. Eye-tracking data from human observers further suggest consistency between predicted fixation regions and early-stage gaze behavior. These findings highlight the value of combining structural and texture-based features for modeling visual search and support the development of perceptually informed observer models.

cs.CV↗

arXiv2025

TUN: Detecting Significant Points in Persistence Diagrams with Deep Learning

Persistence diagrams (PDs) provide a powerful tool for understanding the topology of the underlying shape of a point cloud. However, identifying which points in PDs encode genuine signals remains challenging. This challenge directly hinders the practical adoption of topological data analysis in many applications, where automated and reliable interpretation of persistence diagrams is essential for downstream decision-making. In this paper, we study automatic significance detection for one-dimensional persistence diagrams. Specifically, we propose Topology Understanding Net (TUN), a multi-modal network that combines enhanced PD descriptors with self-attention, a PointNet-style point cloud encoder, learned fusion, and per-point classification, alongside stable preprocessing and imbalance-aware training. It provides an automated and effective solution for identifying significant points in PDs, which are critical for downstream applications. Experiments show that TUN outperforms classic methods in detecting significant points in PDs, illustrating its effectiveness in real-world applications.

cs.CV↗

arXiv2025

Topology Structure Optimization of Reservoirs Using GLMY Homology

Reservoir is an efficient network for time series processing. It is well known that network structure is one of the determinants of its performance. However, the topology structure of reservoirs, as well as their performance, is hard to analyzed, due to the lack of suitable mathematical tools. In this paper, we study the topology structure of reservoirs using persistent GLMY homology theory, and develop a method to improve its performance. Specifically, it is found that the reservoir performance is closely related to the one-dimensional GLMY homology groups. Then, we develop a reservoir structure optimization method by modifying the minimal representative cycles of one-dimensional GLMY homology groups. Finally, by experiments, it is validated that the performance of reservoirs is jointly influenced by the reservoir structure and the periodicity of the dataset.

cs.LG↗

arXiv2025

Collocation and Mass Matrix in Least-squares Isogeometric Analysis

In this paper, we conduct a systematic numerical analysis of the spectral properties of the collocation and mass matrices in the isogeometric least-squares collocation method (IGA-L), for the approximation of the Poisson problem with homogeneous Dirichlet boundary conditions. This study primarily focuses on the spectral properties of the IGA-L collocation and mass matrices in relation to the isogeometric discretization parameters, such as the mesh size, degree, regularity, spatial dimension, and the number and distribution of the collocation points. Through a comprehensive numerical investigation, we provide estimations for the condition number, as well as the maximum and minimum singular values, in relation to the mesh size, degree and regularity. Moreover, in this paper we also study the effect of the number and distribution of the collocation points on the spectral properties of the collocation matrix, providing insights into the optimization of the collocation points for achieving better-conditioned linear systems.

math.NA↗

arXiv2025

Robust Model Reconstruction Based on the Topological Understanding of Point Clouds Using Persistent Homology

Reconstructing models from unorganized point clouds presents a significant challenge, especially when the models consist of multiple components represented by their surface point clouds. Such models often involve point clouds with noise that represent multiple closed surfaces with shared regions, making their automatic identification and separation inherently complex. In this paper, we propose an automatic method that uses the topological understanding provided by persistent homology, along with representative 2-cycles of persistent homology groups, to effectively distinguish and separate each closed surface. Furthermore, we employ Loop subdivision and least squares progressive iterative approximation (LSPIA) techniques to generate high-quality final surfaces and achieve complete model reconstruction. Our method is robust to noise in the point cloud, making it suitable for reconstructing models from such data. Experimental results demonstrate the effectiveness of our approach and highlight its potential for practical applications.

cs.CG↗

arXiv2024

SC3D: Label-Efficient Outdoor 3D Object Detection via Single Click Annotation

LiDAR-based outdoor 3D object detection has received widespread attention. However, training 3D detectors from the LiDAR point cloud typically relies on expensive bounding box annotations. This paper presents SC3D, an innovative label-efficient method requiring only a single coarse click on the bird's eye view of the 3D point cloud for each frame. A key challenge here is the absence of complete geometric descriptions of the target objects from such simple click annotations. To address this issue, our proposed SC3D adopts a progressive pipeline. Initially, we design a mixed pseudo-label generation module that expands limited click annotations into a mixture of bounding box and semantic mask supervision. Next, we propose a mix-supervised teacher model, enabling the detector to learn mixed supervision information. Finally, we introduce a mixed-supervised student network that leverages the teacher model's generalization ability to learn unclicked instances.Experimental results on the widely used nuScenes and KITTI datasets demonstrate that our SC3D with only coarse clicks, which requires only 0.2% annotation cost, achieves state-of-the-art performance compared to weakly-supervised 3D detection methods.The code will be made publicly available.

