Function Representation in Geometric Modeling: Concepts, Implementation and Applications

Function Representation in Geometric Modeling: Concepts, Implementation and Applications
Function Representation in Geometric Modeling:
Concepts, Implementation and Applications
A.Pasko1, V.Adzhiev2, A.Sourin3, V.Savchenko1
1
Shape Modeling Laboratory
The University of Aizu
Tsuruga, Ikki-Machi, Aizu-Wakamatsu City,
Fukushima, 965 Japan
2
Department of Computer Science
University of Warwick
Coventry, CV4 7AL, United Kingdom
3
School of Applied Science
Division of Software Systems
Nanyang Technological University
Nanyang Avenue, Singapore 2263
Geometric modeling using continuous real functions of several variables is
discussed. Modeling concepts include sets of objects, operations and
relations. An object is a closed point set of n-dimensional Euclidean space
with a defining inequality f ( x1 , x 2 ,..., x n ) ≥ 0 . Transformations of a
defining function are described for the set-theoretic operations, blending,
offsetting,
bijective
mapping,
projection,
Cartesian
product
and
metamorphosis. Inclusion, point membership and intersection relations are
described. In the implemented interactive modeling system, we use highlevel geometric language that provides extendibility of the modeling system
by input symbolic descriptions of primitives, operations and predicates. This
approach supports combinations of representational styles, including
constructive geometry, sweeping, soft objects, voxel-based objects,
deformable and other animated objects. Application examples of aesthetic
design, collisions simulation, NC machining, range data processing, and 3D
texture generation are given.
Key words: Geometric modeling - Solid modeling - Real functions Implicit surfaces - R-functions.
1 Introduction
The use of real functions of several variables for defining geometric objects
is quite common in mathematics and computer science. The inequality
f ( x1 , x 2 ,..., x n ) ≥ 0 describes a half space in n-dimensional Euclidean
space. The equation f ( x1 , x 2 ,..., x n ) = 0 specifies an implicit function of
n-1 variables and describes an orientable (n-1)-dimensional surface. The
properties of these geometric objects are studied in algebraic and differential
geometry. In three-dimensional case, an object defined by the abovementioned inequality is usually called a solid (or a volume) and an object
defined by the equation is called an implicit surface. Functionally
represented volumes and surfaces appear to be useful in solid modeling,
computer aided geometric design (CAGD), animation, range data
processing and volume graphics.
Half spaces defined by algebraic inequalities are used as primitives in
constructive solid geometry (CSG) (Requicha 1980, Requicha and
Rossignac 1992). Duff (1992) solves such important problems as collision
detection and rendering on the basis of interval arithmetic for CSG-trees
with primitives bounded by implicit surfaces. The representation of a whole
complex object by a single real function also has attracted an interest. An
attempt to develop a system of the set-theoretic operations closed on this
representation has been made by Ricci (1973). The serious restriction of the
proposed method is that functions defining complex objects contain C 1
discontinuity in their domain. The exact analytical definitions of the settheoretic operations have been proposed in the theory of R-functions
(Rvachev 1963, Rvachev 1974) and applied to solving problems of
mathematical physics on complicated geometric domains (see Shapiro 1988
and Shapiro 1994a for a survey). Shapiro (1994a) uses R-functions to
construct defining functions of regular sets required in CSG. The theory of
R-functions was applied to define several operations on multidimensional
geometric objects in (Pasko 1988, Pasko et al. 1993a).
Wang (1984) pointed out the theoretical possibility of deriving the implicit
description of the surface swept by the moving solid. Symbolic
computations were required to reduce the dimension of variables and to
yield the representation in an implicit form.
Although parametric representation is the most common in CAGD,
attention is also paid to implicit surfaces because of their closure under
some important operations (Hoffmann 1993). Typical operations of this
kind are the offsetting and blending (Ricci 1973, Middleditch and Sears
1985, Hoffmann and Hopcroft 1987, Hoffmann 1989, Rockwood 1989,
Warren 1989).
Sclaroff and Pentland (1991) have generalized the implicit function
representation by introducing deformation defined with a matrix of free
vibrations. Such a generalization provides modeling deformations based on
physical laws and collision detection for animated objects (Essa et al. 1992).
Collision detection for implicit surfaces is also discussed by Baraff (1990),
Gascuel (1993) and Snyder et al. (1993).
Several implicit functions have been proposed on the base of field
generating skeleton (Blinn 1982, Wyvill et al. 1986a, Bloomenthal and
Shoemake 1991). The field defined by superquadrics (Barr 1981) can
increase significantly a complexity of the resultant object. These functions
are useful in interactive modeling (Bloomenthal and Wyvill 1990) and
animation (Wyvill et al. 1986b). Muraki (1991) has applied Blinn's blobby
model to approximate the shape of the object defined by sample surface
points. Modal deformations and displacement map (Sclaroff and Pentland
1991) as well as polynomial functions (Bajaj et al. 1993) have been also
applied for solving the surface reconstruction problem.
Many authors have presented algorithms of polygonization of implicitly
defined surfaces or isosurfaces of trivariate functions (Wyvill et al. 1986a,
Bloomenthal 1988, Pasko et al. 1988, Schmidt 1993). The similar algorithm
was proposed by Lorensen and Cline (1987) for extracting polygonal
surfaces from volumetric data. In fact, objects with implicit surfaces and
voxel-based objects have the unified models. The only difference is that a
tabulated function of three variables is used for the voxel-based objects.
Such a view was applied, for example, by Hughes (1992) to implement
metamorphosis between two voxelized solids.
Thus, the function representation is widely used in geometric modeling and
computer graphics in several different forms. From the other hand, these
different models are not closely related to each other and to such well
known representations as sweeping, B-rep and CSG (Requicha 1980). This
obviously retards further research. The motivation of our work is the need
to fill these gaps by constructing as rich as possible system of operations
closed on functionally represented objects. Dimension independence of
descriptions leads to the possibility of inclusion of 4D and other
multidimensional volumes in the set of treated objects. Application of real
functions of several variables provides a good base for an interactive
modeling system extendable by symbolic descriptions of primitives and
operations.
This paper presents a state of the art report of our project, the main
objectives of which are:
– Categorization and summary of the geometric concepts required in a
functionally based modeling environment;
– Elaboration of a rich system of geometric operations closed on
functionally represented objects;
– Treatment of multidimensional and particularly space-time objects in a
uniform manner;
– Specification, implementation and application of an interactive geometric
modeling system based on the function representation.
In Section 2, we consider geometric concepts as a triple (Objects,
Operations, Relations) with their descriptions in terms of real functions of
several variables and present new descriptions of some geometric
operations. Section 3 deals with interactive geometric modeling based on
the function representation. Section 4 is devoted to the application
examples. Section 5 summarizes the paper and discusses future work.
2 Geometric concepts and function representation
This section provides a summary of geometric concepts of functionally
based modeling environment. We also discuss new descriptions of several
geometric operations that we have recently developed.
