ROS + OpenCV Lane Detection

Real-time lane detection from a USB camera stream and steering signal generation — Teknofest Robotaksi prototype

Project Info

✓Category: Computer Vision / Autonomous Driving
✓Context: Real-time prototype for Teknofest Robotaksi
✓Input: USB camera or ROS camera driver
✓Output: Lane masking & steering angle (PID-ready)
✓Platform: ROS (Ubuntu), OpenCV

ROS OpenCV Python C++ Canny Hough PID

GitHub Repository Back to Projects

Architecture & Algorithm Pipeline

The system targets real-time performance with a classic OpenCV-based lane detection approach. A typical flow consists of the following steps:

ROI (Region of Interest): Masking the road's region of interest (lower triangle/polygon).
Preprocessing: Grayscale, Gaussian blur, dynamic thresholding.
Edge extraction (Canny): Extracting candidate line edges.
HoughLinesP: Finding right/left lane candidates as line segments.
Symmetry & merging: Filtering the lines and merging by slope/continuity.
Heading angle estimation: Computing deviation relative to the lane center and generating steering_angle.
Optional corrections: Temporal-window smoothing, jitter reduction.

Note: Topic names, node names, and parameter files are defined within the repository; the values given here are a general template. They can be customized via the launch and/or param files in the repository.

Example ROS Topics

/camera/image_raw – Input image
/lane/overlay – Detected lane overlay (optional)
/lane/steering_angle – Steering angle (float)
/lane/pid – PID control input (optional)

Example Parameters

Parameter	Description	Example
`canny_low`	Lower edge threshold	50
`canny_high`	Upper edge threshold	150
`hough_threshold`	Hough vote threshold	30
`minLineLength`	Minimum line length	25
`maxLineGap`	Line merging gap	20
`roi_vertices`	Region-of-interest polygon	lower triangle
`smoothing_window`	Angle smoothing window	5

Refer to the repository contents for the exact parameter names.

Setup (Example)

# ROS & dependencies (example: Noetic / Ubuntu 20.04)
sudo apt update
sudo apt install ros-noetic-desktop-full python3-opencv

# Workspace
mkdir -p ~/catkin_ws/src
cd ~/catkin_ws/src

# Add the repository
git clone https://github.com/mr-ozdemir/Lane-detection-with-Opencv-in-Ros.git

# Build (for C++ nodes, if any)
cd ..
catkin_make
source devel/setup.bash

Running (Example)

# Start your camera driver (example)
roslaunch usb_cam usb_cam-test.launch

# Lane detection - if a launch file exists:
roslaunch lane_detection lane_detection.launch

# or run a single node (depending on the repository structure):
rosrun lane_detection lane_node.py
# or
rosrun lane_detection lane_node

Commands may vary depending on the repository structure; adapt them to the README/launch files in the repository.

Notes & Potential Improvements

The ROI and edge/line thresholds should be tuned to the field for different lighting/surface conditions.
FPS/latency targets depend on resolution and hardware; more stable real-time performance is achieved at 640×480.
In the future, the classic method could be combined with segmentation-based approaches (e.g., U-Net, ENet) in a hybrid scheme.
A ROS2 port and hardware acceleration (GPU/OpenCL) can be planned.

Go to Projects