Implements an NDT based spatial index. More...

#include <ndt_map.h>

Inheritance diagram for lslgeneric::NDTMap< PointT >:

Public Member Functions
void	addDistributionToCell (const Eigen::Matrix3d &ucov, const Eigen::Vector3d &umean, unsigned int numpointsindistribution, float r=0, float g=0, float b=0, unsigned int maxnumpoints=1e9, float max_occupancy=1024)
virtual bool	addMeasurement (const Eigen::Vector3d &origin, PointT endpoint, double classifierTh, double maxz, double sensor_noise)
virtual void	addPointCloud (const Eigen::Vector3d &origin, const pcl::PointCloud< PointT > &pc, double classifierTh=0.06, double maxz=100.0, double sensor_noise=0.25, double occupancy_limit=255)
virtual void	addPointCloudMeanUpdate (const Eigen::Vector3d &origin, const pcl::PointCloud< PointT > &pc, const Eigen::Vector3d &localmapsize, unsigned int maxnumpoints=1e9, float occupancy_limit=255, double maxz=100.0, double sensor_noise=0.25)
void	addPointCloudSimple (const pcl::PointCloud< PointT > &pc, double maxz=100.0)
virtual void	computeNDTCells (int cellupdatemode=CELL_UPDATE_MODE_SAMPLE_VARIANCE, unsigned int maxnumpoints=1e9, float occupancy_limit=255, Eigen::Vector3d origin=Eigen::Vector3d(0, 0, 0), double sensor_noise=0.1)
void	computeNDTCellsSimple ()
void	debugToVRML (const char *fname, pcl::PointCloud< PointT > &pc)
virtual std::vector < lslgeneric::NDTCell< PointT > * >	getAllCells () const
virtual std::vector < lslgeneric::NDTCell< PointT > * >	getAllInitializedCells ()
virtual bool	getCellAtPoint (const PointT &refPoint, NDTCell< PointT > *&cell)
	Get the cell for which the point fall into (not the closest cell)
virtual bool	getCellForPoint (const PointT &refPoint, NDTCell< PointT > *&cell, bool checkForGaussian=true) const
NDTCell< PointT > *	getCellIdx (unsigned int idx) const
	return the cell using a specific index (not available for all spatialindexes), will return NULL if the idx is not valid.
virtual std::vector< NDTCell < PointT > * >	getCellsForPoint (const PointT pt, int n_neighbours, bool checkForGaussian=true) const
virtual bool	getCentroid (double &cx, double &cy, double &cz)
double	getDepth (Eigen::Vector3d origin, Eigen::Vector3d dir, double maxDepth=100)
double	getDepthSmooth (Eigen::Vector3d origin, Eigen::Vector3d dir, double maxDepth=20, int n_neigh=1, double weight=5.0, double threshold=0.2, Eigen::Vector3d *hit=NULL)
bool	getGridSizeInMeters (double &cx, double &cy, double &cz)
virtual std::vector< NDTCell < PointT > * >	getInitializedCellsForPoint (const PointT pt) const
virtual double	getLikelihoodForPoint (PointT pt)
SpatialIndex< PointT > *	getMyIndex () const
int	getMyIndexInt () const
	return the spatial index used as an integer
std::string	getMyIndexStr () const
	return the spatial index used as a string
void	guessSize (float cenx, float ceny, float cenz, float sizex, float sizey, float sizez)
void	initialize (double cenx, double ceny, double cenz, double sizex, double sizey, double sizez)
void	loadDepthImage (const cv::Mat &depthImage, DepthCamera< PointT > &cameraParams)
pcl::PointCloud< PointT >	loadDepthImageFeatures (const cv::Mat &depthImage, std::vector< cv::KeyPoint > &keypoints, size_t &supportSize, double maxVar, DepthCamera< PointT > &cameraParams, bool estimateParamsDI=false, bool nonMean=false)
int	loadFromJFF (const char *filename)
virtual void	loadPointCloud (const pcl::PointCloud< PointT > &pc, double range_limit=-1)
void	loadPointCloud (const pcl::PointCloud< PointT > &pc, const std::vector< std::vector< size_t > > &indices)
	each entry in the indices vector contains a set of indices to a NDC cell.
void	loadPointCloudCentroid (const pcl::PointCloud< PointT > &pc, const Eigen::Vector3d &origin, const Eigen::Vector3d &old_centroid, const Eigen::Vector3d &map_size, double range_limit)
	NDTMap ()
	NDTMap (SpatialIndex< PointT > *idx)
	NDTMap (const NDTMap &other)
	NDTMap (SpatialIndex< PointT > *idx, float cenx, float ceny, float cenz, float sizex, float sizey, float sizez)
int	numberOfActiveCells ()
virtual std::vector< NDTCell < PointT > * >	pseudoTransformNDT (Eigen::Transform< double, 3, Eigen::Affine, Eigen::ColMajor > T)
NDTMap< PointT > *	pseudoTransformNDTMap (Eigen::Transform< double, 3, Eigen::Affine, Eigen::ColMajor > T)
void	setMapSize (float sx, float sy, float sz)
void	setMode (bool is3D_)
int	writeCellVectorJFF (FILE *jffout)
int	writeLazyGridJFF (FILE *jffout)
int	writeOctTreeJFF (FILE *jffout)
int	writeToJFF (const char *filename)
void	writeToVRML (const char *filename)
virtual void	writeToVRML (FILE *fout)
virtual void	writeToVRML (FILE *fout, Eigen::Vector3d col)
virtual	~NDTMap ()
Public Attributes
EIGEN_MAKE_ALIGNED_OPERATOR_NEW pcl::PointCloud< PointT >	conflictPoints
	points that were conflicting during update
Protected Attributes
float	centerx
float	centery
float	centerz
bool	guess_size_
SpatialIndex< PointT > *	index_
bool	is3D
bool	isFirstLoad_
float	map_sizex
float	map_sizey
float	map_sizez
std::set< NDTCell< PointT > * >	update_set

