ryanchyshyn / adv_lane_finding

Advanced Lane Finding

This project goal is to write a software pipeline to identify the lane boundaries in a video. This document will describe the algorithm in details.

The Project

The key steps of this project are the following:

• Compute the camera calibration matrix and distortion coefficients given a set of chessboard images.
• Apply a distortion correction to raw images.
• Use color transforms, gradients, etc., to create a thresholded binary image.
• Apply a perspective transform to rectify binary image ("birds-eye view").
• Detect lane pixels and fit to find the lane boundary.
• Determine the curvature of the lane and vehicle position with respect to center.
• Warp the detected lane boundaries back onto the original image.
• Output visual display of the lane boundaries and numerical estimation of lane curvature and vehicle position.

Camera Calibration

I performed camera calibration using the single image and got camera calibration matrix:

Camera calibration matrix: [[ 872.5223319     0.          649.38554509]
 [   0.          853.10470045  213.74930061]
 [   0.            0.            1.        ]]

Original images: Undistorted images:

The difference is not so big but it can be noticable further if not performed right now.

Perspective

To get perspective matrix I performed the following calculations:

    srcPoints = np.float32(
       [[w * 0.18, h * 0.95],
        [w * 0.457, h * 0.63],
        [w * 0.543, h * 0.63],
        [w * 0.82, h * 0.95]])
   dstPoints = np.float32(
       [[w * 0.25, h],
        [w * 0.25, 0],
        [w * 0.75, 0],
        [w * 0.75, h]])
   
   M = cv2.getPerspectiveTransform(srcPoints, dstPoints)
   Minv = cv2.getPerspectiveTransform(dstPoints, srcPoints)

and checked the result using straight image: vertical lines on the first image shows that perspective matrix is correct.

Pipeline

Next steps are performed for every frame in the video.

Warping

Warping is performed using M matrix. The result is displayed on previous image.

Binarization

After many experiments I found the following binarization algorithm is the best:

1. Convert image to HLS color space
2. Get L and S channels:
3. Adjust gamma on L channel:
4. Sum gamma(L) with S images using Max function (i.e. get the brightest pixel from two):
5. Do binary thresholding using values min=110, max=255:

Lines detection

Perform lines detection using provided in the lesson algorithm:

Draw lane highlighting

Draw the polygon on warped clear image

Unwarp the image

Unwarp the image using inverted M matrix

Combine original and highliting image

Draw text on the image

The final image:

The final video