JP | EN

PCL Object Detection

Overview

• A high-performance, component-based object detection package for ROS 2 Jazzy.
• Utilizes the Point Cloud Library (PCL) for rule-based detection.
• Architecture: ROS 2 Lifecycle Nodes with Zero-Copy Intra-Process Communication.
• Modes: Floor, Table, Shelf, Placeable Positions, and 2D-LiDAR Line Detection.

(Back to Top)

Setup

Here, we describe the setup process for this repository.

(Back to Top)

Environment

Below are the system requirements for normal operation.

System	Version
Ubuntu	24.04 (Noble Numbat)
ROS	Jazzy Jalisco

Note

If you need to install Ubuntu or ROS, please check our SOBITS Manual.

(Back to Top)

Installation

1. Move to the src folder of your ROS workspace.

   $ cd ~/colcon_ws/src/

2. Clone this repository.

   $ git clone -b jazzy-devel https://github.com/TeamSOBITS/pcl_object_detection.git

3. Install the dependent packages.

   $ bash install.sh

4. Compile the package.

   $ cd ~/colcon_ws
   $ colcon build --symlink-install

(Back to Top)

Usage

The system runs as a single ComposableNodeContainer to enable zero-copy memory sharing.

Launch the container:

ros2 launch pcl_object_detection pcl_object_detection.launch.py

(Back to Top)

Lifecycle Management

This package uses ROS 2 Lifecycle Nodes. By default, all detection workers are Unconfigured. You must manage their state to begin processing.

Workflow Example (Table Detection)

1. Configure (Allocates memory and loads parameters):

   ros2 lifecycle set /pcl_object_detection/table_detection configure

2. Activate (Starts data subscription and processing):

   ros2 lifecycle set /pcl_object_detection/table_detection activate

3. Deactivate (Stops processing immediately, 0% CPU overhead):

   ros2 lifecycle set /pcl_object_detection/table_detection deactivate

Mode	Node Name	Function
1	table_detection	Detect objects on horizontal surface
2	floor_detection	Detect objects on floor (includes leg filtering)
3	shelf_detection	Detect objects in storage bins
4	placeable_detection	Find empty space for object placement
5	line_detection	Detect lines from 2D LiDAR

※ All parameters are defined in the config directory and can be tuned independently per node.

(Back to Top)

References

• Point Cloud Library

(Back to Top)