Detailed Description

template<typename PointT>
class lslgeneric::NDTMap< PointT >

Implements an NDT based spatial index.

Author:: Jari Saarinen (jari.saarinen@aalto.fi) and Todor Stoyanov (todor.stoyanov@oru.se)

Version:: 2.0

This class contains an interface to a SpatialIndex (custom defined) that contains NDT cells. Provides methods to create from a PointCloud.

This class implements two approaches to NDT mapping - "traditional" NDT approach, as well as a novel NDT Occupancy Map approach.

The "traditional" approach uses only the measurement points and a single scan in order to construct the NDT map.

The NDT-OM fuses incrementally measurement using Recursive Sample Covariance (RSC) approach. It also models the occupancy and free space, it adapts to changes in the cell etc.

Having these two versions combined also means that not all features are working with both. The known NDT-OM issues are

Only Lazy-Grid spatial index is supported

The old interface (e.g. used in registration) loadPointCloud(const pcl::PointCloud<PointT> &pc, double range_limit = -1); works as before: it computes an NDT map using only the samples and without tracking occupancy.

Since version 2.0 the ndt_map integrates also incremental update features. These are accessible through two methods: 1) void addPointCloudSimple(const pcl::PointCloud<PointT> &pc,double maxz=100.0); 2) void addPointCloud(const Eigen::Vector3d &origin, const pcl::PointCloud<PointT> &pc, double classifierTh=0.06, double maxz = 100.0, double sensor_noise = 0.25);

The first one only updates the measurement points and thus is faster, but does not model free space and does not tolerate dynamics The second one uses ray tracing and a number of approaches to model the occupancy as well as adapts to the dynamics.

In all cases the procedure to use the ndt_map is the following: 1) Add the measurement (by above mentioned load or add methods) 2) call computeNDTCells

Afer this the map is updated. There are various methods to access the map elements documented in this header file that you may use.

This class implements now the following papers, which you hopefully cite if you find this useful: Normal Distributions Transform Occupancy Maps: Application to Large-Scale Online 3D Mapping. IEEE International Conference on Robotics and Automation (ICRA 2013), 2013. There is also an implementation of modeling of the dynamics (edata structure in ndt_cell): "Independent Markov chain occupancy grid maps for representation of dynamic environments," in IROS2012 Conference Proceedings, Vilamoura, Algarve, Portugal: IEEE, 2012, pp. 3489-3495.

In addition, this class provide the basis for NDT registration, which is further discussed in the ndt_registration package. The relevant publications are:

Definition at line 100 of file ndt_map.h.

Constructor & Destructor Documentation

template<typename PointT>

lslgeneric::NDTMap< PointT >::NDTMap ( ) [inline]

Definition at line 103 of file ndt_map.h.

template<typename PointT>

lslgeneric::NDTMap< PointT >::NDTMap ( SpatialIndex< PointT > * idx ) [inline]

default constructor. The SpatialIndex sent as a paramter is used as a factory every time that loadPointCloud is called. it can/should be deallocated outside the class after the destruction of the NDTMap

Definition at line 112 of file ndt_map.h.

template<typename PointT>

lslgeneric::NDTMap< PointT >::NDTMap ( const NDTMap< PointT > & other ) [inline]

Definition at line 126 of file ndt_map.h.

template<typename PointT>

lslgeneric::NDTMap< PointT >::NDTMap	(	SpatialIndex< PointT > *	idx,
		float	cenx,
		float	ceny,
		float	cenz,
		float	sizex,
		float	sizey,
		float	sizez
	)		`[inline]`

Construct with given centroid and sizes

Parameters:

cenx,ceny,cenz;	(x,y,z) of the center of the map
sizex,sizey,sizez,:	The size of the map in each respective direction NOTE: Implementation only for the laze grid

Definition at line 141 of file ndt_map.h.

template<typename PointT>

virtual lslgeneric::NDTMap< PointT >::~NDTMap ( ) [inline, virtual]

Default destructor

Definition at line 205 of file ndt_map.h.

Member Function Documentation

template<typename PointT >

void lslgeneric::NDTMap< PointT >::addDistributionToCell	(	const Eigen::Matrix3d &	ucov,
		const Eigen::Vector3d &	umean,
		unsigned int	numpointsindistribution,
		float	r = `0`,
		float	g = `0`,
		float	b = `0`,
		unsigned int	maxnumpoints = `1e9`,
		float	max_occupancy = `1024`
	)

Adds a sample mean and covariance to the map

Parameters:

&ucov	The covariance matrix to be added
&umean	The mean of the normal distribution
numpointsindistribution	The number of points used in computation of the sample mean and covariance
r,g,b	-- optional color parameters
maxnumpoints	-- optional adaptation of the gaussians

Add a distribution to the map

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 327 of file ndt_map.hpp.

template<typename PointT >

bool lslgeneric::NDTMap< PointT >::addMeasurement	(	const Eigen::Vector3d &	origin,
		PointT	endpoint,
		double	classifierTh,
		double	maxz,
		double	sensor_noise
	)		`[virtual]`

Adds one measurement to the map using NDT-OM update step

Returns:: true if an inconsistency was detected

Select the smallest resolution

We only want to check every cell once, so increase the index if we are still in the same cell

Check the validity of the index

< endpoint

< don't accept points further than the measurement

<distance to endpoint

test for distance based sensor noise

This is the probability of max lik point being endpoint

Evidence value for empty - alpha * p(x);

The cell does not have gaussian, so we mark that we saw it empty...

Add fake point to initialize!

Definition at line 896 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::addPointCloud	(	const Eigen::Vector3d &	origin,
		const pcl::PointCloud< PointT > &	pc,
		double	classifierTh = `0.06`,
		double	maxz = `100.0`,
		double	sensor_noise = `0.25`,
		double	occupancy_limit = `255`
	)		`[virtual]`

Add new pointcloud to map - This is the main interface for NDT-OM! Performs raytracing, updates conflicts and adds points to cells computeNDTCells must be called after calling this

Parameters:

&origin	is the position of the sensor, from where the scan has been taken from.
&pc	is the pointcloud to be added
classifierTh	A treshold to judge if the ray passes through a gaussian (obsolete)
maxz	threshold for the maximum z-coordinate value for the measurement point_cloud
sensor_noise	The expected standard deviation of the sensor noise

Adds a new cloud: NDT-OM update step

Select the smallest resolution

We only want to check every cell once, so increase the index if we are still in the same cell

Check the validity of the index

< don't accept points further than the measurement

<distance to endpoint

test for distance based sensor noise

This is the probability of max lik point being endpoint

Evidence value for empty - alpha * p(x);

The cell does not have gaussian, so we mark that we saw it empty...

