Sharp Geometric Features Research Articles

IPUNet: Iterative Cross Field Guided Point Cloud Upsampling.

Point clouds acquired by 3D scanning devices are often sparse, noisy, and non-uniform, causing a loss of geometric features. To facilitate the usability of point clouds in downstream applications, given such input, we present a learning-based point upsampling method, i.e., iPUNet, which generates dense and uniform points at arbitrary ratios and better captures sharp features. To generate feature-aware points, we introduce cross fields that are aligned to sharp geometric features by self-supervision to guide point generation. Given cross field defined frames, we enable arbitrary ratio upsampling by learning at each input point a local parameterized surface. The learned surface consumes the neighboring points and 2D tangent plane coordinates as input, and maps onto a continuous surface in 3D where arbitrary ratios of output points can be sampled. To solve the non-uniformity of input points, on top of the cross field guided upsampling, we further introduce an iterative strategy that refines the point distribution by moving sparse points onto the desired continuous 3D surface in each iteration. Within only a few iterations, the sparse points are evenly distributed and their corresponding dense samples are more uniform and better capture geometric features. Through extensive evaluations on diverse scans of objects and scenes, we demonstrate that iPUNet is robust to handle noisy and non-uniformly distributed inputs, and outperforms state-of-the-art point cloud upsampling methods.

Reconstruction of Machine-Made Shapes from Bitmap Sketches

We propose a method of reconstructing 3D machine-made shapes from bitmap sketches by separating an input image into individual patches and jointly optimizing their geometry. We rely on two main observations: (1) human observers interpret sketches of man-made shapes as a collection of simple geometric primitives, and (2) sketch strokes often indicate occlusion contours or sharp ridges between those primitives. Using these main observations we design a system that takes a single bitmap image of a shape, estimates image depth and segmentation into primitives with neural networks, then fits primitives to the predicted depth while determining occlusion contours and aligning intersections with the input drawing via optimization. Unlike previous work, our approach does not require additional input, annotation, or templates, and does not require retraining for a new category of man-made shapes. Our method produces triangular meshes that display sharp geometric features and are suitable for downstream applications, such as editing, rendering, and shading.

3D mesh saliency from local spiral hop descriptors

Mesh saliency, the process of detecting visually important regions in 3D meshes, is a significant component in computer graphics, that can be used in various applications such as denoising and simplification. In this paper, we propose a new 3D mesh saliency measure that can identify sharp geometric features in meshes. A local normal-based descriptor is built for each vertex thanks to a spiral path within a 2-hop neighborhood. First, a geometric-based saliency is computed as the mean local alignment between the spiral descriptors within a 1-hop, and weighted by a vertex roughness measure. Second, a spectral-based saliency is computed from the spectral energy of each vertex structure tensor with the gradient defined from the spiral descriptor alignments. The final saliency is then defined as a weighted sum of both. This single-scale saliency can be extended to a multi-scale saliency by decimating the mesh at several scales and averaging back the obtained saliencies after mapping them between decimated meshes. The approach presents competitive results with state-of-the-art.

DEF

We propose Deep Estimators of Features (DEFs), a learning-based framework for predicting sharp geometric features in sampled 3D shapes. Differently from existing data-driven methods, which reduce this problem to feature classification, we propose to regress a scalar field representing the distance from point samples to the closest feature line on local patches. Our approach is the first that scales to massive point clouds by fusing distance-to-feature estimates obtained on individual patches. We extensively evaluate our approach against related state-of-the-art methods on newly proposed synthetic and real-world 3D CAD model benchmarks. Our approach not only outperforms these (with improvements in Recall and False Positives Rates), but generalizes to real-world scans after training our model on synthetic data and fine-tuning it on a small dataset of scanned data. We demonstrate a downstream application, where we reconstruct an explicit representation of straight and curved sharp feature lines from range scan data. We make code, pre-trained models, and our training and evaluation datasets available at https://github.com/artonson/def.

