WO2020038981A1 - Procédé de réduction d'un modèle 3d - Google Patents

Procédé de réduction d'un modèle 3d Download PDF

Info

Publication number
WO2020038981A1
WO2020038981A1 PCT/EP2019/072339 EP2019072339W WO2020038981A1 WO 2020038981 A1 WO2020038981 A1 WO 2020038981A1 EP 2019072339 W EP2019072339 W EP 2019072339W WO 2020038981 A1 WO2020038981 A1 WO 2020038981A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
reduced
computing unit
reduction
different
Prior art date
Application number
PCT/EP2019/072339
Other languages
German (de)
English (en)
Inventor
Rebecca Johnson
Original Assignee
Siemens Mobility GmbH
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Siemens Mobility GmbH filed Critical Siemens Mobility GmbH
Publication of WO2020038981A1 publication Critical patent/WO2020038981A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
    • G06T17/205—Re-meshing

Definitions

  • a polygon reduction algorithm causes more visual changes compared to the original mesh when applied to a sphere than when applied to a cube with the same parameters.
  • the respective reduction algorithms often have one or more parameters for which values must be defined before the reduction algorithm can be executed.
  • the Defining these values is also referred to as parameterization; parameterization is therefore a set of values.
  • the parameterization has a major impact on the quality of the reduced 3D model.
  • the present invention is intended to provide an alternative to the prior art.
  • a graphics processing unit renders first images of different views of the 3D model
  • a processing unit executes a reduction algorithm and generates a reduced 3D model from the 3D model, which Graphics processing unit renders second images of the different views of the reduced 3D model
  • a computing unit calculates a similarity value for each second image, which quantifies a similarity between the second image and the corresponding first image
  • a computing unit a qualification from all similarity values Actual value calculated, which quantifies a visual quality of the reduced 3D model.
  • the computing unit is, for example, a microprocessor or microcontroller, a workstation, a server, a computer network or a cloud. It is also possible for several of the computing units mentioned to carry out the same or different calculation steps in parallel.
  • the method uses the comparison of the images to quantify the visual quality of the reduced 3D model.
  • the quality value allows, as a numerical measure, to automate the parameterization of the reduction algorithm and to explore different reduction algorithms, as is described in more detail in the further training.
  • the quality value creates a quantitative measure that allows the visual quality of the reduced 3D model to be calculated and evaluated in a fully automated mechanical process. This enables the system to automatically generate different, reduced 3D models and select the best one.
  • the computing unit executes a plurality of reduction algorithms and each generates a reduced 3D model.
  • the computing unit calculates a quality value for each reduced 3D model.
  • the computing unit executes the reduction algorithm with different parameterizations and in each case generates a reduced 3D model.
  • the computing unit calculates a quality value for each reduced 3D model.
  • the computing unit executes a plurality of reduction algorithms, each with different parameterizations, and in each case generates a reduced ornate 3D model.
  • the computing unit calculates a quality value for each reduced 3D model.
  • the computing unit explores and / or optimizes the parameterization of the reduction algorithm while observing the quality value and / or a file size by means of an optimization method.
  • the computing unit selects a reduced model for output to a user which does not exceed a threshold value for a file size and / or does not fall below a threshold value for the visual quality.
  • the computing unit executes several different reduction algorithms in succession in order to generate the reduced model.
  • the computing unit executes a reduction algorithm iteratively with the same or different parameterization in order to generate the reduced model.
  • the time spent manually searching for the best reduction algorithm is saved. Among other things, it is no longer necessary for humans to wait for the result of the execution of each reduction algorithm and to evaluate it visually. The process is also accelerated This is not the case if the calculations are carried out by high-performance hardware, such as dedicated graphics hardware for rendering the images and a cloud infrastructure for image comparison and the execution of the reduction algorithms, which of course can also run in parallel on different processor cores.
  • high-performance hardware such as dedicated graphics hardware for rendering the images and a cloud infrastructure for image comparison and the execution of the reduction algorithms, which of course can also run in parallel on different processor cores.
  • a computer program is stored on the computer-readable data carrier and executes the method when it is processed in a processor.
  • the computer program is processed in a processor and executes the method.
  • Fig. 1 is a flow chart for the reduction of a 3D model.
  • FIG. 1 shows a flowchart for the automated reduction of a 3D model M.
  • a graphics processing unit first renders first images M_1 ... M_N of different views of the 3D model M, for example from all six directions in parallel with the three spatial axes.
  • the graphics processing unit is, for example, a dedicated graphics processor, a graphics card, a dedicated graphics computer or a cloud.
  • a suitably programmed conventional processor can also be used as the graphics processing unit.
  • the graphics processing unit performs, for example, a ray tracing or radiosity algorithm or a hardware-supported shading method, which can also take textures into account.
  • a computing unit executes a reduction algorithm RA and generates a reduced 3D model RM from the 3D model M.
  • the computing unit is, for example, a microprocessor or microcontroller Workstation, server, computer network or one
  • the graphics processing unit then renders second images RM_1 ... RM_N of the reduced 3D model RM from exactly the same views as before.
  • a computing unit For every second picture RM_1 ... RM_N, a computing unit - the same as before or another - calculates a similarity value AW_1 ... AW_N, which corresponds to a similarity between the second picture RM_1 ... RM_N and the first Image M_1 ... M_N quantified.
  • the computing unit implements suitable programming, for example a neural network NN, which compares the images with one another.
  • the arithmetic unit can also implement the metric for image comparison known from the prior art mentioned at the outset or another suitable algorithm for image comparison.
  • an arithmetic unit - the same as before or another - calculates a quality value QW from all the similarity values AW_1 ... AW_N, which quantifies a visual quality of the reduced 3D model RM.
  • the reduction algorithm RA can be selected from a list of reduction algorithms. In this way, the quality value QW can also be calculated for different reduction algorithms, which allows the most suitable reduction algorithm to be determined.
  • a parameterization is generated, that is to say a set of values for the parameters of the reduction algorithm RA.
  • a quality value can also be calculated for different parameterizations, whereby the most suitable parameterization can be determined.
  • the parameterization consists of only one value.
  • the best reduction algorithm RA and the best parameterization can be determined as part of an optimization process OV.
  • the optimization process can be iterated until a specified termination criterion is reached.
  • the reduced 3D model RM is selected, which has the best properties, for example the smallest possible file size with a quality value QW, which does not fall below a predetermined threshold value.
  • the termination criterion can consist in the fact that no further parameterizations can be generated for the selected reduction algorithm RA, that a predetermined period of time has been exhausted, that a predetermined number of calculations have been carried out, or that a predetermined threshold value for the quality value QW has been undershot.
  • the optimization process OV begins with parameterization, which only brings about a slight reduction, and increases the reduction step by step until a predetermined threshold value for the quality value QW is undershot.
  • the reduction can be canceled as soon as the quality value QW, which can be interpreted as a visual resemblance to the original, falls below 98%.
  • the reduced 3D model RM is output, which was calculated in the penultimate iteration and still has a quality value QW of over 98%.
  • the parameter space can be discretized and systematically explored to generate new parameterizations. Alternatively, a probabilistic sample-based approach can be followed.
  • a numerical value is calculated using the similarity values AW_1 ... AW_N. If the respective pair from the first picture M l ... M_N and second image RM_1 ... RM_N has a high similarity, the respective similarity value turns out to be larger.
  • the quality value QW can be calculated, for example, from the sum of the similarity values AW_1 ... AW_N.
  • the optimization process OV runs here on a level that is superior to the individual reduction algorithms or their parameterizations, since the reduced 3D model, that is to say the end result of the reduction algorithm RA, is evaluated with the quality value QW.
  • the optimization procedure OV selects different reduction algorithms or different parameterizations for the creation of the reduced 3D model RM for one and the same 3D model M and explores which quality values QW are achieved in each case.
  • the optimization method OV can observe whether incremental changes in the parameterization improve or worsen the quality value QW of the reduced 3D model RM, or reduce or enlarge the file size of the reduced 3D model RM until a local optimum is achieved.
  • the OV optimization process can also make major random adjustments to the parameterization in order to test whether a better local optimum can be found.

