GeoWebCad Description

The GEO-WEB-CAD Project:
Description


Contents:
  1. Introduction
  2. Research Objectives
    1. Feautre Extraction and Shape Reconstruction
    2. Local and Random Search Algorithms for 3D Triangulation
    3. Physically Based Modeling and Rendering of Melting Process
    4. Modeling, Rendering and Animation Technology Based on Wavelets
  3. Introduction to Conceptual Model
  4. Publication
Geo-Web-Cad Staff
PI: Fuhua (Frank) Cheng
Department of Computer Science
University of Kentucky
Lexington, KY 40506

Project Assistants:

  • Huai-Jun Wu
  • Zheng Zhang
  • Chris Crockett
  • Co-PI: Brian A. Barsky
    Computer Science Division - EECS
    University of California
    Berkeley, CA 94720


    1. Introduction

    The universal availability of computer networks is rapidly changing many of the ways in which information and goods are developed and delivered. In design and manufacturing, networks provide an environment which allows the integration of activities across the corporation throughout the period of product design and development. In the future, manufacturing will almost certainly rely heavily on access between collaborators and services which is integrated, distributed, and highly networked.

    Advances in computer network methodology also are rapidly changing the methods for the development and dissemination of computer assisted tools and techniques. In a network enriched product development environment, geographically separated researchers should be able to collaborate with more facility and ease.

    Computer aided techniques for modeling and design are rapidly becoming important tools for modern industrial design and manufacturing. In many applications, computer models are replacing physical models because they are cheaper to construct, easier to change, and simpler to analyze. Computer generated images increase productivity by allowing swift and accurate visualization during design. This permits not only rapid prototyping but early detection of errors, which in turn, decreases the fabrication of erroneous components as well as costly redesign during manufacturing. Factory automation is being revolutionized by this methodology.

    Geometric design and modeling is a major research and development area in computer aided design. The automobile, aerospace, and shipbuilding industries rely heavily on applications of many techniques developed in this area. These techniques are also finding their way into many new areas. But, despite these advances, the techniques in geometric design and modeling are nonetheless far from completely developed. Research in this area is constantly extensively conducted so that new computational and visualization techniques can be developed to further facilitate the modeling and design process.

    In this project we propose to develop and deliver CAD/CAM software systems which support such methodologies. These issues require the development of network CAD/CAM interfaces and collaborative design paradigms for the following components of an object-oriented geometric modeling tool kit.

    These should be implemented as platform-independent applications which can be accessed from anywhere over the World Wide Web. With the use of applets written in "Java" [AG96] the research teams from Tsinghua University and the United States will be able to collaborate over the web.


    2. Research Objectives

    Four particular areas of research in solid and geometric modeling included in the above tool kit will be investigated during the tenure of this project. They are:

    Detailed descriptions of the work to be undertaken in each of these areas appears in the sequel.


    2.1 Feature extraction and shape reconstruction of unorganized range data:

    To visualize an object for which we have a set of range data points from its boundary surfaces, a typical approach is to generate a mesh representation of the object and render this mesh representation. This technique is also frequently used in reverse engineering. One of the most important issues that needs to be addressed here is the extraction of special features from the data points. These special features allow the decomposition of the mesh representation into smooth components that can be represented by smooth surfaces. The major objective is to develop a range data based mesh generation technique which preserves all the features of the original object (such as sharp edges and sharp corners), and, for each smooth patch, generate a surface representation.

    The approach is comprised of four steps. The first step is standard: topology mesh construction [HDD92, HDD93, HDD94]. Next, Gaussian mapping is used to develop a technique which identifies special existing features. Under Gaussian mapping, sharp points and edges are denoted as singular and defined as follows.

    Definition. A surface point is called a "singular point" if in each of the point's neighborhoods there is a curve passing through the point whose image (excluding the point) under the Gaussian mapping consists of at least two disjoint connected closed regions of the unit sphere.
    Definition. A surface edge is a "singular edge" if each point of the edge is a singular point.
    It is easy to see that a Gaussian mapping maps a cone, excluding the tip, to a circle on the unit sphere. Each curve through the tip (excluding the tip itself) is mapped to two distinct closed subsets of the circle. Hence the tip of a cone is a singular point. Therefore, by examining the distribution of the unit normal vectors on the unit sphere, one can estimate the number of sharp vertices and sharp edges in each region.

    The third step is to search and identify the location of these special vertices and edges that define special features of the object. An initial estimate may be made by comparing tangent planes for each data point. A refining technique such as the one used by Hoppe, DeRose, et. al [HDD92, HDD93] will be used to refine the location of these vertices and edges, as well as the other data points, while still maintaining the topology of the data points.

    The last step is to generate a smooth surface representation for each patch that is bounded by the sharp edges and boundary edges of the original object. A surface interpolation technique that interpolates a surface through a set of network curves of arbitrary topology will be developed. To avoid having patches of very small sizes, the refining technique used in step three should be capable of removing vertices and edges from the mesh representation without changing its shape or topology.

    It should be pointed out that the second and third tasks can be subsumed by task one or, actually, any visualization process as an alternative input phase for objects whose boundary representation is not available or, simply, to make the input phase more flexible.

    To complete this topic, we must develop algorithms to:

    as well as an object-oriented implementation of this technique as part of a mesh representation generation process.