Rethinking Point Cloud Filtering: A Non-Local Position Based Approach

Existing position based point cloud filtering methods can hardly preserve sharp geometric features. In this paper, we rethink point cloud filtering from a non-learning non-local non-normal perspective, and propose a novel position based approach for feature-preserving point cloud filtering. Unlike normal based techniques, our method does not require the normal information. The core idea is to first design a similarity metric to search the non-local similar patches of a queried local patch. We then map the non-local similar patches into a canonical space and aggregate the non-local information. The aggregated outcome (i.e. coordinate) will be inversely mapped into the original space. Our method is simple yet effective. Extensive experiments validate our method, and show that it generally outperforms position based methods (deep learning and non-learning), and generates better or comparable outcomes to normal based techniques (deep learning and non-learning).

A Robust Rigid Registration Framework of 3D Indoor Scene Point Clouds Based on RGB-D Information

Rigid registration of 3D indoor scenes is a fundamental yet vital task in various fields that include remote sensing (e.g., 3D reconstruction of indoor scenes), photogrammetry measurement, geometry modeling, etc. Nevertheless, state-of-the-art registration approaches still have defects when dealing with low-quality indoor scene point clouds derived from consumer-grade RGB-D sensors. The major challenge is accurately extracting correspondences between a pair of low-quality point clouds when they contain considerable noise, outliers, or weak texture features. To solve the problem, we present a point cloud registration framework in view of RGB-D information. First, we propose a point normal filter for effectively removing noise and simultaneously maintaining sharp geometric features and smooth transition regions. Second, we design a correspondence extraction scheme based on a novel descriptor encoding textural and geometry information, which can robustly establish dense correspondences between a pair of low-quality point clouds. Finally, we propose a point-to-plane registration technology via a nonconvex regularizer, which can further diminish the influence of those false correspondences and produce an exact rigid transformation between a pair of point clouds. Compared to existing state-of-the-art techniques, intensive experimental results demonstrate that our registration framework is excellent visually and numerically, especially for dealing with low-quality indoor scenes.

Urban vegetation segmentation using terrestrial LiDAR point clouds based on point non-local means network

Urban vegetation inventory at city-scale using terrestrial light detection and ranging (LiDAR) point clouds is very challenging due to the large quantity of points, varying local density, and occlusion effects, leading to missing features and incompleteness of data. This paper proposes a novel method, named Point Non-Local Means (PointNLM) network, which incorporates the supervoxel-based and point-wise for automatic semantic segmentation of vegetation from large scale complex scene point clouds. PointNLM captures the long-range relationship between groups of points via a non-local branch cascaded three times to describe sharp geometric features. Simultaneously, a local branch processes the position of scattered feature points and captures the low and high level features. Finally, we proposed a fusion layer of neighborhood max-pooling method to concatenate the long-range features, low level features and high level features for segmenting the trees. The proposed architecture was evaluated on three datasets, including two open access datasets of Semantic3D and Paris-Lille-3D, and an in-house dataset acquired by a commercial mobile LiDAR system. Experimental results indicated that the proposed method provides an efficient and robust result for vegetation segmentation, achieving an Intersection over Union (IoU) of 94.4%, F1-score of 92.7% and overall accuracy of 96.3%, respectively.

Effect of geometric sharpness on translucent material perception.

When judging the optical properties of a translucent object, humans often look at sharp geometric features such as edges and thin parts. An analysis of the physics of light transport shows that these sharp geometries are necessary for scientific imaging systems to be able to accurately measure the underlying material optical properties. In this article, we examine whether human perception of translucency is likewise affected by the presence of sharp geometry, by confounding our perceptual inferences about an object's optical properties. We use physically accurate simulations to create visual stimuli of translucent materials with varying shapes and optical properties under different illuminations. We then use these stimuli in psychophysical experiments, where human observers are asked to match an image of a target object by adjusting the material parameters of a match object with different geometric sharpness, lighting, and three-dimensional geometry. We find that the level of geometric sharpness significantly affects perceived translucency by observers. These findings generalize across a few illumination conditions and object shapes. Our results suggest that the perceived translucency of an object depends on both the underlying material's optical parameters and the three-dimensional shape of the object. We also find that models based on image contrast cannot fully predict the perceptual results.

