Nearest neighbor search based classification of PCL point type features. FLANN is used to identify a neighborhood, based on which different scoring schemes can be employed to obtain likelihood values for a specified list of classes. More...

#include <nn_classification.h>

Public Types
typedef std::pair< std::vector < std::string >, std::vector < float > >	Result
	Result is a list of class labels and scores.
typedef boost::shared_ptr< Result >	ResultPtr
Public Member Functions
ResultPtr	classify (const PointT &p_q, double radius, float gaussian_param, int max_nn=INT_MAX)
	Utility function for the default classification process.
ResultPtr	getGaussianBestScores (float gaussian_param, std::vector< int > &k_indices, std::vector< float > &k_sqr_distances)
	Computes a score exponentially decreasing with the distance for each class given a neighborhood.
int	getKNearestExemplars (const PointT &p_q, int k, std::vector< int > &k_indices, std::vector< float > &k_sqr_distances)
	Search for k-nearest neighbors for the given query point.
ResultPtr	getLinearBestScores (std::vector< int > &k_indices, std::vector< float > &k_sqr_distances)
	Computes a score that is inversely proportional to the distance to each class given a neighborhood.
int	getSimilarExemplars (const PointT &p_q, double radius, std::vector< int > &k_indices, std::vector< float > &k_sqr_distances, int max_nn=INT_MAX)
	Search for all the nearest neighbors of the query point in a given radius.
boost::shared_ptr< std::vector < float > >	getSmallestSquaredDistances (std::vector< int > &k_indices, std::vector< float > &k_sqr_distances)
	Gets the smallest square distance to each class given a neighborhood.
bool	loadTrainingFeatures (std::string file_name, std::string labels_file_name)
	Load the list of training examples and corresponding labels.
	NNClassification ()
bool	saveTrainingFeatures (std::string file_name, std::string labels_file_name)
	Save the list of training examples and corresponding labels.
void	setTrainingFeatures (const typename pcl::PointCloud< PointT >::ConstPtr &features)
	Setting the training features.
void	setTrainingLabelIndicesAndLUT (const std::vector< std::string > &classes, const std::vector< int > &labels_idx)
	Updating the labels for each training example.
void	setTrainingLabels (const std::vector< std::string > &labels)
	Setting the labels for each training example. The unique labels from the list are stored as the class labels, and for each training example an index pointing to these labels is stored.
Private Attributes
std::vector< std::string >	classes_
	List of class labels.
std::vector< int >	labels_idx_
	The index in the class labels list for all the training examples.
pcl::KdTree< PointT >::Ptr	tree_

Detailed Description

template<typename PointT>
class pcl::NNClassification< PointT >

Nearest neighbor search based classification of PCL point type features. FLANN is used to identify a neighborhood, based on which different scoring schemes can be employed to obtain likelihood values for a specified list of classes.

Author:: Zoltan Csaba Marton

Definition at line 59 of file nn_classification.h.

Member Typedef Documentation

template<typename PointT>

typedef std::pair<std::vector<std::string>, std::vector<float> > pcl::NNClassification< PointT >::Result

Result is a list of class labels and scores.

Definition at line 75 of file nn_classification.h.

template<typename PointT>

typedef boost::shared_ptr<Result> pcl::NNClassification< PointT >::ResultPtr

Definition at line 76 of file nn_classification.h.

Constructor & Destructor Documentation

template<typename PointT>

pcl::NNClassification< PointT >::NNClassification ( ) [inline]

Definition at line 72 of file nn_classification.h.

Member Function Documentation

template<typename PointT>

ResultPtr pcl::NNClassification< PointT >::classify	(	const PointT &	p_q,
		double	radius,
		float	gaussian_param,
		int	max_nn = `INT_MAX`
	)		`[inline]`

Utility function for the default classification process.

Parameters:

p_q	the given query point
radius	the radius of the sphere bounding all of p_q's neighbors
gaussian_param	influences the width of the Gaussian by specifying where the 36.78 score should be: score = exp(-distance/gaussian_param)
max_nn	if given, bounds the maximum returned neighbors to this value

Returns:: pair of label and score for each training class from the neighborhood

Definition at line 185 of file nn_classification.h.

template<typename PointT>

ResultPtr pcl::NNClassification< PointT >::getGaussianBestScores	(	float	gaussian_param,
		std::vector< int > &	k_indices,
		std::vector< float > &	k_sqr_distances
	)		`[inline]`

Computes a score exponentially decreasing with the distance for each class given a neighborhood.

Parameters:

[in]	gaussian_param	influences the width of the Gaussian: score = exp(-distance/gaussioan_param)
[out]	k_indices	the resultant indices of the neighboring points
[out]	k_sqr_distances	the resultant squared distances to the neighboring points

Returns:: pair of label and score for each training class from the neighborhood

Definition at line 280 of file nn_classification.h.

template<typename PointT>

int pcl::NNClassification< PointT >::getKNearestExemplars	(	const PointT &	p_q,
		int	k,
		std::vector< int > &	k_indices,
		std::vector< float > &	k_sqr_distances
	)		`[inline]`

Search for k-nearest neighbors for the given query point.

