open3d::t::pipelines::registration::TransformationEstimationForColoredICP Class Reference

#include <TransformationEstimation.h>

Inheritance diagram for open3d::t::pipelines::registration::TransformationEstimationForColoredICP:

Public Member Functions
	~TransformationEstimationForColoredICP () override
	TransformationEstimationForColoredICP (double lambda_geometric=0.968, const RobustKernel &kernel=RobustKernel(RobustKernelMethod::L2Loss, 1.0, 1.0))
	Constructor.
TransformationEstimationType	GetTransformationEstimationType () const override
double	ComputeRMSE (const geometry::PointCloud &source, const geometry::PointCloud &target, const core::Tensor &correspondences) const override
	Computes RMSE (double) for ColoredICP method, between two pointclouds, given correspondences.
core::Tensor	ComputeTransformation (const geometry::PointCloud &source, const geometry::PointCloud &target, const core::Tensor &correspondences, const core::Tensor &current_transform=core::Tensor::Eye(4, core::Float64, core::Device("CPU:0")), const std::size_t iteration=0) const override
	Estimates the transformation matrix for ColoredICP method, a tensor of shape {4, 4}, and dtype Float64 on CPU device.
Public Member Functions inherited from open3d::t::pipelines::registration::TransformationEstimation
	TransformationEstimation ()
	Default Constructor.
virtual	~TransformationEstimation ()

Data Fields
double	lambda_geometric_ = 0.968
RobustKernel	kernel_ = RobustKernel(RobustKernelMethod::L2Loss, 1.0, 1.0)
	RobustKernel for outlier rejection.

Detailed Description

This is implementation of following paper J. Park, Q.-Y. Zhou, V. Koltun, Colored Point Cloud Registration Revisited, ICCV 2017.

Class to estimate a transformation matrix tensor of shape {4, 4}, dtype Float64, on CPU device for colored-icp method.

Constructor & Destructor Documentation

~TransformationEstimationForColoredICP()

open3d::t::pipelines::registration::TransformationEstimationForColoredICP::~TransformationEstimationForColoredICP ( )

inlineoverride

TransformationEstimationForColoredICP()

open3d::t::pipelines::registration::TransformationEstimationForColoredICP::TransformationEstimationForColoredICP ( double  lambda_geometric = 0.968, const RobustKernel &  kernel = RobustKernel(RobustKernelMethod::L2Loss, 1.0, 1.0)  )

inlineexplicit

Constructor.

Parameters

lamda_geometric	λ ∈ [0,1] in the overall energy λEG + (1−λ)EC. Refer the documentation of Colored-ICP for more information.
kernel	(optional) Any of the implemented statistical robust kernel for outlier rejection.

Member Function Documentation

ComputeRMSE()

double open3d::t::pipelines::registration::TransformationEstimationForColoredICP::ComputeRMSE ( const geometry::PointCloud &  source, const geometry::PointCloud &  target, const core::Tensor &  correspondences  ) const

overridevirtual

Computes RMSE (double) for ColoredICP method, between two pointclouds, given correspondences.

Parameters

source	Source pointcloud. (Float32 or Float64 type).
target	Target pointcloud. (Float32 or Float64 type). It must contain normals, colors and color_gradients of the same shape and dtype as the positions.
correspondences	Tensor of type Int64 containing indices of corresponding target points, where the value is the target index and the index of the value itself is the source index. It contains -1 as value at index with no correspondence.

Implements open3d::t::pipelines::registration::TransformationEstimation.

ComputeTransformation()

core::Tensor open3d::t::pipelines::registration::TransformationEstimationForColoredICP::ComputeTransformation ( const geometry::PointCloud &  source, const geometry::PointCloud &  target, const core::Tensor &  correspondences, const core::Tensor &  current_transform = core::Tensor::Eye(4, core::Float64, core::Device("CPU:0")), const std::size_t  iteration = 0  ) const

overridevirtual

Estimates the transformation matrix for ColoredICP method, a tensor of shape {4, 4}, and dtype Float64 on CPU device.

Parameters

source	Source pointcloud. (Float32 or Float64 type).
target	Target pointcloud. (Float32 or Float64 type). It must contain normals, colors and color_gradients of the same shape and dtype as the positions.
correspondences	Tensor of type Int64 containing indices of corresponding target points, where the value is the target index and the index of the value itself is the source index. It contains -1 as value at index with no correspondence.
current_transform	The current pose estimate of ICP.
iteration	The current iteration number of the ICP algorithm.

Returns

transformation between source to target, a tensor of shape {4, 4}, type Float64 on CPU device.

Implements open3d::t::pipelines::registration::TransformationEstimation.

GetTransformationEstimationType()

TransformationEstimationType open3d::t::pipelines::registration::TransformationEstimationForColoredICP::GetTransformationEstimationType ( ) const

inlineoverridevirtual

Implements open3d::t::pipelines::registration::TransformationEstimation.

Field Documentation

kernel_

RobustKernel open3d::t::pipelines::registration::TransformationEstimationForColoredICP::kernel_ = RobustKernel(RobustKernelMethod::L2Loss, 1.0, 1.0)

RobustKernel for outlier rejection.

lambda_geometric_

double open3d::t::pipelines::registration::TransformationEstimationForColoredICP::lambda_geometric_ = 0.968

---

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

• cpp/open3d/t/pipelines/registration/TransformationEstimation.h
• cpp/open3d/t/pipelines/registration/TransformationEstimation.cpp