Let us describe the geometric concepts as a triple:
a M , Φ, W f
where M is a set of geometric objects, Φ is a set of geometric operations,
and W is a set of relations for the set of objects. Mathematically this triple is
an algebraic system. Its application to multidimensional and time-dependent
geometric modeling have been studied by Pilyugin et al. (1988), Pasko
(1988) and Sourin (1988). Below, we consider the main parts of this triple.
2.1 Objects
We consider geometric objects as closed subsets of n-dimensional Euclidean
space E n with the definition:
f ( x1 , x 2 ,..., x n ) ≥ 0
(1)
where f is a real continuous function defined on E n . We call f a defining
function. The inequality (1) we call a function representation (or F-rep) of
a geometric object. In the three-dimensional case, the boundary of such an
object is a so-called "implicit surface". Note that the definition of an object
is the inequality (1) with the explicit function of n variables
f = f ( x1 , x 2 ,..., x n ) but not the implicit function of n-1 variables
f ( x1 , x 2 ,..., x n ) = 0 . The function can be defined analytically, or with a
function evaluation algorithm, or with tabulated values and an appropriate
interpolation procedure. The major requirement to the function is to have at
least C 0 continuity. There is a classification of points in E n associated with
the closed n-dimensional object
with
function
representation.
If
X = ( x1 , x 2 ,..., x n ) is a point in E n , then:
and
f(X) > 0
if X is inside the object;
f(X) = 0
if X is on the boundary of the object;
f(X) < 0
if X is a point outside the object.
Here, X = ( x1 , x 2 ,..., x n ) is a point in E n . However, after applying
operations to such an object, the constructed function may have points with
f=0 inside and on the boundary of the object. In some practical applications
it can require special treatment. On the other hand, it gives much more
freedom for designing operations on objects. Generally speaking, geometric
objects defined by the inequality (1) are not reqularized solids required in
CSG. The object can have the boundary with dangling portions that are not
adjacent to the object's interior. We define objects in multidimensional space
for choosing a space of arbitrary dimension in each specific case. For
example, if n=4 then ( x1 , x 2 , x 3 ) can be space coordinates and x 4 can be
interpreted as the time.
Two major types of elements of the set M are basic geometric objects
(primitives) and complex geometric objects. A geometric primitive is
described by a specific instance of a function chosen from a finite set of
possible types. A complex geometric object is a result of operations on
primitives. In the modeling system, the finite set of primitives can be
defined. However, the possibility of the extension of this set in a symbolic
manner should be provided. Actually, this approach allows the modeling
system to be initially "empty" and make the user to be responsible for an
application oriented filling of the primitive set. We consider this flexibility as
one of the major advantages of the F-rep based geometric modeling system.
2.2 Operations
The set of geometric operations Φ includes such operations as:
Φ i : M 1 + M 2 + ... + M n → M
where n is a number of operands of an operation.
We consider only unary and binary operations in this paper. The result of
the operations is also an object from the set M that ensures the closure
property of the function representation. Let object G1 has the definition
f1 ( X ) ≥ 0 . For unary operations the object G2 is said to be derived from G1
as G2 = Φ i ( G1 ) and is defined by f 2 = Ψ( f1 ( X )) ≥ 0 , where Ψ is a
continuous real function of one variable. The examples of unary operations
are the bijective mapping, affine mapping, projection, offsetting. For binary
operations the object G3 is said to be derived from G1 and G2 as
G3 = Φi (G1 , G2 ) and is defined by f 3 = Ψ( f1 ( X ), f 2 ( X )) ≥ 0 , where Ψ is a
continuous real function of two variables. The examples of binary
operations are the set-theoretic operations, blending operations, Cartesian
product, metamorphosis. Like with the objects, the user of the F-rep based
modeling system is able to introduce any desirable operation by its analytical
or procedural description in symbolic form and thus extend the list of
operations. In this connection, we shall not attempt to specify the complete
set of possible operations on objects, but only introduce the most commonly
used types.
The transformations of the function representation associated with
operations on an object are described below. We follow here the logic of
step by step extension of modeling environment from well-known settheoretic operations to less familiar metamorphosis.
2.2.1 Set-theoretic operations
The analytical definitions of the set-theoretic operations on functionally
described objects have been introduced and studied by Rvachev (1963,
1974) for solving problems of mathematical physics on areas with complex
shapes. Rvachev proposed these definitions in order to transform settheoretic operations on areas with the description (1) in to operations on the
defining functions. The resultant object will have the defining function as
follows:
f 3 = f1 | f2
for the union;
f 3 = f1 & f 2
for the intersection;
f 3 = f1 \ f 2
for the subtraction,
where f1 and f2 are defining functions of initial objects and |,& ,\ are signs
of so-called R-functions. One of the possible analytical descriptions of
R-functions is as follows:
1
f1 | f2 =
( f1 + f 2 + f12 + f22 − 2 αf1 f 2 )
1+ α
1
f1 & f 2 =
( f1 + f 2 − f12 + f 22 − 2 αf1 f 2 )
1+ α
(2)
where α = α ( f1 , f 2 ) is an arbitrary continuous function satisfying the
following conditions:
b
g
b
g b
g b
g b
−1 < α f1 , f2 ≤ 1, α f1 , f 2 = α f 2 , f1 = α − f1 , f2 = α f1 , − f 2
g
The expression for the subtraction operation is
f1 \ f2 = f1 & ( − f2 )
Note that with this definition of the subtraction, the resultant object includes
its boundary.
If α=1, the functions (2) become
f1 | f 2 = min ( f1 , f2 )
f1 & f2 = max ( f1 , f2 )
(3)
This is the particular case described by Ricci (1973). The functions (3) are
very convenient for calculations but have C 1 discontinuity when f1 = f 2 . If
α=0, the functions (2) take the most useful in practice form:
= f1 + f 2 +
f12 + f22
f1 & f2 = f1 + f2 −
f12 + f 22
f1 | f 2
(4)
The functions (4) have C 1 discontinuity only in points where both
arguments are equal to zero. If C m continuity is to be provided, one may
use another set of R-functions:
f1 | f 2
f1 & f 2
e
=ef + f
= f1 + f 2 +
1
2
−
f12 + f22
f +f
2
1
2
2
jc f
jc f
2
1
+ f22
2
1
+f
2
2
h
h
m
2
m
2
(5)
Generally, R-functions correspond to standard but not regularized settheoretic operations. They can result in "interior zeroes" of a defining
function. Thus in a general case one can not distinguish between a point of
"interior zero" and a boundary point. If regularity of resulting objects is
required, the special method of construction of defining functions can be
applied (see Shapiro 1994a).
Note that the R-functions can be used for describing geometric primitives.
For example, the description of a segment in E 1 can be obtained from the
descriptions of two rays as follows:
f ( x ) = ( x1 − b1 )& ( b2 − x1 )
The plot of this function for the expressions (3) is shown in Fig. 1a with
α=1, Fig. 1b corresponds to (4) with α=0, and Fig. 1c corresponds to (5)
with m=1. It is important to point out that the function in Figs. 1b and 1c
does not have points in its domain where the derivative is discontinuous.
