Auto Canny T265 - ROS2 Package

📌 Overview

This ROS2 package integrates the Intel RealSense T265 Tracking Camera with OpenCV for real-time motion tracking and fisheye image processing. The package provides:

• C++ Node (t265_node): Publishes odometry, IMU, and fisheye images.
• Python Node (auto_canny_node): Applies Auto-Canny edge detection on fisheye images.
• Launch File: Starts both nodes in ROS2.

📸 Features

✔ Real-time Odometry & IMU Data
✔ Fisheye Image Publishing
✔ TF Broadcast for Localization
✔ Adaptive Canny Edge Detection on Fisheye Images
✔ Seamless ROS2 Integration (Humble)

---

🚀 Installation & Setup

Clone the Repository

Ensure you have the Librealsense2 library compatible with your system

cd ~/ros2_ws/src
git clone git@github.com:RubenCasal/auto_canny_t265.git
cd ~/ros2_ws
colcon build --packages-select auto_canny_t265
source install/setup.bash

🏁 Running the Package

Run Nodes Separately

Start the T265 Node (C++)

ros2 run auto_canny_t265 t265_node

Start the Auto-Canny Node (Python)

ros2 run auto_canny_t265 auto_canny_node

Run with Launch File

To start both nodes together:

ros2 launch auto_canny_t265 canny_detection_launch.py

📡 Published Topics

This package publishes multiple topics for navigation and image processing.

Topic Name	Message Type	Description
/rs_t265/odom	nav_msgs/msg/Odometry	Odometry data (position & pose).
/rs_t265/imu	sensor_msgs/msg/Imu	IMU data (gyro & acceleration).
/rs_t265/fisheye_left	sensor_msgs/msg/Image	Left fisheye image from the T265.
/rs_t265/fisheye_right	sensor_msgs/msg/Image	Right fisheye image from the T265.
/rs_t265/canny_edge_detection	sensor_msgs/msg/Image	Processed edge-detected fisheye image.
/tf	tf2_msgs/msg/TFMessage	TF transformations for localization.

🎥 Visualizing Results in RViz2

1️⃣ Launch RViz2

Start RViz2:

rviz2

2️⃣ Add Required Displays

Once RViz2 is open:

  Click "Add" → "By Topic".

  Select:
      /rs_t265/odom → Odometry (for trajectory visualization).
      /rs_t265/imu → IMU (for orientation data).
      /rs_t265/fisheye_left → Image (for raw camera feed).
      /rs_t265/canny_edge_detection → Image (for edge-detected images).
      /tf → TF (for viewing the transform frames).

  Adjust settings as needed and view real-time sensor data.

Python Node: auto_canny_node.py

The Auto-Canny Node enhances edge detection by dynamically adjusting thresholds based on image characteristics. It processes fisheye images from the Intel RealSense T265, making edge detection more adaptive and robust than the traditional Canny method.

🛠 How Auto-Canny Works

1️⃣ Bilateral Filtering: Noise Reduction Without Losing Edges

Unlike Gaussian Blur, Bilateral Filtering smooths the image while preserving edges, reducing noise without blurring important details.

2️⃣ Adaptive Canny Edge Detection: Smart Threshold Selection

Traditional Canny requires fixed thresholds, making it unreliable under different lighting. Auto-Canny solves this by:

• Computing the median intensity of the image.
• Defining adaptive lower and upper thresholds based on the median.
• Applying Canny edge detection dynamically.

🎯 Why Auto-Canny?

✔ No manual tuning – Adapts automatically to different images.
✔ Better edge detection – Works well in low-contrast conditions.
✔ More reliable in real-world applications – Handles lighting variations effectively.

This makes Auto-Canny ideal for robotics, SLAM, and vision-based tasks, where edge clarity matters despite changing environments. 🚀