Research Overview
Fine Digital Engineering
Mission:
conduct quality research in the filed of 3D digital engineering which can be disseminated to change the future industry, and to develop competent human resources through research-based education.
Vision:
implement innovative design/manufacturing processes by information technologies and tools to make the industry "IT-armed" to win the global competition.
Approach: 3D Discrete Geometric Modeling
Proliferation of 3D digital modeling tools all over the design/manufacturing activities based on discrete geometric modeling technologies.
See also Our research flyers.
Convergence Engineering
Convergence Engineering is a concept of a new horizon of digital engineering systems. While the guiding principle of the conventional engineering systems is "virtual manufacturing" where products, their manufacturing processes and behaviors are represented and evaluated without making real objects and experiments as far as possible. However the physical world is not so simple to be faithfully represented into computers. Thus the difference between the virtual world and the physical world must be filled up so as to improve quality and performance of products. For this, it becomes crucial to capture real objects into computers using modern scanning technologies such as laser scanners and X-ray CT scanners.
The movies in the below show one example. First we scan a large transmission case (a part of an automotive transmission system) using a high-power X-rays CT machine to obtain a complete CT data of this part. The second top image shows its volumetric rendering image of the CT data. From this CT data we generate a triangular mesh model (in right third top image) since the original CT image data cannot be used in their its form in digital engineering systems. For instance from this mesh model an FEM mesh can be generated. We are also developing a Reverse Engineering system which automatically convert a triangular mesh model into 3D CAM model of trimmed NURBS surfaces.
All those processes have been very time and labor intensive in industry and thus hinder spread of scanning technologies. We wish to make those processes as much efficient as possible. This challenge pushes us to investigate many difficult but interesting research topics.
The Sinogram Polygonizer
The sinogram polygonizer is a novel method for directly reconstructing 3D shapes from sinograms (i.e., raw output from X-ray CT scanners consisting of projection image sequences showing a rotating object). To obtain a polygon mesh approximating the surface of a scanned object, a grid-based isosurface polygonizer such as Marching Cubes has conventionally been applied to the CT volume reconstructed from a sinogram. In contrast, the proposed method treats CT values as a continuous function and directly extracts a triangle mesh based on tetrahedral mesh deformation. This deformation involves quadratic error metric minimization and optimal Delaunay triangulation for the generation of accurate, highquality meshes. Thanks to analytical gradient estimation of CT values, sharp features are well approximated even though the generated mesh is very coarse. Moreover, this approach eliminates aliasing artifacts on triangle meshes.
Beam Hardening Correction
In industrial X-ray CT application, it is particularly important to remove beam hardening artifacts in reconstructed images in order to achieve accurate surface extraction for dimensional inspection of mechanical parts. This paper proposes a new method in the same category as the IAR (iterative artifact reduction) technique with the added advantage of not requiring repeated image reconstruction. In general, beam hardening artifacts are basically reduced to create a non-linear characteristic curve representing the relationship between the projection value and the transmission length of the ray. In the proposed approach, this curve is based on the property that the sum of the transmission lengths must be preserved for all projection angles (views) – a characteristic known as the volume conservation corollary of the Radon transform. Solving an optimization problem for determining the characteristic curve based on this property is very little overhead of the ordinary CT reconstruction.
Medial Surface Generation for Thin Plate Structure
In one of our projects for Convergence Engineering, we are studying thing plate structures such as car body structures. We have succeeded to generate medial surface meshes from CT data of thin plate structures. The medial surface is useful for CAD/CAM/CAE applications, though existing surfacing methods cannot generate medial surfaces. Our method also evaluates thickness distribution. From experiments, we showed that the accuracy of our medial surfaces are almost equal to that of the boundary surface. We are now extending this method to deal with more complex structures.
In order to extract a medial surface we need to compute a distance field of an object represented as a binary volumetric model. One serious problem here is that sizes of CT data is increasing very fast as performance of XCT scanners has been improved rapidly. We are studying Out of Core approach for extracting medial voxels to process a large volumetric model. The above picture shows an example of a cylinder head of an automotive engine.