Such points can cause problems in subsequent operations on objects,
especially when blending is used (see Rockwood 1989). Fig.2 illustrates the
process of construction of 3D solid with the union and intersection
operations.
2.2.2 Blending
Although blending was studied by many researchers, the need in the
continuous analytical description of the set-theoretic operations with
controllable blend shape existed. Pasko and Savchenko (1994a) have
proposed to consider a blending surface as a boundary of an object obtained
by the modified set-theoretic operations with the description based on the
technique of R-functions. This approach is based on the observation of
contours behavior shown in Fig. 3. Note that we use expressions (4) here to
define the set-theoretic operation. That is why f=C gives an exact definition
of the corner for C=0 and smooth contour lines for other C values. The
blending set-theoretic operation based on R-functions can be defined as
F ( f1 , f2 ) = R( f1 , f2 ) + d ( f1 , f 2 )
(6)
where R is a corresponding R-function, d is a displacement function that has
a maximal absolute value d(0,0) and asymptotically approximates a zero
value with increasing absolute values of the arguments. We have found that,
for example, the following simple form of the displacement function is
suitable for blending:
a0
d ( f1 , f2 ) =
2
f1
f
1+
+ 2
a1
a2
FG IJ FG IJ
H K H K
(7)
2
It is assumed that the defining functions for both objects have the distance
property. The proposed displacement function is not the only one possible.
Other Gaussian-like functions can be designed for specific applications.
Applying equations (4), (6) and (7) we can get, for example, the final
description form of the blending intersection operation:
a0
F ( f1 , f2 ) = f1 + f2 + f12 + f 22 +
2
2
f1
f2
1+
+
a1
a2
FG IJ FG IJ
H K H K
(8)
This definition provides highly intuitive shape control of added material,
subtracted material, and variable radius blends. It was used to generate
aesthetic blends defined by hand-drawn strokes (see Section 4). Constant-
radius blending is connected with offsetting operation and is discussed
below. Fig.4a illustrates an application of the blending set-theoretic
operations. The basic block is not defined as a single primitive but as an
intersection of six halfspaces to control the shape of edges in further
blending. Different values of the displacement function parameters were
assigned to the different edges. The prominent front edge of the resultant
object was defined by added material blending. Other edges were defined by
subtracted material blending. Multiple blending operations were applied in
the corners. The small cylindrical hole was made after blending to show that
no problem occurs when applying the set-theoretic operations to the
blended object. This is ensured by the continuity of the function (8).
2.2.3 Offsetting
To generate expanded or contracted versions of an initial object one can
apply the positive and negative offsetting operation respectively. The
descriptions of the following three offsetting operations have been proposed
in (Pasko and Savchenko 1994b):
1. Iso-valued offsetting with F=f(X)+C, where negative constant C defines
the negative offset, and positive C defines the positive offset. Fig.4b
shows an application of this operation to a 3D solid.
2. Offsetting along the normal with F=f(X+DN) for the positive and
F=f(X-DN) for the negative offsetting, where D is the given distance
value, and N is a gradient vector of the function f in the point with X
coordinates.
3. Constant-radius offsetting with F=max(f(X')) for the positive and
F=min(f(X')) for the negative offsetting, where X' is vector of
coordinates of points belonging to the sphere of the given radius D and
the center at X. Note that continuity and distance property of the
defining function of the object are also used here. The procedure of the
constant-radius blending of the object's convex edges includes
consequently applied negative and positive constant-radius offsetting
(Rossignac and Requicha 1984). This procedure applied to 2D solid is
illustrated in Fig.4c.
2.2.4 Bijective mapping
Let Φ i be defined by coordinate transformations
x j ' = ϕ j ( x1 , x 2 ,..., x n ), j = 1,..., n
where ϕ j are continuous real functions. We assume also that inverse
functions ϕ −1j exist. Then the resultant object is described as
G2 : f1 ( ϕ 1−1 ( x1 ' , x 2 ' ,..., x n ' ),..., ϕ −n−11 ( x1 ' , x 2 ' ,..., x n ' ) ≥ 0
(9)
The examples of such a bijective mapping are the tapering, twisting, bending
(Barr 1984) and modal deformations (Sclaroff and Pentland 1991). The
twisting of a solid constructed by the set-theoretic operations is shown in
Fig.5a. Another example is the mapping of an object, defined in an arbitrary
coordinate system, to a Cartesian coordinate system. The inverse functions
ϕ −1
j from (9) for the mapping of an object from the cylindrical coordinate
system to the Cartesian one will be defined as:
x1 = ϕ 1−1 ( x1 ' , x 2 ' , x 3 ' ) = x1 ' 2 + x 2 ' 2
x 2 = ϕ 2−1 ( x1 ' , x 2 ' , x 3 ' ) = arctan( x 2 ' x1 ' )
(10)
x 3 = ϕ 3−1 ( x1 ' , x 2 ' , x 3 ' ) = x 3 '
where
x1 = ρ, x 2 = θ, x 3 = z
are cylindrical coordinates
x1 ' = x , x 2 ' = y , x 3 ' = z are Cartesian coordinates
This mapping can be applied to describe an object defined by rotational
sweeping (see the Cartesian product operation described below).
2.2.5 Affine mapping
The affine mapping is an important specific case of the bijective mapping.
Let Φ i be the affine mapping defined by the equality X' = AX + C , where
X = ( x1 , x 2 ,..., x n ) T , X' = ( x1 ' , x 2 ' ,..., x n ' ) T , C = ( c1 , c2 ,..., c n ) T ,
and A = {aij } is the matrix of transformation with dimensionality n×n and
det A ≠ 0. Then the definition (9) is changed to:
G2 : f1 ( A −1 ( X' − C ))
(11)
In general case, elements of the matrix A and the vector C can be functions
of coordinates X. For example, if they are time dependent, then some timedependent affine mapping is defined. It can be the movement along a line,
rotation and scaling in time, delay and bringing forward, complex movement
with time mapping, etc. The unification of mappings of space coordinates
and the time allows an effective description of complex geometric processes
unifying models used in a description of geometric volumes and complex
movements in animation.
Next we consider the operation of projection where detA=0.
2.2.6 Projection
R-functions can also be applied for approximated describing the projection
operation from E n to E n−1 that does not have an inverse transformation
Φ −1 . Let us define
G1 ⊂ E n : f1 ( x1 , x 2 ,..., x i ,..., x n ) ≥ 0
G2 ⊂ E n −1 : f 2 ( x1 , x 2 ,..., x i −1 , x i +1 ..., x n ) ≥ 0
and G2 is a projection of G1 to E n−1 . The object G2 can be defined as a
union
of
sections
of
G1
by
hyperplanes
x j = Cj
where
C j +1 = C j + ∆x i , j = 1, N and C1 = X i min .
Let f1 j = f1 ( x1 , x 2 ,..., x i −1 , C j , x i +1 ,..., x n ) be a defining function for a
section. Then the defining function for the projection with ∆x i → 0 can be
expressed as:
f 2 = f11 | f12 |...| f1 j |...| f1N
(12)
Numerical procedures for this function evaluation are discussed elsewhere
(Pasko 1988). Fig.5b illustrates a projection from E 3 to E 2 . One of the
applications of this operation is a description of a volume swept by a
moving solid as a projection of 4D object to E 3 .
