Scribd
Upload
20 views23 pages

FC 46

Uploaded by

alex.muravev
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
20 views23 pages

FC 46

Uploaded by

alex.muravev
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 23

Computational Geometry

Algorithms Library
Pierre Alliez
INRIA
Mission Statement
“Make the large body of geometric
algorithms developed in the field of
computational geometry available for
industrial applications”

CGAL Project Proposal, 1996


Algorithms and Datastructures
CGAL in Numbers

500,000 lines of C++ code


10,000 downloads/year (+ Linux distributions)
3,500 manual pages
3,000 subscribers to cgal-announce
1,000 subscribers to cgal-discuss
120 packages
90 commercial users
20 active developers
12 months release cycle
2 licenses: Open Source and commercial
Some Commercial Users

Image
Processing
CAD/CAM
VLSI

Digital maps
Medical

Scientific
visualization

GIS
Geophysics
(Oil&Gas)

Telecom
CGAL Open Source Project
Project = « Planned Undertaking »

• Institutional members make a long term commitment:


Inria, MPI, Tel-Aviv U, Utrecht U, Groningen U,
ETHZ, GeometryFactory, FU Berlin, Forth, U Athens
• Editorial Board
– Steers and animates the project
– Reviews submissions
• Development Infrastructure
– Gforge: svn, tracker, nightly testsuite,...
– 120p developer manual and mailing list
– Two 1-week developer meetings per year
Contributions
• Submission of specifications of new
contributions
• Review and decision by the Editorial Board

• Value for contributor


– Integration in the CGAL community
– Gain visibility in a mature project
– Publication value for accepted
contributions
Exact Geometric Computing
Predicates and Constructions

Predicates Constructions

orientation in_circle intersection circumcenter


Robustness Issues
• Naive use of floating-point arithmetic causes
geometric algorithms to:
– Produce [slightly] wrong output
– Crash after invariant violation
– Infinite loop

• There is a gap between


– Geometry in theory
– Geometry with floating-point arithmetic
Geometry in Theory
ccw(s,q,r) & ccw(p,s,r) & ccw(p,q,s)  ccw(p,q,r)

Correctness proofs of algorithms rely on such


theorems
Demo: The Trouble with Double
orientation(p,q,r) = sign((px-rx)(qy-ry)-(py-ry)(qx-rx))‫‏‬

negative zero positive


Exact Geometric Computing [Yap]

Make sure that the control flow in the


implementation corresponds to the control
flow with exact real arithmetic

orientation(p,q,r)‫‏‬

<0 >0
=0
Filtered Predicates
• Generic functor adaptor Filtered_predicate<>
– Try the predicate instantiated with intervals
– In case of uncertainty, evaluate the predicate
with multiple precision arithmetic

• Refinements:
– Static error analysis
– Progressively increase precision
Filtered Constructions
Lazy number = interval and arithmetic Lazy object = approximated object and
expression tree geometric operation tree
b
s1
l
(3.2 + 1.5) * 13 i
m p
a
s2

m' ([], [])‫‏‬ midpoin


n' []
* t

[] + 13 i' ([], [])‫‏‬ intersect p' ([], [])‫‏‬ projection

3.2 1.5
s1 s2 b l

Test that may trigger an exact re-evaluation:


if ( n' < m' )‫‏‬ if (collinear(a',m',b'))‫‏‬
The User Perspective
• Convenience Kernels
– Exact_predicates_inexact_constructions_kernel
– Exact_predicates_exact_constructions_kernel
– Exact_predicates_exact_constructions_kernel_with_sqrt
• Number Types
– double, float
– CGAL::Gmpq (rational), Core (algebraic)‫‏‬
– CGAL::Lazy_exact_nt<ExactNT>
• Kernels
– CGAL::Cartesian<NT>
– CGAL::Filtered_kernel<Kernel>
– CGAL::Lazy_kernel<NT>

CGAL manual
Merits and Limitations
• Ultimate robustness inside the black box

• The time penalty is reasonable, e.g. 10% for


3D Delauny triangulation of 1M random points

• Limitations of Exact Geometric Computing


– Topology preserving rounding is non-trivial
– Construction depth must be reasonable
– Cannot handle trigonometric functions
Generic Programming
STL Genericity
template <class Key, class Less>
class set {
Less less;

insert(Key k)‫‏‬
{
if (less(k, treenode.key))‫‏‬
insertLeft(k);
else
insertRight(k);
}
};

STL manual
CGAL Genericity
template < class Geometry >
class Delaunay_triangulation_2 {
Geometry::Orientation orientation;
Geometry::In_circle in_circle;

void insert(Geometry::Point t) {
...
if(in_circle(p,q,r,t)) {...}
...
if(orientation(p,q,r){...}
}
};
CGAL Genericity Demo
Without explicit conversion to points in the plane
• Triangulate the terrain in an xy-plane
• Triangulate the faces of a Polyhedron

Courtesy: IPF,Vienna University


of Technology & Inpho GmbH
Summary: Overview
• Open Source project

• Clear focus on geometry


• Interfaces with de facto standards/leaders:
STL, Boost, GMP, Qt, blas

• Robust and fast through exact geometric


computing
• Easy to integrate through generic programming

You might also like