Mesh Generation from Multi-Material Parts
Metal parts produced with casting process is one of the major targets scanned by CT device. Here we deal with a part made of different kinds of materials. For instance, this engine block made of Iron and Aluminum. It was impossible to generate a mesh for such a multi-material part. We developed a method to contour boundary meshes between different materials. The generated meshes are of a non-manifold structure. The most difficult point in this research is to generate meshes around the portion where more than three materials meet. Conventional contouring algorithms cannot be applied because they rely on thresholds between only two materials.
The pictures below show a case for a pedal of a motor bike. It contains 5 materials.
doi:10.1115/1.3569830
Mesh Generation from Point Set
We are also interested in geometric processing of point sets obtained by scanning objects with optical scanners. We are particularly interested in an approach based on implicit function. Dr. Ohtake who joined our group from 2007 is very famous for this approach.
Benefits of the implicit function approach are enormous. For instance it is very robust for noisy data and lack of data which are very common in practice and head ache for users.
Smoothing of Partition of Unity Implicit Surfaces for Noise Robust Surface Reconstruction
We propose a novel method for smoothing partition of unity (PU) implicit surfaces consisting of sets of nonconforming linear functions with spherical supports. We derive new discrete differential operators and Laplacian smoothing using a spherical covering of PU as a grid-like data structure. These new differential operators are applied to the smoothing of PU implicit surfaces. First, Laplacian smoothing is performed for the vector field defined by the gradient of the PU implicit surface, which is then updated to reflect the smoothing of the gradient field. This process achieves a method for noise robust surface reconstruction from scattered points.
Scanned data ((a), top) for a terra-cotta Shiisa object((a), bottom); (b) model reconstructed using MPU; (c) Poisson surface reconstruction; (d) reconstruction using our algorithm
doi: 10.1111/j.1467-8659.2009.01511.x
4DCT Image Processing
4D CT device can capture CT volumetric image at real time rate of a few frames per second. We use Toshiba Aquilion-One to scan "slime" which has high viscosity to flow very slowly. We are now working for analyzing such 4D CT images.
The movie is of triple speed and the bright dots are small metal balls mingled to show the flow.
Fully Automatic Reverse Engineering System
The major object representation in engineering fields is B-rep solid model usually generated by CAD systems. "Reverse Engineering" is a system to generate solid model involving NURBS surfaces from the scanned data of an object. It is very useful for engineers because they can use a CAD model of an real object.
In the past Reverse Engineering was used mainly for style design to generate surface models by scanning a clay model. But today its application areas have been expanding rapidly in Japan. For instance, in some company they apply Reverse Engineering to generate CAD models for purchased parts whose CAD data of course they do not have. These CAD models are combined with their own CAD models to build a complete product model and used for conducting simulation of the total system.
Though there are so many commercial software available, they are still not sufficient for satisfying the needs. We are developing a system can automatically generate a CAD model of moderate quality (high quality surfaces cannot be automatically generated anyway).
CT Image Segmentation using Structural Analysis
CT images are often blurred in interference regions, and are hard to be segmented into individual parts as there is no certain threshold to separate them. In this paper, we propose a segmentation method for blurred and low-resolution CT images. The basic idea of our research is simple: two objects can be easily separated in areas of structural weakness. Given CT images of an object, we assign a physical property such as Young's modulus to each voxel and create functional images (e.g., von Mises strain at the voxel). We then remove the voxel with the largest value in the functional image, and these steps are reiterated until the input model is decomposed into multiple parts. This simple and unique approach provides various advantages over conventional segmentation methods, including preciousness and noise robustness. This paper also demonstrates the efficiency of our approach using the results of various types of CT images, including biological representations and those of engineering objects.
doi:10.1371/journal.pone.0031116
Sealed Decomposition of a Triangular Mesh with Tetrahedral Mesh Segmentation
We propose a method for decomposing a closed 2-manifold triangular mesh (tri-mesh) into a set of sub-meshes. This approach has two novel features. First, the decomposed tri-meshes are "sealed". This type of subset tri-mesh generally has boundaries, but our method automatically fills these boundaries to close the subset mesh in a technique known as "sealed decomposition". Second, the mesh is decomposed along its concave features; the outcome tends to be a convex mesh decomposition, which is useful for a variety of applications. In addition, we smooth those parts of the set of decomposed sub-meshes not belonging to the closed 2-manifold tri-mesh using the umbrella algorithm.