Robust and Fast 3-D Saliency Mapping for Industrial Modeling Applications

New generation 3-D scanning technologies are expected to create a revolution at the Industry 4.0, facilitating a large number of virtual manufacturing tools and systems. Such applications require the accurate representation of physical objects and/or systems achieved through saliency estimation mechanisms that identify certain areas of the 3-D model, leading to a meaningful and easier to analyze representation of a 3-D object. 3-D saliency mapping is, therefore, guiding the selection of feature locations and is adopted in a large number of low-level 3-D processing applications including denoising, compression, simplification, and registration. In this article, we propose a robust and fast method for creating 3-D saliency maps that accurately identifies sharp and small-scale geometric features in various industrial 3-D models. An extensive experimental study using a large number of 3-D scanned and CAD models verifies the effectiveness of the proposed method as compared to other recent and relevant approaches despite the constraints posed by complex geometry patterns or the presence of noise.

An efficient Poisson solver for complex embedded boundary domains using the multi-grid and fast multipole methods

We present an efficient method to solve the Poisson equation in embedded boundary (EB) domains. The original problem is divided into an inhomogeneous problem without the effects of EB and a homogeneous problem that imposes the effects of EB. The inhomogeneous problem is efficiently solved through a geometric multi-grid (GMG) solver and the homogenous problem is solved through a boundary element method (BEM) utilizing the free space Green's function. Our method is robust and can handle sharp geometric features without any special treatment. Analytical expressions are presented for the boundary and the domain integrals in BEM to reduce the computational cost and integration error relative to numerical quadratures. Furthermore, a fast multipole method (FMM) is employed to evaluate the boundary integrals in BEM and reduce the computational complexity of BEM. Our method inherits the complementary advantages of both GMG and FMM and presents an efficient alternative with linear computational complexity even for problems involving complex geometries. We observe that the overall computational cost is an order of magnitude lower compared with a stand-alone FMM and is similar to that of an ideal GMG solver.

PDE-Based Multidimensional Extrapolation of Scalar Fields over Interfaces with Kinks and High Curvatures

We present a PDE-based approach for the multidimensional extrapolation of smooth scalar quantities across interfaces with kinks and regions of high curvature. Unlike the commonly used method of [T. Aslam, J. Comput. Phys., 193 (2004), pp. 349--355], in which normal derivatives are extrapolated, the proposed approach is based on the extrapolation and weighting of Cartesian derivatives. As a result, second- and third-order accurate extensions in the $L^\infty$ norm are obtained with linear and quadratic extrapolations, respectively, even in the presence of sharp geometric features. The accuracy of the method is demonstrated on a number of examples in two and three spatial dimensions and compared to the approach of [T. Aslam, J. Comput. Phys., 193 (2004), pp. 349--355]. The importance of accurate extrapolation near sharp geometric features is highlighted on an example of solving the diffusion equation on evolving domains.

The Level-Set Method for Multi-Material Wet Etching and Non-Planar Selective Epitaxy

We present numerical methods to enable accurate and robust level-set based simulation of anisotropic wet etching and non-planar epitaxy for semiconductor fabrication. These fabrication techniques are characterized by highly crystal orientation-dependent etch/growth rates, which lead to non-convex Hamiltonians in their description by the level-set equation. As a consequence, instable surface propagation may emerge, leading to unphysical results. We propose a calibration-free Stencil Lax-Friedrichs scheme and an advanced adaptive time-stepping approach, tailored to the level-set speed functions associated with anisotropic etching and epitaxy. The scheme calculates the numerical dissipation based on information about the local geometry and the nature of the etch rates/growth function, which enables an optimized trade-off between overly rounding of sharp geometric features and stable surface propagation. Furthermore, we introduce the deposition top layer method, which allows for robust handling of multiple material regions in non-planar epitaxy simulations. Both methods are demonstrated in a prototypical implementation, which is used to validate the capability and accuracy of our approaches. In particular, two-dimensional wet etching and three-dimensional epitaxy simulations are performed and characteristic geometry parameters are compared to the ideally expected values, showing robustness and high accuracy.

