\( \newcommand{\E}{\mathrm{E}} \) \( \newcommand{\A}{\mathrm{A}} \) \( \newcommand{\R}{\mathrm{R}} \) \( \newcommand{\N}{\mathrm{N}} \) \( \newcommand{\Q}{\mathrm{Q}} \) \( \newcommand{\Z}{\mathrm{Z}} \) \( \def\ccSum #1#2#3{ \sum_{#1}^{#2}{#3} } \def\ccProd #1#2#3{ \sum_{#1}^{#2}{#3} }\)
CGAL 5.0.4 - Shape Detection
CGAL::Shape_detection::Efficient_RANSAC< Traits > Class Template Reference

#include <CGAL/Shape_detection/Efficient_RANSAC/Efficient_RANSAC.h>

Definition

Shape detection algorithm based on the RANSAC method.

Given a point set in 3D space with unoriented normals, sampled on surfaces, this class enables to detect subsets of connected points lying on the surface of primitive shapes. Each input point is assigned to either none or at most one detected primitive shape. The implementation follows [2].

Template Parameters
Traitsmust be a model of EfficientRANSACTraits.
Examples:
Shape_detection/efficient_RANSAC_and_plane_regularization.cpp, Shape_detection/efficient_RANSAC_basic.cpp, Shape_detection/efficient_RANSAC_with_callback.cpp, Shape_detection/efficient_RANSAC_with_custom_shape.cpp, Shape_detection/efficient_RANSAC_with_parameters.cpp, and Shape_detection/efficient_RANSAC_with_point_access.cpp.

Classes

struct  Parameters
 Parameters for the shape detection algorithm. More...
 

Types

typedef Traits::Input_range Input_range
 Model of the concept Range with random access iterators, providing input points and normals through the following two property maps. More...
 
typedef Traits::Point_map Point_map
 Property map to access the location of an input point.
 
typedef Traits::Normal_map Normal_map
 Property map to access the unoriented normal of an input point.
 
typedef Shape_base< Traits > Shape
 Shape type.
 
typedef Plane< Traits > Plane_shape
 Plane shape type.
 
typedef Iterator_range< Point_index_iterator > Point_index_range
 

Initialization

 Efficient_RANSAC (Traits t=Traits())
 Constructs an empty shape detection object.
 
 ~Efficient_RANSAC ()
 Releases all memory allocated by this instance including shapes.
 
const Traits & traits () const
 Retrieves the traits class.
 
const Point_mappoint_map () const
 Retrieves the point property map.
 
const Normal_mapnormal () const
 Retrieves the normal property map.
 
Input_iterator input_iterator_first () const
 
Input_iterator input_iterator_beyond () const
 
void set_input (Input_range &input_range, Point_map point_map=Point_map(), Normal_map normal_map=Normal_map())
 Sets the input data. More...
 
template<class Shape_type >
void add_shape_factory ()
 Registers the shape type ShapeType in the detection engine that must inherit from Shape_base. More...
 
bool preprocess ()
 Constructs internal data structures required for the shape detection. More...
 

Memory Management

void clear_shape_factories ()
 Removes all shape types registered for detection.
 
void clear_octrees ()
 Frees memory allocated for the internal search structures but keeps the detected shapes. More...
 
void clear ()
 Calls clear_octrees() and removes all detected shapes. More...
 

Detection

bool detect (const Parameters &options=Parameters(), const std::function< bool(double)> &callback=std::function< bool(double)>())
 Performs the shape detection. More...
 

Access

Shape_range shapes () const
 Returns an Iterator_range with a bidirectional iterator with value type boost::shared_ptr<Shape> over the detected shapes in the order of detection. More...
 
Plane_range planes () const
 Returns an Iterator_range with a bidirectional iterator with value type boost::shared_ptr<Plane_shape> over only the detected planes in the order of detection. More...
 
std::size_t number_of_unassigned_points () const
 Number of points not assigned to a shape.
 