Computer Aided Paper Crafting: Unfolding Free From Surface Mesh
An unfolded patterns are generated from a mesh model of a free form surface. The mesh models have several ten thousands of triangles. We developed an algorithm to convert such a mesh model into a set of wide triangle strips. Since any triangle strip can be unfolded on a plane, we can unfold the mesh model. A strip generation is based on a segmentation algorithm.
Digital Human Modeling
"Mass Customization" is considered as a key to success in "BtoC" market. For the mass customization, we need a database to represent precise shape information of each costumer. Such database is also useful for ergonomic design of products.
In this project, we collaborate with AIST Digital Human Lab to develop a modelling method to automatically extract major features of human shapes which are captured by 3D scanning technology.
201 foot models
CAD Model Viewer using Subdivision Surface Approximation
Today collaboration among engineers at distant locations are common. For engineering, it is desired to share CAD models among those engineers. Though the band width of internet becomes broader, it is still difficult to transmit or share CAD data because of their size. So the compression must be made. We apply a concept of subdivision surfaces to this problem. We developed a method to approximate model generated with CAD systems to a set of subdivision surfaces. The highest level of subdivision surfaces is very coarse and thus very suitable for transmitting the data. And they can be further subdivided to generate fine image
Surface Modeling from Design Sketch
Aesthetic characteristics of products have become more important for such products as automotive cars and electric appliances, because their functionality has been too matured to distinguish one product from other. So the needs of computer tools for industrial designers to produce better aesthetic design becomes large.
In this project we use conventional design sketch as input to the system because they are still the most useful method for designers to quickly represent their concepts. From the sketch, we extract 3D wire frame and then generate surface models by applying so called "subdivision surface."
Triangular Mesh Reconstruction from Measured Data Points
A method for reconstructing surface meshes from measured data points is proposed. The points are obtained by laser scanning 3D shape measurement equipment. The method is based on "Implicit Function Reconstruction" method, in which the mesh is generated by iso-surfacing a potential field in 3D voxel space. A new approach for defining the potential field from the measured data points is used. Using this approach, a prototype system is developed for reconstructing meshes for complex objects.
An Approach for Remeshing Irregular Triangular Meshes
The pattern of triangular meshes are important both for modeling and for rendering. It is desired that a user can change the mesh pattern without damaging its original shape. We propose an interactive remeshing method by which a user can specify the region to be remeshed and the pattern of triangles to be filled in the region. The proposed method is based of DSI (Discrete Smooth Interpolation) algorithm.
G1 Stitching Method at Boundary Curve between Surface Patches in Computer Graphics
The purpose of this research is to suggest the method that interactively re-constructs approximate $G^1$ continuity for B-spline surface models, considering used particular in the field of computer graphics. It is for non-uniform B-spline surface model, once B-spline surface is subdivided to Bezier patches and whose control points near the B-spline boundary curve are modified. Later, B-spline control points are re-calculated with minimum square approximation method. Since each step is resolved itself into problem of linear equations without non-linear optimizing method, this method performs in short time.
Metamorphosis of Arbitrary Triangular Meshes with User-Specified Correspondence
Recently, animations with deforming objects have been frequently used in various computer graphics applications. Metamorphosis (or morphing) of three-dimensional objects is one of the techniques which realizes shape transformation between two or more existing objects. In this paper, we present an efficient framework for metamorphosis between two topologically equivalent, arbitrary meshes with the control of surface correspondences by the user. The basic idea of our method is to partition meshes according to the reference shapes specified by the user, whereby vertex-to-vertex correspondences between the two meshes can be specified. Each of the partitioned meshes is embedded into a polygonal region on the plane with harmonic mapping. Those embedded meshes have the same graph structure as their original meshes. By overlapping those two embedded meshes, we can establish correspondence between them. Based on this correspondence, metamorphosis is achieved by interpolating the corresponding vertices from one mesh to the other. We demonstrate that the minimum control of surface correspondences by the user generates sophisticated results of the interpolation be-tween two meshes.
