Chapter 14
3D Spherical Geometry Kernel

Pedro Machado Manhães de Castro, Frédéric Cazals, Sébastien Loriot, and Monique Teillaud

Table of Contents

14.1 Introduction
14.2 Spherical Kernel Objects
14.3 Software Design
14.4 Examples
14.5 Design and Implementation History

14.1   Introduction

The goal of the 3D spherical kernel is to offer to the user a large set of functionalities on spheres, circles and circular arcs, in the 3D space or restricted on a given sphere. These functionalities require computing on algebraic numbers, which motivates the creation of a new kernel concept extending the Cgal Kernel concept, that is restricted to objects and functionality in a FieldNumberType.

All the choices (interface, robustness, representation, and so on) made here are consistent with the choices made in the Cgal kernel, for which we refer the user to the kernel manual (Chapter 11).

14.2   Spherical Kernel Objects

New main geometric objects are introduced by Spherical_kernel_3: circular arcs ((model of SphericalKernel::CircularArc_3), points of circular arcs (model of SphericalKernel::CircularArcPoint_3), and line segments (model of SphericalKernel::LineArc_3) whose endpoints are points of this new type.

SphericalKernel::CircularArcPoint_3 is used in particular for endpoints of arcs and intersection points between spheres, circles or arcs. The coordinates of these points are algebraic numbers of degree two. Therefore, general predicates offered by the Kernel on Point_3, which have coordinates in a FieldNumberType, would require heavy algebraic computations in algebraic extensions of higher degrees and thus are not provided on them, which explains the need for a new point type.

A consistent set of predicates and constructions is offered on these new types.

General functionalities  

The spherical kernel currently implements a set of fundamental functionalities like intersection, comparisons, inclusion, etc. More might be provided in the future, as long as only algebraic numbers of degree two are used.

Functionalities relative to a sphere  

The interface of the underlying objects is extended by providing additional operations that make sense only if the objects are considered on the same sphere. For example, the result of the comparison of the cylindrical or spherical coordinates of two points is well-defined only when looking at them on a given common sphere. The presentation of these operations requires the following definitions:

Coordinate system. Let consider a sphere with center c and radius r. Using the Cartesian frame centered at c, we define a cylindrical coordinate system (,z) on that sphere, with [ 0,2π) and z [ -r,r ]. is given in radian and measured in the xy-plane around the z-axis, starting from x>0, y=0. The z-extremal points of a sphere are its North and South poles defined as (,r) and (,-r) respectively, for any value of . Observe that each point on the sphere different from a pole corresponds to a unique pair (,z).

Definition of a meridian. Given a sphere and its associated cylindrical coordinate system, a meridian of that sphere is a circular arc consisting of the points having the same theta-coordinate (the poles are the end points). A plane containing the two poles of that sphere defines two meridians, one on each side of the line passing through the poles. A vector M whose direction is different from that of the latter line defines a unique meridian on that sphere. The plane of that meridian is defined by the direction of M and the two poles. The sense of M disambiguates the choice among the pair of meridians thus defined. On Fig. 14.1, the normal vectors n0 and n1 define two meridians of S: the circular arcs A0 and A1 respectively.

Definition of meridians.

Figure 14.1:  Definition of two meridians on S, a sphere of center c. The intersection of the plane P (passing through the two poles of S) and the sphere S is a circle. The two poles of S split that circle into two circular arcs A0 and A1, each being a meridian of S. The -coordinates of meridians A0 and A1 are 0 and 1= 0 + π respectively.

Types of circles on a sphere. Given a sphere, a circle on that sphere is termed polar if it goes through only one pole, bipolar if it goes through the two poles of that sphere and threaded if it separates the sphere into two connected components, each containing one pole. Any other circle is termed normal. These definitions are illustrated on Fig. 14.2.

The four types of circles on a sphere.

Figure 14.2:  The four types of circles on a sphere. Black dots are the -extremal points.

-extremal points. Given a sphere one has: a -extremal point of a normal circle is a point of tangency between the circle and a meridian anchored at the poles of that sphere. Each normal circle defines two such points; the -extremal point of a polar circle is the pole the circle goes through. No such point is defined on a bipolar or a threaded circle. These definitions are illustrated on Fig. 14.2. Notice that the -extremal points should not be confused with the endpoints of an arbitrary arc on a sphere.

The -coordinate of a -extremal point of a normal circle on a sphere is well defined. For a polar circle on a sphere, the plane containing the two poles and which is tangent to that circle contains two different meridians. The -values of these meridians are the two -coordinates associated to the same -extremal point of a polar circle.

-monotone circular arcs. An arc on a sphere is said to be -monotone if any meridian on that sphere intersects that arc in at most one point. With this definition, a circular arc on a threaded circle is always -monotone, and an arc on a polar or normal circle is -monotone if it does not contain a -extremal point, unless it is an endpoint. No such arc is defined on a bipolar circle.

14.3   Software Design

The design of Spherical_kernel_3 is similar to the design of Circular_kernel_2 (see Chapter 13). It has two template parameters:

The 3D spherical kernel uses the extensibility scheme presented in the kernel manual (see Section 11.5). The types of Kernel are inherited by the 3D spherical kernel and some types are taken from the AlgebraicKernelForSpheres parameter. Spherical_kernel_3 introduces new geometric objects as mentioned in Section 14.2.

In fact, the spherical kernel is documented as a concept, SphericalKernel and two models are provided:

14.4   Examples

The first example shows how to construct spheres and compute intersections on them using the global function.