2.2.7 Cartesian product
Let G1 ⊂ E k and G2 ⊂ E m .
We define G3 as a Cartesian product of G1 and G2 :
G3 = G1 × G2 = {( x1 , x 2 ,..., x n )|( x1 , x 2 ,..., x k ) ∈ G1 ,( x k +1 , x k + 2 ,..., x n ) ∈ G2 }
where G3 ⊂ E n and n=k+m. The defining function for G3 can be obtained
using R-functions:
f 3 ( x1 , x 2 ,..., x n ) = f1 ( x1 , x 2 ,..., x k )& f2 ( x k +1 , x k + 2 ,..., x n )
(13)
Swept objects can be defined with help of the Cartesian product and the
bijective mapping. A 3D object defined by the rotational sweeping is shown
in Fig.5c. Firstly, the 3D object was constructed as Cartesian product of 2D
solid and 1D segment. Then bijective mapping (10) was applied to this 3D
object. In other words, its x-coordinate was interpreted as the angle and
y-coordinate as the radius of a cylindrical coordinate system. Another
application is a definition of time dependent geometric objects on some time
interval by the Cartesian product of a static object and a time segment or a
ray. For example,
G1 ⊂ E 3 : f1 ( x , y , z ) ≥ 0
is a static object in E 3
G2 ⊂ E 1 :( t − t1 )& ( t 2 − t ) ≥ 0
is a time segment [ t1 , t 2 ]
(14)
G3 ⊂ E 4 , G3 = G1 × G2 : f1 ( x , y , z )& (( t − t1 )& ( t 2 − t )) ≥ 0
G3 is a time dependent geometric object being activated at the time t1 and
terminated at the time t 2 in space E 4 with coordinates (x,y,z,t).
2.2.8 Metamorphosis
We consider the metamorphosis as a binary operation on two objects G1
and G2 defined in E n−1 . The resultant object G3 is defined in E n and
described as
f 3 ( x1 , x 2 ,..., x n ) =
f1 ( x1 , x 2 ,..., x n −1 ) ⋅ (1 − g ( x n )) + f2 ( x1 , x 2 ,..., x n −1 ) ⋅ g ( x n )
(15)
where g ( x n ) is a positive continuous function, g ( x n 0 ) = 0 , and g ( x n 1 ) = 1 .
It means that G1 is a section of G3 by the hyperplane x n = x n 0 and G2 is a
section of G3 by the hyperplane x n = x n 0 in E n . For n=4, an object G3 can
be thought as a time dependent object reconstructed from its two instances
at different time moments. Fig.5d shows several time steps of
metamorphosis of a 3D solid to another one. For n=3, this operation
generates a 3D solid reconstructed from its two planar cross-sections that
can be useful in tomography and range data processing.
2.3 Relations
We will consider only binary relations as the subsets of the set M 2 . The
examples of binary relations are the inclusion, point membership,
interference or collision. Similarly, like with the objects and operations, we
give the user a possibility to extend the set of relations by symbolic
definition of their predicates.
2.3.1 Inclusion relation
This relation is described as G2 ⊂ G1 and means that the object G2 is
included in G1 . If G2 is the point P the relation can be described by the
following bivalued predicate:
0 , if f1 ( X ) < 0 for P ∉ G1
S2 ( P, G1 ) =
1, if f1 ( X ) ≥ 0 for P ∈ G1
RS
T
(16)
2.3.2 Point membership relation
Let iG1 be the interior of G1 and bG1 be the boundary of G1 . The point
membership relation is described by the 3-valued predicate:
R|0, if f ( X) < 0
S ( P , G ) = S1, if f ( X ) = 0
|T2, if f ( X ) > 0
1
3
1
1
1
for P ∉ G1
for P ∈ bG1
for P ∈ iG1
(17)
This predicate can be correctly evaluated for G1 with no interior zeroes.
Note that the set-theoretic operations with R-functions correspond to the
operations of 3-valued logic over predicates S3 but not to the Boolean logic
over predicates S2 .
2.3.3 Intersection relation
The relation is defined by the bivalued predicate:
0 , if G1 ∩ G2 = ∅
Sc ( G1 , G2 ) =
1, if G1 ∩ G2 ≠ ∅
RS
T
(18)
A function f 3 ( X ) = f1 ( X )& f 2 ( X ) defining the result of the intersection
can be used to evaluate Sc . It can be stated that Sc = 0 if f 3 ( X ) < 0 for any
point of E n (Rvachev 1967). This definition leads to the collision detection
algorithm discussed in Section 4.
3 Interactive geometric modeling based on the function
representation
3.1 A machine representation and a user representation
In the previous section we introduced F-rep as a mathematical notion of an
abstract nature allowing to define basic geometric concepts - objects,
operations, and relations - in terms of real functions of several variables.
This section is devoted to describing our experience of building an
interactive geometric modeling system based on the F-rep notion.
Accordingly, the very notion "representation" will be treated in more
applied sense as concerned with the computer's and user's manipulation and
interpretation of it in our current modeling system.
We share Snyder's belief that "the representation to be the part of a
geometric modeling system which most determines its quality" (Snyder
1992) and consider F-rep as an essentially "user representation" allowing to
specify user's geometric model in the symbolic form. We explicitly
distinguish it from a lower-level "machine representation" which can be
present inside the system. This lower level can be in the form of a
generalized CSG-type tree with, in turn, other levels, such as collections of
polygons. The corresponding set of procedural tools forms a kernel
geometric modeler. Pasko et al. (1993a) give the VDM-specification of
such a modeler which formally defines principal data structures of a
"machine representation" as well as operations over them. This modeler
together with a visualization subsystem serve as a basis for building an
interactive geometric modeling system.
3.2 Geometric language and the example of modeling
F-rep as a user representation serves as a base for a high-level geometric
language which is the user's instrument for modeling specification. To
realize how significant features of F-rep are reflected in the geometric
language, let us consider the corresponding modeling program (Fig.6). This
program creates a model of cycled metamorphosis between the following
four geometric objects which are normally modeled through different
representational styles:
1. Constructive object defined with help of the set-theoretic operations on
primitives;
2. Swept object defined by the Cartesian product and the subsequent
bijective mapping;
3. Voxel-based object built by manual sculpting similar to (Galyean and
Hughes 1991) with the subsequent trilinear interpolation providing C 0
continuity to make this volumetric tea-pot to be a "legal" F-rep object;
4. "Blobby" object as a representative of objects with analytically defined
implicit surfaces.
Finally, metamorphosis itself is modeled with using the objects described as
"key volumes" and eventual getting necessary intermediate volumes in
accordance with (15). Fig.7 represents the frames of the computer film
corresponding to the specification of the resultant 4D geometric object
gob_metamorp from Fig.6.
