#include <loop_closing.h>

Public Member Functions
void	addClusterToQueue (Cluster cluster)
	Add a cluster to the queue of clusters.
void	finalize ()
	Finalizes the loop closing class.
	LoopClosing ()
	Class constructor.
void	run ()
	Starts graph.
void	setGraph (Graph *graph)
	Set the graph object.
Protected Member Functions
bool	checkNewClusterInQueue ()
	Check if there are clusters in the queue.
bool	closeLoopWithCluster (Cluster candidate)
	Tries to close a loop between two clusters.
void	drawLoopClosure (vector< int > cand_kfs, vector< int > cand_matchings, vector< int > inliers, vector< int > definitive_inliers_per_pair, vector< vector< int > > definitive_cluster_pairs, vector< cv::Point2f > matched_query_kp_l, vector< cv::Point2f > matched_cand_kp_l)
	Draw and publish a loop closure image with all the correspondences between current keyframe and all the loop closing keyframes.
void	getCandidates (int cluster_id, vector< pair< int, float > > &candidates)
	Get the best candidates to close a loop by hash.
void	processNewCluster ()
	Processes the new cluster.
Cluster	readCluster (int id)
	Read cluster data from file.
void	searchByHash ()
	Searches a loop closing between current cluster and all other clusters using the hash.
void	searchByProximity ()
	Searches a loop closing between current cluster and its closest neighbors.
Private Attributes
Cluster	c_cluster_
image_geometry::PinholeCameraModel	camera_model_
	> Publishes the image with the loop closure matchings
vector< pair< int, int > >	cluster_lc_found_
	> Hash table: stores a hash for every image. This is the unique variable that grows with the robot trajectory
list< Cluster >	cluster_queue_
	> Current cluster to be processed
string	execution_dir_
	> Stores the number of loop closures
Graph *	graph_
	> Directory where images of loop closures will be stored
haloc::Hash	hash_
	> Mutex for the insertion of new clusters
vector< pair< int, vector < float > > >	hash_table_
	> Hash object
string	loop_closures_dir_
	> Execution directory where all image information will be stored
mutex	mutex_cluster_queue_
	> Clusters queue to be inserted into the graph
int	num_loop_closures_
	> Stores all the loop closures (between clusters) found in order to do not repeat them
ros::Publisher	pub_lc_matchings_
	> Publishes the loop closing queue size
ros::Publisher	pub_num_keyframes_
	> Graph pointer
ros::Publisher	pub_num_lc_
	> Publishes the number of keyframes
ros::Publisher	pub_queue_
	> Publishes the number of loop closings

Detailed Description

Definition at line 30 of file loop_closing.h.

Constructor & Destructor Documentation

slam::LoopClosing::LoopClosing ( )

Class constructor.

Definition at line 14 of file loop_closing.cpp.

Member Function Documentation

void slam::LoopClosing::addClusterToQueue ( Cluster cluster )

Add a cluster to the queue of clusters.

Parameters:

The	cluster to be inserted

Definition at line 95 of file loop_closing.cpp.

bool slam::LoopClosing::checkNewClusterInQueue ( ) [protected]

Check if there are clusters in the queue.

Returns:: true if clusters queue is not empty.

Definition at line 101 of file loop_closing.cpp.

bool slam::LoopClosing::closeLoopWithCluster ( Cluster candidate ) [protected]

Tries to close a loop between two clusters.

Returns:: true if loop closing

Parameters:

Candidate cluster

Definition at line 174 of file loop_closing.cpp.

void slam::LoopClosing::drawLoopClosure	(	vector< int >	cand_kfs,
		vector< int >	cand_matchings,
		vector< int >	inliers,
		vector< int >	definitive_inliers_per_pair,
		vector< vector< int > >	definitive_cluster_pairs,
		vector< cv::Point2f >	matched_query_kp_l,
		vector< cv::Point2f >	matched_cand_kp_l
	)		`[protected]`

Draw and publish a loop closure image with all the correspondences between current keyframe and all the loop closing keyframes.

Parameters:

The	loop closing keyframe identifiers
The	loop closing cluster identifiers
The	inlier indices
The	number of inliers for every loop closure cluster pair
The	cluster pairs vector
All	matched keypoints of the current keyframe
All	matched keypoints of the candidate keyframes