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Computer Graphics (AREA)
  • Geometry (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Processing Or Creating Images (AREA)

Abstract

Pour la réduction d'un modèle 3D (M), des premières images (M_1...M_N) de différentes vues du modèle 3D sont reproduites avant qu'une unité de calcul exécute un algorithme de réduction (RA) et génère à partir du modèle 3D un modèle 3D réduit (RM). Ensuite, des secondes images (RM_1...RM_N) des différentes vues du modèle 3D réduit sont reproduites. Puis, pour chaque seconde image est calculée une valeur de similarité (AW_1...AW_N) qui quantifie une similarité entre la seconde image et la première image correspondante. Finalement, à partir de toutes les valeurs de similarité est calculée une valeur de qualité (QW) qui quantifie une qualité visuelle du modèle 3D réduit. La valeur de qualité, en tant que cote numérique, permet d'automatiser le paramétrage de l'algorithme de réduction ainsi que d'explorer des algorithmes de réduction différents, comme cela est décrit plus précisément dans les modes de réalisation perfectionnés de la présente invention. Ce procédé d'optimisation (OV) permet de générer automatiquement différents modèles 3D réduits et d'en sélectionner le meilleur. Par ailleurs, aucune connaissance d'expert n'est requise puisque le procédé se déroule sans surveillance. La présente invention permet d'économiser le temps de travail pour la recherche manuelle du meilleur algorithme de réduction.
PCT/EP2019/072339 2018-08-23 2019-08-21 Procédé de réduction d'un modèle 3d WO2020038981A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
DE102018214247.5A DE102018214247A1 (de) 2018-08-23 2018-08-23 Verfahren zur Reduktion eines 3D-Modells
DE102018214247.5 2018-08-23