Metamorphosis
Interactive Mesh Dragging with Adaptive Remeshing Technique
We propose a 3D mesh-dragging method useful for intuitive, ancient geometric modeling of free-form polygonal models. With our method, the user can drag a part of a triangular mesh and change its position and orientation. This method is based on an adaptive remeshing procedure which evaluates the deformation of faces by dragging and properly modifies them by deleting or splitting with local topological operations. Therefore the mesh is automatically adjusted for dragging, and irregularity caused by dragging onto the mesh is no longer a concern. In addition, this method is local to the dragged portion, so its computation is ancient In this paper we describe our adaptive remeshing method and demonstrate some dragging examples.
Interactive Mesh Fusion Based on Local 3D Metamorphosis
This paper proposes a new mesh modeling scheme, called mesh fusion, based on three-dimensional (3D) mesh-based metamorphosis. We establish the attachment from a part of one mesh to a part of another with smooth boundaries, employing the traditional cutting and pasting operation in conjunction with a combination of meshes, applying the idea of 3D metamorphosis. We also offer an algorithm for adjusting two boundaries by using the combination of three geometrical operations rigid transformation, scaling and deformation. Our schematic offers a computation time swift enough that the user can create various shapes with interactive speed.
Fitting Subdivision Surface to a Range of Points
The objective of this research is to apply the subdivision surface for surface fitting problems when generating surfaces from data points or polyhedral models. The basic idea is to use the subdivision limit position (SLP) to adapt the control mesh of the subdivision surface to the data points. This method is not a time consuming process involving global optimization. It does, however, fail to capture local characteristics of data points. Consequently, the proposed method is not suitable for generating a surface that precisely interpolates the data points, but it will be useful for quickly generating a surface that captures the overall shape constituted by the data points. A prototype system has been developed to demonstrate several examples for purposes of evaluating the proposed method.
Detailed Triangular Mesh Modeling based on Local Surface Fitting Subdivision Method
The purpose of this research is to develop methods to rapidly create smooth and detailed shape models from a rough triangular mesh (a polyhedron with triangular faces) for free-form shape design. In our previous work, a subdivision approach using the theory of local fitting a quadric surface to a triangular mesh was proposed. This approach can create smooth surface models but cannot add features such as valley and fillet to initial mesh. By ex tending this approach, new subdivision method which is able to both create smooth surface models and add detailed feature shapes is presented in this paper. New method consists of two steps, topological operations and decisions of vertex positions, and combinations of two steps are possible to create various feature shapes, including smooth shape.
3D Sketch: Sketch-Based Model Reconstruction and Rendering
We propose a new modeling and rendering system that enables users to construct 3D models with an interface that seems no different from sketching by hand, and that displays models in a sketch-like style, preserving the features of the user's strokes. We call this system 3D SKETCH. To reconstruct 3D objects from sketches, we limit the domain of renderable sketches and prepare a template for interpreting sketches. As long as a sketch can be matched to such a template, the system can reconstruct a mesh model from the sketch. The system collects information about strokes made, and uses that information for our rendering scheme. When that is accomplished, the designer can view the sketched object from different directions. Our goal is to make users feel as though they can draw their own sketches directly as 3D models.
A Method for Generating Developments of Triangular Polyhedral Models
We study a method for generating developments of polyhedral models, which are easily constructed with paper sheet. The easiness is evaluated in terms of the total length of edges which must be cut, the area of a rectangle circumscribing the development and the number of parts of development. Our goal is not for generating the optimal development but for generating a acceptable development quickly. Basically, generating a development corresponds to finding a spanning tree of face-edge graph of the polyhedral model. We propose two algorithms for traversing the face-edge graph, and four methods for setting costs representing the easiness of manual construction. We implemented an application which generates developments using this method and supports a user to construct paper models.