Skip to content

Tracking

Antonio Terpin edited this page May 27, 2021 · 1 revision

Tracking parameters

This page summarizes the purpose of the parameters that affect the tracking module. Please, consider the linked references for further details on how the parameters are used by the tracking algorithm.

If you plan to modify or extend the tracking module, consider having a look at the documentation.

Table of contents

  1. Basic parameters
  2. Trajectory estimation
  3. LKSE3
  4. [OPTIONAL] Robustness
  5. [OPTIONAL] Visualizations
  6. Further references

Parameter Description
discard_events_when_idle whether not to collect events when idle

Trajectory estimation

Parameter Description
max_event_rate events processed are randomly sampled so that the rate is below this value
events_per_kf events required for a new keyframe
frame_size window of events considered
step_size minimum number of new events to wait before a pose update
noise_rate if events rate is below this value, the frame is skipped
pose_mean_filter_size median filter size (poses are median filtered)

LKSE3 parameters

Parameter Description
batch_size batch-gradient descent batch size
max_iterations maximum number of iterations in the optimization
map_blur sigma of the gaussian filter applied to the reprojected map
pyramid_levels number of pyramid levels used in the KLT process
weight_scale_translation weight used in the translation error
weight_scale_rotation weight used in the rotation error

Please refer to the documentation to go deeper in the details of how these parameters are used in our implementation of the Lucas-Kanade tracker based on the SE3 group, or consider reading the linked references.

[OPTIONAL] Robustness

To further improve the robustness of the pipeline, it is possible to introduce triggering strategies when specific conditions are met.

Parameter Description
min_map_size minimum number of map points to proceed with the update
min_n_keypoints minimum number of extracted keypoints (LKSE3::keypoints_) required for a reliable tracking

[OPTIONAL] Visualizations

These parameters are used to customize the images juxtaposing the reprojected map and the aggregated events frame.

Parameter Description
event_map_overlap_rate publishing rate of the visualizations
max_depth maximum expected depth in the scene (it is recommended to use voxel grid size)
min_depth minimum expected depth in the scene (it is recommended to use voxel grid size)

The following parameters describes some of these conditions.

Parameter Description
min_map_size minimum number of map points to proceed with the update
min_n_keypoints minimum number of extracted keypoints (LKSE3::keypoints_) required for a reliable tracking

Further references

To fully understand how the parameters are used in the tracking module, please refer to: