Detailed Description

Defines Point Cloud ICP module to align point clouds using iterative closest point algorithms.

Note: SW Release Applicability: These APIs are available in both NVIDIA DriveWorks and NVIDIA DRIVE Software releases.

Data Structures
struct	dwPointCloudICPParams
	Defines point cloud icp parameter. More...

struct	dwPointCloudICPResultStats
	Some result stats about the most recent ICP run. More...

Typedefs
typedef bool(*	dwPointCloudICPConvergenceCheck) (const dwTransformation3f prevSrc2Tgt, const dwTransformation3f newSrc2Tgt, void *userData)
	Callback function to be executed by ICP module allowing user to overwrite default convergence criteria method. More...

typedef struct dwPointCloudICPObject *	dwPointCloudICPHandle_t

Enumerations
enum	dwPointCloudICPType { DW_POINT_CLOUD_ICP_TYPE_DEPTH_MAP }
	Defines point cloud icp type. More...

Functions
DW_API_PUBLIC dwStatus	dwPointCloudICP_bindInput (const dwPointCloud sourcePCD, const dwPointCloud targetPCD, const dwTransformation3f *sourceToTarget, dwPointCloudICPHandle_t obj)
	Binds Input buffers to point cloud icp. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_bindOutput (dwTransformation3f *pose, dwPointCloudICPHandle_t obj)
	Binds output buffer to point cloud icp. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_getCUDAStream (cudaStream_t *stream, dwPointCloudICPHandle_t obj)
	Gets CUDA stream of point cloud icp. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_getDefaultParams (dwPointCloudICPParams *params)
	Gets default point cloud icp parameters. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_getLastResultStats (dwPointCloudICPResultStats *resultStats, dwPointCloudICPHandle_t obj)
	Get stats about latest point cloud ICP run, returns the costs for last pose. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_getMaximumDepthMapSize (uint32_t *maxDepthMapSize)
	Get the maximum size of the depth map in number of points supported by the ICP implementation. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_initialize (dwPointCloudICPHandle_t obj, const dwPointCloudICPParams params, dwContextHandle_t ctx)
	Initializes point cloud icp. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_process (dwPointCloudICPHandle_t obj)
	Estimates the transformation aligns two PointClouds. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_release (dwPointCloudICPHandle_t obj)
	Releases point cloud icp. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_reset (dwPointCloudICPHandle_t obj)
	Resets point cloud icp. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_setConvergenceCriteriaCallback (dwPointCloudICPConvergenceCheck callback, void *userData, dwPointCloudICPHandle_t obj)
	Set callback to be executed to test convergence. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_setConvergenceTolerance (float32_t angleTol, float32_t distanceTol, dwPointCloudICPHandle_t obj)
	Set tolerances used by the default ICP convergence criteria method. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_setCUDAStream (const cudaStream_t stream, dwPointCloudICPHandle_t obj)
	Sets CUDA stream of point cloud icp. More...

DW_API_PUBLIC dwStatus	dwPointCloudICP_setMaxIterations (uint16_t maxIterations, dwPointCloudICPHandle_t obj)
	Set maximum number of iterations which need to be executed. More...

Data Structure Documentation

◆ dwPointCloudICPParams

struct dwPointCloudICPParams

Data Fields
float32_t	angleConvergenceTol	Angle convergence tolerance, change in radians between two consecutive iteration steps.
dwVector2ui	depthmapSize	If icpType is DW_POINT_CLOUD_ICP_TYPE_DEPTH_MAP, this defines the size of depthmap.
float32_t	distanceConvergenceTol	Distance convergence tolerance, change in units between two consecutive iteration steps.
dwPointCloudICPType	icpType	Type of the ICP implementation to be used.
uint16_t	maxIterations	Maximum number of iterations which need to be executed.
uint32_t	maxPoints	Maximum number of points that will be pushed to ICP optimization.
bool	usePriors	Controls whether or not ICP uses the initialization pose as a prior in the optimization.

◆ dwPointCloudICPResultStats

struct dwPointCloudICPResultStats

Data Fields
uint16_t	actualNumIterations	How many iterations were actually performed.
float32_t	inlierFraction	Fraction of points which are inliers to the final ICP pose.
uint32_t	numCorrespondences	Number of 3D points which qualify as valid correspondences.
float32_t	rmsCost	RMS (weighted) cost after last ICP iteration.

Typedef Documentation

◆ dwPointCloudICPConvergenceCheck

typedef bool(* dwPointCloudICPConvergenceCheck) (const dwTransformation3f *prevSrc2Tgt, const dwTransformation3f *newSrc2Tgt, void *userData)

Callback function to be executed by ICP module allowing user to overwrite default convergence criteria method.

The method will receive original and new transformations as computed by the ICP module

Definition at line 99 of file PointCloudICP.h.

◆ dwPointCloudICPHandle_t

typedef struct dwPointCloudICPObject* dwPointCloudICPHandle_t

Definition at line 58 of file PointCloudICP.h.

Enumeration Type Documentation

◆ dwPointCloudICPType

enum dwPointCloudICPType

Defines point cloud icp type.

Enumerator

DW_POINT_CLOUD_ICP_TYPE_DEPTH_MAP

Grid based depthmap representation for the lidar point cloud.

It is assumed that the structured point cloud passed for alignment was generated from a spinning lidar

Definition at line 63 of file PointCloudICP.h.

Function Documentation

◆ dwPointCloudICP_bindInput()

DW_API_PUBLIC dwStatus dwPointCloudICP_bindInput	(	const dwPointCloud *	sourcePCD,
		const dwPointCloud *	targetPCD,
		const dwTransformation3f *	sourceToTarget,
		dwPointCloudICPHandle_t	obj
	)

Binds Input buffers to point cloud icp.

Parameters

[in]	sourcePCD	Pointer to source point cloud input buffer. Must be in cartesian coordinate space.
[in]	targetPCD	Pointer to target point cloud input buffer. Must be in cartesian coordinate space.
[in]	sourceToTarget	Pointer to the initial transformation that transforms from source to target point cloud
[in]	obj	Handle to point cloud icp

Returns: DW_SUCCESS
DW_INVALID_ARGUMENT

◆ dwPointCloudICP_bindOutput()

DW_API_PUBLIC dwStatus dwPointCloudICP_bindOutput	(	dwTransformation3f *	pose,
		dwPointCloudICPHandle_t	obj
	)

Binds output buffer to point cloud icp.

Parameters

[out]	pose	Pointer to the transformation that aligns the source point cloud with target point cloud
[in]	obj	Handle to point cloud icp module

Returns: DW_SUCCESS
DW_INVALID_ARGUMENT

◆ dwPointCloudICP_getCUDAStream()

DW_API_PUBLIC dwStatus dwPointCloudICP_getCUDAStream	(	cudaStream_t *	stream,
		dwPointCloudICPHandle_t	obj
	)

Gets CUDA stream of point cloud icp.

Parameters

[out]	stream	Pointer to CUDA stream handle
[in]	obj	Handle to point cloud icp

Returns: DW_SUCCESS
DW_INVALID_ARGUMENT

◆ dwPointCloudICP_getDefaultParams()

DW_API_PUBLIC dwStatus dwPointCloudICP_getDefaultParams ( dwPointCloudICPParams * params )

Gets default point cloud icp parameters.

Parameters

[out] params Pointer to point cloud icp parameters

Returns: DW_SUCCESS
DW_INVALID_ARGUMENT

◆ dwPointCloudICP_getLastResultStats()

DW_API_PUBLIC dwStatus dwPointCloudICP_getLastResultStats	(	dwPointCloudICPResultStats *	resultStats,
		dwPointCloudICPHandle_t	obj
	)

Get stats about latest point cloud ICP run, returns the costs for last pose.

Parameters

[out]	resultStats	Struct with stats about latest ICP run
[in]	obj	Handle to point cloud icp

Returns: DW_BAD_CAST - if the provided ICP handle is invalid.
DW_INVALID_ARGUMENT - if provided parameters are null or invalid
DW_SUCCESS

◆ dwPointCloudICP_getMaximumDepthMapSize()

DW_API_PUBLIC dwStatus dwPointCloudICP_getMaximumDepthMapSize ( uint32_t * maxDepthMapSize )

Get the maximum size of the depth map in number of points supported by the ICP implementation.

Width x height of the depth map has to be less than this number.

Parameters

[out] maxDepthMapSize - return maximal size in number of points supported by the ICP implementation.

Returns: DW_INVALID_ARGUMENT - if provided parameters are null or invalid
DW_SUCCESS

◆ dwPointCloudICP_initialize()

DW_API_PUBLIC dwStatus dwPointCloudICP_initialize	(	dwPointCloudICPHandle_t *	obj,
		const dwPointCloudICPParams *	params,
		dwContextHandle_t	ctx
	)

Initializes point cloud icp.

Parameters

[out]	obj	Handle to point cloud icp
[in]	params	Pointer to point cloud icp parameters
[in]	ctx	Handle to the context

Returns: DW_SUCCESS
DW_INVALID_ARGUMENT

◆ dwPointCloudICP_process()

DW_API_PUBLIC dwStatus dwPointCloudICP_process ( dwPointCloudICPHandle_t obj )

Estimates the transformation aligns two PointClouds.