Add fake point to initialize!

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 436 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::addPointCloudMeanUpdate	(	const Eigen::Vector3d &	origin,
		const pcl::PointCloud< PointT > &	pc,
		const Eigen::Vector3d &	localmapsize,
		unsigned int	maxnumpoints = `1e9`,
		float	occupancy_limit = `255`,
		double	maxz = `100.0`,
		double	sensor_noise = `0.25`
	)		`[virtual]`

Add new pointcloud to map - Updates the occupancy using the mean values of a local map generated from an observation

Performs raytracing, updates conflicts and adds points to cells computeNDTCells must be called after calling this

Parameters:

&origin	is the position of the sensor, from where the scan has been taken from.
&pc	is the pointcloud to be added
&localmapsize	The dimensions of the local map used for computing the gaussians
maxnumpoints	Defines the forgetting factor (default 100000) the smaller the value the faster the adaptation
occupancy_limit	Clamping threshold for log-odds value
maxz	threshold for the maximum z-coordinate value for the measurement point_cloud
sensor_noise	The expected standard deviation of the sensor noise

fprintf(stderr,"UPDATING\n");

Select the smallest resolution

Lets first create a local ndmap (this really works only if we have lazy grid as well)

FIXME:: fixed max length

Use Student-T

Use Sample variance

TODO: This returns now copies --- should be pointers?

Lets update negative even though the measurement was too high

We only want to check every cell once, so increase the index if we are still in the same cell

Check the validity of the index

< don't accept points further than the measurement

<distance to endpoint

test for distance based sensor noise

This is the probability of max lik point being endpoint

Evidence value for empty - alpha * p(x);

The cell does not have gaussian, so we mark that we saw it empty...

Add fake point to initialize!

<FIXME:: local implementation can be faster?

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 694 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::addPointCloudSimple	(	const pcl::PointCloud< PointT > &	pc,
		double	maxz = `100.0`
	)

This interface updates only the end points into the map without raytracing

Just adds the points, without raytracing and such

Definition at line 283 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::computeNDTCells	(	int	cellupdatemode = `CELL_UPDATE_MODE_SAMPLE_VARIANCE`,
		unsigned int	maxnumpoints = `1e9`,
		float	occupancy_limit = `255`,
		Eigen::Vector3d	origin = `Eigen::Vector3d(0,0,0)`,
		double	sensor_noise = `0.1`
	)		`[virtual]`

Computes the NDT-cells after a measurement has been added

Parameters:

cellupdatemode	Defines the update mode (default CELL_UPDATE_MODE_SAMPLE_VARIANCE)
maxnumpoints	Defines the forgetting factor (default 100000) the smaller the value the faster the adaptation

Helper function, computes the NDTCells

Process the conflict points

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 1298 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::computeNDTCellsSimple ( )

Computes the Normaldistribution parameters without erasing the points

Computes the normaldistribution parameters and leaves the points a

Definition at line 1398 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::debugToVRML	(	const char *	fname,
		pcl::PointCloud< PointT > &	pc
	)

Definition at line 2094 of file ndt_map.hpp.

template<typename PointT >

std::vector< lslgeneric::NDTCell< PointT > * > lslgeneric::NDTMap< PointT >::getAllCells ( ) const [virtual]

Returns all computed cells from the map This method gives all the vectors that contain a gaussian within a cell (hasGaussian is true).