Point_index_range indices_of_unassigned_points ()
 Returns an Iterator_range with a bidirectional iterator with value type std::size_t as indices into the input data that has not been assigned to a shape.
 

Member Typedef Documentation

◆ Input_range

template<class Traits >
typedef Traits::Input_range CGAL::Shape_detection::Efficient_RANSAC< Traits >::Input_range

Model of the concept Range with random access iterators, providing input points and normals through the following two property maps.

Member Function Documentation

◆ add_shape_factory()

template<class Traits >
template<class Shape_type >
void CGAL::Shape_detection::Efficient_RANSAC< Traits >::add_shape_factory ( )

Registers the shape type ShapeType in the detection engine that must inherit from Shape_base.

For example, for registering a plane as detectable shape, you should call ransac.add_shape_factory< Shape_detection::Plane<Traits> >();. Note that if your call is within a template, you should add the template keyword just before add_shape_factory: ransac.template add_shape_factory< Shape_detection::Plane<Traits> >();.

◆ clear()

template<class Traits >
void CGAL::Shape_detection::Efficient_RANSAC< Traits >::clear ( )

Calls clear_octrees() and removes all detected shapes.

All internal structures are cleaned, including formerly detected shapes. Thus iterators and ranges retrieved through shapes(), planes() and indices_of_unassigned_points() are invalidated.

◆ clear_octrees()

template<class Traits >
void CGAL::Shape_detection::Efficient_RANSAC< Traits >::clear_octrees ( )

Frees memory allocated for the internal search structures but keeps the detected shapes.

It invalidates the range retrieved using unassigned_points().

◆ detect()

template<class Traits >
bool CGAL::Shape_detection::Efficient_RANSAC< Traits >::detect ( const Parameters options = Parameters(),
const std::function< bool(double)> &  callback = std::function<bool(double)>() 
)

Performs the shape detection.

Shape types considered during the detection are those registered using add_shape_factory().

Parameters
optionsparameters for shape detection
callbackcan be omitted if the algorithm should be run without any callback. It is called regularly when the algorithm is running: the current advancement (between 0.0 and 1.0) is passed as parameter. If it returns true, then the algorithm continues its execution normally; if it returns false, the algorithm is stopped. Note that this interruption may leave the class in an invalid state.
Returns
true if shape types have been registered and input data has been set. Otherwise, false is returned.

◆ planes()

template<class Traits >
Plane_range CGAL::Shape_detection::Efficient_RANSAC< Traits >::planes ( ) const

Returns an Iterator_range with a bidirectional iterator with value type boost::shared_ptr<Plane_shape> over only the detected planes in the order of detection.

Depending on the chosen probability for the detection, the planes are ordered with decreasing size.

◆ preprocess()

template<class Traits >
bool CGAL::Shape_detection::Efficient_RANSAC< Traits >::preprocess ( )

Constructs internal data structures required for the shape detection.

These structures only depend on the input data, i.e. the points and normal vectors. This method is called by detect(), if it was not called before by the user.

◆ set_input()

template<class Traits >
void CGAL::Shape_detection::Efficient_RANSAC< Traits >::set_input ( Input_range input_range,
Point_map  point_map = Point_map(),
Normal_map  normal_map = Normal_map() 
)

Sets the input data.

The range must stay valid until the detection has been performed and the access to the results is no longer required. The data in the input is reordered by the methods detect() and preprocess(). This function first calls clear().

Parameters
input_rangeRange of input data.
point_mapProperty map to access the position of an input point.
normal_mapProperty map to access the normal of an input point.

◆ shapes()

template<class Traits >
Shape_range CGAL::Shape_detection::Efficient_RANSAC< Traits >::shapes ( ) const

Returns an Iterator_range with a bidirectional iterator with value type boost::shared_ptr<Shape> over the detected shapes in the order of detection.

Depending on the chosen probability for the detection, the shapes are ordered with decreasing size.