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.
The project is done using python 3.6.2.
OpenCV version 3.1.0 is used.
Python version less than 3.6.0 will not work properly because in my code I used f-string which is a new feature.
1. The images for camera calibration are stored in the folder called camera_cal.
2. The images in test_images are for testing pipeline on single frames.
3. The outputs after finding the chase board corner are in camera_cal_output.
4. The binary outputs after color and gradient transformation of test_images
are in test_images_transformed folder.
5. The binary outputs after bird's eye view perspective transform of test_images
are in test_images_wraped folder.
6. The test_output folder contains all the test images after applying final pipeline function.
7. Resources folder contains some images for purpose of writing README.md file.
8. The output video after applying the pipeline to project_video.mp4 is output.mp4.
Note: All codes are self explanatory. Comments and function documentation has been given where needed.
The code for this step is contained in the fourth code cell of the IPython notebook advanced_lane_finding.ipynb.
I start by preparing "object points", which will be the (x, y, z) coordinates of the chessboard corners in the world. Here I am assuming the chessboard is fixed on the (x, y) plane at z=0, such that the object points are the same for each calibration image. Thus, objp is just a replicated array of coordinates, and objpoints will be appended with a copy of it every time cv2.findChessboardCorners() detects all chessboard corners in a test image. imgpoints will be appended with the (x, y) pixel position of each of the corners in the image plane with each successful chessboard detection.
Then draw the chess board corners by cv2.drawChessboardCorners(). Here is an example of chessboard corner drawn .
Chess Board Corners
undistort_image() function is used to undistort a image, the previous output objpoints and imgpoints have been used
to compute the camera calibration and distortion coefficients using the cv2.calibrateCamera() function. I applied this distortion correction to the input image using the cv2.undistort() function .
Undistorted Chess board image
To demonstrate this step, I will describe how I apply the distortion correction to one of the test images like this one:
Undistorted test image
I used a combination of color and gradient thresholds to generate a binary image .
1. compute the sobel magnitude then take a thresolding range b/w 20 and 255.
2. select_yellow() and select_white() aretwo functions to extract the yellow
and white lane lines the combine it by bitwise or.
3. Then combine it with sobel magnitude calculated.
Transformed image after applying gradient and color thresolding
I used a wrap() function to get a birds eye view using perspective transform. It takes an image , source points and destination points and returns the wraped image transformation matrix and it's inverse .
| Source | Destination |
|---|---|
| 710, 460 | 990,0 |
| 1090, 720 | 990, 720 |
| 200, 720 | 300, 720 |
| 580, 460 | 300, 0 |
I verified that my perspective transform was working as expected by drawing the src and dst points onto a test image and its warped counterpart to verify that the lines appear parallel in the warped image.
1. Divide the thresolded binary transformed image into n=9 horizontal strips of equal height.
2. Compute the histogram of each strip using np.sum().
3. Identify two peaks where histogram computed are maximums.
Birds eye view of a test transformed image and histogram drawn
4. Get the pixels in that horizontal strip that have x coordinates close to the two
peaks of x coordinates.
Rectangles are drawn where lane line pixels are detected
5. Fit a second order polynomial to each lane line using np.polyfit() function.
Polynomial fitted to birds-eye-view image
left_curverad = ((1 + (2*left_fit_cr[0]*y_eval*ym_per_pix + left_fit_cr[1])**2)**1.5)\
/ np.absolute(2*left_fit_cr[0])
right_curverad = ((1 + (2*right_fit_cr[0]*y_eval*ym_per_pix + right_fit_cr[1])**2)**1.5)\
/ np.absolute(2*right_fit_cr[0])
curvature = (left_curverad + right_curverad)/2
The final pipeline_func() contains all the necessary steps for finding and plotting the lane lines. Here is an example of my result on a test image:
I applied my image_pipeline() function to the project_video. It detect's lane lines reasonably well but under a
curved shadowed path it once showing some deviation.
Here's the link of my Output Video
This pipeline will not actually work rather than this project video. The source and destination
points have been chosen manually for birds eye view perspective transformation. Which indeed
is the most important step for finding lane line pixels.
The color and gradient thresolding and identifying lane lines is a computing intensive work
and a maximum speed of 1.5 iteration/s achieved in these case will not work in realtime.
But this pipeline detects lane lines moderately well on different colored , shaded highway path.