563 research outputs found
Visual Computing in Materials Sciences (Dagstuhl Seminar 19151)
No full textVisual computing has become highly attractive for boosting research endeavors in the materials science domain. Using visual computing, a multitude of different phenomena may now be studied, at various scales, dimensions, or using different modalities. This was simply impossible before. Visual computing techniques generate novel insights to understand, discover, design, and use complex material systems of interest. Its huge potential for retrieving and visualizing (new) information on materials, their characteristics and interrelations as well as on simulating the material's behavior in its target application environment is of core relevance to material scientists. This Dagstuhl seminar on Visual Computing in Materials Sciences thus focuses on the intersection of both domains to guide research endeavors in this field. It targets to provide answers regarding the following four challenges, which are of imminent need:
-The Integrated Visual Analysis Challeng identifies standard visualization tools as insufficient for exploring materials science data in detail. What is required are integrated visual analysis tools, which are tailored to a specific application area and guide users in their investigations. Using linked views and other interaction concepts, these tools are required to combine all data domains using meaningful and easy to understand visualization techniques. Especially for the analysis of spatial and temporal data in dynamic processes (e.g., materials tested under load or in different environmental conditions) or multimodal, multiscale data, these tools and techniques are highly anticipated. Only integrated analysis concepts allow to make the most out of all the data available.
- The Quantitative Data Visualization Challenge centers around the design and implementation of tailored visual analysis systems for extracting and analyzing derived data (e.g., computed from extracted features over spatial, temporal or even higher dimensional domains). Therefore, feature extraction and quantification techniques, segmentation techniques, or clustering techniques, are required as prerequisites for the targeted visual analysis. As the quantification may easily end up in 25 or more properties to be computed per feature, clustering techniques allow to distinguish features of interest into feature classes. These feature classes may then be statistically evaluated to visualize the properties of the individual features as well as the properties of the different classes. Information visualization techniques will be of special interest for solving this challenge.
- The Visual Debugger Challenge is an idea which uses visual analysis to remove errors in the parametrization of a simulation or a data acquisition process. Similarly, to a debugger in computer programming, identifying errors in the code and providing hints to improve, a visual debugger in the domain of visual computing for materials science should show the following characteristics: It should indicate errors and identify wrongly used algorithms in the data analysis. Such a tool should also identify incorrect parameters, which either show no or very limited benefit or even provide erroneous results. Furthermore, it should give directions on how to improve a targeted analysis and suggest suitable algorithms or pipelines for specific tasks.
- The Interactive Steering Challenge uses visual analysis tools to control a running simulation or an ongoing data acquisition process. Respective tools monitor costly processes and give directions to improve results regarding the respective targets. For example, in the material analysis domain, this could be a system which provides settings for improved data acquisition based on the current image quality achieved: If the image quality does no more fulfill the target requirements, the system influences all degrees of freedom in the data acquisition to enhance image quality. The same holds for the materials simulation domain. Visual analysis can help to steer target material properties in a specific application environment by predicting tendencies of costly simulation runs, e.g., using cheaper surrogate models
Learning Multiple-Scattering Solutions for Sphere-Tracing of Volumetric Subsurface Effects
No full textCode for reproducing the results in "Learning Multiple-Scattering Solutions for Sphere-Tracing of Volumetric Subsurface Effects", Leonard, L., Höhlein, K. and Westermann, R. (Eurographics, 2021
khoehlein/Permutation-invariant-Postprocessing: Preliminary code release with DOI batch
No full textCode for paper "Postprocessing of Ensemble Weather Forecasts Using Permutation-invariant Neural Networks" by Kevin Höhlein, Benedikt Schulz, Rüdiger Westermann and Sebastian Lerch
Code "Postprocessing of Ensemble Weather Forecasts Using Permutation-invariant Neural Networks"
No full textCode for paper "Postprocessing of Ensemble Weather Forecasts Using Permutation-invariant Neural Networks" by Kevin Höhlein, Benedikt Schulz, Rüdiger Westermann and Sebastian Lerch
Nutrient management guidelines for Russet Burbank potatoes
No full textBulletin no. 840 Moscow, Idaho :University of Idaho, College of Agriculture, Cooperative Extension System, 2004-10-01. Author(s): Stark, Jeff; Westermann, Dale; Hopkins, Bryan
A Globally Conforming Lattice Structure for 2D Stress Tensor Visualization
No full textWe present a visualization technique for 2D stress tensor fields based on the construction of a globally conforming lattice. Conformity ensures that the lattice edges follow the principal stress directions and the aspect ratio of lattice elements represents the stress anisotropy. Since such a lattice structure cannot be space-filling in general, it is constructed from multiple intersecting lattice beams. Conformity at beam intersections is ensured via a constrained optimization problem, by computing the aspect ratio of elements at intersections so that their edges meet when continued along the principal stress lines. In combination with a coloring scheme that encodes relative stress magnitudes, a global visualization is achieved. By introducing additional constraints on the positional variation of the beam intersections, coherent visualizations are achieved when external loads or material parameters are changed. In a number of experiments using non-trivial scenarios, we demonstrate the capability of the proposed visualization technique to show the global and local structure of a given stress field.Accepted Author ManuscriptMaterials and Manufacturin
Echtzeit-Simulation und -Visualisierung deformierbarer Körper
No full textIn this thesis, I present a framework for physical simulation and visualization of deformable volumetric bodies in real time. Based on the implicit finite element method a multigrid approach for the efficient numerical simulation of elastic materials has been developed. Due to the optimized implementation of the multigrid scheme, 200,000 elements can be simulated at a rate of 10 time steps per second. The approach enables realistic and numerically stable simulation of bodies that are described by tetrahedral or hexahedral grids. It can efficiently simulate heterogeneous bodies---i.e., bodies that are composed of material with varying stiffness---and includes linear as well as non-linear material laws.
