Interface class for the ICP pipeline using the Power Method to estimate the rotation and considering weighted residual errors. More...

#include <algorithms.hpp>

Collaboration diagram for cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >:

[legend]

Public Types
enum	Memory : uint8_t { Memory::H_IN_F, Memory::H_IN_M, Memory::H_IO_T, Memory::D_IN_F, Memory::D_IN_M, Memory::D_IO_T }
	Enumerates the memory objects handled by the class. More...

Public Member Functions
	ICPStep (clutils::CLEnv &_env, clutils::CLEnvInfo< 1 > _infoRBC, clutils::CLEnvInfo< 1 > _infoICP)
	Configures an OpenCL environment as specified by `_info`. More...

cl::Memory &	get (ICPStep::Memory mem)
	Returns a reference to an internal memory object. More...

void	init (unsigned int _m, unsigned int _nr, float _a=1e2f, float _c=1e-6f, Staging _staging=Staging::IO)
	Configures kernel execution parameters. More...

void	write (ICPStep::Memory mem=ICPStep::Memory::D_IN_F, void ptr=nullptr, bool block=CL_FALSE, const std::vector< cl::Event > events=nullptr, cl::Event *event=nullptr)
	Performs a data transfer to a device buffer. More...

void *	read (ICPStep::Memory mem=ICPStep::Memory::H_IO_T, bool block=CL_TRUE, const std::vector< cl::Event > events=nullptr, cl::Event event=nullptr)
	Performs a data transfer to a staging buffer. More...

void	buildRBC (const std::vector< cl::Event > events=nullptr, cl::Event event=nullptr)
	Builds the RBC data structure. More...

void	run (const std::vector< cl::Event > events=nullptr, cl::Event event=nullptr, bool config=false)
	Executes the necessary kernels. More...

float	getAlpha ()
	Gets the scaling parameter \( \alpha \) involved in the distance calculations of the `RBC` data structure. More...

void	setAlpha (float _a)
	Sets the scaling parameter \( \alpha \) involved in the distance calculations of the `RBC` data structure. More...

float	getScaling ()
	Gets the scaling factor c used when computing the `S` matrix. More...

void	setScaling (float _c)
	Sets the scaling factor c used when computing the `S` matrix. More...

template<typename period >
double	run (clutils::GPUTimer< period > &timer, const std::vector< cl::Event > *events=nullptr, bool config=false)
	Executes the necessary kernels. More...

Public Attributes
cl_float *	hPtrInF

cl_float *	hPtrInM

cl_float *	hPtrIOT

Eigen::Matrix3f	Rk

Eigen::Quaternionf	qk

Eigen::Vector3f	tk

cl_float	sk

Eigen::Matrix3f	R

Eigen::Quaternionf	q

Eigen::Vector3f	t

cl_float	s

Protected Attributes
clutils::CLEnv &	env

clutils::CLEnvInfo< 1 >	infoRBC

clutils::CLEnvInfo< 1 >	infoICP

cl::Context	context

cl::CommandQueue	queue

Staging	staging

ICPReps	fReps

RBC::RBCConstruct< RBC::KernelTypeC::KINECT_R, RBC::RBCPermuteConfig::GENERIC >	rbcC

ICPTransform< ICPTransformConfig::QUATERNION >	transform

RBC::RBCSearch< RBC::KernelTypeC::KINECT_R, RBC::RBCPermuteConfig::GENERIC, RBC::KernelTypeS::KINECT >	rbcS

ICPWeights	weights

ICPMean< ICPMeanConfig::WEIGHTED >	means

ICPDevs	devs

ICPS< ICPSConfig::WEIGHTED >	matrixS

ICPPowerMethod	powMethod

cl_float *	Tk

float	a

float	c

unsigned int	m

unsigned int	nr

unsigned int	d

unsigned int	bufferFMSize

unsigned int	bufferTSize

cl::Buffer	hBufferInF

cl::Buffer	hBufferInM

cl::Buffer	hBufferIOT

cl::Buffer	dBufferInF

cl::Buffer	dBufferInM

cl::Buffer	dBufferIOT

Detailed Description

template<>
class cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >

Interface class for the ICP pipeline using the Power Method to estimate the rotation and considering weighted residual errors.

