ROS publisher/subscriber for rgbd image and its point cloud

Example of usage:

$ rosrun ros_pub_and_sub_rgbd_and_cloud pub_rgbd_and_cloud.py \
    --base_dir $(rospack find ros_pub_and_sub_rgbd_and_cloud) \
    --config_file config/rgbd_pub_config.yaml

For quick unit test, run: bash run_test_case.sh

Introduction

This package has three nodes:

1. pub_rgbd_and_cloud.py: For publishing rgbd images and their corresponding point clouds to ROS topics as set by config/rgbd_pub_config.yaml.
2. sub_rgbd_and_cloud.py: Example code of subscribing these data.

You can publish any combination of the following 4 data types by setting is_publish to True in the configuration file:

1. color image (read from a folder).
2. depth image (read from a folder).
3. camera_info (read from a file, without distortion).
4. point cloud (generated from color, depth, and camera_info)

Notes about data paths:

• If publishing both color and depth images, their filename should be the same in their respective folder (e.g.: color/001.png and depth/001.png.)
• If publishing point cloud, data paths of all color and depth and camera_info need to be valid, because the point cloud is generated from them.
• If a data path is a relative path (e.g.: folder: "data/color/"), then the command line argument base_dir will be prepended to create an absolute path.

Test data are already included in the data/ folder. You may clone this repo to your ROS catkin workspace, and then run following commands to test this package.

Usage instructions

(0) Check configurations in config/rgbd_pub_config.yaml. Make ROS nodes executable:

chmod a+x pub_rgbd_and_cloud.py
chmod a+x sub_rgbd_and_cloud.py
roscore # Start ROS.

(1) Set camera pose (Not necessary):

$ rosrun tf static_transform_publisher 0 0 0 0 0 0 1 map head_camera 10

This step is just for visualization in rviz. The camera's frame_id head_camera is set in the config file.

(2) Publish data:

$ rosrun ros_pub_and_sub_rgbd_and_cloud pub_rgbd_and_cloud.py \
    --base_dir $(rospack find ros_pub_and_sub_rgbd_and_cloud) \
    --config_file config/rgbd_pub_config.yaml

(3) View result in RVIZ: (If you change the topic names in config file, you also need to change it in rviz.)

$ roslaunch ros_pub_and_sub_rgbd_and_cloud run_rviz.launch 

(4) Run steps (1)~(3) by using roslaunch (You need stop the above 3 commands first):

$ roslaunch ros_pub_and_sub_rgbd_and_cloud run_publisher.launch

(5) Subscribe data:

$ rosrun ros_pub_and_sub_rgbd_and_cloud sub_rgbd_and_cloud.py \
    --base_dir $(rospack find ros_pub_and_sub_rgbd_and_cloud) \
    --config_file config/rgbd_pub_config.yaml

Other Notes

• About point cloud. The point cloud is generated from the rgbd image and camera info by using Open3D library.

• About tests. See tests/testing_commands.txt.

Install Dependencies

The package is written under: Ubuntu 18.04, ROS melodic, Python 2.7.

The main dependencies are:

• opencv
• pcl (for ROS point cloud message type.)
• open3d (for creating point cloud from rgbd images.)

Installation commands:

• Install pcl (Might be installed by default when installing ROS melodic): http://wiki.ros.org/pcl_ros

  sudo add-apt-repository ppa:v-launchpad-jochen-sprickerhof-de/pcl  
  sudo apt-get update  
  sudo apt-get install libpcl-all  
  sudo apt-get install ros-melodic-pcl-ros  
  sudo apt-get install ros-melodic-pcl-conversions    

• Open3D
  See two official installation tutorials: this and this.

  pip install --user open3d