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 2218 of file ndt_map.hpp.

template<typename PointT >

std::vector< lslgeneric::NDTCell< PointT > * > lslgeneric::NDTMap< PointT >::getAllInitializedCells ( ) [virtual]

Returns all cells that have been initialized (including ones that do not contain gaussian at the moment). This is useful if you want to use the empty cells or dynamic cells

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 2243 of file ndt_map.hpp.

template<typename PointT >

bool lslgeneric::NDTMap< PointT >::getCellAtPoint	(	const PointT &	refPoint,
		NDTCell< PointT > *&	cell
	)		`[virtual]`

Get the cell for which the point fall into (not the closest cell)

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 388 of file ndt_map.hpp.

template<typename PointT >

bool lslgeneric::NDTMap< PointT >::getCellForPoint	(	const PointT &	refPoint,
		NDTCell< PointT > *&	cell,
		bool	checkForGaussian = `true`
	)		const `[virtual]`

returns the closest cell to refPoint Does not work with NDT-OM

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 2067 of file ndt_map.hpp.

template<typename PointT >

NDTCell< PointT > * lslgeneric::NDTMap< PointT >::getCellIdx ( unsigned int idx ) const

return the cell using a specific index (not available for all spatialindexes), will return NULL if the idx is not valid.

Definition at line 2207 of file ndt_map.hpp.

template<typename PointT >

std::vector< NDTCell< PointT > * > lslgeneric::NDTMap< PointT >::getCellsForPoint	(	const PointT	pt,
		int	n_neighbours,
		bool	checkForGaussian = `true`
	)		const `[virtual]`

Returns all the cells within radius Does not work with NDT-OM

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 2048 of file ndt_map.hpp.

template<typename PointT>

virtual bool lslgeneric::NDTMap< PointT >::getCentroid	(	double &	cx,
		double &	cy,
		double &	cz
	)		`[inline, virtual]`

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 416 of file ndt_map.h.

template<typename PointT >

double lslgeneric::NDTMap< PointT >::getDepth	(	Eigen::Vector3d	origin,
		Eigen::Vector3d	dir,
		double	maxDepth = `100`
	)

Computes a maximum likelihood depth from the map, given a position and a view vector

Get estimated depth This goes through now the cells always to max depth and is not the most efficient solution

Definition at line 1051 of file ndt_map.hpp.

template<typename PointT >

double lslgeneric::NDTMap< PointT >::getDepthSmooth	(	Eigen::Vector3d	origin,
		Eigen::Vector3d	dir,
		double	maxDepth = `20`,
		int	n_neigh = `1`,
		double	weight = `5.0`,
		double	threshold = `0.2`,
		Eigen::Vector3d *	hit = `NULL`
	)

Definition at line 1097 of file ndt_map.hpp.

template<typename PointT>

bool lslgeneric::NDTMap< PointT >::getGridSizeInMeters	(	double &	cx,
		double &	cy,
		double &	cz
	)		`[inline]`

Definition at line 423 of file ndt_map.h.

template<typename PointT >

std::vector< NDTCell< PointT > * > lslgeneric::NDTMap< PointT >::getInitializedCellsForPoint ( const PointT pt ) const [virtual]

Returns all the cells within radius

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 2033 of file ndt_map.hpp.

template<typename PointT >

double lslgeneric::NDTMap< PointT >::getLikelihoodForPoint ( PointT pt ) [virtual]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 1836 of file ndt_map.hpp.

template<typename PointT>

SpatialIndex<PointT>* lslgeneric::NDTMap< PointT >::getMyIndex ( ) const [inline]

Definition at line 358 of file ndt_map.h.

template<typename PointT >

int lslgeneric::NDTMap< PointT >::getMyIndexInt ( ) const

return the spatial index used as an integer

returns the current spatial index as an integer (debugging function)

Definition at line 1813 of file ndt_map.hpp.

template<typename PointT >

std::string lslgeneric::NDTMap< PointT >::getMyIndexStr ( ) const

return the spatial index used as a string

returns the current spatial index as a string (debugging function)

Definition at line 1790 of file ndt_map.hpp.

template<typename PointT>

void lslgeneric::NDTMap< PointT >::guessSize	(	float	cenx,
		float	ceny,
		float	cenz,
		float	sizex,
		float	sizey,
		float	sizez
	)		`[inline]`

Parameters:

guess_size try to guess the size based on point cloud. Otherwise use pre-set map size

Definition at line 433 of file ndt_map.h.

