Elsevier

Computers & Graphics

Volume 56, May 2016, Pages 11-19
Computers & Graphics

Special Section on SIGRAD 2015
Complex multi-material approach for dynamic simulations

https://doi.org/10.1016/j.cag.2016.02.001Get rights and content

Highlights

  • We propose a method for the definition of multiple materials.

  • The method is convenient for dynamic simulations.

  • The method utilizes binary space partitions (BSP).

  • An automated method for the generation of BSP is proposed.

Abstract

Triangle meshes are the most common representation of an object in the field of computer graphics. Recently, they have found their way into the field of erosion simulation, where volumetric representation used to prevail. Real-life erosion scenes are usually formed of multiple materials and so a reliable means of material definition is needed. Unfortunately, coupling the material information with a triangle mesh is not as straightforward as in the volumetric case. This paper proposes an approach for multiple material definition based on space subdivision. Binary space partitioning (BSP) is used to simulate complex multi-material scenes. The approach allows the definition of a nontrivial scene composed of several materials, including the definition of a gradually changing material. A method for an automated creation of the BSP tree from input volumetric data is proposed. The construction algorithm extracts a triangle mesh as an intermediate product and uses its faces as the splitting planes of the BSP tree.

Introduction

When representing a heterogeneous object or scene in 3D graphics simulations, multiple materials can be assigned to it to achieve better results or visual quality. Material properties can describe its visual appearance or define its structure, which is necessary, e.g., for applications working with haptic feedback. The way the material properties can be assigned depends greatly on the data structure used for the representation of the object.

A heterogeneous scene is commonly represented as a set of non-intersecting triangle meshes, one for each material present in the scene. This representation is efficient for static scenes with several separated materials; however, it may not be sufficient to describe a dynamic scene containing several materials blending into each other, e.g., during an erosion simulation. During such a dynamic simulation, the surface mesh evolves and a more sophisticated way to describe the material properties is necessary. A volumetric approach is another common method of addressing the problem. This approach gives satisfactory results in many situations, but the memory requirements grow very fast with the increasing size of the scene.

As triangle meshes are the most commonly used data structure in 3D graphics, a multiple-material mesh-based approach for dynamic simulations is needed. In this paper, a material description approach suitable for such purposes is proposed. The method is an extension of the approach by Skorkovská and Kolingerová [1]. In [1], the scene is represented with a surface mesh for each individual object in the scene, regardless of the material of the object. The material is then assigned through a separate binary space partitioning (BSP) structure. As the BSP tree is independent of the triangle meshes, it can be constructed once in the preprocessing step and used throughout the simulation. The method also allows an optional definition of a distance function as a simulation of a continuous change of the material.

The main contributions of this paper are

  • An extension of the BSP method [1] to allow general implicit surfaces as splitting functions in the BSP tree;

  • An automated method for the construction of the BSP tree from an input volumetric data to eliminate the need for the manual definition of the BSP tree.

These extensions of the original method allow much wider range of uses in the field of dynamic simulations.

The structure of the paper is as follows. Section 2 summarizes the related work. In Section 3, several simple multiple-material definition approaches are proposed. Section 4 describes a more complex approach using BSP. In Section 5, an automated method for the generation of a BSP tree is proposed. Results are presented in Section 6 and Section 7 concludes the paper.

Section snippets

Related work

Most of the research on multiple-material scenes focuses on extracting a correct and consistent mesh for each of the materials present in the volumetric data obtained, e.g., from a medical scan. Wu and Sullivan [2] enhanced the marching cubes algorithm to reconstruct multiple material meshes. Zhang et al. [3] generate the mesh using an octree-based isocontouring method. Wang [4] generates the mesh surfaces using a ray representation of a solid as an intermediate structure. These approaches are

Definition of multiple materials

A real-life scene is composed of objects made of different materials which are eroded in different ways; hard and resistant materials are eroded slowly, while the erosion of soft materials is happening much faster. To be able to simulate such phenomena, we need a means to consistently describe the material of an object. A common way of representation is to have a separate mesh for each material present in the scene. This approach is suitable for the simulation of static scenes, where the

Binary space partitioning

The methods of multiple-material definition described in the previous section can be used to describe the material distribution of a simple scene containing only two distinct materials. To simulate more complex distribution of materials, a more sophisticated approach is needed. We are using a binary space partitioning (BSP) [12] approach that can be seen as an extension of the method described in Section 3.2, allowing the splitting up of the scene into more parts using multiple division planes.

Automated generation of a BSP tree

The binary space subdivision as described in Section 4 can be used as a memory-efficient means of material distribution description in a scene. However, for big and complex scenes, the manual definition of splitting functions would be complicated and time-consuming.

We have proposed an automated method for the generation of the BSP tree. Typical material distribution information comes in the form of volumetric data. To construct the BSP tree, a triangle mesh needs to be extracted from the

Splitting planes

We have tested the proposed approach in various scenarios to test its applicability to the problem of material spatial distribution. Fig. 10 shows an example of the basic BSP material definition. Fig. 10(a) shows an initial un-eroded model; the same initial model was used in all the erosion simulations presented in Fig. 10. The scene has been divided into 16 vertical spatial cells; each one has been assigned a unique random material. The darker color of the mesh marks the regions made of a

Conclusion

Erosion simulation is a very important topic of computer graphics. More and more often, triangle meshes are used to represent the eroded terrain. A means to simulate the erosion of scenes composed of various materials is needed, but the commonly used approach of integrating the material properties with the vertices of the mesh may not be appropriate. Volumetric approaches are capable of storing the necessary information at the cost of high memory requirements.

A multiple material definition

Acknowledgements

This work has been supported by the European Regional Development Fund (ERDF) – project NTIS (New Technologies for the Information Society), The European Centre of Excellence, CZ.1.05/1.1.00/0.2.0090, by the project SGS-2013-029 – Advanced Computing and Information Systems and by the project SGS-2016-013 – Advanced Graphical and Computing Systems.

References (16)

There are more references available in the full text version of this article.

Cited by (0)

View full text