Performs one ICP iteration. It accepts two sets (fixed and moving) of landmarks, estimates the relative homogeneous transformation \( (\dot{q}_k,t_k) \) between them, and transforms the moving set according to this transformation.

Note: All computation is done on the GPU. The rotation computation is also solved on the GPU with the Power Method.; The class creates its own buffers. If you would like to provide your own buffers, call get to get references to the placeholders within the class and assign them to your buffers. You will have to do this strictly before the call to init. You can also call get (after the call to init) to get a reference to a buffer within the class and assign it to another kernel class instance further down in your task pipeline.

The following input/output OpenCL memory objects are created by a ICPStep<ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED> instance:

Name	Type	Placement	I/O	Use	Properties	Size
H_IN_F	Buffer	Host	I	Staging	CL_MEM_READ_WRITE	\(m*sizeof\ (cl\_float8)\)
H_IN_M	Buffer	Host	I	Staging	CL_MEM_READ_WRITE	\(m*sizeof\ (cl\_float8)\)
H_IO_T	Buffer	Host	IO	Staging	CL_MEM_READ_WRITE	\(2*sizeof\ (cl\_float4)\)
D_IN_F	Buffer	Device	I	Processing	CL_MEM_READ_ONLY	\(m*sizeof\ (cl\_float8)\)
D_IN_M	Buffer	Device	I	Processing	CL_MEM_READ_ONLY	\(m*sizeof\ (cl\_float8)\)
D_IO_T	Buffer	Device	IO	Processing	CL_MEM_READ_WRITE	\(2*sizeof\ (cl\_float4)\)

Member Enumeration Documentation

enum cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::Memory : uint8_t

strong

Enumerates the memory objects handled by the class.

Note: H_* names refer to staging buffers on the host.; D_* names refer to buffers on the device.

Enumerator
H_IN_F	Input staging buffer for the fixed set of landmarks.
H_IN_M	Input staging buffer for the moving set of landmarks.
H_IO_T	Input-output staging buffer for the quaternion and the translation vector. It is loaded with an initial estimation of the transformation and gets refined with every ICP iteration. The first `cl_float4` element contains the quaternion, \( \dot{q} = \left[ \begin{matrix} q_x & q_y & q_z & q_w \end{matrix} \right]^T \), and the second `cl_float4` element contains the translation vector, \( t = \left[ \begin{matrix} t_x & t_y & t_z & 1 \end{matrix} \right]^T \). If scaling is desired, the factor should be placed in the last element of the translation vector, \( t = \left[ \begin{matrix} t_x & t_y & t_z & s \end{matrix} \right]^T \).
D_IN_F	Input buffer for the fixed set of landmarks.
D_IN_M	Input buffer for the moving set of landmarks.
D_IO_T	Input-output buffer for the quaternion and the translation vector. It is loaded with an initial estimation of the transformation and gets refined with every ICP iteration. The first `cl_float4` element contains the quaternion, \( \dot{q} = \left[ \begin{matrix} q_x & q_y & q_z & q_w \end{matrix} \right]^T \), and the second `cl_float4` element contains the translation vector, \( t = \left[ \begin{matrix} t_x & t_y & t_z & 1 \end{matrix} \right]^T \). If scaling is desired, the factor should be placed in the last element of the translation vector, \( t = \left[ \begin{matrix} t_x & t_y & t_z & s \end{matrix} \right]^T \).

Constructor & Destructor Documentation

cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::ICPStep	(	clutils::CLEnv &	_env,
		clutils::CLEnvInfo< 1 >	_infoRBC,
		clutils::CLEnvInfo< 1 >	_infoICP
	)

Configures an OpenCL environment as specified by _info.

Parameters

[in]	_env	opencl environment.
[in]	_infoRBC	opencl configuration for the `RBC` classes. It specifies the context, queue, etc, to be used.
[in]	_infoICP	opencl configuration for the `ICP` classes. It specifies the context, queue, etc, to be used.

Member Function Documentation

void cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::buildRBC	(	const std::vector< cl::Event > *	events = `nullptr`,
		cl::Event *	event = `nullptr`
	)

Builds the RBC data structure.

Note: Call buildRBC after the D_IN_F buffer has been written, and before any calls to run (for each registration).

cl::Memory & cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::get ( ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::Memory mem )

Returns a reference to an internal memory object.

This interface exists to allow CL memory sharing between different kernels.

Parameters

[in] mem enumeration value specifying the requested memory object.

Returns: A reference to the requested memory object.

float cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::getAlpha ( )

Gets the scaling parameter \( \alpha \) involved in the distance calculations of the RBC data structure.

Returns: The scaling parameter \( \alpha \).

float cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::getScaling ( )

Gets the scaling factor c used when computing the S matrix.

Note: The scaling factor c multiplies the points (deviations) before processing in order to deal with floating point arithmetic issues.

Returns: The scaling factor c.

void cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::init	(	unsigned int	_m,
		unsigned int	_nr,
		float	_a = `1e2f`,
		float	_c = `1e-6f`,
		Staging	_staging = `Staging::IO`
	)

Configures kernel execution parameters.

Sets up memory objects as necessary, and defines the kernel workspaces.

Note: If you have assigned a memory object to one member variable of the class before the call to init, then that memory will be maintained. Otherwise, a new memory object will be created.

Parameters

[in]	_m	number of points in the sets.
[in]	_nr	number of fixed set representatives.
[in]	_a	factor scaling the results of the distance calculations for the geometric \( x_g \) and photometric \( x_p \) dimensions of the \( x\epsilon\mathbb{R}^8 \) points. That is, \( \\|x-x'\\|_2^2= f_g(a)\\|x_g-x'_g\\|_2^2+f_p(a)\\|x_p-x'_p\\|_2^2 \). For more info, look at `euclideanSquaredMetric8` in kernels/rbc_kernels.cl.
[in]	_c	scaling factor for dealing with floating point arithmetic issues when computing the `S` matrix.
[in]	_staging	flag to indicate whether or not to instantiate the staging buffers.

void * cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::read	(	ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::Memory	mem = `ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::Memory::H_IO_T`,
		bool	block = `CL_TRUE`,
		const std::vector< cl::Event > *	events = `nullptr`,
		cl::Event *	event = `nullptr`
	)

Performs a data transfer to a staging buffer.

The transfer happens from a device buffer to the associated (specified) staging buffer on the host.

Parameters

[in]	mem	enumeration value specifying an output staging buffer.
[in]	block	a flag to indicate whether to perform a blocking or a non-blocking operation.
[in]	events	a wait-list of events.
[out]	event	event associated with the read operation to the staging buffer.

void cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::run	(	const std::vector< cl::Event > *	events = `nullptr`,
		cl::Event *	event = `nullptr`,
		bool	config = `false`
	)

Executes the necessary kernels.

The function call is non-blocking.

Parameters

[in]	events	a wait-list of events.
[out]	event	event associated with the kernel execution.
[in]	config	flag. If true, configures the `RBC search` process. Set to true once, when the `RBC` data structure is reset.

template<typename period >

double cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::run	(	clutils::GPUTimer< period > &	timer,
		const std::vector< cl::Event > *	events = `nullptr`,
		bool	config = `false`
	)

inline

Executes the necessary kernels.

This run instance is used for profiling.

Parameters

[in]	timer	`GPUTimer` that does the profiling of the kernel executions.
[in]	events	a wait-list of events.
[in]	config	flag. If true, configures the `RBC search` process. Set to true once, when the `RBC` data structure is reset.

Returns: Τhe total execution time measured by the timer.

void cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::setAlpha ( float _a )

Sets the scaling parameter \( \alpha \) involved in the distance calculations of the RBC data structure.

Updates the kernel arguments for the scaling parameter \( \alpha \).

Parameters

[in] _a scaling parameter \( \alpha \).

void cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::setScaling ( float _c )

Sets the scaling factor c used when computing the S matrix.

Updates the kernel argument for the scaling factor c.

Note: The scaling factor c multiplies the points (deviations) before processing in order to deal with floating point arithmetic issues.

