Matching

Low-level functions for matching the map pair. More...

Functions
	map_merge_3d::ENUM_CLASS (EstimationMethod, MATCHING, SAC_IA)
Eigen::Matrix4f	map_merge_3d::estimateTransform (const PointCloudPtr &source_points, const PointCloudPtr &source_keypoints, const LocalDescriptorsPtr &source_descriptors, const PointCloudPtr &target_points, const PointCloudPtr &target_keypoints, const LocalDescriptorsPtr &target_descriptors, EstimationMethod method, bool refine, double inlier_threshold, double max_correspondence_distance, int max_iterations, size_t matching_k, double transform_epsilon)
	Estimate transformation between two pointclouds. More...
Eigen::Matrix4f	map_merge_3d::estimateTransformFromCorrespondences (const PointCloudPtr &source_keypoints, const PointCloudPtr &target_keypoints, const CorrespondencesPtr &correspondences, CorrespondencesPtr &inliers, double inlier_threshold)
	Estimates transformation between source and target pointcloud based on correspondences. More...
Eigen::Matrix4f	map_merge_3d::estimateTransformFromDescriptorsSets (const PointCloudPtr &source_keypoints, const LocalDescriptorsPtr &source_descriptors, const PointCloudPtr &target_keypoints, const LocalDescriptorsPtr &target_descriptors, double min_sample_distance, double max_correspondence_distance, int max_iterations)
	Estimates transformation between source and target pointcloud based on descriptors. More...
Eigen::Matrix4f	map_merge_3d::estimateTransformICP (const PointCloudPtr &source_points, const PointCloudPtr &target_points, const Eigen::Matrix4f &initial_guess, double max_correspondence_distance, double outlier_rejection_threshold, int max_iterations=100, double transformation_epsilon=0.0)
	Use ICP to estimate transform between grids. More...
CorrespondencesPtr	map_merge_3d::findFeatureCorrespondences (const LocalDescriptorsPtr &source_descriptors, const LocalDescriptorsPtr &target_descriptors, size_t k=5)
	Finds correspondences between two sets of feature descriptors. More...
double	map_merge_3d::transformScore (const PointCloudPtr &source_points, const PointCloudPtr &target_points, const Eigen::Matrix4f &transform, double max_distance)
	Computes euclidean distance between two pointclouds. More...

Detailed Description

Low-level functions for matching the map pair.

Low-level functions to find correspondences between two pointclouds and match them together. These functions operate on two pointclouds.

Function Documentation

ENUM_CLASS()

map_merge_3d::ENUM_CLASS	(	EstimationMethod	,
		MATCHING	,
		SAC_IA
	)

estimateTransform()

Eigen::Matrix4f map_merge_3d::estimateTransform	(	const PointCloudPtr &	source_points,
		const PointCloudPtr &	source_keypoints,
		const LocalDescriptorsPtr &	source_descriptors,
		const PointCloudPtr &	target_points,
		const PointCloudPtr &	target_keypoints,
		const LocalDescriptorsPtr &	target_descriptors,
		EstimationMethod	method,
		bool	refine,
		double	inlier_threshold,
		double	max_correspondence_distance,
		int	max_iterations,
		size_t	matching_k,
		double	transform_epsilon
	)

Estimate transformation between two pointclouds.

Uses extracted features to estimate rigid transformation. First the initial transformation is estimated using selected method, then the transformation may be selectively refined using ICP.

Parameters

source_points	Source pointcloud
source_keypoints	Keypoints of source pointcloud
source_descriptors	Descriptors for keypoints of source pointcloud
target_points	Target pointcloud
target_keypoints	Keypoints of target pointcloud
target_descriptors	Descriptors for keypoints of target pointcloud
method	Method for estimating initial transformation
refine	Whether to refine initial transformation with ICP.
inlier_threshold	Threshold for inliers in RANSAC during initial estimation.
max_correspondence_distance	Maximum distance for a matched points to be considered the same point
max_iterations	maximum iterations for RANSAC
matching_k	number of nearest descriptors to consider for matching
transform_epsilon	the smallest change allowed until ICP convergence.

