Advances in Geometric Modeling

Ravi Bhatta
bhattar@mail.gvsu.edu
1-410-292-9578

A Term Paper for EGR 611

Computer Aided Design and Engineering

Fall 2014

Instructor: Dr. Sung Hwan Joo

Abstract
In the last few decades, the advances in computing have led to huge improvements in geometric
modeling, making CAD synonymous to geometric modeling. Its application in manufacturing,
virtual reality, product modeling, rapid prototyping, etc. has made it one of the emerging fields.
This paper presents the advances that have been made in geometric modeling by illustrating
some modifications and some recent developments in geometric modeling. In the last few
decades, the 2D wireframe model has evolved to parametric modeling with much richer features
and greater analytical power. There have been modifications on the simpler geometric modelling
techniques to make it more accurate and flexible. For example the modification of Hermite curve
to represent complex plot using the trigonometric blending functions and tension parameters for
local control, implicit 3D volume fitting using radial Hermite function and use of the least energy
method to modify B-spline. New geometric modeling methods are being developed and used
continuously. Reverse engineering has gained wide popularity due to the advent of sophisticated
modeling techniques which are becoming more automatic with less a priori information and
higher accuracy. The rendering of 3D image from a 2D photograph, which seemed impossible
previously, is a reality these days. The reason for such a rapid advancement in geometric
modeling is due to the multi-disciplinary areas as botany, arts, zoology, etc. that are also
benefited by geometric modeling.

Keywords: geometric modeling, CAD, parametric modeling, Hermite curve, B-spline curve,
reverse engineering, a priori

1. Introduction
Geometric modeling is a method of representing the geometrical shapes of an object in
mathematical description. It consists of processes as data capture, preprocessing, segmentation
and surface fitting and finally creation of a geometric model. Mortenson (1985) defines
geometric modeling as "the technique we use to describe the shape of an object...Geometric
modeling provides a description or model that is analytical, mathematical and abstract rather than
concrete" 1. It may be graphical, mathematical or verbal representation of the object1.
Geometry has been used in Egypt and Babylon since ancient times. The projection theory
developed during the Renaissance and technical drawing and descriptive geometry developed in
the 18th century makes the base for geometric modeling. However, with the progress in
engineering, geometric modeling has been advancing at a faster rate than ever before and now is
an integral part of any engineering and graphics course. Due to the wide use of geometric models
in 2D and 3D model creations, the geometric modeling is usually associated with Computer
Aided Design (CAD).
The Geometric Modeling has numerous applications. Some of them are briefly described
below.1,2
a. Virtual Reality/Animation: Accurate rendering for the purpose of visualization and accurate
engineering analysis. These models can be used by architects to create a virtual model and
simulate a walk-through of a new design and for animation purposes.
b. Product Modeling: Kinematic and dynamics analysis as well as finite element analysis.
c. Manufacturing: Simulation, tool path generation, tolerance analysis and tooling design
d. Production: Geometric modeling stores information in a secure way that can be used for
maintenance and product performance optimization.
e.

Rapid Prototyping: The concept of 3D printing has only been possible due to the geometric
modeling of the object.

f. Other fields: Botany, zoology, arts, sculpture, etc.

2. Objective
The paper aims to present the advances made in the field of geometric modeling by illustrating
some modifications made in simpler modeling techniques and introducing some new advanced
modeling techniques.

3. Evolution in Geometric Modeling

Geometric modeling in CAD applications has evolved through a series of phases to improve the
geometric representation of physical artifacts. Early CAD system, before 1970s, used the three
projections of an object for geometric modeling. The 3D wireframe modeling came in the 1970s
that showed the whole object in Euclidean 3D space. But it could not represent the complex
geometries as found in forged and molded parts, sculptured surfaces and transitions. The surface
modeling developed later was able to show the complete description of the object and perform
the mass properties analysis. But it lacked the spatial integrity of the solid model. In the 1980s,
the solid model was introduced and these problems were solved. The solid models can represent
the mass and the boundary conditions and can be used in design, architecture, finite element
methods and other analysis. Solid modeling can be Constructive Solid Geometry (CSG),
Boundary Representation (B-Rep) or parametric modeling. CSG uses the basic models as
rectangular prism, cylinder, cone, sphere torus, wedge and frustum and a solid model is created
by combining these primitive shapes but has problem in representing complex sculptured
surfaces. Boundary representation method uses the bounding surfaces of a solid object and
performs extrusion, revolving, sweeping and blending of 2-D profile curves to create 3D model
and are used in finite element analysis. Most of the geometric modeling in the early 1990s was
done via hybrid B-Rep and CSG method. The feature based modeling were developed later than
uses form features. The evolution of dimension driven parametric equation helped in making a
constraint based system that uses variables or parameters to specify the dimension of entities and
this made iterative designing fast and trouble free. Parametric modeling is found in many CAD
packages. Assembly modeling uses the same tree and data structure and constraints. Today,
geometric modeling along with the product modeling that adds information to develop, produce
and support for the whole life cycle of the product1.
The geometric model used today are far more advanced than what they were a decade earlier.
The sophistication of representation and the flexibility of modeling has drastically increased.
One of the reasons for this is the modification of the previous simple functions so that it can
represent a curve or a surface more precisely. The following section states some of the
modifications that were made in classical geometric modeling.

4. Modifications in basic parametric representation

The classical approach to define a complicated curve by cubic splines, Hermite, Bezier curves
have gone modification leading to advanced surfaces and curves representation. For example Bspline is a better modification of Bezier curve and is the principle type of the spline used in CAD
softwares as SolidWorks and even this B-spline has a lot of modification. The following are
some of the modifications used in these basic parametric curves used in modern geometric
modeling.

a.

The use of trigonometric blending functions and tension parameters in Hermite spline curve

Hermite curves are one of the oldest and most frequently used parametric curve. Being an
interpolating curve, this curve is useful in reverse engineering. But being a polynomial, there are
shortcoming in defining shapes and also there is no local control. But due to integration of quasiinterpolants based on trigonometric by Lyche3, the drawback due to polynomial was removed.
Sheng Min developed a new method for construction of a new type of interpolation curve and
surfaces that is C2 continuous and degree of spline is variable with tension parameters 1 and 2
and can have local control. With the tension parameters and interpolation points chosen properly,
the trigonometric polynomial curves can be used to represent straight lines, parabola, circular
arcs and some transcendental curves precisely as compared to basic Hermite curve4.

b.

Implicit 3D volume fitting using radial Hermite basis functions

An implicit 3D volume model can be created from a scattered point cloud of data points with
associated normal vectors with an aim of minimizing the error between the measured point and
the point in the volume. This model is known as radial Hermite model as all the weight functions
are the function of distance from an arbitrary point and the model involves normal vectors. The
knots are selected via a VoxelGridCentroids i.e. taking the centroid of the data points lying in
their respective Theissen/Dirichlet regions5. Due to the aim of minimizing error, the volume
generated is accurate.

c.

Modifications in Geometric modeling by B-spline using the least energy method

For the use in image processing, automotive design, etc. curves with minimum energy are
usually desired. The curve internal energy functions are the stretch energy (measure of length of
the curve), the strain energy (measure of the bent of curve) and the jerk energy (curvature
variation), which typically depend on geometric information of the planar curve such as the
length, curvature and variation of curvatures. The simple Bezier curve developed from the
Bernstein polynomial can be employed to get the optimum shape that can minimize these energy
and give the desired curve which is better than the normal B-spline curve6.