Parameters

[in] _c scaling factor.

void cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::write	(	ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::Memory	mem = `ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::Memory::D_IN_F`,
		void *	ptr = `nullptr`,
		bool	block = `CL_FALSE`,
		const std::vector< cl::Event > *	events = `nullptr`,
		cl::Event *	event = `nullptr`
	)

Performs a data transfer to a device buffer.

The transfer happens from a staging buffer on the host to the associated (specified) device buffer.

Parameters

[in]	mem	enumeration value specifying an input device buffer.
[in]	ptr	a pointer to an array holding input data. If not NULL, the data from `ptr` will be copied to the associated staging buffer.
[in]	block	a flag to indicate whether to perform a blocking or a non-blocking operation.
[in]	events	a wait-list of events.
[out]	event	event associated with the write operation to the device buffer.

Member Data Documentation

float cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::a

protected

unsigned int cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::bufferFMSize

protected

unsigned int cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::bufferTSize

protected

float cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::c

protected

cl::Context cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::context

protected

unsigned int cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::d

protected

cl::Buffer cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::dBufferInF

protected

cl::Buffer cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::dBufferInM

protected

cl::Buffer cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::dBufferIOT

protected

ICPDevs cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::devs

protected

clutils::CLEnv& cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::env

protected

ICPReps cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::fReps

protected

cl::Buffer cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::hBufferInF

protected

cl::Buffer cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::hBufferInM

protected

cl::Buffer cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::hBufferIOT

protected

cl_float* cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::hPtrInF

Mapping of the input staging buffer for the fixed set of points.

cl_float* cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::hPtrInM

Mapping of the input staging buffer for the moving set of points.

cl_float* cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::hPtrIOT

Mapping of the input-output staging buffer for the estimated quaternion and translation vector.

clutils::CLEnvInfo<1> cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::infoICP

protected

clutils::CLEnvInfo<1> cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::infoRBC

protected

unsigned int cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::m

protected

ICPS<ICPSConfig::WEIGHTED> cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::matrixS

protected

ICPMean<ICPMeanConfig::WEIGHTED> cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::means

protected

unsigned int cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::nr

protected

ICPPowerMethod cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::powMethod

protected

Eigen::Quaternionf cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::q

Represents the rotation estimation up to iteration k, given in quaternion representation, \( \dot{q} = \left[ \begin{matrix} q_x & q_y & q_z & q_w \end{matrix} \right]^T \).

Eigen::Quaternionf cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::qk

Represents the incremental development in the rotation estimation in iteration k, given in quaternion representation, \( \dot{q}_k = \left[ \begin{matrix} q_x & q_y & q_z & q_w \end{matrix} \right]^T \).

cl::CommandQueue cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::queue

protected

Eigen::Matrix3f cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::R

Represents the rotation estimation up to iteration k, given in rotation matrix representation, \( R \).

RBC::RBCConstruct<RBC::KernelTypeC::KINECT_R, RBC::RBCPermuteConfig::GENERIC> cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::rbcC

protected

RBC::RBCSearch<RBC::KernelTypeC::KINECT_R, RBC::RBCPermuteConfig::GENERIC, RBC::KernelTypeS::KINECT> cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::rbcS

protected

Eigen::Matrix3f cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::Rk

Represents the incremental development in the rotation estimation in iteration k, given in rotation matrix representation, \( R_k \).

cl_float cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::s

Represents the scale estimation up to iteration k, given as a scalar, \( s \).

cl_float cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::sk

Represents the incremental development in the scale estimation in iteration k, given as a scalar, \( s_k \).

Staging cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::staging

protected

Eigen::Vector3f cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::t

Represents the translation estimation up to iteration k, given as a vector in 3-D, \( t \).

Eigen::Vector3f cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::tk

Represents the incremental development in the translation estimation in iteration k, given as a vector in 3-D, \( t_k \).

cl_float* cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::Tk

protected

ICPTransform<ICPTransformConfig::QUATERNION> cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::transform

protected

ICPWeights cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >::weights

protected

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

Public Types

Public Member Functions

Public Attributes

Protected Attributes

Detailed Description

template<> class cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >

Member Enumeration Documentation

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation

template<>
class cl_algo::ICP::ICPStep< ICPStepConfigT::POWER_METHOD, ICPStepConfigW::WEIGHTED >