cs.CV↗

arXiv2024

Quantum Signal Processing and Quantum Singular Value Transformation on $U(N)$

Quantum signal processing and quantum singular value transformation are powerful tools to implement polynomial transformations of block-encoded matrices on quantum computers, and has achieved asymptotically optimal complexity in many prominent quantum algorithms. We propose a framework of quantum signal processing and quantum singular value transformation on $U(N)$, which realizes multiple polynomials simultaneously from a block-encoded input, as a generalization of those on $U(2)$ in the original frameworks. We provide a comprehensive characterization of achievable polynomial matrices and give recursive algorithms to construct the quantum circuits that realize desired polynomial transformations. As three example applications, we propose a framework to realize bi-variate polynomial functions, demonstrate $N$-interval decision achieving $O(d)$ query complexity with a $\log_2 N$ improvement over iterative $U(2)$-QSP requiring $O(d\log_2 N)$ queries, and present a quantum amplitude estimation algorithm achieving the Heisenberg limit without adaptive measurements.

quant-ph↗

arXiv2024

Topology-Aware Blending Method for Implicit Heterogeneous Porous Model Design

Porous structures are materials consisting of minuscule pores, where the microstructure morphology significantly impacts their macroscopic properties. Integrating different porous structures through a blending method is indispensable to cater to diverse functional regions in heterogeneous models. Previous studies on blending methods for porous structures have mainly focused on controlling the shape of blending regions, yet they have fallen short in effectively addressing topological errors in blended structures. This paper introduces a new blending method that successfully addresses this issue. Initially, a novel initialization method is proposed, which includes distinct strategies for blending regions of varying complexities. Subsequently, we formulate the challenge of eliminating topological errors as an optimization problem based on persistent homology. Through iterative updates of control coefficients, this optimization problem is solved to generate a blended porous structure. Our approach not only avoids topological errors but also governs the shape and positioning of the blending region while remaining unchanged in the structure outside blending region. The experimental outcomes validate the effectiveness of our method in producing high-quality blended porous structures. Furthermore, these results highlight potential applications of our blending method in biomimetics and the design of high-stiffness mechanical heterogeneous models.

cs.GR↗

arXiv2024

The Gestalt Computational Model by Persistent Homology

Widely employed in cognitive psychology, Gestalt theory elucidates basic principles in visual perception. However, the Gestalt principles are validated mainly by psychological experiments, lacking quantitative research supports and theoretical coherence. In this paper, we utilize persistent homology, a mathematical tool in computational topology, to develop a unified computational model for Gestalt principles, addressing the challenges of quantification and computation. On the one hand, the Gestalt computational model presents quantitative supports for Gestalt theory. On the other hand, it shows that the Gestalt principles can be uniformly calculated using persistent homology, thus developing a coherent theory for Gestalt principles in computation. Moreover, it is anticipated that the Gestalt computational model can serve as a significant computational model in the field of computational psychology, and help the understanding of human being visual perception.

cs.CG↗

arXiv2024

Analyzing Singular Patterns in Discrete Planar Vector Fields via Persistent Path Homology

Analyzing singular patterns in vector fields is a fundamental problem in theoretical and practical domains due to the ability of such patterns to detect the intrinsic characteristics of vector fields. In this study, we propose an approach for analyzing singular patterns from discrete planar vector fields. Our method involves converting the planar discrete vector field into a specialized digraph and computing its one-dimensional persistent path homology. By analyzing the persistence diagram, we can determine the location of singularities, and the variations of singular patterns can also be analyzed. The experimental results demonstrate the effectiveness of our method in analyzing the singular patterns of noisy real-world vector fields and measuring the variations between different vector fields.

cs.CG↗

arXiv2023

Asymptotically Optimal Quantum Amplitude Estimation by Generalized Qubitization

We first show that the standard deviation error of quantum amplitude estimation is asymptotically lower bounded by approximately $1.28 L^{-1}$, where $L$ is the number of queries. Then we propose a generalized qubitization that can block-encode several polynomial functions simultaneously, and show how it can help estimating quantum amplitude to achieve the optimal asymptotic accuracy, so the bound is tight.

quant-ph↗

arXiv2023

Random-depth Quantum Amplitude Estimation

The maximum likelihood amplitude estimation algorithm (MLAE) is a practical solution to the quantum amplitude estimation problem with Heisenberg limit error convergence. We improve MLAE by using random depths to avoid the so-called critical points, and do numerical experiments to show that our algorithm is approximately unbiased compared to the original MLAE and approaches the Heisenberg limit better.

quant-ph↗

arXiv2022

Reasonable thickness determination for implicit porous sheet structure using persistent homology

Porous structures are widely used in various industries because of their excellent properties. Porous surfaces have no thickness and should be thickened to sheet structures for further fabrication. However, conventional methods for generating sheet structures are inefficient for porous surfaces because of the complexity of the internal structures. In this study, we propose a novel method for generating porous sheet structures directly from point clouds sampled on a porous surface. The generated sheet structure is represented by an implicit B-spline function, which ensures smoothness and closure. Moreover, based on the persistent homology theory, the topology structure of the generated porous sheet structure can be controlled, and a reasonable range of the uniform thickness of the sheet structure can be calculated to ensure manufacturability and pore existence. Finally, the implicitly B-spline represented sheet structures are sliced directly with the marching squares algorithm, and the contours can be used for 3D printing. Experimental results show the superiority of the developed method in efficiency over the traditional methods.

cs.CG↗