    2.2 Local and random search algorithms for 3D triangulation:

    Triangulations play an important role in finite element mesh analysis, approximation theory, numerical analysis and computer-aided geometric design. Delaunay triangulation is often used but does not optimally satisfy angle criteria. Thus there may be tetrahedra of poor shape in a 3D Delaunay triangulation. Local transformations such as those of Joe [Joe98] have proved acceptable and desirable, but could be improved.

    Beginning with either Joe's triangulations or standard Delaunay triangulations, final solutions will be formulated using:

    For each method,
    a) representations for feasible solutions,
    b) search techniques for neighborhoods, and
    c) evaluation methods for gains in cost
    need to be defined. Then search mechanisms need to be developed for the local search along the lines of Kernighan and Lin's classic min-cut heuristic [KL70]. Next, these can be applied to simulated annealing triangulation algorithms.

    At this point, genetic operators, such as adaptive polyhedron crossover and mutation will be developed for use in genetic algorithms.

    All algorithms will be tuned to existing commercial data sets and compared with the standard triangulation techniques.


    2.3 Physically based modeling and rendering of melting process:

    First, we will develop modeling and rendering techniques for visualizing the melting process of an object in 3D space. The problem will be studied in a general setting so that objects with different melting points, absorptivity, and viscosity can be considered. The causes of the melting can be point-heat sources, line-heat sources, face-heat sources, or ambient temperature. Boundaries of objects are defined by piecewise surfaces (for example: including polyhedra and free-form surfaces).

    Modeling of the deformation of an object during the melting process focuses on understanding and simulating the physics of the melting process and comprises the initial portion of this investigation. The second part, rendering of a time-dependent shape representation, emphasizes the development of efficient time-dependent surface evaluation algorithms.

    Object deformation during melting involves:

    1. building the heating model,
    2. building the heat-deformation relation,
    3. developing a surface representation scheme for
      • continuous deformation and
      • discrete deformation, and
    4. developing a mathematical model for the flowing of melted object.
    An idea similar to Terzopoulos and Qin's dynamic NURBS [TQ94] will be used in the development of the surface representation scheme for continuous deformation.

    At present, the most efficient and numerically stable NURB surface rendering technique is a two-stage Cox-de Boor recurrence based tessellation approach [LC93, LC96]. However, for time-dependent shape representations, it is not known if the incremental factor is also considered. Thus, rendering time-dependent shape representations will incorporate developing an incremental method for the surface scheme for continuous deformation, as well as performing a comparison on forward differencing [LSP87, SC88], knot insertion [RHD89], and the two-stage Cox-de Boor recurrence based tessellation [LC93] for time-dependent shape representations.

    In order to complete this topic, we first need:

    Then, an object-oriented implementation that will must be developed and delivered.


    2.4 Modeling, rendering and animation technology based on wavelets:

    In contemporary modeling systems, curves are usually represented by spline forms, and surfaces are often represented as triangular meshes or spline surfaces. For a complex scene, or an object defined by data from a laser scanning system, there could be a mesh of extreme complexity or a vast amount of control vertices which make it is too expensive to store, transmit, or render. We propose developing modeling, rendering and animation techniques for such complex scenes or objects.

    Our investigation of this problem is divided into three sections.

    1. Compression or Simplification. Any multiresolution mesh can be compressed by removing small wavelet coefficients. A proper threshold for removal can be chosen such that the resulting approximation is guaranteed to be within a specified error tolerance of the original mesh. For integral B-splines or NURBS (periodic or non-periodic), we will also develop a unified approach to compress the data of control vertices.

    2. Editing and smoothing. In geometric modeling, we frequently need the ability to directly manipulate curves and surfaces. Direct manipulation of curves and surfaces often leads to underdetermined problems since there are generally many possible curves or surfaces satisfying users' constrains. A common approach is to find the curve or surface with minimum-energy. To avoid solutions to an ill- conditioned system of equations, we propose the use of wavelet constructions.

      For this we need to :

      • edit the curves and surfaces,
      • smooth them using a minimum-energy solution; and
      • smooth them using a least-squares approximation.

      We shall develop a method which extends Finkelstein and Salesin's multiresolution editing for endpoint-interpolating B-spline curves to that of uniform , nonuniform, and NURBS B-spline curves or surfaces. Then we shall develop an algorithm for multiresolution smoothing for these curves or surfaces based upon arbitrary meshes.

    3. Rendering and Animation. In order to efficiently display a complex scene or a surface represented by a large number of triangular meshes, multiresolution models would seem to be a good choice. In addition, wavelet radiosity has efficiently been used for solving the global illumination problem in Lambertian diffuse environments. We propose to develop an algorithm for multiresolution wavelet radiosity to display surfaces consisting of a vast amount of triangular meshes. We also must preprocess the data for animation because radiosity is independent of view points.


    3. Introduction to Conceptual Model

    Computer Supported Cooperative Work (CSCW) has been studied intensely in recent years. Infrastructure level in enabling techniques, such as data sharing, concurrency control and coordination control, etc. have been addressed frequently. The system architecture aspect, however, is often left out. While the enabling techniques are general, they ought to be organized under a proper system architecture to fulfill the requirement of collaboration on a specific application.

    In the field of collaborative engineering, including collaborative CAD, efforts have been made to develop the the enabling techniques suitable to the engineering context.

    We will analyze the requirements and properties of a collaborative CAD system from the system architecture point of view and present a conceptual model that fulfills these requirements. Issues that should be considered in the implementation of this architecture are also discussed.