Real-time Lane Detection

We will train and construct a model that helps detect lanes on straight and curved roads based on the raw pixels captured by the single front-facing camera.

Requirements

To install all the dependencies in this project run

  pip install -r requirements.txt

Deploymnt

To run this project, you will need to add the following command

git clone https://github.com/rahularepaka/lane-detection.git

cd lane-detection

python main.python

Libraries

• OpenCV
• Numpy
• Tensorflow
• Keras

Dataset

This project demonstrates lane detection using a single image from a road dataset. The lanes are marked by a solid white line (on the right) and alternating short line segments with dots (on the left).

Link to Dataset : Lane Line Detection : Kaggle

Computer Vision Techniques

Canny Edge Detection

  def canny(img):
    if img is None:
        cap.release()
        cv.destroyAllWindows()
        exit()
    gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
    kernel = 5
    blur = cv.GaussianBlur(gray, (kernel, kernel), 0)
    canny = cv.Canny(gray, 50, 150)
    return canny

def edge_detection(image):
    edges = cv.Canny(image, 80, 200)
    cv.namedWindow("edges", cv.WINDOW_NORMAL)
    cv.resizeWindow("edges", 500, 500)
    cv.imshow('edges', edges)
    return edges

Masking using HSV Channel

  def red_white_masking(image):
    hsv = cv.cvtColor(image, cv.COLOR_BGR2HSV)

    lower_y = np.array([10, 130, 120], np.uint8)
    upper_y = np.array([40, 255, 255], np.uint8)
    mask_y = cv.inRange(hsv, lower_y, upper_y)
    cv.namedWindow("mask_y", cv.WINDOW_NORMAL)
    cv.resizeWindow("mask_y", 500, 500)
    cv.imshow('mask_y', mask_y)
    
    lower_w = np.array([0, 0, 212], np.uint8)
    upper_w = np.array([170, 200, 255], np.uint8)
    mask_w = cv.inRange(hsv, lower_w, upper_w)
    cv.namedWindow("mask_w", cv.WINDOW_NORMAL)
    cv.resizeWindow("mask_w", 500, 500)
    cv.imshow('mask_w', mask_w)
    
    mask = cv.bitwise_or(mask_w, mask_y)
    cv.namedWindow("mask", cv.WINDOW_NORMAL)
    cv.resizeWindow("mask", 500, 500)
    cv.imshow('mask', mask)
    
    masked_bgr = cv.bitwise_and(image, image, mask=mask)
    cv.namedWindow("masked_bgr", cv.WINDOW_NORMAL)
    cv.resizeWindow("masked_bgr", 500, 500)
    cv.imshow('masked_bgr', masked_bgr)
    
    return masked_bgr

Image Filtering

def filtered(image):
    kernel = np.array([[-1, 0, 1], [-1, 0, 1], [-1, 0, 1]])
    filtered_image = cv.filter2D(image, -1, kernel)
    cv.namedWindow("filtered_image", cv.WINDOW_NORMAL)
    cv.resizeWindow("filtered_image", 500, 500)
    cv.imshow('filtered_image', filtered_image)
    return filtered_image

Average slope intercept

  def average_slope_intercept(image, lines):
    left_fit = []    # list for all multiple lines found in left lane
    right_fit = []   # list for all multiple lines found in right lane
    global l
    global r
    for line in lines:
        x1, y1, x2, y2 = line.reshape(4)

        parameters = np.polyfit((x1, x2), (y1, y2), 1)
        slope = parameters[0]
        intercept = parameters[1]
        if slope < 0:
            left_fit.append((slope, intercept))
        else:
            right_fit.append((slope, intercept))
    if left_fit != []:
        left_fit_average = np.average(left_fit, axis=0)
        left_line = create_coordinates(image, left_fit_average)
        l = left_fit_average
    else:
        left_line = create_coordinates(image, l)
    if right_fit != []:
        right_fit_average = np.average(right_fit, axis=0)
        right_line = create_coordinates(image, right_fit_average)
        r = right_fit_average
    else:
        right_line = create_coordinates(image, r)
    return np.array([left_line, right_line])

Current Progress on newer datasets

Testing with Simulators

• CARLA Simulator

• lGSVL Simulator

• Udacity Car Simulator by Unity

• 

Test Videos (Kinda Failed)

Testing Video on Udacity Simulator

test1_out.mp4

Testing on Indian Road

dashcam_video_trim_output.mp4

Ways to contribute

1. Fork the repo on GitHub
2. Clone the project to your own machine
3. Commit changes to your own branch
4. Push your work back up to your fork
5. Submit a Pull request so that we can review your changes

Please abide by Contributing Guidelines and Code of Conduct.

Contributors

• Rahul Arepaka
• Avula Vijaya Koushik
• Cherukuru Swaapnika Chowdary
• Lohit Garje
• Naga Tharun Makkena

References

• Canny Edge Detection
• Building a lane detection system
• Advanced Lane Detection
• Road Lane Detection using OpenCV (Hough Lines Transform Explained)

Feedback

If you have any feedback, please reach out to us at rahul.arepaka@gmail.com

License

MIT