File: examples/Circular_kernel_3/intersecting_spheres.cpp
#include <CGAL/Exact_spherical_kernel_3.h>
#include <CGAL/Random.h>

typedef CGAL::Exact_spherical_kernel_3         Spherical_k;

typedef CGAL::Point_3<Spherical_k>             Point_3;
typedef CGAL::Sphere_3<Spherical_k>            Sphere_3;

int main() {

  CGAL::Random generatorOfgenerator;
  int random_seed = generatorOfgenerator.get_int(0, 123456);
  CGAL::Random theRandom(random_seed);
  int count = 0;

  std::cout << "We will compute the approximate probability that 3 spheres wit"
  << "h radius 1 intersect on a 5x5x5 box, it might take some time." << std::endl;

  for(int i=0; i<10000; i++) {

    double x1 = theRandom.get_double(0.0,5.0);
    double y1 = theRandom.get_double(0.0,5.0);
    double z1 = theRandom.get_double(0.0,5.0);
    double r = 1.0;
    double x2 = theRandom.get_double(0.0,5.0);
    double y2 = theRandom.get_double(0.0,5.0);
    double z2 = theRandom.get_double(0.0,5.0);
    double x3 = theRandom.get_double(0.0,5.0);
    double y3 = theRandom.get_double(0.0,5.0);
    double z3 = theRandom.get_double(0.0,5.0);

    Sphere_3 s1 = Sphere_3(Point_3(x1,y1,z1), r);
    Sphere_3 s2 = Sphere_3(Point_3(x2,y2,z2), r);
    Sphere_3 s3 = Sphere_3(Point_3(x3,y3,z3), r);

    std::vector< CGAL::Object > intersecs;
    CGAL::intersection(s1, s2, s3, std::back_inserter(intersecs));
    if(intersecs.size() > 0) count++;

  std::cout << "The approximate probability that 3 spheres with radius 1"
            << std::endl;
  std::cout << "choosen (uniformly) randomly on a 5x5x5 box intersect is: "
            << ((double)count)/((double)(10000)) << std::endl;

  return 0;

The second example illustrates the use of a functor.

File: examples/Circular_kernel_3/functor_has_on_3.cpp
#include <CGAL/Exact_spherical_kernel_3.h>
#include <CGAL/Random.h>

typedef CGAL::Exact_spherical_kernel_3         Spherical_k;

typedef CGAL::Point_3<Spherical_k>             Point_3;
typedef CGAL::Circular_arc_3<Spherical_k>      Circular_arc_3;

int main()
  int n = 0;
  Circular_arc_3 c = Circular_arc_3(Point_3(10,10,0), Point_3(5,5,5), Point_3(0, 0, 0));
  for(int i = 0; i <= 10; i++) {
    for(int j = 0; j <= 10; j++) {
      for(int k = 0; k <= 10; k++) {
        Point_3 p = Point_3(i, j, k);
        if(Spherical_k().has_on_3_object()(c,p)) {
          std::cout << "(" << i << "," << j << "," << k << ")" << std::endl;

  std::cout << "There are " << n << " points in the " 
            << "[0,..,10]x[0,..,10]x[0,...,10] " 
            << "grid on the circular" << std::endl
            << " arc defined by the points (10,10,0), (5,5,5), (0,0,0)" 
            << std::endl << "See the points above." << std::endl;
  return 0;

The third example illustrates the use of a functor on objects on the same sphere. The intersection points of two circles on the same sphere are computed and their cylindrical coordinates are then compared.

File: examples/Circular_kernel_3/functor_compare_theta_3.cpp
#include <CGAL/Exact_spherical_kernel_3.h>

typedef CGAL::Exact_spherical_kernel_3               SK;

int main(){
  //construction of 3 spheres from their centers and squared radii
  SK::Sphere_3 s1(SK::Point_3(0,0,0),2);
  SK::Sphere_3 s2(SK::Point_3(0,1,0),1);
  SK::Sphere_3 s3(SK::Point_3(1,0,0),3);

  //construct two circles lying on sphere s1
  SK::Circle_3 C1(s1,s2);
  SK::Circle_3 C2(s1,s3);
  SK::Intersect_3 inter;
  //create a functor to compare theta-coordinates on sphere s1
  SK::Compare_theta_z_3 cmp(s1);
  std::vector< CGAL::Object > intersections;

  //unsigned integer indicates multiplicity of intersection point 
  std::pair<SK::Circular_arc_point_3,unsigned> p1=
    CGAL::object_cast< std::pair<SK::Circular_arc_point_3,unsigned> >(intersections[0]);
  std::pair<SK::Circular_arc_point_3,unsigned> p2=
    CGAL::object_cast< std::pair<SK::Circular_arc_point_3,unsigned> >(intersections[1]);

  SK::Circular_arc_point_3 t_extreme[2];
  //Compute theta extremal points of circle C1 on sphere s1
  //The theta coordinates of theta extremal points of C1 enclose that of each intersection point.
  return 0;

14.5   Design and Implementation History

This package follows the 2D circular kernel package (see Chapter 13), which induced the basic choices of design.

Julien Hazebrouck and Damien Leroy participated in a first prototype.

The first version of the package was co-authored by Pedro Machado Manhães de Castro and Monique Teillaud, and integrated in CGAL 3.4. Frédéric Cazals and Sébastien Loriot extended the package by providing functionalities restricted on a given sphere [dCCLT09].

Sylvain Pion is acknowledged for helpful discussions.

This work was partially supported by the IST Programme of the 6th Framework Programme of the EU as a STREP (FET Open Scheme) Project under Contract No IST-006413 (ACS - Algorithms for Complex Shapes).