The program being interactively introduced during a modeling session
consists of the following parts:
–
Geometric model
–
Geometric types
–
Environment
Each part can be defined and changed irrespective of other ones in the
process of interactive modeling work. The brief description of these parts is
given below.
3.2.1 Geometric Model
This is actually a parametrized specification of geometric objects themselves
which is intimately related with F-rep. Geometric objects are given by their
defining functions. In keeping with the mathematical framework described
in Section 2, each function defining complex geometric object gob_<name>
is built as a rather traditional mathematical expression with using symbols of
coordinate variables x i , geometric primitives pob_<name>, numerical
constants and parameters, arithmetic operations, standard algebraic
functions (sin, cos, log, min, max, etc.), and built-in set-theoretic operations
implemented by R-functions ( | - for union, & - for intersection, \ - for
subtraction, and @ - for Cartesian product). One can make choice of the
type of R-functions by setting r_alpha parameter. If built-in pob_block_3D
was defined with help of "min-max" system of R-functions with r_alpha=1
(3), the user can introduce new one defined with help of the system of Rfunctions (4) with better continuity properties. Such operators of structured
programming as "if-then-else" and "while-do" together with complete set of
logical functions provide a proper programming flexibility allowing to define
complex and non-traditional geometric transformations and relations.
Note, besides built-in conventional primitives and operations one can
introduce new ones during a modeling session. For instance, if there is a
need in a "block" with blended edges, one can define the following
"blending intersection" transformation based on (8):
tr_bl_int( (gob|pob) g1, (gob|pob) g2, (real)a1, (real)a2, (real)a3 ) =
g1+g2-sqrt(g12+g22)+a1/(1+(g1/a2)2+(g2/a3)2);
Then this operation can be applied to half-spaces to get a block with smooth
edges and corners.
3.2.2 Geometric Types
Each coordinate variable x i can be associated with a certain "geometric
type". These types establish conventions governing their semantics by giving
geometric interpretation of x i . This interpretation can be important during
exploring the geometric model, in particular in visualization. There are the
following geometric types in the current version of our system:
–
"constant" : Variables of this type are assigned numerical values to
define a single cross-section, such as x i = const ;
–
"g" : These variables define a group of constants corresponding to
several cross-sections;
–
"x", "y", "z" : These are coordinate variables corresponding to axes in
the 3D Cartesian coordinate system;
–
"t" : Variables of this type model a course of time with a possibility of its
incremental or decremental changing that can be used in animation.
Time-dependent geometric objects are called geometric processes in our
system;
–
"v", "w" : These correspond to the additional V-axis and W-axis for
building a geometric spreadsheet with elementary images in cells to
support
so-called
inductive
approach
to
multivariate
function
visualization (Pasko et al. 1992);
–
"c" : Variables of this type are used for mapping to colors within a
spectral range.
Note, that geometric types "constant" and "g" can also be assigned to the
defining function of the resultant geometric object. Normally, the zero value
sets boundaries of a geometric object. If "g" type is assigned, visualization
of corresponding isolines and isosurfaces can be very useful for exploring
the features of the defining function (see Fig.3 as an example).
3.2.3 Environment
This section is intended for concretizing those parameters being present in
the "Geometric Model" which were denoted by their abstract names. The
first sub-section defines ranges [ x i min , x i max ] of coordinate variables to set
boundaries of a modeling space. Then an incremental interval xi_delta for
coordinate variable with geometric type "t" must be defined. In the example,
the value xi_delta=0.5 lets us get one intermediate frame between each pair
of "key volumes". The next sub-section deals with assigning values of
numerical parameters including arrays. The user can specify necessary
visualization parameters that essentially depend on the assignment of
geometric types to coordinate variables.
3.3. Advantages of Function Representation as a Base for Interactive
Modeling System
We use (with some modifications) the following three criteria for evaluation
of F-rep that were introduced by Snyder (1992) for a user representation as
a special adaptation of the categories from [Requicha 1980].
1) Ease of specification.
This is the first and the most crucial criterion for evaluation of the quality of
a user representation which basically assesses how efficiently the users can
define and change their geometric model.
The function representation is closed, meaning that further operations can
be applied to results of previous ones. This representation is also uniform in
the sense that it supports combinations of representational styles to describe
objects of traditionally different nature. These two properties were
substantiated above.
It results in the conclusion about the higher abstraction level as regards
many other representations. Uniform definition for objects in spaces of
various dimensions as well as uniform representation of static and timedependent objects are especially attractive. Moreover, multidimensionality
can be interpreted in the interesting and natural for this representation way
through the concept of geometric types.
It is obvious that F-rep, because of its analytical nature, provides such
important for user's modeling work categories as compactness and accuracy
of a model.
Just because of thousand-year tradition of analytical description of
geometry, we consider F-rep as natural for users we are oriented towards.
Even elementary knowledge of analytical geometry and brief training let the
users connect the way they think about geometric shapes with symbolic
descriptions in the form of analytical expressions. The very nature of F-rep
lets the users easily change some parameters in the analytical model and
observe the visual or computational effect setting the understandable
correspondence between analytical description and model's behavior. So,
editability and controllability of description are naturally provided.
At last but not at least one should mention extendibility meaning the
possibility of introducing new primitives and operations not only by
procedural defining but also by symbolic one during a modeling session. It
lets creating a specific modeling system for particular application areas and
even for particular users who are able to change the system in some item if
they want. This is a base for "Empty Case" technology of geometric
modeling (Pasko et al. 1993b) which supposes working process of
"absolutely first user" who has to create his personal geometric modeling
system defining necessary primitives and transformations in a symbolic
manner. This seems to be especially valuable in educational perspective, and
corresponding project is being realized at the University of Aizu.
2) Renderability.
We consider F-rep as a quite suitable for providing fast visual feedback
given to the user. It is achieved through conversion to a polygonal mesh.
The original algorithm of polygonization of an isosurface f ( x1 , x 2 , x 3 ) = 0
(Pasko et al. 1988) was used to generate the frames in Fig.7. This algorithm
can be easily decomposed in order to map on parallel computer architecture.
Higher quality rendering is also possible with ray casting which also
undergoes parallelization very well. Compactness of F-rep allows to run
rendering software even on parallel computers with not very large size of
distributed memory (e.g. transputer networks).
3) Analyzability.
Geometric queries such as point inclusion are simple for the function
representation. It helps to compute physical quantities about the shape
(volume, moments of inertia) with well-known algorithms. Collision
detection can be realized by a maximum search procedure (see Section 2.3
and the example in Section 4). On the other hand, finding curves of
intersection of surfaces require slow numerical algorithms.
3.4 Perspective interactive environment
The function representation as a high-level user representation fits very well
for "exploratory geometric modeling". It means not simply describing
geometric model whose properties and features are preliminary given, but
rather introducing new geometric objects and transformations with
subsequent exploring their characteristics and behaviour in an interactive
manner. This creative process is similar to a traditional scientific
investigation of a physical phenomenon when experimenting with the model
created and observing its behaviour under changeable conditions are
performed.