Parameters:

p_q	the given query point
k	the number of neighbors to search for
k_indices	the resultant indices of the neighboring points (does not have to be resized to k a priori!)
k_sqr_distances	the resultant squared distances to the neighboring points (does not have be resized to k a priori!)

Returns:: number of neighbors found

Definition at line 202 of file nn_classification.h.

template<typename PointT>

ResultPtr pcl::NNClassification< PointT >::getLinearBestScores	(	std::vector< int > &	k_indices,
		std::vector< float > &	k_sqr_distances
	)		`[inline]`

Computes a score that is inversely proportional to the distance to each class given a neighborhood.

Note:: Scores will sum up to one.

Parameters:

k_indices	the resultant indices of the neighboring points
k_sqr_distances	the resultant squared distances to the neighboring points

Returns:: pair of label and score for each training class from the neighborhood

Definition at line 249 of file nn_classification.h.

template<typename PointT>

int pcl::NNClassification< PointT >::getSimilarExemplars	(	const PointT &	p_q,
		double	radius,
		std::vector< int > &	k_indices,
		std::vector< float > &	k_sqr_distances,
		int	max_nn = `INT_MAX`
	)		`[inline]`

Search for all the nearest neighbors of the query point in a given radius.

Parameters:

p_q	the given query point
radius	the radius of the sphere bounding all of p_q's neighbors
k_indices	the resultant indices of the neighboring points
k_sqr_distances	the resultant squared distances to the neighboring points
max_nn	if given, bounds the maximum returned neighbors to this value

Returns:: number of neighbors found in radius

Definition at line 218 of file nn_classification.h.

template<typename PointT>

boost::shared_ptr<std::vector<float> > pcl::NNClassification< PointT >::getSmallestSquaredDistances	(	std::vector< int > &	k_indices,
		std::vector< float > &	k_sqr_distances
	)		`[inline]`

Gets the smallest square distance to each class given a neighborhood.

Parameters:

k_indices	the resultant indices of the neighboring points
k_sqr_distances	the resultant squared distances to the neighboring points

Returns:: a square distance to each training class

Definition at line 230 of file nn_classification.h.

template<typename PointT>

bool pcl::NNClassification< PointT >::loadTrainingFeatures	(	std::string	file_name,
		std::string	labels_file_name
	)		`[inline]`

Load the list of training examples and corresponding labels.

Parameters:

file_name	PCD file containing the training features
labels_file_name	the class label for each training example

Returns:: true on success, false on failure (read error or number of entries don't match)

Definition at line 138 of file nn_classification.h.

template<typename PointT>

bool pcl::NNClassification< PointT >::saveTrainingFeatures	(	std::string	file_name,
		std::string	labels_file_name
	)		`[inline]`

Save the list of training examples and corresponding labels.

Parameters:

file_name	file name for writing the training features
labels_file_name	file name for writing the class label for each training example

Returns:: true on success, false on failure (write error or number of entries don't match)

Definition at line 162 of file nn_classification.h.

template<typename PointT>

void pcl::NNClassification< PointT >::setTrainingFeatures ( const typename pcl::PointCloud< PointT >::ConstPtr & features ) [inline]

Setting the training features.

Parameters:

[in] features the training features

Definition at line 84 of file nn_classification.h.

template<typename PointT>

void pcl::NNClassification< PointT >::setTrainingLabelIndicesAndLUT	(	const std::vector< std::string > &	classes,
		const std::vector< int > &	labels_idx
	)		`[inline]`

Updating the labels for each training example.

Parameters:

classes	the class labels
labels_idx	the index in the class labels list for each training example

Definition at line 98 of file nn_classification.h.

template<typename PointT>

void pcl::NNClassification< PointT >::setTrainingLabels ( const std::vector< std::string > & labels ) [inline]

Setting the labels for each training example. The unique labels from the list are stored as the class labels, and for each training example an index pointing to these labels is stored.

Note:: See the setTrainingLabelIndicesAndLUT method for easily re-labeling.

Parameters:

labels the class label for each training example

Definition at line 112 of file nn_classification.h.

Member Data Documentation

template<typename PointT>

std::vector<std::string> pcl::NNClassification< PointT >::classes_ [private]

List of class labels.

Definition at line 66 of file nn_classification.h.

template<typename PointT>

std::vector<int> pcl::NNClassification< PointT >::labels_idx_ [private]

The index in the class labels list for all the training examples.

Definition at line 68 of file nn_classification.h.

template<typename PointT>

pcl::KdTree<PointT>::Ptr pcl::NNClassification< PointT >::tree_ [private]

Definition at line 63 of file nn_classification.h.

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

nn_classification.h

Public Types

Public Member Functions

Private Attributes

Detailed Description

template<typename PointT> class pcl::NNClassification< PointT >

Member Typedef Documentation

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation

template<typename PointT>
class pcl::NNClassification< PointT >