Object Detection and Tracking in ROS 2 with TurtleBot4

This project demonstrates an object detection and tracking pipeline in ROS 2 using YOLOv8 and OpenCV. The system processes a live camera feed from a simulated TurtleBot4 running in Ignition, detects objects (e.g., persons, bottles, etc.), tracks them over time using a simple centroid tracker, and publishes detection and tracking information as ROS 2 topics.

Features

• Object Detection: Uses a pre-trained YOLOv8 (nano) model to detect objects in a live camera feed.
• Object Tracking: Implements a simple centroid-based tracker to maintain consistent IDs for detected objects.
• ROS 2 Integration: Publishes detected object data (class labels and bounding boxes) and tracking information as ROS 2 topics.
• Simulation Environment: Demonstrates simulation with TurtleBot4 in Gazebo Classic (or Ignition) with camera feed.
• Teleoperation Support (Optional): Allows for manual control of TurtleBot4 via keyboard or joystick teleop nodes.

Prerequisites

• Operating System: Ubuntu 22.04 or later
• ROS 2: Humble Hawksbill (or a compatible ROS 2 distribution)
• Gazebo Classic: (e.g., Gazebo 11) – if using Gazebo Classic simulation
• TurtleBot4 Packages:
  • ros-humble-turtlebot4-description
  • (Optionally) ros-humble-turtlebot4-simulator if using provided simulation packages
• Python Dependencies:
  • opencv-python
  • numpy
  • ultralytics
    Install them using pip:

  pip install opencv-python numpy ultralytics
  
  

Running the Simulation

ros2 launch turtlebot4_ignition_bringup turtlebot4_ignition.launch.py
ros2 run object_detection detect


Use rqt_image_view to see the annotated image:

ros2 run rqt_image_view rqt_image_view