Part geometry and conduction-based laser power control for powder bed fusion additive manufacturing

Laser powder bed fusion (LPBF) uses a focused, high power laser to repeatedly scan geometric patterns on thin layers of metal powder, which build up to a final, solid three-dimensional (3D) part. This process is somewhat limited in that the parts tend to have poorer surface finish (compared to machining or grinding) and distortion due to residual stress, as well as multiple other deficiencies. Typical laser scan strategies are relatively simple and use constant laser power levels. This elicits local variations in the melt pool size, shape, or temperature, particularly near sharp geometric features or overhang structures due to the relatively higher thermal conductivity of solid metal compared to metal powder. In this paper, we present a new laser power control algorithm, which scales the laser power to a value called the geometric conductance factor (GCF). The GCF is calculated based on the amount of solid vs. powder material near the melt pool. The algorithm for calculating GCF is presented along with some basic examples for clarification. Then, we detail the hardware and software implementation on the National Institute of Standards and Technology (NIST) additive manufacturing metrology testbed (AMMT), which includes co-axial melt pool monitoring using a high-speed camera. Six parts were fabricated out of nickel superalloy 625 (IN625) with the same nominal laser power, but with varying GCF algorithm parameters. We demonstrate the effect of tailored laser power on reducing the variability of melt pool intensity measured throughout the 3D build. Finally, we contrast the difference between the ‘optimized’ part vs. the standard build parameters, including the deflection of the final part top surface near the overhang and the variation of surface finish on the down-facing surfaces. Ultimately, the improvements to the in-situ process monitoring and part qualities demonstrate the utility and future potential tuning and optimizing more complex laser scan strategies.

Robust Three-Dimensional Level-Set Method for Evolving Fronts on Complex Unstructured Meshes

With a purpose to evolve the surfaces of complex geometries in their normal direction at arbitrarily defined velocities, we have developed a robust level-set approach which runs on three-dimensional unstructured meshes. The approach is built on the basis of an innovative spatial discretization and corresponding gradient-estimating approach. The numerical consistency of the estimating method is mathematically proven. A correction technology is utilized to improve accuracy near sharp geometric features. Validation tests show that the proposed approach is able to accurately handle geometries containing sharp features, computation regions having irregular shapes, discontinuous speed fields, and topological changes. Results of the test problems fit well with the reference results produced by analytical or other numerical methods and converge to reference results as the meshes refine. Compared to level-set method implementations on Cartesian meshes, the proposed approach makes it easier to describe jump boundary conditions and to perform coupling simulations.

Eulerian-on-lagrangian cloth simulation

We resolve the longstanding problem of simulating the contact-mediated interaction of cloth and sharp geometric features by introducing an Eulerian-on-Lagrangian (EOL) approach to cloth simulation. Unlike traditional Lagrangian approaches to cloth simulation, our EOL approach permits bending exactly at and sliding over sharp edges, avoiding parasitic locking caused by over-constraining contact constraints. Wherever the cloth is in contact with sharp features, we insert EOL vertices into the cloth, while the rest of the cloth is simulated in the standard Lagrangian fashion. Our algorithm manifests as new equations of motion for EOL vertices, a contact-conforming remesher, and a set of simple constraint assignment rules, all of which can be incorporated into existing state-of-the-art cloth simulators to enable smooth, inequality-constrained contact between cloth and objects in the world.

Shape and Size Control of Substrate-Grown Gold Nanoparticles for Surface-Enhanced Raman Spectroscopy Detection of Chemical Analytes

Anisotropic gold nanoparticles often exhibit superior optical properties compared to spherical ones, in part due to intense electric field localization near sharp geometric features and a broadly tunable localized surface plasmon resonance. As a result, anisotropic nanoparticles are attractive building blocks for surface-enhanced Raman spectroscopy (SERS) substrates. To unlock the full potential of such substrates, one should be able to (1) generate a sufficient number of SERS hotspots with structures of controlled shape and size and (2) remove ligands so that analytes can easily access nanoparticle surface sites. Here, we develop a synthetic strategy for the shape- and size-controlled anisotropic growth of gold nanoparticles (concave rhombic dodecahedra and concave cubes, 70–135 nm characteristic length) from spherical seeds anchored on a structurally complex surface (common filter paper) and subsequently combine electrodynamics and electronic structure calculations with experiment to systematically char...

Maskless 3D Ablation of Precise Microhole Structures in Plastics Using Femtosecond Laser Pulses.