Publications (1)

Publication Number Publication Date
WO2020038981A1 true WO2020038981A1 (fr) 2020-02-27

Family

ID=67953736

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2019/072339 WO2020038981A1 (fr) 2018-08-23 2019-08-21 Procédé de réduction d'un modèle 3d

Country Status (2)

Country Link
DE (1) DE102018214247A1 (fr)
WO (1) WO2020038981A1 (fr)

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140111510A1 (en) * 2012-10-19 2014-04-24 Donya Labs Ab Method for optimized polygon reduction of computer graphics

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
P LINDSTROM ET AL: "Image-driven mesh optimization", 1 June 2000 (2000-06-01), XP055109036, Retrieved from the Internet <URL:http://www.osti.gov/servlets/purl/15005455-wEoRJM/native/> [retrieved on 20140319] *
ROB ROTHFARB: "Mixing Realities to Connect People, Places, and Exhibits Using Mobile Augmented-Reality Applications", MUSEUMS AND THE WEB 2011: PROCEEDINGS, 31 March 2011 (2011-03-31), pages 1 - 15, XP055277153 *

Also Published As

Publication number Publication date
DE102018214247A1 (de) 2020-02-27

Similar Documents

Publication Publication Date Title
EP3657440B1 (fr) Procédé et système de reconstruction tridimensionnelle d&#39;une tête humaine à partir d&#39;une pluralité d&#39;images
DE112018006047T5 (de) Deformation von aufwandsfunktionen bei der quanten-näherungsoptimierung
DE202010017612U1 (de) Verbesserter genetischer Algorithmus mit mehreren Zielen für die Bauteildesignoptimierung
EP3336800B1 (fr) Détermination d&#39;une fonction d&#39;apprentissage destinée à produire des images d&#39;apprentissage annotées
DE112018004693T5 (de) Verbessern der effizienz eines neuronalen netzes
DE112018002822T5 (de) Klassifizieren neuronaler netze
DE102011016347A1 (de) Hierarchisches Umgrenzen von verschobenen parametrischen Flächen
DE102005010169A1 (de) Aktives Polyeder für 3D-Bildsegmentierung
DE10144932A1 (de) Visualisierung von Werkstücken bei der Simulation von Fräsprozessen
Hel-Or et al. Relaxed parametric design with probabilistic constraints
EP3355214A1 (fr) Procédé, ordinateur et programme informatique destinés à préparer un modèle de bord de monture
EP1529249B1 (fr) Procede et ensemble permettant la conception d&#39;un systeme technique
EP3188053A1 (fr) Procede de configuration d&#39;une co-simulation pour systeme entier
DE4326487C2 (de) Adaptives Filter
DE202016009145U1 (de) Deep-Learning-basierte funktionale Korrelation von volumetrischen Designs
EP1437685A2 (fr) Procédé pour la segmentation de structures tridimensionelles
DE112020005732T5 (de) Erzeugen von trainingsdaten zur objekterkennung
WO2020064761A1 (fr) Procédé, produit-programme informatique et dispositif informatique pour l&#39;évaluation de modèles de subdivision volumétrique
EP1869620B1 (fr) Procede de traitement de donnees-image et de donnees volumiques sur la base de modeles statistiques
WO2020038981A1 (fr) Procédé de réduction d&#39;un modèle 3d
WO2006061185A1 (fr) Procede pour realiser des dessins techniques a partir de modeles tridimensionnels comportant au moins deux elements tridimensionnels en collision
DE112014002120T5 (de) System und Verfahren zur Durchführung von Operationen zum Rückgängigmachen und Wiederherstellen während der Bearbeitungssimulation
DE10023377C2 (de) Verfahren zur Erhöhung der Leistungsfähigkeit einer Computereinrichtung bei Finite-Elemente-Simulationen und eine solche Computereinrichtung
EP2490181A1 (fr) Procédé et dispositif de reconstruction d&#39;objets 3D à partir de nuages de points
WO2006128421A1 (fr) Procede pour caracteriser des objets

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19768719

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19768719

Country of ref document: EP

Kind code of ref document: A1