Returns

estimated rigid transform between source and target pointclouds or zero matrix if the transformation could not be estimated

estimateTransformFromCorrespondences()

Eigen::Matrix4f map_merge_3d::estimateTransformFromCorrespondences	(	const PointCloudPtr &	source_keypoints,
		const PointCloudPtr &	target_keypoints,
		const CorrespondencesPtr &	correspondences,
		CorrespondencesPtr &	inliers,
		double	inlier_threshold
	)

Estimates transformation between source and target pointcloud based on correspondences.

Uses RANSAC to find inliers fitting a rigid transformation model. Final transformation is found using SVD on inliers set.

Parameters

	source_keypoints	Keypoints of source pointcloud
	target_keypoints	Keypoints of target pointcloud
	correspondences	Correspondences between keypoints
[out]	inliers	Estimated inliers from RANSAC
	inlier_threshold	threshold for considering a point as inlier in RANSAC

Returns

estimated rigid transformation between source and target or zero matrix if transformation could not be estimated

estimateTransformFromDescriptorsSets()

Eigen::Matrix4f map_merge_3d::estimateTransformFromDescriptorsSets	(	const PointCloudPtr &	source_keypoints,
		const LocalDescriptorsPtr &	source_descriptors,
		const PointCloudPtr &	target_keypoints,
		const LocalDescriptorsPtr &	target_descriptors,
		double	min_sample_distance,
		double	max_correspondence_distance,
		int	max_iterations
	)

Estimates transformation between source and target pointcloud based on descriptors.

Use SampleConsensusInitialAlignment to find a rough alignment from the source to the target

Parameters

source_keypoints	Keypoints of source pointcloud
source_descriptors	descriptors for source_keypoints
target_keypoints	Keypoints of target pointcloud
target_descriptors	descriptors for target_keypoints
min_sample_distance	The minimum distance between any two random samples
max_correspondence_distance	Maximum distance for a matched points to be considered the same point
max_iterations	The number of RANSAC iterations to perform

Returns

estimated rigid transformation between source and target or zero matrix

estimateTransformICP()

Eigen::Matrix4f map_merge_3d::estimateTransformICP	(	const PointCloudPtr &	source_points,
		const PointCloudPtr &	target_points,
		const Eigen::Matrix4f &	initial_guess,
		double	max_correspondence_distance,
		double	outlier_rejection_threshold,
		int	max_iterations = 100,
		double	transformation_epsilon = 0.0
	)

Use ICP to estimate transform between grids.

Parameters

source_points	source point cloud
target_points	target point cloud.
initial_guess	Initial transformation to start with
max_correspondence_distance	A threshold on the distance between any two corresponding points. Any corresponding points that are further apart than this threshold will be ignored when computing the source-to-target transformation
outlier_rejection_threshold	A threshold used to define outliers during RANSAC
max_iterations	maximum iterations for RANSAC
transformation_epsilon	The smallest iterative transformation allowed before the algorithm is considered to have converged

Returns

estimated rigid transformation between source and target pointclouds

findFeatureCorrespondences()

CorrespondencesPtr map_merge_3d::findFeatureCorrespondences	(	const LocalDescriptorsPtr &	source_descriptors,
		const LocalDescriptorsPtr &	target_descriptors,
		size_t	k = 5
	)

Finds correspondences between two sets of feature descriptors.

Uses cross-matching algorithm with first-k selection

Parameters

source_descriptors	Feature descriptors of source pointcloud
target_descriptors	Feature descriptors of target pointcloud
k	number of nearest descriptors to consider for matching

Returns

correspondences source -> target

transformScore()

double map_merge_3d::transformScore	(	const PointCloudPtr &	source_points,
		const PointCloudPtr &	target_points,
		const Eigen::Matrix4f &	transform,
		double	max_distance
	)

Computes euclidean distance between two pointclouds.

Computes a euclidean score for an estimated transformation. Because we expect only some parts of the maps overlapping, only points closer than max_distance will be include in the score.

Parameters

source_points	Source pointcloud
target_points	Target pointcloud
transform	estimated transformation between source and target
max_distance	Maximum distance between two points to be included in the score.

Returns

transformation euclidean score