Parameters

[in] obj Handle to point cloud icp

Returns: DW_SUCCESS
DW_INVALID_ARGUMENT
DW_CALL_NOT_ALLOWED - input and out buffers are not bound

◆ dwPointCloudICP_release()

DW_API_PUBLIC dwStatus dwPointCloudICP_release ( dwPointCloudICPHandle_t obj )

Releases point cloud icp.

Parameters

[in] obj Handle to point cloud icp

Returns: DW_SUCCESS
DW_INVALID_ARGUMENT

◆ dwPointCloudICP_reset()

DW_API_PUBLIC dwStatus dwPointCloudICP_reset ( dwPointCloudICPHandle_t obj )

Resets point cloud icp.

Parameters

[in] obj Handle to point cloud icp

Returns: DW_SUCCESS
DW_INVALID_ARGUMENT

◆ dwPointCloudICP_setConvergenceCriteriaCallback()

DW_API_PUBLIC dwStatus dwPointCloudICP_setConvergenceCriteriaCallback	(	dwPointCloudICPConvergenceCheck	callback,
		void *	userData,
		dwPointCloudICPHandle_t	obj
	)

Set callback to be executed to test convergence.

If nullptr, the default test will be used.

Parameters

[in]	callback	Method to execute to test for convergence
[in]	userData	User data to be passed to the convergence callback
[in]	obj	Handle to point cloud icp

Returns: DW_BAD_CAST - if the provided ICP handle is invalid.
DW_SUCCESS

◆ dwPointCloudICP_setConvergenceTolerance()

DW_API_PUBLIC dwStatus dwPointCloudICP_setConvergenceTolerance	(	float32_t	angleTol,
		float32_t	distanceTol,
		dwPointCloudICPHandle_t	obj
	)

Set tolerances used by the default ICP convergence criteria method.

Parameters

[in]	angleTol	Change in radians between two consecutive iteration steps
[in]	distanceTol	Change in units between two consecutive iteration steps
[in]	obj	The initialized pointcloud ICP Module.

Returns: DW_BAD_CAST - if the provided point cloud ICP handle is invalid.
DW_SUCCESS

Note: both tolerances has to be fulfilled, to declare convergence (i.e. it is a binary AND).

◆ dwPointCloudICP_setCUDAStream()

DW_API_PUBLIC dwStatus dwPointCloudICP_setCUDAStream	(	const cudaStream_t	stream,
		dwPointCloudICPHandle_t	obj
	)

Sets CUDA stream of point cloud icp.

Parameters

[in]	stream	Handle to CUDA stream
[in]	obj	Handle to point cloud icp

Returns: DW_SUCCESS
DW_INVALID_ARGUMENT

◆ dwPointCloudICP_setMaxIterations()

DW_API_PUBLIC dwStatus dwPointCloudICP_setMaxIterations	(	uint16_t	maxIterations,
		dwPointCloudICPHandle_t	obj
	)

Set maximum number of iterations which need to be executed.

Note that ICP might converge earlier, due to the tolerances set in dwPointCloudICP_setConvergenceTolerance().

Parameters

[in]	maxIterations	Maximal number of iterations to execute.
[in]	obj	The initialized pointcloud ICP Module.

Returns: DW_BAD_CAST - if the provided ICP handle is invalid.
DW_SUCCESS

Detailed Description

Data Structures

Typedefs

Enumerations

Functions

Data Structure Documentation

◆ dwPointCloudICPParams

◆ dwPointCloudICPResultStats

Typedef Documentation

◆ dwPointCloudICPConvergenceCheck

◆ dwPointCloudICPHandle_t

Enumeration Type Documentation

◆ dwPointCloudICPType

Function Documentation

◆ dwPointCloudICP_bindInput()

◆ dwPointCloudICP_bindOutput()

◆ dwPointCloudICP_getCUDAStream()

◆ dwPointCloudICP_getDefaultParams()

◆ dwPointCloudICP_getLastResultStats()

◆ dwPointCloudICP_getMaximumDepthMapSize()

◆ dwPointCloudICP_initialize()

◆ dwPointCloudICP_process()

◆ dwPointCloudICP_release()

◆ dwPointCloudICP_reset()

◆ dwPointCloudICP_setConvergenceCriteriaCallback()

◆ dwPointCloudICP_setConvergenceTolerance()

◆ dwPointCloudICP_setCUDAStream()

◆ dwPointCloudICP_setMaxIterations()