ICP Class Reference

#include <ImFusion/Vision/ICP.h>

ICP-based point cloud alignment. More...

Detailed Description

ICP-based point cloud alignment.

Public Types
enum class	MatchingStrategy { NearestNeighbor , Projective }
	Matching strategy for point association. More...

Public Member Functions
	ICP (bool useGPU=true)
	Default constructor. If useGPU is true, OpenCL will be used for alignment, otherwise the CPU will be used. GPU support is only available for projective matching.
bool	loadPointClouds (const std::vector< PointCloud * > &pointClouds)
	Load point clouds to GPU and make them available for later use without prior upload.
bool	addPointClouds (const std::vector< PointCloud * > &pointClouds)
	Add more point clouds in addition to existing ones.
bool	setPointClouds (int input, int target)
	Set point cloud from set of previously loaded point clouds - align 1 to 1 case.
bool	setPointClouds (int input, const std::vector< int > &target)
	Set point clouds from set of previously loaded point clouds - align 1 to n case.
bool	setPointClouds (PointCloud *input, PointCloud *target)
	Set point clouds.
bool	setPointClouds (PointCloud *input, const std::vector< PointCloud * > &targets)
	Set point clouds Deletes all preloaded point clouds.
PointCloud *	inputPointCloud () const
std::vector< PointCloud * >	targetPointClouds () const
void	setMatchingStrategy (MatchingStrategy s)
	Set matching strategy.
MatchingStrategy	matchingStrategy () const
	Get matching strategy.
void	setMinMatches (int minMatches)
	Set minimum number of required matches.
int	minMatches () const
	Get minimum number of required matches.
void	setDistanceThreshold (double t)
	Set distance threshold for point matching.
double	distanceThreshold () const
	Get distance threshold for point matching.
void	setAngleThreshold (double t)
	Set angle threshold in degrees for normals for point matching.
double	angleThreshold () const
	Get angle threshold in degrees for normals for point matching.
void	setOverlapRatio (double o)
	Set (known) overlap ratio between point clouds. Currently not used.
double	overlapRatio () const
	Get (known) overlap ratio between point clouds. Currently not used.
void	setAbortIterations (int iter)
	Abort after a certain number of cost function evaluations.
int	abortIterations () const
	Get the number of iterations after which to abort.
void	setAbortTime (double sec)
	Abort after a certain elapsed number of seconds.
double	abortTime () const
	Get the number of seconds after which to abort.
void	setAbortParTol (double val)
	Abort if change in parameter values becomes too small.
double	abortParTol () const
	Get the threshold for aborting if change in parameter values becomes too small.
double	rmsError (int *matchedPoints=nullptr, std::vector< std::vector< int > > *matches=nullptr, bool weighted=true) const
	Compute alignment RMS error using the current point cloud pose.
bool	align (mat4 *inputT=nullptr, std::pair< double, int > *rms=nullptr)
	Align point clouds.

Member Enumeration Documentation

MatchingStrategy

enum class MatchingStrategy

strong

Matching strategy for point association.

Enumerator
NearestNeighbor	Use nearest neighbor to point. Works for all kinds of point clouds.
Projective	Project point to other view and take point at that position. This mode requires a dense point cloud and the camera intrinsics for the target point cloud.

Member Function Documentation

loadPointClouds()

bool loadPointClouds

(

const std::vector< PointCloud * > &

pointClouds

)

Load point clouds to GPU and make them available for later use without prior upload.

Ownership is not transferred. To use preloaded point clouds you must use the align method with the input indices as parameters. Calling setPointClouds deletes all preloaded point clouds.

Parameters

pointClouds

Point clouds to preload

setPointClouds() [1/4]

bool setPointClouds	(	int	input,
		int	target )

Set point cloud from set of previously loaded point clouds - align 1 to 1 case.

Parameters

input	Input point cloud index. Must have normals.
target	Point cloud index to which to align input point cloud. Normals are optional.

setPointClouds() [2/4]

bool setPointClouds	(	int	input,
		const std::vector< int > &	target )

Set point clouds from set of previously loaded point clouds - align 1 to n case.

Parameters

input	Input point cloud index. Must have normals.
target	Point cloud indeces to which to align input point cloud. Normals are optional.

setPointClouds() [3/4]

bool setPointClouds	(	PointCloud *	input,
		PointCloud *	target )

Set point clouds.

Deletes all preloaded point clouds.

Parameters

input	Input point cloud. Must have normals.
target	Point cloud to which to align input point cloud. Normals are optional.

setPointClouds() [4/4]

bool setPointClouds	(	PointCloud *	input,
		const std::vector< PointCloud * > &	targets )

Set point clouds Deletes all preloaded point clouds.

Parameters

input	Input point cloud. Must have normals.
target	Point clouds to which to align input point cloud. Normals are optional.

align()

bool align	(	mat4 *	inputT = nullptr,
		std::pair< double, int > *	rms = nullptr )

Align point clouds.

If the alignment succeeds the matrix of the input point cloud will be updated. If inputT has been specified the point cloud matrix will not be updated and the new matrix will be returned in inputT.

---

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

• ImFusion/Vision/ICP.h