We think the corresponding interactive environment can be built on the
basis of the "definitive-based" programming paradigm (Beynon 1989) and
the agent-oriented framework which provides easy interactive way for
modifying the specification (both parameters in defining functions and
functions themselves) with indivisible propagation of changing any
dependent entities.
The corresponding implementation work is in progess. Adzhiev et al. (1994)
propose LSD-specification of a geometric modeling system. This
specification describes within agent-oriented modeling framework both the
interactions of the user with the system and the interactions between the
principal components of the system itself. This specification can serve as a
basis for parallel implementation of the interactive modeling system with an
advanced graphical user interface.
4 Application examples
Here we give several examples of application of geometric modeling
software based on the function representation principles.
4.1 Aesthetic design
Although the blending operations described in 2.2 provide easy shape
control, they seem too indirect for the aesthetic blending. Rather than to
change numeric parameters, a designer prefers to define the shape of an
aesthetic blend with a single hand-drawn stroke. These points are then used
to estimate values of parameters which define the blend as close to this
stroke as possible. This is illustrated in Fig.8. The stroke belongs to a
certain plane. Points of hand-drawn strokes are assumed to belong to the
blending surface. These constraints determine a system of nonlinear
equations:
F ( f1 ( x i , yi , z i ), f2 ( x i , y i , z i ), a0 , a1 , a2 ) = 0 , i = 1, N
where f1 and f2 are defining functions of initial objects, F defines the
blending set-theoretic operation (6), and N is a number of points. To find
out the best estimation of a0 , a1 and a2 in the sense of least squares, we
applied a simple quadratic search using random points for the initial
estimate.
4.2 Simulation of collisions
Several application problems (e.g., air and water quality control, collision
dynamics of bodies in celestial mechanics, computer games) deal with
irregularly shaped interacting solids. Fig.9 illustrates the simulation of
colliding particles sticking to each others. It presents collisions of the noisy
block, small spheres, torus and noisy ellipsoid. The spheres having collision
point with the noisy block are stuck to it by the blending union operation.
The surfaces of the block and the ellipsoid were generated using solid noise
that is discussed below. All bodies have changeable orientation in the space
and interact each other as rigid bodies. The collision detection algorithm is
based on the intersection relation defined in 2.3. To find out a collision point
of two particles we apply maximum searching algorithm to the defining
function of particles intersection. The admissible domain is detected using
bounding spheres. Then we use spiral quadratic search within this domain to
detect a point with positive or zero function value. The simulation algorithm
was implemented on transputers T805 in OCCAM-2 (INMOS 1988) and is
described in Savchenko and Pasko (1993). It has simulated 1320 time steps
for 35 small spheres. Steps 1020 and 1023 in Fig.9 illustrate the collision
event between the torus and the newly formed object.
4.3 Modeling NC machining
The set-theoretic operations between moving solids can be used to model a
process of NC machining. The result can be defined with the set-theoretic
subtraction between the workpiece model and the swept model of the
moving tool. Sourin and Pasko (1994) have proposed the procedural
function representation of a swept solid with an envelope surface for a
parametrically defined trajectory. Fig.10 illustrates its application to
modeling the time-dependent set-theoretic operations.
4.4 Reconstruction of solids from cross-sections
Tomography, range data processing, and other applications need
reconstruction of a solid from its given cross-sections. The metamorphosis
operation (15) describes a solid using defining functions of two parallel
cross-sections. Generation of a defining spline function of a cross-section
given by its contour points is proposed by Savchenko et al. (1994).
Reconstruction process is illustrated by Fig.11. Note that this approach is
capable of generating highly concave and branching solids automatically.
4.5 Three-dimensional texture generation
To obtain 3D textures on constructive solids, Pasko and Savchenko (1993)
have proposed to apply the set-theoretic and other operations to a solid
defined by a "solid noise" function. Fig.12a shows an irregularly shaped
vase. The initial vase was designed using aesthetic blending (see above) of
several ellipsoids. Irregular shape was obtained by metamorphosis between
the initial vase and solid noise. Fig.12b shows fur obtained by offsetting and
set-theoretic intersection applied to an initial solid and fur strands defined
using solid noise.
5 Summary and future work
In conclusion, we would like to summarize the main advantages of F-rep
and to discuss future research developments.
The function representation for geometric objects offers a number of
advantages:
–
Higher abstraction level as regards other known representations is
provided.
Combinations
of
representational
styles,
including
constructive geometry, sweeping, soft (blobby) objects, and voxel-based
objects, are supported.
–
Possibility of the symbolic definition of new primitives, operations and
predicates is naturally provided. A symbolic description of a complex
geometric object as a result of modeling can be generated too.
–
Uniform representation for objects defined in spaces of various
dimensions. Dimension increasing (Cartesian product) and dimension
decreasing (projection) operations are supported. Static objects and
time-dependent geometric processes are described uniformly with the
time concerned as one of coordinates.
–
Convenience of designing application algorithms especially for parallel
computers. Compactness of the representation allows to implement
application algorithms even on parallel computers with not very large
size of distributed memory, for example, on a transputer network.
There are several problems of using F-rep. Evaluation of defining function
in a given point is a time consuming task. Moreover, if R-functions with
square roots are applied, calculations become slower. The halftone images
presented in the paper have been produced on Silicon Graphics Indigo2
using ray-casting. Average time for 200x200 image calculation in double
precision is 20-90 sec. One can suppose that numerical stability of nested
square roots (see Eqs.2) is questionable. Our numerical experiment has
shown that R-union (4) applied to 10000 different spheres (calculated in
double precision) gives error 0.2*10sup(-15) for zero value of the defining
function in a boundary point. Although graphic workstations provide
acceptable response time even for ray-casting, we see the final solution in
parallel computing and special hardware. Parallel implementation of the
polygonization algorithm (Savchenko and
Pasko
1994)
improved
performance of computations and practically linearly scaled with the number
of processors.
In practical systems, conversion from boundary representations may be
required. Shapiro (1994b) has proposed that the way to convert B-rep to
F-rep is first to convert it to a constructive representation using standard
(non-regularized) set operations, and then to F-rep using R-functions. The
problem of converting B-rep to the standard set operations is similar to
B-rep/CSG conversion (Shapiro and Vossler 1993) but not the same. This
problem needs further investigation.
Now, there is not direct connection between F-rep and parametric
representations. Because parametric surfaces are very suitable for
interactive geometric design, we try to incorporate these models in the
function representation. Spline controlled deformations of constructive
solids are also very attractive.
The function representation requires users to define a model in a highly
abstract way. Coordinate variables, numerical constants and parameters,
arithmetic operations and standard algebraic functions are always needed to
be explicitly defined by the users. It can be difficult for general users to
define objects in this way.
Although symbolic description is a powerful modeling tool, adequate
graphical user interface has to be specified for F-rep based modeling.
However, the possibility to extend a modeling system with symbolic
descriptions of new elements has to be preserved. Interrelations with
computer algebra systems will be under research. Now we are applying the
function representation to virtual reality applications.