template<typename PointT>

void lslgeneric::NDTMap< PointT >::initialize	(	double	cenx,
		double	ceny,
		double	cenz,
		double	sizex,
		double	sizey,
		double	sizez
	)		`[inline]`

Initilize with known values - normally this is done automatically, but in some cases you want to influence these - call only once and before calling any other function

Definition at line 177 of file ndt_map.h.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::loadDepthImage	(	const cv::Mat &	depthImage,
		DepthCamera< PointT > &	cameraParams
	)

Definition at line 1192 of file ndt_map.hpp.

template<typename PointT >

pcl::PointCloud< PointT > lslgeneric::NDTMap< PointT >::loadDepthImageFeatures	(	const cv::Mat &	depthImage,
		std::vector< cv::KeyPoint > &	keypoints,
		size_t &	supportSize,
		double	maxVar,
		DepthCamera< PointT > &	cameraParams,
		bool	estimateParamsDI = `false`,
		bool	nonMean = `false`
	)

Definition at line 1200 of file ndt_map.hpp.

template<typename PointT >

int lslgeneric::NDTMap< PointT >::loadFromJFF ( const char * filename )

method to load NDT maps from .jff files USAGE: create NDTMap with desired index and PointType (index type is checked, but Point type is NOT checked) via e.g.

lslgeneric::NDTMap<pcl::PointXYZ> nd1( new lslgeneric::LazyGrid<pcl::PointXYZ>(0.4)); --> (*)

and then call

nd1.loadFromJFF("map0027.jff");

) use this constructor so index is not initialized and attributes can be set manually

Definition at line 1696 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::loadPointCloud	(	const pcl::PointCloud< PointT > &	pc,
		double	range_limit = `-1`
	)		`[virtual]`

loadPointCloud - You can call this if you are only interested in dealing with one scan without need for fusing several ones or representing empty space and occupancy

Otherwise you should always call addPointCloud (or if you don't want occupancy then addPointCloudSimple)

Parameters:

pc	the PointCloud that is to be loaded

Note:: every subsequent call will destroy the previous map!

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 31 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::loadPointCloud	(	const pcl::PointCloud< PointT > &	pc,
		const std::vector< std::vector< size_t > > &	indices
	)

each entry in the indices vector contains a set of indices to a NDC cell.

NOTE: These load functions are not supported by occupancy mapping This function loads a point cloud around specific indeces for usage in NFT-feature based mapping.

Parameters:

&pc	the Point Cloud to use as input
&indices	a vector of the indeces that will be added for each cell. We add indices.size() number of cells, each cell c[i] contains the points, indexed by the vector indices[i]

Definition at line 1171 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::loadPointCloudCentroid	(	const pcl::PointCloud< PointT > &	pc,
		const Eigen::Vector3d &	origin,
		const Eigen::Vector3d &	old_centroid,
		const Eigen::Vector3d &	map_size,
		double	range_limit
	)

loadPointCloudCentroid - A special load function to enable the matching of centroids (create alligned maps) This is more efficient than the standard, but needs also the origin and size as parameters

Parameters:

&pc	the PointCloud that is to be loaded
&origin	The desired origin of the map (will be fitted acording to old_centroid)
&old_centroid	The centroid to which we want to align
&map_size	The size of the new map
range_limit	The maximum range value for measurements

Note:: every subsequent call will destroy the previous map!

loadPointCloudCentroid - A special load function to enable the matching of centroids (create alligned maps)

Parameters:

pc	the PointCloud that is to be loaded

Note:: every subsequent call will destroy the previous map!

Should never happen!

How many cell to each direction is the new origin from old one

Definition at line 197 of file ndt_map.hpp.

template<typename PointT >

int lslgeneric::NDTMap< PointT >::numberOfActiveCells ( )

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 2267 of file ndt_map.hpp.

template<typename PointT >

std::vector< NDTCell< PointT > * > lslgeneric::NDTMap< PointT >::pseudoTransformNDT ( Eigen::Transform< double, 3, Eigen::Affine, Eigen::ColMajor > T ) [virtual]

Returns a transformed NDT as a vector of NDT cells

NOTE: The rotation of the covariance fixed by Jari 6.11.2012

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 2173 of file ndt_map.hpp.