Femtosecond laser ablation is a robust tool for the fabrication of microhole structures. This technique has several advantages compared to other microfabrication strategies for reliably preparing microhole structures of high quality and low cost. However, few studies have explored the use of femtosecond laser ablation in plastic materials because of the lack of controllability over the fabrication process in plastics. In particular, the depth profile of microhole structures prepared by conventional laser ablation techniques in plastics cannot be precisely and reproducibly controlled. In this paper, a novel three-dimensional femtosecond laser ablation technique was developed for the rapid fabrication of precise microhole structures in multiple plastics in air. Using a three-step fabrication scheme, microholes demonstrated extremely clean and sharp geometric features. This new technique also enables the precise creation of arbitrary-shaped microwell structures in plastic substrates through a rapid single-step ablation process, without the need for any masks. As a proof of concept for practical applications, precise microhole structures prepared by this novel femtosecond laser ablation technique were exploited for robust resistive-pulse sensing of microparticles.

A high‐order element based adaptive mesh refinement strategy for three‐dimensional unstructured grid

SummaryAdaptive mesh refinement (AMR) shows attractive properties in automatically refining the flow region of interest, and with AMR, better prediction can be obtained with much less labor work and cost compared to manually remeshing or the global mesh refinement. Cartesian AMR is well established; however, AMR on hybrid unstructured mesh, which is heavily used in the high‐Reynolds number flow simulation, is less matured and existing methods may result in degraded mesh quality, which mostly happens in the boundary layer or near the sharp geometric features. User intervention or additional constraints, such as freezing all boundary layer elements or refining the whole boundary layer, are required to assist the refinement process. In this work, a novel AMR strategy is developed to handle existing difficulties. In the new method, high‐order unstructured elements are first generated based on the baseline mesh; then the refinement is conducted in the parametric space; at last, the mesh suitable for the solver is output. Generating refined elements in the parametric space with high‐order elements is the key of this method and this helps to guarantee both the accuracy and robustness. With the current method, 3‐dimensional hybrid unstructured mesh of huge size and complex geometry can be automatically refined, without user intervention nor additional constraints. With test cases including the 2‐dimensional airfoil and 3‐dimensional full aircraft, the current AMR method proves to be accurate, simple, and robust.

Smart octrees: Accurately integrating discontinuous functions in 3D

This paper presents an efficient and accurate method for the integration of discontinuous functions on a background mesh in three dimensions. This task is important in computational mechanics applications where internal interfaces are present in the computational domain. The proposed method creates boundary-conforming integration subcells for composed numerical quadrature, even in the presence of sharp geometric features (e.g. edges or vertices). Similar to the octree procedure, the algorithm subdivides the cut elements into eight octants. However, the octant nodes are moved onto the interface, which allows for a robust resolution of the intersection topology in the element while maintaining algorithmic simplicity. Numerical examples demonstrate that the method is able to deliver highly accurate domain integrals with a minimal number of quadrature points. Further examples show that the proposed method provides a viable alternative to standard octree-based approaches when combined with the Finite Cell Method.

Simulating sharp geometric features in six degrees-of-freedom haptic rendering.

It is a challenging problem to achieve six degree-of-freedom (DoF) haptic simulation of subtle force feelings caused by contacts at sharp geometric features in multi-region contact scenarios. We design a novel sphere-tree model for objects with sharp features and use a configuration-based optimization method to compute feedback force and torque. Given a triangle mesh of an object, a sphere-tree model is created based on dihedral angles between pairs of adjacent triangles. The model consists of a hierarchical sphere-tree for global shape and a linked-list of spheres for local areas with sharp features. In each local area with a sharp edge, we first identify those spheres with radii greater than an upper limit determined by the dihedral angle of the edge. Those spheres are further divided into a linear list of smaller spheres by a splitting method. The experiment results from a cylinder-cube interaction validate that the proposed method can simulate subtle force direction changes when an object slides across sharp edges. Perception-based experiments and a haptic-to-vision shape matching task are also used to compare the performance between our proposed method and other rendering methods. The comparison results show that our method is more effective in simulating sharp features both in terms of measured force signals and human subjective evaluation. Non-penetration among objects is maintained for multi-region contact scenarios. The haptic rendering rate is about 1kHz, and the subtle force feeling of sliding along sharp features can be stably simulated.