Acknowledgements
Our sincere thanks to Dr. V.V.Pilyugin from MEPI (Russia) and
Dr. W.M.Beynon from University of Warwick (UK) for fruitful theoretical
discussions. We wish to thank Dr. V.A.Galatenko and V.K.Nikolaev from
SRISA RAS (Russia) for software support. The authors also wish to
acknowledge referees for their helpful comments.
References
Adzhiev VD, Beynon WM, Pasko AA (1994) Interactive geometric
modeling based on R-functions: an agent-oriented approach, Settheoretic Solid Modeling:
Techniques and Applications, CSG'94
Conference Proceedings, Information Geometers, Winchester, UK,
253-272.
Bajaj C, Ihm I, Warren J (1993) Higher-order interpolation and leastsquares approximation using implicit algebraic surfaces, ACM
Transactions on Graphics, 12(4):327-347.
Barr A (1981) Superquadrics and angle-preserving transforms, IEEE
Computer Graphics and Applications, 1(1):11-23.
Barr AH (1984) Global and local deformations of solid primitives,
Computer Graphics, 18(3):21-30.
Baraff D (1990) Curved surfaces and coherence for non-penetrating rigid
body simulation, Computer Graphics, 24(4):19-28.
Beynon WM (1989) Evaluating definitive principles for interaction in
graphics, New Advances in Computer Graphics, Proceedings of CG
International'89, 291-303.
Blinn JF (1982) A generalization of algebraic surface drawing, ACM
Transactions on Graphics, 1(3):235-256.
Bloomenthal J (1988) Polygonization of implicit surfaces, Computer Aided
Geometric Design, 5(4):341-355.
Bloomenthal J, Wyvill B (1990) Interactive techniques for implicit
modeling, Computer Graphics, 24(2):109-116.
Bloomenthal J, Shoemake K (1991) Convolution surfaces, Computer
Graphics, 25(4):251-256.
Duff T (1992) Interval arithmetic and recursive subdivision for implicit
functions and constructive solid geometry, Computer Graphics,
26(2):131-138.
Essa I.A., Sclaroff S., Pentland A. (1992) A unified approach for physical
and geometric modeling for graphics and animation, Computer
Graphics Forum, 11(3):129-138.
Galyean TA, Hughes JF (1991) Sculpting: an interactive volumetric
modeling technique, Computer Graphics, 25(4):267-274.
Gascuel MP (1993) An implicit formulation for precise contact modeling
between flexible solids, Computer Graphics Proceedings, Annual
Conference Series, 313-320.
Hoffmann C, Hopcroft J (1987) The potential method for blending surfaces
and corners, Geometric Modeling, G.Farin, Ed., SIAM, Philadelphia,
347-365.
Hoffmann CM (1989) Geometric and solid modeling, Morgan Kaufmann,
San Mateo, California.
Hoffmann CM (1993) Implicit curves and surfaces in CAGD, IEEE
Computer Graphics and Applications, 13(1):79-88.
Hughes JF (1992) Scheduled Fourier volume morphing, Computer
Graphics, 26(2):43-46.
INMOS (1988) OCCAM-2. Reference manual, INMOS Limited, UK.
Lorensen WE, Cline HE (1987) Marching cubes: a high resolution 3D
surface construction algorithm, Computer Graphics, 21(4):163-169.
Middletich AE, Sears KH (1985) Blend surfaces for set theoretic volume
modeling systems, Computer Graphics, 19(3):161-170.
Muraki S (1991) Volumetric shape description of range data using Blobby
model, Computer Graphics, 25(4):227-235.
Pasko A (1988) Conceptual and instrumental tools for a direct method of
multidimensional geometric problems solving with a computer, Ph.D.
dissertation, Moscow Engineering Physics Institute, Russia.
Pasko A, Pilyugin V, Pokrovskiy V (1988) Geometric modeling in the
analysis of trivariate functions, Computers & Graphics, 12(3/4):457465.
Pasko A, Adzhiev V, Prostakov I (1992) Multivariate functions
visualization:
the inductive approach, Proceedings of Third
Eurographics Workshop on Visualization in Scientific Computing,
Viareggio (Italy) 27-29 April 1992, 303-316.
Pasko A, Savchenko V, Adzhiev V, Sourin A (1993a) Multidimensional
geometric modeling and visualization based on the function
representation of objects, TR 93-1-008, The University of Aizu, Japan.
Pasko AA, Savchenko VV, Adzhiev VD (1993b) Educational project:
Empty Case technology of geometric modeling, EDUGRAPHICS'93
and COMPUGRAPHICS'93, Combined Proceedings, Ed. H.P.Santo,
ACM, Portugal, 6-11.
Pasko AA, Savchenko VV (1993) Solid noise in the constructive solid
geometry, EDUGRAPHICS'93 and COMPUGRAPHICS'93, Combined
Proceedings, Ed. H.P.Santo, ACM, Portugal, 351-357.
Pasko AA, Savchenko VV (1994a) Blending operations for the functionally
based
constructive
geometry,
Set
Theoretic
Solid
Modeling:
Techniques and Applications, CSG 94 Conference Proceedings,
Information Geometers, Winchester, UK, 151-161.
Pasko AA, Savchenko VV (1994b) Offsetting operations for the
functionally based constructive geometry, 6th International Conference
on Engineering Computer Graphics and Descriptive Geometry, Tokyo,
Japan (to appear)
Pilyugin VV, Adzhiev VD, Pasko AA, Sourin AI (1988) Theoretical issues
of dynamic and multidimensional geometric modeling with a computer,
Problems of informational systems, No.5, 1988, ICSTI, Moscow, 3-35.
Requicha AAG (1980) Representations for rigid solids: theory, methods,
and systems, Computing Surveys, 12(4):437-464.
Requicha AAG, Rossignac JR (1992) Solid modeling and beyond, IEEE
Computer Graphics and Applications, 12(9):31-44.
Ricci A.
(1973) A constructive geometry for computer graphics, The
Computer Journal, 16(2):157-160.
Rockwood AP (1989) The displacement method for implicit blending
surfaces in solid models, ACM Transactions on Graphics, 8(4):279297.
Rvachev VL (1963) On the analytical description of some geometric
objects, Reports of Ukrainian Academy of Sciences, 153(4):765-767.
Rvachev VL (1974) Methods of logic algebra in mathematical physics,
Naukova Dumka, Kiev.
Savchenko VV, Pasko AA (1993) Simulation of dynamic interaction
between rigid bodies with time-dependent implicitly defined surfaces,
Parallel Computing and Transputers, ed. D.Arnold, IOS Press,122-129.
Savchenko VV, Pasko AA, Okunev OG, Kunii TL (1994) Function
representation of solids reconstructed from scattered surface points and
contours, TR 94-1-032, The University of Aizu, Japan.
Savchenko VV, Pasko AA (1994) Parallel polygonization of implicit
surfaces on transputers, Transputer Research and Applications 7, Ed.
H.Arabnia, IOS Press, pp.22-30.
Schmidt MFW (1993) Cutting cubes - visualizing implicit surfaces by
adaptive polygonization, Visual Computer, 10(2):101-115.