5. Some modern methods of Geometric Modeling

Geometric modeling is an ever changing field. The complexity and flexibility are increasing,
making it possible to do more analysis than ever and more realistic representation than ever. The
entire volumetric modeling can be done, which was quite tedious before. There are two types of
methods for volumetric mapping, a priori method and adaptive method7. The examples of these
two methods are shortly described below.

a. A Priori Method
In this method, a discretization is generated independent of the geometric model to map and
approximate. Usually, it involves some human interference to say that the surface is a cylinder or
other features. Some of the examples where priori methods are used are as follows.

i. Geometric modeling in Reverse Engineering

Reverse engineering is a rapidly growing area. The process consists of capturing data points of a
physical objects, preprocessing the data, creating a geometric model and finally developing a
CAD model. Data acquisition is done by either contact or non-contact methods which can be
scattered or regular data. Triangulated models are very popular for creating the geometric model
for computer graphics and animation. The exact type of model created depends on the intended
use of the model. For animation and other purposes, the planar facets and higher order surface
patches are created without enforcing any connectivity and assuming that the bulk of data is
compactly covered. But applications as mechanical engineering has features as holes that needs
to be replicated exactly and irregularity will cause problems. One of the ways to achieve this is to
perform a boundary representation of the outer skin of the object and then use a priori method for

important features. This saves a lot of computation and provides better feature. For developing
surface, they are divided into primary and secondary surfaces as shown in Figure 1.

Figure 1: Hierarchy of surfaces8

After defining the type of surface, the surfaces are segmented and fitted using a variety of
methods. Method can be global approximating surfaces (choosing a rough boundary so that the
points of interest all fall in the boundary region and using least square method to minimize error
of surface point and measured points by iterations), curve network (where the characteristics
curves are first drawn and then user helps to patch the surfaces), arbitrary topology surfaces
(mixture of both global approximating and curve network methods),

and functionally

decomposed surfaces, (which looks at the function of a components or structure for plotting).
After getting data for these surfaces from one view, multiple views are taken with accurate
rotation and translation registered with sufficient overlap of information. Finally all the models
are collected and final model is created with appropriate vertex and edge blending functions as
required8.

ii. 3D object reconstruction from 2D image

Alyn P Rockwood and Jim Winget developed means to create 3D models from 2D images of a
body by using energy functions. Energy functions are defined between the objects photo and
corresponding images of an articulated mesh in three dimensions. With proper lighting, viewing,
scene parameters, the intensity of points in an image is used to compute the local orientation of
the point in the object with a priori data. The process uses shape from shading technology to

define the cost functions and try to minimize the error between the photo and image and
ultimately make the model by iterations9.

b. Adaptive Method
In this method, accumulation of points can be dynamically varied using weights on points and
thus obtain the discretization from the geometric model to map and approximate the object7.
Some of the examples where these methods are used are as follows.

i. Automated Geometric modeling of textile structures

The extensive use of textiles and especially synthetic textile fibers in automotive airbag and other
critical engineering application has led to the need of a separate automated geometric model of
these structures which can handle the mechanical and physical properties of the textile and the
nonlinear response of textile in solid mechanics. TexGen is one of the commercial software
available for this purpose. The principle of this geometric modeling is to represent the yarn path
by a spline (Bezier, natural or periodic spline) and to define the 2D yarn cross section (elliptical
for warp yarn and lenticular for weft yarn) and develop a parametric yarn surface based on the
yarn path and the cross section10.

ii. Geometric modeling of a leaf

The modeling and analysis of leaf shape is important in many fields of study as systematics and
paleobotany. The mathematical representation of leaf boundary was modelled by Y.T. Chi et al 11
by using Bezier curve. The significant data reduction by using two simple Bezier curves with the
classification rate was 95.1 % for the type of leaf identification11 is a good achievement,
although the Bezier curve is not able to replicate the intricate leaf patterns and will create
problems in leaf identification with similar outer boundary and different wavy edges.
iii. 3D Surface construction using shapes
Most of the parametric fitting methods requires a priori information. This method, however uses
shapes and constructive manifolds to reconstruct surface of arbitrary topology. shapes are a
generalization of the convex hull of the point set. This method first estimates the topological

structure of the unknown surface by constructing a simple surface (a polygon mesh with
triangular faces). The correct value of is chosen as per the finite set of points and then a
parametric surface is fitted on it. The importance of the correct value of shapes is demonstrated
by Figure 2. The fitting algorithm embeds into the three dimensional Euclidian space, an abstract
manifold, which is built with topological information extracted from the simple surface. The
embedding is a C2 Smooth mapping derived from B-splines basis functions13. Although, this is a
novel concept without a priori information, the computation of O(n2) times for shapes makes
this process expensive than a priori method.

