object_tracker

This package allows for the tracking of moving (rotating) objects and the estimation of the relevant motion parameters. In the case of the rotation the estimated parameters are the center of rotation, the rotation axis and the rotation speed.

How it works

The object_tracker package uses the the output provided by the ROS Object Recognition Kitchen. This package examines subsequent detection results looking for moving objects and, if found, is able to determine the relevant rotation parameters, i.e. the center and axis of rotation and the angular speed. Some statistical indicators (covariances) are also output to give an idea of the stability in time and spatial location of the estimated parameters.

Inputs

There is only one input needed, in the form of a topic of type object_recognition_msgs/RecognizedObjectArray The topic name is by default recognized_object_array

Outputs

This package has various outputs:

• TF data: the center of rotation determines a moving (rotating) reference frame, each tracked object in addition is associated to a static reference frame whose coordinates are expressed in relation to the rotating frame expressed by the center of rotation.
• rotating_objects_markers: topic of type visualization_msgs/MarkerArray, contains a spherical marker for each of the objects currently being tracked.
• rotating_objects: topic of type object_tracker/RotatingObjects; the message contains the current rotation parameters, the covariances and a list of the tracked objects including their respective radii and phases.
• recognized_rotating_objects: topic of type object_recognition_msgs/RecognizedObjectArray, contains the same data as the input recognized_object_array but with the object poses expressed using the rotating reference frame instead of the original TF frame.

Installation

###Dependencies This package depends on the Object Recognition Kitchen hence you need to have in your ROS setup the following packages:

• object_recognition_core
• object_recognition_msgs
• object_recognition_ros

You also need an object recognition algorithm; while you can use any algorithm of your choice the suggested one is object_recognition_tabletop.

Setup

Inside your Catkin workspace clone this repository:

$ git clone https://github.com/wg-perception/object_tracker.git

Then simply run

$ catkin_make

to generate the message, service and dynamic_reconfigure files.

Launch

To launch the tracker with the default parameters and having it launch for you the tabletop detector simply run:

$ roslaunch object_tracker track.launch

Alternatively, if you prefer launching a different object detector or you want the maximum flexibility regarding topic names you can do this:

$ rosrun object_recognition_ros server.py -c /path/to/recognition/config/file
$ rosrun object_tracker estimate_rotation_server.py
$ rosrun object_tracker multi_object_tracker.py

Parameters

There are various parameters that can be set to fine tune the rotation model estimation; the default values should be good for a variety of practical cases but, if you want to change the behavior, you can do so by running:

$ rosrun dynamic_reconfigure reconfigure_gui /objects_tracker

To understand the meaning of each parameter simply hover your mouse on top of its name or have a look inside the [cfg/RotatingObjectTracker.cfg] file.