Sclaroff S, Pentland A (1991) Generalized implicit functions for computer
graphics, Computer Graphics, 25(4):247-250.
Shapiro V (1988) Theory of R-functions and applications: a primer, TR
CPA88-3, Cornell University.
Shapiro V, Vossler DL (1993) Separation for boundary to CSG conversion,
ACM Transactions on Graphics, 12(1):35-55.
Shapiro V (1994a) Real functions for representation of rigid solids,
Computer Aided Geometric Design, 11(2):153-175.
Shapiro V (1994b) On B-rep to F-rep conversion, private communication
Snyder JM (1992) Generative modeling for computer graphics and CAD,
Academic Press.
Snyder JM, Woodbury AR, Fleischer K, Currin B, Barr AH (1993) Interval
methods for multi-point collisions between time-dependent curved
surfaces, Computer Graphics Proceedings, Annual Conference Series,
321-334.
Sourin AI (1988) Mathematics and software for problems of timedependent geometry, Ph.D. dissertation , Moscow Engineering Physics
Institute, Russia.
Sourin AI, Pasko AA (1994) Geometric modeling for NC machining with
functionally represented moving solids, IEEE TENCON'94 Conference,
Singapore (to appear)
Wang WP (1984) Solid geometric modeling for mold design and
manufacture, TR 44, Cornell University, Ithaca, New York.
Warren J (1989) Blending algebraic surfaces, ACM Transactions on
Graphics, 8(4):263-278.
Wyvill G, McPheeters C, Wyvill B (1986a) Data structure for soft objects,
The Visual Computer, 2(4):227-234.
Wyvill B, McPheeters C, Wyvill G (1986b) Animating soft objects, Visual
Computer, 2(4):235-242.
{ Metamorphosis }
{ "Geometric model" section }
{ Setting the system of R-functions ("|", "&", "\", "@") }
r_alpha = 1.
{ Primitive "block" }
pob_block_3D(a0,b0,c0,a,b,c,)= (x1-a0 & a0+a-x1)& (x2-b0 & b0+b-x2)& (x3-c0 &c0+c-x3);
{ CSG object }
{ Basic block }
gob_bl1 = p_block_3D(-9.,-8.,-9.,18.,16.,18.);
{ Vertical cylinder }
gob_cyl1 = (r1**2-x1**2-x3**2) & 10.5-x2) & x2+10.5;
{ Horizontal infinite cylinder }
gob_cyl2 = r2**2-x1**2-x2**2;
{ Smaller block with infinite x1-dimensionality }
gob_bl2 = (x2+d1 & d1-x2) & (x3+d1 & d1-x3);
{ final CSG object }
gob_csg = ( ( gob_bl1 | gob_cyl1 ) \ gob_cyl2 ) \ gob_bl2;
{ Swept object built with help of cartesian product and bijective mapping }
gob_swept = ( r3**2 - (x1-c1*SIN(-1+x2/c2)**2 - (x3-c1*COS(-1+x2/c2)**2 ) @ ( x2+10. & 10.-x2);
{ Voxel-based object defined by 3D array in a file with subsequent interpolation }
gob_vox = INTERPOLATE("trilinear", "teapot.dat", x1,x2,x3);
{ blobby object }
gob_blobby = b;
WHILE ( i <= 8 ) DO
gob_blobby = gob_blobby + a[i]*EXP(-SQRT( (x1-px[i])**2+ (x2-py[i])**2+ (x3-pz[i])**2);
{Metamorphosis between all objects }
gob_metamorp =
IF ( 0 <= x4 < = 1 )
THEN gob_csg*(1-x4) + gob_swept*x4;
ELSE IF ( 1 < x4 <= 2 )
THEN gob_swept*(2-x4) + gob_vox*(x4-1);
ELSE IF ( 2 < x4 <= 3 )
THEN gob_vox*(3-x4) + gob_blobby*(x4-2);
ELSE IF ( 3 < x4 <= 4 )
THEN gob_blobby*(4-x4) + gob_csg*(x4-3);
{ "Geometric types" section }
x1 : X;
x2 : Y;
x3 : Z;
x4 : T;
{ "Environment" section }
{ modeling space }
x1min = -11.;
x1max = 11.;
x2min = -11.;
x2max = 11.;
x3min = -11;
x3max = 11.;
x4min = 0.;
x4max = 4.;
{ incremental interval for x4 with geometric type "T" }
x4_delta = 0.5;
{ "Parameters" }
r1 = 4.;
r2 = 6.;
d1 = 5.;
r3 = 3.5;
e1 = -8.;
e2 = 2.5;
b = -0.07;
a = [1.5, 1., 1., 0.8, 0.5, 0.3, 0.4, 0.35 ];
px = [0., 6., -4., -5., 5., 6., 5., 8. ];
py = [0., 6., -4., 3., -5., -7., -6., -9. ];
pz = [0., 6., -4., 7., 5., 9., -3., -4. ];
Captions
Fig.1.
A segment in E 1 described as intersection of two rays. The
intersection operation is defined by different R-functions: (a) α=1,
(b) α=0, (c) m=1.
Fig.2.
Constructing of "CSG" solid. Three initial solids "C", "S", and "G"
are defined as a union of blocks. The final solid is defined as an
intersection ("C"∩"S")∩"G".
Fig.3.
The contour map of the intersection of two 2D halfspaces x≥0 and
y≥0 represented by the R-function with α=0. The contour with f=0
is drawn with the bold line.
Fig.4a. Blending followed by the set-theoretic subtraction (a cylindrical
hole).
Fig.4b. Negative and positive offset solids obtained by the iso-valued
offsetting operation.
Fig.4c. Rounding convex vertices of a 2D solid by the blending based on
the constant-radius offsetting operations.
Fig.5a. Twisting a constructive solid with the bijective mapping.
Fig.5b. A union of three and a projection of the object to a plane
orthogonal to an axes of one of tori.
Fig.5c. The application of the Cartesian product and bijective mapping:
a 3D solid defined by the rotational sweeping.
Fig.5d. Several steps of time dependent metamorphosis between 3D solids.
Fig.6. Metamorphosis modeling program in high-level geometric
language
Fig.7. Frames of metamorphosis process between "key volumes"
reflecting different representational styles: constructive geometry,
sweeping, soft objects and voxel-based objects.
Fig.8
a. The body and the bottom of a wine glass to be connected with
an aesthetic blend defined by the hand-drawn stroke;
b. The result of blending union with the estimated parameters.
Fig.9.
Simulation of colliding particles sticking to each other.
Fig 10. Application of the set-theoretic operations for NC machining:
a. The set-theoretic subtraction between two moving solids;
b. The object cut achieved as a result of the subtraction of the solid
swept by a linearly moving cutter from the rotating workpiece.
Fig.11. Two
functionally
represented
cross-sections
and
a
solid
reconstructed using metamorphosis operation.
Fig.12a. Metamorphosis between a vase and solid noise produces a noisy
vase.
Fig.12b.Furry object defined by the offsetting and intersection with fur
strands described with solid noise.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement