voting_table Class Reference

Array being used in geometric hashing. More...

#include <voting_table.hh>

Inheritance diagram for voting_table:

[legend]Collaboration diagram for voting_table:

[legend]List of all members.

Public Member Functions
	voting_table (const calibration_ptr &_calib, const mimas::mm_xml_reference< XalanElement > &geometricVotingNode, int _samplingRate, int _fringe, short int minGradient, geometric_map_ptr _hash_map, bool _verbose)
	Configure table.
std::vector< std::pair< mi_pose, double > >	recognise (const mimas::mm_image< bool > &scene_features, const std::vector< boost::numeric::ublas::vector< int > > &points, const mi_pose &poseInv, geometric_voting_statistics *statistic) const
	Generate match for a point-pair.
double	get_z_resolution (void) const
	Z-Resolution.
geometric_map_ptr	get_hash_map (void) const
	Get pointer to object for generating hash-functions.
Protected Member Functions
mi_pose	pose2d (const mi_pose &pose, int x, int y) const
	Convert 4-DOF transformation to two-dimensional transformation.
int	pose_depth (const mi_pose &pose) const
	Extract stack-level from 4-DOF transformation.
mi_pose	pose3d (const mi_pose &pose, int x, int y, int z) const
	Convert two-dimensional transformation to transformation with 4-DOF.
bool	verification (const mimas::mm_image< bool > &scene_features, const mi_pose &objectPose, geometric_voting_statistics *statistic) const
	Verify a hypothesis.
Protected Attributes
calibration_ptr	calib
	Calibration data.
int	sampling_rate
	Sampling-rate.
int	fringe
	Fringe.
double	min_hits
	Minimum fraction of matching features for a hypothesis.
double	max_misses
	Maximum fraction of unmatched features for a hypothesis.
int	skip
	Number of images being skipped from shared focus stack.
double	z_resolution
	Z-Resolution of focus stack.
std::vector< std::vector< boost::numeric::ublas::vector< int > > >	points_stack
	Feature lists.
int	max_num_features
	Maximum number of features.
boost::multi_array< std::map< int, int >, 2 >	hash_table
	Table with votes.
std::vector< mi_pose >	candidates
	Candidate list.
geometric_map_ptr	hash_map
	Hash functions.
bool	verbose
mimas::mm_image_qtoutput< mimas::mm_rgba< unsigned char > >	display
	Display for debugging purposes.

---

Detailed Description

Array being used in geometric hashing.

An object of this class holds the table for recognizing one object with four degrees of freedom.

See also:

geometric_voting

Author:

Jan Wedekind (j.wedekind@shu.ac.uk)

---

Constructor & Destructor Documentation

voting_table::voting_table (  const calibration_ptr &  _calib, const mimas::mm_xml_reference< XalanElement > &  geometricVotingNode, int  _samplingRate, int  _fringe, short int  minGradient, geometric_map_ptr  _hash_map, bool  _verbose )

Configure table. Creates the table with votes for a single type of micro-object. Parameters: geometricVotingNode  XML-node referencing a node with X-path database/object-type/recognition/geometric-voting. _calib  Calibration data. geometricVotingNode  Root-node of part of XML-document for initialisation. _samplingRate  Sampling rate for sub-sampling the images. _fringe  Size of fringe to be discarded from image. minGradient  Threshold to select gradient-points. _hash_map  Object for computing hash-functions. _verbose  Set to true, if you want to get verbose information from the vision-algorithms.

---

Member Function Documentation

geometric_map_ptr voting_table::get_hash_map (  void   )  const [inline]

Get pointer to object for generating hash-functions. See also: geometric_map

double voting_table::get_z_resolution (  void   )  const [inline]

Z-Resolution. Get distance between the focussed planes of two consecutive images of the focus stack. Returns: z-resolution of focus stack.

mi_pose voting_table::pose2d (  const mi_pose &  pose, int  x, int  y )  const [protected]

Convert 4-DOF transformation to two-dimensional transformation. This does the inverse of the pose3d method. Parameters: pose  3-dimensional transformation. x  x-offset (pixels) to be added to coordinates. y  y-offset (pixels) to be added to coordinates. See also: pose3d

mi_pose voting_table::pose3d (  const mi_pose &  pose, int  x, int  y, int  z )  const [protected]

Convert two-dimensional transformation to transformation with 4-DOF. The transformation-matrix and -vector are made 3-dimensional. The 4th degree of freedom (z-translation) is added. Parameters: pose  2-dimensional transformation. x  x-offset (pixels) to be subtracted from coordinates. y  y-offset (pixels) to be subtracted from coordinates. z  Additional unit-less translation in z-direction. See also: Units pose2d

int voting_table::pose_depth (  const mi_pose &  pose  )  const [protected]

Extract stack-level from 4-DOF transformation. The following condition always should hold: pose_depth( pose3d( anyPose, z ) ) == z

std::vector< std::pair< mi_pose, double > > voting_table::recognise (  const mimas::mm_image< bool > &  scene_features, const std::vector< boost::numeric::ublas::vector< int > > &  points, const mi_pose &  poseInv, geometric_voting_statistics *  statistic )  const

Generate match for a point-pair. This method finds the best match for a single point-pair. The number of votes for all possible poses is accumulated and the best is returned. Parameters: scene_features  Image of features. points  List of the 2-D features. poseInv  randomly generated pose for using as hash-function. maxHitsReached  This is used for generating verbose output only. minHitsCounter  This is used for generating verbose output only. minMissesReached  This is used for generating verbose output only. maxMissesCounter  This is used for generating verbose output only. Returns: Vector containing poses and fit-values of the matches with sufficient hits. See also: min_hits

bool voting_table::verification (  const mimas::mm_image< bool > &  scene_features, const mi_pose &  objectPose, geometric_voting_statistics *  statistic )  const [protected]

Verify a hypothesis. After a hypothesis was formed, it is verified by projecting the object's features on the image and checking, wether most of them are there. Parameters: scene_features  Image of scene-features. objectPose  4-DOF pose of object. minMissesReached  This is used for generating verbose output only. maxMissesCounter  This is used for generating verbose output only.

---

Member Data Documentation

std::vector< mi_pose > voting_table::candidates [protected]

Candidate list. Each candidate is a different possible pose of the object (relative to a randomly selected point of pairs).

geometric_map_ptr voting_table::hash_map [protected]

Hash functions. Functionality for generating transformations from features.

boost::multi_array< std::map< int, int >, 2 > voting_table::hash_table [protected]

Table with votes. The table is two-dimensional and each element is a map, assigning mapping candidate-indices on number of votes. See also: candidates

double voting_table::max_misses [protected]

Maximum fraction of unmatched features for a hypothesis. The maximum allowed number of features not being found in the scene. This number is used in the verification step. See also: verification

int voting_table::max_num_features [protected]

Maximum number of features. Number of features of the stack-image, with has the most.

double voting_table::min_hits [protected]

Minimum fraction of matching features for a hypothesis. A match with a lower number of hits has to be discarded immediately.

std::vector< std::vector< boost::numeric::ublas::vector< int > > > voting_table::points_stack [protected]

Feature lists. The features of each image of the micro-object's focus-stack is stored as an element of this list.

int voting_table::skip [protected]

Number of images being skipped from shared focus stack. An additional offset has to be taken into account, if only a part of the original focus stack is being used.

double voting_table::z_resolution [protected]

Z-Resolution of focus stack. Distance between the focussed planes of two consecutive images of the focus stack.

---

The documentation for this class was generated from the following file:

• voting_table.hh

---