TargetAdaptiveTracking

What Is This

This node finds the nearest plane from input pose and publish pose aligned to that plane.

Subscribing Topic

• ~input_obj_cloud (sensor_msgs/PointCloud2)

  Object point cloud used for initialization.

• ~input_bkgd_cloud (sensor_msgs/PointCloud2)

  Background point cloud used for initialization.

• ~input_obj_pose (geometry_msgs/PoseStamped)

  Object pose used for initialization.

Note that ~input_obj_cloud, ~input_bkgd_cloud and ~input_obj_pose should be synchronized.

• ~input_cloud (sensor_msgs/PointCloud2)

  Target point cloud.

• ~input_pose (geometry_msgs/PoseStamped)

  Object pose used for updating motion displacement.

Note that ~input_cloud and ~input_pose shoule be synchronized.

Publishing Topic

• /selected_pointcloud (sensor_msgs/PointCloud2)

  Template point cloud to track.

• /target_adaptive_tracking/output/object_pose (geometry_msgs/PoseStamped)

  Result pose of tracking.

• /target_adaptive_tracking/output/cloud (sensor_msgs/PointCloud2)

  Relay of ~input_cloud.

• /target_adaptive_tracking/output/normal (sensor_msgs/PointCloud2)

  Point cloud of centroid with normal field.

• /target_adaptive_tracking/output/inliers (sensor_msgs/PointCloud2)

  Point cloud of inliers.

• /target_adaptive_tracking/output/centroids (sensor_msgs/PointCloud2)

  Point cloud of estimated position.

• /target_adaptive_tracking/output/probability_map (sensor_msgs/PointCloud2)

  Point cloud representing probability.

• /target_adaptive_tracking/supervoxel/cloud (sensor_msgs/PointCloud2)

• /target_adaptive_tracking/supervoxel/indices (jsk_recognition_msgs/ClusterPointIndices)

  Result point cloud / cluster point indices of supervoxel segmentation.

• /target_adaptive_tracking/supervoxel/tdp_indices (jsk_recognition_msgs/ClusterPointIndices)

  Filtered cluster point indices of supervoxel segmentation.

Parameters

• ~use_tf (Bool, default: False)

  Whether to look up transform between ~parent_frame_id and ~child_frame_id to update transform.

• ~parent_frame_id (String, default: /track_result)

  Parent frame ID.

  This parameter is used only when ~use_tf is True.

• ~child_frame_id (String, default: /camera_rgb_optical_frame)

  Child frame ID used for /target_adaptive_tracking/output/cloud and /target_adaptive_tracking/output/object_pose.

• ~color_importance (Float, default: 0.2)

• ~spatial_importance (Float, default: 0.4)

• ~normal_importance (Float, default: 1.0)

  Importances used for supervoxel segmentation.

• ~use_transform (Bool, default: True)

  This parameter is not used.

• ~seed_resolution (Float, default: 0.1)

  Average size in meters of resulting supervoxels.

• ~voxel_resolution (Float, default: 0.008)

  Resolution in meters of voxel used.

• ~min_cluster_size (Int, default: 20)

  Minimum number of supervoxels.

• ~threshold (Float, default: 0.4)

  Probability threshold.

• ~bin_size (Int, default: 18)

  Local structural rpy bin.

• ~eps_distance (Float, default: 0.03)

• ~eps_min_samples (Int, default: 3)

  These parameters are not used now.

• ~update_tracker_reference (Bool, default: False)

  Whether to update tracking model.

• ~vfh_scaling (Float, default: 0.7)

  Likelihood scaling factor for vfh matching.

• ~color_scaling (Float, default: 0.5)

  Likelihood scaling factor for color matching.

• ~structure_scaling (Float, default: 0.0)

  Likelihood scaling factor for local voxel adjacency.

• ~update_filter_template (Bool, default: False)

  Whether to update the particle filter tracking template.

• ~history_window_size (Int, default: 5)

  Number of frame after which the unmatched voxel is discarded.

Sample

roslaunch jsk_pcl_ros sample_target_adaptive_tracking.launch