Advanced Lane Finding Project

The goals / 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.

Navigating this directory

• Project pipelines are in 'main.ipynb'

Project Outline:

The code for each process is named in each section of 'main.ipynb'.

I. Camera Calibration

Computing the camera matrix

• Read the calibration images of chess board
• Generate object points (points to map the chessboard corners to in the undistorted image)
• Find the image points (chessboard corners) using 'cv2.findChessboardCorners'.
• The function returns Camera matrix and distortion coefficients which are used for Correcting the distortion
• made mtx, dist, nx, ny as global variables as these variables will be used later ####Example of camera calibration image.

II. Correcting the distortion

Performing the distortion correction on camera calibration image

• Using cv2.Undistort and vairbales mtx, dist i performed the camera_undistort ####Camera undistortion on test images

III. Perspective Transform

function that takes an image, number of x and y points

• my source points are ([570, 465], [707, 465], [258, 682], [1049, 682])
• destination points are ([450,0],[w-450,0],[450,h],[w-450,h])
• Need to work on selection of these points as they can improve the algorithm
• The funtion returns unwarped_image, Matrix and Inverse Matrix

Perspective transform

IV. Thresholding images

Applying sobel threshold

• I applied sobel threshold in x direction as the lanes are in y direction
• The function returns a binary_output based on threshold limits

Sobel gradient on a test image

Magnitude Threshold

• I changed the kernel size to 9

Direction threshold

• I did not find this function useful as it is giving me the exact opposite result of what i desired

s channel

• I am using S channel of hls as it is invariant to white and yellow lines
• h channel is also used as it is seems to help the pipeline in detecting the lanes better

V. Pipeline

• I am using s_channel and h_channel of hls and sobelAbs and sobel Magnitude in finding lanes
• the pipeline is not working that good on project hard challenges
• The function returns combo, Minv as we need it to reverse the Obtained lane Detection in image plane back to the real world

VI. Sliding window

• After detecting the two lane lines using histogram
• I am using sliding_window as mentioned in the tutorials
• The function returns left_fit, right_fit, left_lane_indexes, right_lane_indexes, and Data which contains, the reactangle and histogram data

Example Image , Its Binary image and Plot of histogram of lanes

VII. polyfit using prev fit

• The function is used for finding for finding lane lines using previous frame data. If lane cannot be realised for present data then the data uses previous lane frame

VIII. Radius of curvature and center_distance

• Here i am using an extra variable to find center distance
• Jeremy mentioned that based on his previous destination points and source points he choose these particular values

Finding lanes and visualizing

Discussion

Here I'll talk about the approach I took, what techniques I used, what worked and why, where the pipeline might fail and how I might improve it if I were going to pursue this project further.

I implemented camera calibration and undistorting the images as mentioned in the class. Next for designing the pipeline. I used hls image, sobel gradient, sobel magnitude The debugging part helped me a lot in solving threshold binary images. But for some test images, it is not working that properly as it is recongizing the lightness and shadows of the image. I have to work more on the pipeline color values to make the pipeline more robust. I want to implement the convolutional algorithm as mentioned in self driving Advanced lane Detection in Q/A for lane detection. I have increased the confidence level for fits by having threshold limit set or right and left fit. I want to improve this algorithm in the future as i learn much more baout computational geometry.

Project video

Challege video

Harder challenge video