Offers functions that are serial CPU implementations of the relevant algorithms in the ICP pipeline. More...

Functions
bool	setProfilingFlag (int argc, char **argv)
	Checks the command line arguments for the profiling flag, `--profiling`. More...

template<typename T >
uint64_t	nextPow2 (T num)
	Returns the first power of 2 greater than or equal to the input. More...

template<typename T >
void	printBuffer (const char title, T ptr, uint32_t width, uint32_t height)
	Prints an array of an integer type to standard output. More...

template<typename T >
void	printBufferF (const char title, T ptr, uint32_t width, uint32_t height, uint32_t prec)
	Prints an array of floating-point type to standard output. More...

template<typename T >
void	cpuReduce (T in, T out, uint32_t cols, uint32_t rows, std::function< bool(T, T)> func)
	Reduces each row of an array to a single element. More...

template<typename T >
void	cpuReduceSum (T in, T out, uint32_t cols, uint32_t rows)
	Reduces each row of an array to a single element (sum). More...

template<typename T >
void	cpuInScan (T in, T out, uint32_t width, uint32_t height)
	Performs an inclusive scan operation on the columns of an array. More...

template<typename T >
void	cpuExScan (T in, T out, uint32_t width, uint32_t height)
	Performs an exclusive scan operation on the columns of an array. More...

template<typename T >
void	cpuICPLMs (T in, T out)
	Samples a point cloud for 16384 (128x128) landmarks. More...

template<typename T >
void	cpuICPReps (T in, T out, uint32_t nr)
	Samples a set of 16384 (128x128) landmarks for representatives. More...

template<typename T >
void	cpuICPWeights (rbc_dist_id D, T W, cl_double *SW, uint32_t n)
	Computes weights for pairs of points in the fixed and moving sets, and also reduces them to get their sum. More...

template<typename T >
void	cpuICPMean (T F, T M, T *mean, uint32_t n)
	Computes the mean on the xyz dimensions of the set of 8-D points. More...

template<typename T >
void	cpuICPMeanWeighted (T F, T M, T MEAN, T W, uint32_t n)
	Computes the weighted mean on the xyz dimensions of the set of 8-D points. More...

template<typename T >
void	cpuICPDevs (T F, T M, T DF, T DM, T *mean, uint32_t n)
	Computes the deviations of a set of points from their mean. More...

template<typename T >
void	cpuICPS (T DM, T DF, T *S, uint32_t m, float c)
	Calculates the S matrix and the constituents of the scale factor s. More...

template<typename T >
void	cpuICPSw (T M, T F, T W, T S, uint32_t m, float c)
	Calculates the S matrix and the constituents of the scale factor s. More...

template<typename T >
void	cross_product (T a, T b, T *c)
	Performs a cross product. More...

template<typename T >
void	cpuICPTransformQ (T M, T tM, T *D, uint32_t m)
	Performs a homogeneous transformation on a set of points using a quaternion and a translation vector. More...

template<typename T >
void	cpuICPTransformQ2 (T M, T tM, T *D, uint32_t m)
	Performs a homogeneous transformation on a set of points using a quaternion and a translation vector. More...

template<typename T >
void	cpuICPTransformM (T M, T tM, T *D, uint32_t m)
	Performs a homogeneous transformation on a set of points using a transformation matrix. More...

template<typename T >
T	cpuLength (T *x)
	Computes the vector length ( \(\ell_2\) norm). More...

template<typename T >
T	cpuDistance (T x1, T x2)
	Computes the vector distance ( \(\ell_2\) norm). More...

template<typename T >
void	cpuNormalize (T *x)
	Normalizes a vector. More...

template<typename T >
void	cpuProd (T N, T x, T *x_new)
	Computes a matrix-vector product, \( x_{new}=Nx \). More...

template<typename T >
T	cpuICPPowerMethod (T Sij, T means, T *Tk)
	Computes the quantities that represent the incremental development in the transformation estimation in iteration `k`. More...

Variables
std::function< unsigned char()>	rNum_0_255 = std::bind (distribution1, generator)
	Uniform number generator in the range \([0, 255]\). More...

std::function< unsigned short()>	rNum_0_10000 = std::bind (distribution2, generator)
	Uniform number generator in the range \([0, 10000]\). More...

std::function< float()>	rNum_R_0_1 = std::bind (distributionR1, generator)
	Uniform number generator in the range \([0.0, 1.0)\). More...

std::function< float()>	rNum_R_1_255_E__6 = std::bind (distributionR2, generator)
	Uniform number generator in the range \([1e-6, 255*1e-6)\). More...

Detailed Description

Offers functions that are serial CPU implementations of the relevant algorithms in the ICP pipeline.

Function Documentation

template<typename T >

T ICP::cpuDistance	(	T *	x1,
		T *	x2
	)

Computes the vector distance ( \(\ell_2\) norm).

Template Parameters

T	type of the data to be handled.

Parameters

[in]	x1	4-vector.
[in]	x2	4-vector.

Returns: The vector distance.

template<typename T >

void ICP::cpuExScan	(	T *	in,
		T *	out,
		uint32_t	width,
		uint32_t	height
	)

Performs an exclusive scan operation on the columns of an array.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	in	input data.
[out]	out	output (scan) data.
[in]	width	width of the array.
[in]	height	height of the array.

template<typename T >

void ICP::cpuICPDevs	(	T *	F,
		T *	M,
		T *	DF,
		T *	DM,
		T *	mean,
		uint32_t	n
	)

Computes the deviations of a set of points from their mean.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	F	fixed set of 8-D elements.
[in]	M	moving set of 8-D elements.
[out]	DM	array (moving set deviations) of 4-D elements.
[out]	DF	array (fixed set deviations) of 4-D elements.
[in]	mean	fixed and moving set means.
[in]	n	number of points in the sets.

template<typename T >

void ICP::cpuICPLMs	(	T *	in,
		T *	out
	)

Samples a point cloud for 16384 (128x128) landmarks.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	in	input data.
[out]	out	output (LM) data.

template<typename T >

void ICP::cpuICPMean	(	T *	F,
		T *	M,
		T *	mean,
		uint32_t	n
	)

Computes the mean on the xyz dimensions of the set of 8-D points.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	F	fixed set.
[in]	M	moving set.
[out]	mean	output data.
[in]	n	number of points in the array.

template<typename T >

void ICP::cpuICPMeanWeighted	(	T *	F,
		T *	M,
		T *	MEAN,
		T *	W,
		uint32_t	n
	)

Computes the weighted mean on the xyz dimensions of the set of 8-D points.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	F	fixed set.
[in]	M	moving set.
[out]	MEAN	output data.
[in]	W	weights.
[in]	n	number of points in the array.

template<typename T >

T ICP::cpuICPPowerMethod	(	T *	Sij,
		T *	means,
		T *	Tk
	)

Computes the quantities that represent the incremental development in the transformation estimation in iteration k.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	Sij	sums of products.
[in]	means	fixed and moving set means.
[out]	Tk	the unit quaternion \( \dot{q_k} = q_w + q_x i + q_y j + q_z k = \left[ \begin{matrix} q_x & q_y & q_z & q_w \end{matrix} \right]^T \), and the translation vector \( t_k=\left[ \begin{matrix} t_x & t_y & t_z & 1 \end{matrix} \right]^T \). The scale is placed in the last element of the translation vector. That is, \( t_k = \left[ \begin{matrix} t_x & t_y & t_z & s_k \end{matrix} \right]^T \).

template<typename T >

void ICP::cpuICPReps	(	T *	in,
		T *	out,
		uint32_t	nr
	)

Samples a set of 16384 (128x128) landmarks for representatives.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	in	input data.
[out]	out	output (LM) data.
[in]	nr	number of representatives.

template<typename T >

void ICP::cpuICPS	(	T *	DM,
		T *	DF,
		T *	S,
		uint32_t	m,
		float	c
	)

Calculates the S matrix and the constituents of the scale factor s.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	DM	input array (moving set deviations) of 4-D elements.
[in]	DF	input array (fixed set deviations) of 4-D elements.
[out]	S	output (sums of products) matrix S.
[in]	m	number of points in the sets.
[in]	c	scaling factor.

template<typename T >

void ICP::cpuICPSw	(	T *	M,
		T *	F,
		T *	W,
		T *	S,
		uint32_t	m,
		float	c
	)

Calculates the S matrix and the constituents of the scale factor s.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	M	input array (moving set deviations) of 4-D elements.
[in]	F	input array (fixed set deviations) of 4-D elements.
[in]	W	input array (weights).
[out]	S	output (sums of products) matrix S.
[in]	m	number of points in the sets.
[in]	c	scaling factor.

template<typename T >

void ICP::cpuICPTransformM	(	T *	M,
		T *	tM,
		T *	D,
		uint32_t	m
	)

Performs a homogeneous transformation on a set of points using a transformation matrix.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	M	input array (set of points) of 8-D elements.
[out]	tM	output array (transformed points) of 8-D elements.
[in]	D	transformation parameters.
[in]	m	number of points in the set.

template<typename T >

void ICP::cpuICPTransformQ	(	T *	M,
		T *	tM,
		T *	D,
		uint32_t	m
	)

Performs a homogeneous transformation on a set of points using a quaternion and a translation vector.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	M	input array (set of points) of 8-D elements.
[out]	tM	output array (transformed points) of 8-D elements.
[in]	D	transformation parameters.
[in]	m	number of points in the set.

template<typename T >

void ICP::cpuICPTransformQ2	(	T *	M,
		T *	tM,
		T *	D,
		uint32_t	m
	)

Performs a homogeneous transformation on a set of points using a quaternion and a translation vector.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	M	input array (set of points) of 8-D elements.
[out]	tM	output array (transformed points) of 8-D elements.
[in]	D	transformation parameters.
[in]	m	number of points in the set.

template<typename T >

void ICP::cpuICPWeights	(	rbc_dist_id *	D,
		T *	W,
		cl_double *	SW,
		uint32_t	n
	)

Computes weights for pairs of points in the fixed and moving sets, and also reduces them to get their sum.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	D	input data (distances).
[out]	W	output data (weights).
[out]	SW	output data (sum of weights).
[in]	n	number of elements in the input array.

template<typename T >

void ICP::cpuInScan	(	T *	in,
		T *	out,
		uint32_t	width,
		uint32_t	height
	)

Performs an inclusive scan operation on the columns of an array.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	in	input data.
[out]	out	output (scan) data.
[in]	width	width of the array.
[in]	height	height of the array.

template<typename T >

T ICP::cpuLength ( T * x )

Computes the vector length ( \(\ell_2\) norm).

Template Parameters

T	type of the data to be handled.

Parameters

[in] x 4-vector.

Returns: The vector length.

template<typename T >

void ICP::cpuNormalize ( T * x )

Normalizes a vector.

Template Parameters

T	type of the data to be handled.

Parameters

x 4-vector.

template<typename T >

void ICP::cpuProd	(	T *	N,
		T *	x,
		T *	x_new
	)

Computes a matrix-vector product, \( x_{new}=Nx \).

Template Parameters

T	type of the data to be handled.

Parameters

[in]	N	`4x4` matrix.
[in]	x	4-vector.
[out]	x_new	4-vector.

template<typename T >

void ICP::cpuReduce	(	T *	in,
		T *	out,
		uint32_t	cols,
		uint32_t	rows,
		std::function< bool(T, T)>	func
	)

Reduces each row of an array to a single element.

It is just a naive serial implementation.

Parameters

[in]	in	input data.
[out]	out	output (reduced) data.
[in]	cols	number of columns in the input array.
[in]	rows	number of rows in the input array.
[in]	func	function supporting the requested operation.

template<typename T >

void ICP::cpuReduceSum	(	T *	in,
		T *	out,
		uint32_t	cols,
		uint32_t	rows
	)

Reduces each row of an array to a single element (sum).

It is just a naive serial implementation.

Parameters

[in]	in	input data.
[out]	out	output (reduced) data.
[in]	cols	number of columns in the input array.
[in]	rows	number of rows in the input array.

template<typename T >

void ICP::cross_product	(	T *	a,
		T *	b,
		T *	c
	)

Performs a cross product.

It is just a naive serial implementation.

Template Parameters

T	type of the data to be handled.

Parameters

[in]	a	first input vector.
[in]	b	second input vector.
[out]	c	output vector.

template<typename T >

uint64_t ICP::nextPow2 ( T num )

Returns the first power of 2 greater than or equal to the input.

Parameters

[in] num input number.

Returns: The first power of 2 >= num.

template<typename T >

void ICP::printBuffer	(	const char *	title,
		T *	ptr,
		uint32_t	width,
		uint32_t	height
	)

Prints an array of an integer type to standard output.

Template Parameters

T	type of the data to be printed.

Parameters

[in]	title	legend for the output.
[in]	ptr	array that is to be displayed.
[in]	width	width of the array.
[in]	height	height of the array.

template<typename T >

void ICP::printBufferF	(	const char *	title,
		T *	ptr,
		uint32_t	width,
		uint32_t	height,
		uint32_t	prec
	)

Prints an array of floating-point type to standard output.

Template Parameters

T	type of the data to be printed.

Parameters

[in]	title	legend for the output.
[in]	ptr	array that is to be displayed.
[in]	width	width of the array.
[in]	height	height of the array.
[in]	prec	the number of decimal places to print.

bool ICP::setProfilingFlag	(	int	argc,
		char **	argv
	)

Checks the command line arguments for the profiling flag, --profiling.

Parameters

[in]	argc	command line argument count
[in]	argv	command line arguments

Returns: A flag to indicate whether a command line argument for profiling was provided.

Variable Documentation

std::function< unsigned short()> ICP::rNum_0_10000 = std::bind (distribution2, generator)

Uniform number generator in the range \([0, 10000]\).

std::function< unsigned char()> ICP::rNum_0_255 = std::bind (distribution1, generator)

Uniform number generator in the range \([0, 255]\).

std::function< float()> ICP::rNum_R_0_1 = std::bind (distributionR1, generator)

Uniform number generator in the range \([0.0, 1.0)\).

std::function< float()> ICP::rNum_R_1_255_E__6 = std::bind (distributionR2, generator)

Uniform number generator in the range \([1e-6, 255*1e-6)\).

Functions

Variables

Detailed Description

Function Documentation

Variable Documentation