template<typename PointT >

NDTMap< PointT > * lslgeneric::NDTMap< PointT >::pseudoTransformNDTMap ( Eigen::Transform< double, 3, Eigen::Affine, Eigen::ColMajor > T )

Returns a transformed NDT as an NDT map with a CellVector data structure

NOTE: The rotation of the covariance fixed by Jari 6.11.2012

Definition at line 2139 of file ndt_map.hpp.

template<typename PointT>

void lslgeneric::NDTMap< PointT >::setMapSize	(	float	sx,
		float	sy,
		float	sz
	)		`[inline]`

Set the map size in meters - Must be called before first addPointCloud call if you want to set the size - otherwise it is automatically determined

Definition at line 226 of file ndt_map.h.

template<typename PointT>

void lslgeneric::NDTMap< PointT >::setMode ( bool is3D_ ) [inline]

Definition at line 217 of file ndt_map.h.

template<typename PointT >

int lslgeneric::NDTMap< PointT >::writeCellVectorJFF ( FILE * jffout )

Definition at line 1570 of file ndt_map.hpp.

template<typename PointT >

int lslgeneric::NDTMap< PointT >::writeLazyGridJFF ( FILE * jffout )

Definition at line 1635 of file ndt_map.hpp.

template<typename PointT >

int lslgeneric::NDTMap< PointT >::writeOctTreeJFF ( FILE * jffout )

Definition at line 1603 of file ndt_map.hpp.

template<typename PointT >

int lslgeneric::NDTMap< PointT >::writeToJFF ( const char * filename )

output methods for saving the map in the jff format

Definition at line 1534 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::writeToVRML ( const char * filename )

Stuff for saving things

output method, saves as vrml the oct tree and all the ellipsoids

Definition at line 1442 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::writeToVRML ( FILE * fout ) [virtual]

helper output method

Definition at line 1478 of file ndt_map.hpp.

template<typename PointT >

void lslgeneric::NDTMap< PointT >::writeToVRML	(	FILE *	fout,
		Eigen::Vector3d	col
	)		`[virtual]`

Definition at line 1506 of file ndt_map.hpp.

Member Data Documentation

template<typename PointT>

float lslgeneric::NDTMap< PointT >::centerx [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 465 of file ndt_map.h.

template<typename PointT>

float lslgeneric::NDTMap< PointT >::centery [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 465 of file ndt_map.h.

template<typename PointT>

float lslgeneric::NDTMap< PointT >::centerz [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 465 of file ndt_map.h.

template<typename PointT>

EIGEN_MAKE_ALIGNED_OPERATOR_NEW pcl::PointCloud<PointT> lslgeneric::NDTMap< PointT >::conflictPoints

points that were conflicting during update

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 471 of file ndt_map.h.

template<typename PointT>

bool lslgeneric::NDTMap< PointT >::guess_size_ [protected]

Definition at line 466 of file ndt_map.h.

template<typename PointT>

SpatialIndex<PointT>* lslgeneric::NDTMap< PointT >::index_ [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 460 of file ndt_map.h.

template<typename PointT>

bool lslgeneric::NDTMap< PointT >::is3D [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 459 of file ndt_map.h.

template<typename PointT>

bool lslgeneric::NDTMap< PointT >::isFirstLoad_ [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 461 of file ndt_map.h.

template<typename PointT>

float lslgeneric::NDTMap< PointT >::map_sizex [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 462 of file ndt_map.h.

template<typename PointT>

float lslgeneric::NDTMap< PointT >::map_sizey [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 463 of file ndt_map.h.

template<typename PointT>

float lslgeneric::NDTMap< PointT >::map_sizez [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 464 of file ndt_map.h.

template<typename PointT>

std::set<NDTCell<PointT>*> lslgeneric::NDTMap< PointT >::update_set [protected]

Reimplemented in lslgeneric::NDTMapHMT< PointT >.

Definition at line 467 of file ndt_map.h.

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

Public Member Functions

Public Attributes

Protected Attributes

Detailed Description

template<typename PointT> class lslgeneric::NDTMap< PointT >

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation

template<typename PointT>
class lslgeneric::NDTMap< PointT >