Computational Geometry Algorithms Library

http://parasol.tamu.edu
•  Find the closest emergency phone on campus?
•  How I can go from HRBB to Reed arena?

http://parasol.tamu.edu/CampusNavigator/
•  Find the closest emergency phone on campus?
•  How I can go from HRBB to Reed arena?
•  I want go to HEB, but I don't have car, what I
need to do?
•  Is the marked point inside
of the polygon?
 Computational Geometry
•  A branch of computer science devoted to the
study of algorithms which can be stated in
terms of geometry.[1]

•  Systematic study of algorithms and data

structures for geometric objects, with a focus
on exact algorithms that are asymptotically
fast.[2]

1. http://en.wikipedia.org/wiki/Computational_geometry
2. http://www.cs.uu.nl/geobook/
 Applications of Computational
Geometry
•  Computer Graphics
•  Computer-Aided Design and http://irrlicht.sourceforge.net/

Manufacturing
http://www.gsd.harvard.edu/inside/

•  Robotics cadcam/

•  Geographic Information
Systems
•  Integrated Circuit design http://en.wikipedia.org/

•  Computer-Aided
Engineering
Convex Hull
Convex Hull
Convex Hull
Triangulation
Triangulation
Delaunay Triangulation
Delaunay Triangulation
Delaunay Triangulation
Voronoi diagram
Voronoi diagram
Voronoi diagram
 CGAL (Computational Geometry
Algorithms Library)
  The goal of CGAL Open Source project, is to
provide easy access to efficient and reliable
geometric algorithms in the form of C++
libraries.
  Offers data structures and algorithms, all of
these operate on geometric objects like points
and segments, and perform geometric tests on
them.

http://www.cgal.org/
 Structure of CGAL

Basic Library
Algorithm and Data Structures
Support
Library

Kernel Visualization
Geometric Objects File
Geometric Operations I/O
NumberTypes
Generators
...
Core Library
Configurations, assertions, ...

http://www.loria.fr/~lazard/CGAL/2004/intro.pdf
 CGAL Kernel
  Elementary geometric objects
  Elementary computation of them

  Primitives 2D, 3D, dD

  Point, Vector, Triangle, Iso_Rectangle,Circle, …
  Predicates
  Comparison, Orientation, InSphere, …
  Constructions
  Intersection, square distance, ...
http://www.loria.fr/~lazard/CGAL/2004/intro.pdf
 Kernel and NumberTypes
  Cartesian   Homogeneous
representation representation
Point: (x,y) Point: (hx,hy,hw)

CGAL::Cartesian<FT> CGAL::Homogeneous<RT>
double, Quotient<Gmpz>, int, Gmpz, double
leda_real
typedef double NumberType;
typedef Cartesian< NumberType > Kernel;
typedef Kernel::Point_2 Point;

http://www.loria.fr/~lazard/CGAL/2004/intro.pdf
 Kernel and NumberTypes
typedef CGAL::Cartesian<NT> Kernel;
NT sqrt2 = sqrt( NT(2) );
Kernel::Point_2 p(0,0), q(sqrt2,sqrt2);
Kernel::Circle_2 C(p,2);
assert( C.has_on_boundary(q) );

Assertion if NT doesn't gives an exact sqrt

http://www.loria.fr/~lazard/CGAL/2004/intro.pdf
 CGAL Predicates & Constructions

Predicates Constructions
s
r
r

r
q

Orientation in_circle Intersection circumcenter

http://www.loria.fr/~lazard/CGAL/2004/intro.pdf
 CGAL Basic Library

http://www.loria.fr/~lazard/CGAL/2004/intro.pdf
 CGAL Basic Library
  Convex Hull (5 alg. In 2D and 3 alg. In 3D)
  2D/3D Triangle/Tetrahedron based data-
structure
  2D/3D Delaunay triangulation
  2D/3D Regular Triangulation
  2D Voronoi Diagram
  2D Meshes

http://www.loria.fr/~lazard/CGAL/2004/intro.pdf
 CGAL Basic Library
  Polyhedra data-structure
  Half-edge
  Polyhedral surface
  2D Nef polygons
  3D Nef polyhedra
  Geometric Optimizations
  Smallest enclosing circle and ellipse in 2D
  Smallest enclosing sphere in dD
  Largest empty rectangle
  ...
http://www.loria.fr/~lazard/CGAL/2004/intro.pdf
 CGAL Basic Library
  Line segments or polylines
  Search structures
  Range-tree, Segment-tree,kD-tree
  Window query
  Approximate nearest neighbors
  ...

http://www.loria.fr/~lazard/CGAL/2004/intro.pdf
 Installation prerequisites
  Compilers
  GNU g++ 4.0, 4.1, 4.2, 4.3
  MS Visual C++ 8.0, 9.0
  cmake: a cross-platform ``makefile
generator'' (http://www.cmake.org/)
  Boost Library: (http://www.boost.org/ or http://
www.boostpro.com/products/free for Windows
installer)
 Prerequisites
Optional
  GMP+MPFR and LEDA: Exact arithmetic
(http://gmplib.org/, http://www.mpfr.org/, http://
www.algorithmic-solutions.com/leda/)
  Visualization
  Trolltech's cross-plaform GUI toolkits Qt3 or
Qt4
  Geomview
 Installation
  The CGAL library can be downloaded from
http://www.cgal.org/download.html