Figure 2: Two tori. The points are randomly generated on the surface of two linked tori. Six
different shapes for values of decreasing from top to bottom and left to right are shown. First
one is at = and last one is =012
6. Conclusion
As discussed in paper, the advances in geometric modeling concludes that there has been
impressive advances in graphic modeling along with the development of computers. The
modification made in Hermite and B-spline curves and surfaces has led to added accuracy and
better rendering. The newer methods for automatic curve fitting are useful but require a lot of
computations as shapes and a priori method require extra information to operate. The newer
adaptive methods can be the future of reverse engineering which still needs to be worked upon to
get a precise modeling without any a priori information. The extension of this paper could be
investigation on automatic geometric modeling methods that can accurately represent object for
scientific and engineering purposes.

7. References
1. Ault, H. K. (1999), 3-D geometric modeling for the 21st century, Engineering Design
Graphics Journal, Vol. no.63, pp 33,42
2. Raja, Vinesh, (2008) Introduction to Reverse Engineering , University of Warwick, UK,
Retrieved from web 10/12/2014, http://www.springer.com/computer/information+systems
+and+applications/book/978-1-84628-855-5
3. T. Lyche, L.L. Schumaker, S. Stanley, (1998) Quasi-interpolants Based on Trigonometric
Splines, Journal of Approximation Theory, vol. 95, pp. 280-309
4. Sheng Min; Su Benyue, (2012) Geometric Modeling by Blending Hermite Interpolation,
Digital Home (ICDH), 2012 Fourth International Conference on , vol., no., pp.195,200, 2325 doi: 10.1109/ICDH.2012.54
5. Nielson, G. M., Hagen, H., & Lee, K. (2007). Implicit fitting of point cloud data using radial
hermite basis functions. Computing, 79(2), 301-307. doi:10.1007/s00607-006-0206-y
6. Xu, G., Wang, G., & Chen, W. (2011). Geometric construction of energy-minimizing bezier
curves. Science China Information Sciences, 54(7), 1395-1406. doi:10.1007/s11432-0114294-8
7. Umberto Cugini and Michael Wozny (Eds.). (2002) From Geometric Modeling to Shape
Modeling. Kluwer Academic Publishers, Norwell, MA, USA.
8. Vrady, T., Martin, R. R., & Cox, J. (1997). Reverse engineering of geometric modelsan
introduction. Computer-Aided Design, 29(4), 255-268. doi:10.1016/S0010-4485(96)00054-1
9. Rockwood, A. P., & Winget, J. (1997). Three-dimensional object reconstruction from twodimensional images. Computer-Aided Design, 29(4), 279-285. doi:10.1016/S00104485(96)00056-5
10. Lin, H., Zeng, X., Sherburn, M., Long, A. C., & Clifford, M. J. (2012). Automated geometric
modeling of textile structures. Textile Research Journal, 82(16), 1689-1702.
11. Chi, Y., Chien, C., & Lin, T. (2003). Leaf shape modeling and analysis using geometric
descriptors derived from bezier curves. Transactions of the ASAE, 46(1), 175-185.
12. Edelsbrunner, H., & Mcke, E. (1994). Three-dimensional alpha shapes. ACM Transactions
on Graphics (TOG), 13(1), 43-72. doi:10.1145/174462.156635
13. Guo, B. (1997). Surface reconstruction: From points to splines. Computer-Aided Design,
29(4), 269-277. doi:10.1016/S0010-4485(96)00055-3