Awesome

principal_component_analysis_tracking

Configuration for the drl perception pipeline for performing 6 DoF tracking of an object within a Region of Interest (ROI) using Principal Component Analysis (PCA).

This package was tested for tracking the top section of a box moving on top of a conveyor, as seen in the figure below.

Check the documentation of the dynamic_robot_localization package for customizing the perception pipeline for your specific use case.

Usage

For starting all the main launch files:

roslaunch principal_component_analysis_tracking bringup.launch

For starting each of the main launch files separately, first run the 3D sensor driver with its extrinsic calibration:

roslaunch principal_component_analysis_tracking intel_realsense.launch
roslaunch principal_component_analysis_tracking intel_realsense_dynamic_reconfigure.launch
roslaunch principal_component_analysis_tracking tfs.launch

Then, run the PCA tracking node:

roslaunch principal_component_analysis_tracking principal_component_analysis_tracking.launch

For monitoring the tracking state, run:

roslaunch principal_component_analysis_tracking rviz.launch

Calibration of coordinate system associated with the ROI

The PCA tracking is performed within a calibrated workspace, in which a ROI is defined for segmenting the region in which the target object will be moving (box on top of a conveyor). For recalibrating it, start the charuco_detector:

roslaunch principal_component_analysis_tracking charuco_detector.launch

Then, adjust the tf publisher in launch/tfs.launch with the updated coordinate system given by:

rostopic echo /camera/color/image_raw_charuco_pose