To visualize deformable bodies, a novel rendering method has been developed on programmable graphics hardware. It includes the efficient rendering of surfaces as well as interior volumetric structures. Both the physical simulation framework and the rendering approach have been integrated into a single simulation support system. Thereby, available communication bandwidths have been efficiently exploited. To enable the use of the system in practical applications, a novel approach for collision detection has been included. This approach allows one to handle geometries that are deformed on the graphical subsystem.In dieser Arbeit präsentiere ich ein Framework für die physikalische Simulation und Visualisierung von deformierbaren volumetrischen Körpern in Echtzeit. Basierend auf der Methode der impliziten finiten Elemente wurde ein Mehrgitteransatz zur effizienten numerischen Simulation elastischer Materialien entwickelt. Durch die optimierte Implementierung des Mehrgitterverfahrens können 200.000 Elemente mit einer Rate von 10 Zeitschritten pro Sekunden simuliert werden. Der Ansatz ermöglicht die Simulation von heterogenen Körpern und berücksichtigt lineare sowie nicht-lineare Materialgesetze.
Das Simulationssystem wurde um eine neuartige Rendering Methode zur Visualisierung deformierbarer Körper - einschließlich interner volumetrischer Strukturen - auf programmierbarer Graphikhardware erweitert. Ein neuer Ansatz zur Kollisionserkennung ermöglicht die Handhabung von Gittern, die auf dem grafischen Subsystem deformiert werden
Interactive Rendering of Medical Volumetric Data in Extended Reality
No full textTomographic imaging techniques like computed tomography (CT) or magnetic resonance imaging (MRI) play an increasingly important role not only for medical diagnosis, but also for surgery planning, patient-doctor communication, medical teaching and forensics. Physically-based volumetric path tracing turns the resulting slice image stack into photo-realistic images of the insides of the human body, enabling non-radiologists like surgeons and even medically inexperienced users like patients, students or legal experts to make use of those imaging modalities. The benefit of three-dimensional renderings of medical data can be further increased by presenting them to the viewer in extended reality (XR), thus either fully immersing the user into data exploration (for virtual reality, VR) or enhancing objects from the physical world, e.g. the skin of a patient, with digital content (for augmented reality, AR). However, for an XR application to be useful for the viewer, it has to be interactive in two ways: The user needs to be provided both with intuitive interaction methods to explore the medical data and with interactive frame rates to experience the effects of those interactions in real-time.
As a step towards the first goal (intuitive interaction with medical volumetric data), we present a novel projection-based AR approach to surgery planning by the example of a specific surgical intervention, DIEP flap breast reconstruction. We propose a view-dependent volumetric rendering overlay for the abdominal skin of the patient and a controller-driven, two-step point selection method for finding and labeling internal points of interest directly in situ on the patient. As the labeling of patient-specific landmark points on the skin serving as a preoperative planning step therefore does not depend on classical viewing software for CT/MRI image stacks and the user's spatial imagination anymore, the resulting labels exhibit greatly improved accuracy and the method can easily be utilized by medical laypersons, as we show in a pilot user study.
For rendering photo-realistic images of medical volumetric data from tomography using physically-based Monte Carlo path tracing, a large number of light paths needs to be traced in order to reach a smooth, converged image without Monte Carlo noise. This renders interactivity infeasible as only a single path per pixel and frame can be traced in real-time. As a step towards the second goal (interactivity in terms of high frame rates), we therefore propose Adaptive Temporal Sampling to reduce noise in single-sample path tracing, thus bringing physically-based volume rendering to the interactive domain. To this end, we transfer the idea of temporal antialiasing to volume rendering and enable temporal reprojection for volumetric instead of surface data and Monte Carlo instead of shader noise by introducing a representative depth for each pixel, a combined depth and direction check, an improved weighting scheme for arithmetic averaging and an error accumulation method to downweight less appropriate older samples as ghosting countermeasures from temporal antialiasing cannot be applied. Furthermore, we adaptively determine the number of newly generated samples for each pixel using reprojection information, thus dedicating more computational effort to noisier regions. Building upon this original method, we propose an additional learning-based post-processing pass to eradicate remaining noise. We reach interactive frame rates similar to the ones observed in noisy single-sample rendering and nearly noise-free images in an exemplary VR application.
Using the proposed methods both for intuitive interaction (projection-based surgery planning prototype) and for interactivity through drastically reduced frame times (Adaptive Temporal Sampling with learning-based extension), we open the door for a more extensive use of interactive XR-based volume rendering applications for non-radiologists and all kinds of medically inexperienced users
- …
