The rosbag2_to_pcd package provides functionality to convert point cloud data from a ROS 2 rosbag file to PCD (Point
Cloud Data) files.
For each point cloud message in the specified topic, the package creates an individual PCD file.
Alternative (much more capable) tool:
To install, follow the ROS 2 Humble Installation Guide and then execute:
mkdir -p ~/rosbag2pcd_ws/src
cd ~/rosbag2pcd_ws/src
git clone https://github.com/xmfcx/rosbag2_to_pcd.git
cd ~/rosbag2pcd_ws
sudo apt update
rosdep init
rosdep update
rosdep install -y --from-paths src --ignore-src --rosdistro $ROS_DISTRO
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_EXPORT_COMPILE_COMMANDS=1The node accepts the following parameters:
path_bag: The absolute path to the rosbag2 file that you want to convert.- Example:
/home/mfc/bags/rosbag2_2023_08_15-18_42_55which contains:/home/mfc/bags/rosbag2_2023_08_15-18_42_55/metadata.yaml/home/mfc/bags/rosbag2_2023_08_15-18_42_55/rosbag2_2023_08_15-18_42_55_0.mcap
- Example:
topic_cloud: The topic that contains the point cloud data in the rosbag2 file.
By default, the parameters are sourced from the rosbag2_to_pcd.param.yaml file in the config directory.
Modify this file's default values or provide your own parameter file as needed.
Execute the following commands:
source ~/rosbag2pcd_ws/install/setup.bash
ros2 launch rosbag2_to_pcd rosbag2_to_pcd.launch.xmlOnce launched, the node will process the rosbag file specified by the path_bag parameter.
It will generate a PCD file for each point cloud message, storing them in a directory named <rosbag-name>_pcds adjacent to the rosbag2 folder.
The resulting PCD files use the timestamp of the point cloud message for naming, e.g., 1610472435-452138256.pcd.
To visually check which point cloud corresponds to which timestamp:
- Launch Rviz2 to view the point cloud.
- In a separate terminal, run:
ros2 topic echo /your_point_cloud_topic --field header.stamp
- Play your bag file.
- Pause the bag file playing by pressing
spacekey when needed. - Match the point cloud message's timestamp with the corresponding PCD file name.
import open3d as o3d
import os
def clean_ply_files(input_dir, output_dir):
os.makedirs(output_dir, exist_ok=True)
for filename in os.listdir(input_dir):
if filename.endswith(".pcd"):
# Load with NaN removal
pcd = o3d.io.read_point_cloud(
os.path.join(input_dir, filename),
remove_nan_points=True,
remove_infinite_points=True
)
# Additional cleaning
pcd.remove_non_finite_points()
# Save cleaned version
o3d.io.write_point_cloud(
os.path.join(output_dir, filename),
pcd
)
clean_ply_files("output_pcd_folder", "output_cleaned_ply_folder")