ADAS-Undergraduate-Research

Lane detection, steering control, obstacle recognition, and traffic sign detection.

Project Description & Objective

This project showcases a lane detection system for vehicles with front-facing cameras, a crucial part of advanced driver assistance systems (ADAS) in autonomous and semi-autonomous vehicles. It detects lanes, measures curve radius, and monitors center offset, enhancing safety and comfort. The repository provides dashcam footage for testing and development.

Demo Video:

Get This Project Up and Running

Setting up a Python Virtual Environment for "ADAS-Undergraduate-Research"

To maintain a clean and isolated Python environment for your "ADAS-Undergraduate-Research" project, follow these steps in your terminal:

Navigate to the Project Directory

cd ADAS-Undergraduate-Research/

Create a Virtual Environment

python -m venv venv

Activate the Virtual Environment

On Windows:

.\<env_name>\Scripts\activate

On macOS and Linux:

source venv/bin/activate

Install Project Dependencies from requirements.txt

pip install -r requirements.txt

This project follows a simplistic approach in terms of how to run it. Only 2 files are needed in order to get it running.

• laneDetection is the only file to perform image processing nad detecting lanes
• drive is the video file on which image processing is performed.

Simply clone this repository to desired path by launching command prompt and running

https://github.com/Jamshid-Ganiev/ADAS-Undergraduate-Research

Make sure both files (laneDetection.py & drive.mp4) are in the same directory.

How the Code Works

The Python script laneDetection.py processes a provided dashcam footage of a car cruising on the highway. The code follows a modular approach with several functions for lane detection.

Image Processing

• readVideo(): Reads the video file drive.mp4 from the same directory.

• processImage(): Applies HLS color filtering, grayscale conversion, thresholding, blurring, and edge extraction to isolate white lane lines.

• perspectiveWarp(): Performs a perspective warp by defining 4 points around the lane area, creating a bird's-eye view for accurate lane curvature detection. Note that the warp parameters may need adjustment for different camera angles or videos.

Lane Detection, Curve Fitting & Calculations

plotHistogram()

Generates a histogram for the bottom half of the image to identify the starting positions of the left and right lanes. The peaks in the histogram indicate these positions.

slide_window_search()

Uses a sliding window approach to detect lanes and their curvature. Starting from the histogram results, it positions boxes on the lanes, moving upward to the top of the frame. It fits a second-degree polynomial to obtain a curve in pixel space.

general_search()

Building upon the results from slide_window_search(), this function fills an area around the detected lanes and performs a second-degree polynomial fit. A yellow line is drawn accurately representing the lanes. This line aids in measuring lane curvature, which is crucial for steering angle prediction.

measure_lane_curvature()

Uses np.polyfit() with pixel-to-meter conversion factors (xm_per_pix and ym_per_pix) to convert lane data from pixel space to meter space for curvature calculations.

Visualization and Main Function

draw_lane_lines()

Visualizes the detected lanes by filling them with green color and representing the lane center with a yellowish color. This function also calculates the vehicle's lateral offset.

offCenter()

Calculates the vehicle's offset using the pts_mean variable and displays it in meter space.

addText()

Adds text to the final image, providing relevant information.

main()

The main function orchestrates the sequence of function calls, including the video processing loop.