  Unpack source code:

tar xzf CGAL-3.8.tar.gz
  Windows installer: CGAL-3.8-Setup.exe

  CGAL manual must be downloaded separately

from http://www.cgal.org/download.html
 Directories
  config: configuration files for install script
  cmake/modules: modules for finding and
using libraries
  demo: demo programs (most of them need
QT, geomview or other third-party products)
  doc_html: Installation documentation (HTML)
  examples: example programs
  Include: header files
  scripts: some useful scripts (e.g. for creating
CMakeLists.txt files)
 Configure and Build CGAL
  Using cmake GUI
cd CGAL-3.8
cmake-gui .
make
  Using cmake command line

cd CGAL-3.8
cmake .
make
 Installing CGAL
  After configure and build
make install
  By default: /usr/local

  Change CMAKE_INSTALL_PREFIX to the

preferred location to install
 Compile examples
  Compile the demos and examples
cd examples/Core
cmake .
make
  To generate verbose makefile (show executed

command) use
cmake -DCMAKE_VERBOSE_MAKEFILE=ON .
 Simple example
#include <iostream>

#include <CGAL/Exact_predicates_inexact_constructions_kernel.h>

#include <CGAL/convex_hull_2.h>
#include <vector>

typedef CGAL::Exact_predicates_inexact_constructions_kernel K;

typedef K::Point_2 Point_2;

int main() {

Point_2 points[5] = { Point_2(0,0), Point_2(10,0), Point_2(10,10), Point_2(6,5),

Point_2(4,1) };

std::vector<Point_2> result;

CGAL::convex_hull_2( points, points+5, std::back_inserter(result) );

std::cout << result.size() << " points on the convex hull\n";

return 0;

}
 Compile simple example
  Without cmake:
g++ -frounding-math -O3
-o simple_example -lCGAL
simple_example.cpp
  Using cmake:

  Create CmakeLists.txt file using CGAL-3.8/scripts/

cgal_create_cmake_script
  cmake .
  make simple_example
 Timing your experiments
  Dedicated machine, minimal background processes
  Measure “wallclock” time

  ctime
  gettimeofday
  PAPI counters
  ...
  Measure within clock resolution
  Average multiple executions

  Choose appropriate input size range

  Use optimization level -O3

https://parasol.tamu.edu/people/amato/Courses/620/cgal/simple-cgal-demo.tgz
 Determine number of iterations
Confidence interval
  m: number of samples (iterations)
  : sample mean
  s: sample standard deviation
  r: tolerance/accuracy (e.g., 0.05)
  z: confidence coefficient (95% is 1.96)
  Conditional (require compute mean, s and m on each
iteration):

While continue iterating

  Precomputed (require compute a et of initial samples):

 Timing your experiments
// Computes the difference between two times. Result in seconds
double diff_time(timeval t0, timeval t1) {
return (t1.tv_sec-t0.tv_sec) + (double(t1.tv_usec-t0.tv_usec)/1000000);
}
int main(int argc, char const **argv) {
// Initialize set of points
….
timeval t0,t1;
double elapse = 0;
for(int i=0; i < num_iter; ++i) {
gettimeofday(&t0,NULL);
CGAL::convex_hull_2( points.begin(), points.end(), std::back_inserter(result) );
gettimeofday(&t1,NULL);
elapse += diff_time(t0,t1);
}
std::cout << result.size() << " points on the convex hull in " << (elapse/num_iter) << " sec" << std::endl;
 Finding Big-O constant
A function T(n) is O(F(n)) if for some constant c and for all
values of n greater than some value

Where T(n) is the exact complexity (execution time) for

the input size n

Experimentally determining c and n0

•  Let F(n) be the asymptotic complexity and let T(n) be the
measured time of your algorithm
•  Run experiments for a wide range of values of n and plot T(n)/F
(n) versus n. The value of n where T(n)/F(n) levels off is n0 and the
y value that it converges to is the constant c
 Finding Big-O constant
Example: computing ratio of expected time F(n)
and measured time T(n)
n time (sec) T(n)/F(n)

100 1.84E-006 1.8435E-008

1000 8.38E-006 8.3787E-009

10000 7.51E-005 7.5094E-009

100000 0.000736748 7.36748E-009

1000000 0.00757946 7.57946E-009

 Finding Big-O constant
Example: graphically computing c and n0

c
 More information
•  CSCE-620 Computational Geometry
http://parasol.tamu.edu/people/amato/Courses/620/

•  CGAL
–  Tutorials: http://www.cgal.org/Tutorials/
–  Manuals: http://www.cgal.org/Manual/
–  Videos: http://www.cgal.org/videos.html