ros2_opencv_tracker

Real-time colour-based object tracking and autonomous pillar counting for ROS2 Jazzy using OpenCV.

Nodes

1. Colour Tracker (colour_tracker)

General-purpose colour tracking with depth estimation. Detects the largest object matching a configurable HSV range, computes its bounding box and centroid, and publishes the 3D position.

2. Pillar Counter (pillar_counter)

Autonomous coloured pillar counter. Detects green and blue pillars via HSV filtering, estimates their world position using depth + robot odometry, and counts unique pillars (avoids double-counting by tracking world positions).

Combined with ros2_slam_explorer, the robot autonomously explores an unknown environment while counting coloured pillars.

Architecture

┌──────────────┐                    ┌──────────────────┐
│ SLAM Explorer│──Nav2 goals───────▶│     Nav2         │──▶ Robot moves
└──────────────┘                    └──────────────────┘
                                           │
┌──────────────┐  /colour_image     ┌──────▼──────────┐
│   Camera     │───────────────────▶│ Pillar Counter  │──▶ /pillar_counter/summary
│   + Depth    │───────────────────▶│                 │──▶ /pillar_counter/debug_image
└──────────────┘  /depth_image      │ - HSV filtering │
                                    │ - Depth lookup  │
┌──────────────┐  /odom             │ - World position│
│   Odometry   │───────────────────▶│ - Unique count  │
└──────────────┘                    └─────────────────┘

Features

• Multi-colour detection — Tracks green and blue pillars independently
• World position estimation — Depth + odometry to compute pillar locations
• Unique counting — Only counts a pillar once using distance-based deduplication
• Debug visualization — Annotated camera feed with bounding boxes, depth, and live count
• SLAM integration — Combined launch with autonomous frontier exploration
• Configurable — HSV ranges, minimum area, unique distance via ROS2 parameters

Dependencies

• ROS2 Jazzy
• OpenCV (python3-opencv)
• cv_bridge (ros-jazzy-cv-bridge)
• ros2_slam_explorer (for combined SLAM + counting)
• TurtleBot3 Gazebo (ros-jazzy-turtlebot3-gazebo)

Installation

cd ~/ros2_ws/src
git clone https://github.com/manavsikkas/ros2_opencv_tracker.git
git clone https://github.com/manavsikkas/ros2_slam_explorer.git  # for combined mode
cd ..
colcon build
source install/setup.bash

Usage

Colour tracker only:

export TURTLEBOT3_MODEL=waffle
ros2 launch ros2_opencv_tracker tracker.launch.py

SLAM exploration + pillar counting:

export TURTLEBOT3_MODEL=waffle
ros2 launch ros2_opencv_tracker slam_counter.launch.py

View the live count and detections:

ros2 topic echo /pillar_counter/summary
ros2 run rqt_image_view rqt_image_view /pillar_counter/debug_image

Parameters (Pillar Counter)

Parameter	Default	Description
min_area	300	Minimum contour area (pixels) to consider a detection
unique_distance	1.0	Minimum distance (meters) between unique pillars

Built-in colour ranges:

Colour	Hue	Saturation	Value
Green	35-85	100-255	50-255
Blue	100-130	150-255	50-255

Topics

Topic	Type	Description
/pillar_counter/summary	String	Live count: "Pillars — green: 3, blue: 2, total: 5"
/pillar_counter/debug_image	Image	Annotated camera feed with detections
/tracker/position	PointStamped	Tracked object 3D position (colour_tracker)
/tracker/debug_image	Image	Annotated image (colour_tracker)

License

MIT