This repository contains a simple demonstration of using ROS and OpenCV to track a ball with a mobile robot via a camera feed.
The associated tutorial is available here.
Thanks very much to Tiziano Fiorenzani for his ROS 1 tutorial and corresponding code on which this repo is based (code reproduced with permission).
One other thing to be wary of: this is a ROS/ament Python package, which means the launch and config files contained will NOT be linked with the --symlink-install command for colcon, and will need to be rebuilt after changes.
Remove warning triggers regarding setuptools by downgrading the version:
pip install setuptools==58.2.0
A launch file and example parameters are provided.
Clone this repo into your ~/dev_ws/src/ folder:
git clone git@github.com:joshnewans/ball_tracker.git
Preliminary tuning mode (remember to source foxy and workspace):
ros2 run ball_tracker detect_ball --ros-args -p tuning_mode:=true -r image_in:=camera/image_raw
rviz2: add Image -> Topic: /image_out
Record those parameters!
Get detect ball's point measurements:
ros2 topic echo /detected_ball
3D ball location estimation:
ros2 run ball_tracker detect_ball_3d
rviz2: add Marker -> Topic: /ball_3d_marker
Follow the ball:
ros2 run ball_tracker follow_ball --ros-args -r cmd_vel:=cmd_vel_tracker
Copy ball_tracker_params_example.yaml into your main ros package config folder.
Rename it to ball_tracker_params_sim.yaml with the appropriate tuning parameters.
Then you can launch:
ros2 launch ball_tracker ball_tracker.launch.py params_file:=src/<main_ros_package_name>/config/ball_tracker_params_sim.yaml enable_3d_track:=true
TODO Document them
- Can't easily detect red due to hue wrap. If you need to detect red robustly you'll need to go in and mess with the HSV filtering.