In this project, you will use the tools you learned about in the lesson to identify lane lines on the road. You can develop your pipeline on a series of individual images, and later apply the result to a video stream (really just a series of images). Check out the video clip "raw-lines-example.mp4" (also contained in this repository) to see what the output should look like after using the helper functions below.
Once you have a result that looks roughly like "raw-lines-example.mp4", you'll need to get creative and try to average and/or extrapolate the line segments you've detected to map out the full extent of the lane lines. You can see an example of the result you're going for in the video "P1_example.mp4". Ultimately, you would like to draw just one line for the left side of the lane, and one for the right.
Let's have a look at our first image called 'test_images/solidWhiteRight.jpg'. Run the 2 cells below (hit Shift-Enter or the "play" button above) to display the image.
Note If, at any point, you encounter frozen display windows or other confounding issues, you can always start again with a clean slate by going to the "Kernel" menu above and selecting "Restart & Clear Output".
The tools you have are color selection, region of interest selection, grayscaling, Gaussian smoothing, Canny Edge Detection and Hough Tranform line detection. You are also free to explore and try other techniques that were not presented in the lesson. Your goal is piece together a pipeline to detect the line segments in the image, then average/extrapolate them and draw them onto the image for display (as below). Once you have a working pipeline, try it out on the video stream below.
Your output should look something like this (above) after detecting line segments using the helper functions below
Your goal is to connect/average/extrapolate line segments to get output like this
In [1]:
#importing some useful packages
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
from sklearn.linear_model import ElasticNet
import numpy as np
import cv2
%matplotlib inline
%config InlineBackend.figure_format = 'retina'
In [2]:
#reading in an image
image = mpimg.imread('test_images/solidWhiteRight.jpg')
#printing out some stats and plotting
print('This image is: {0} with dimensions {1}'.format(type(image), image.shape))
plt.imshow(image) #call as plt.imshow(gray, cmap='gray') to show a grayscaled image
Out[2]:
Some OpenCV functions (beyond those introduced in the lesson) that might be useful for this project are:
cv2.inRange() for color selection
cv2.fillPoly() for regions selection
cv2.line() to draw lines on an image given endpoints
cv2.addWeighted() to coadd or overlay two images
cv2.cvtColor() to grayscale or change color
cv2.imwrite() to output images to file
cv2.bitwise_and() to apply a mask to an image
Check out the OpenCV documentation to learn about these and discover even more awesome functionality!
Helper functions and lane detection pipeline
In [3]:
import math
def grayscale(img):
"""Applies the Grayscale transform
This will return an image with only one color channel
but NOTE: to see the returned image as grayscale
you should call plt.imshow(gray, cmap='gray')"""
return cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# Or use BGR2GRAY if you read an image with cv2.imread()
# return cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
def canny(img, low_threshold, high_threshold):
"""Applies the Canny transform"""
return cv2.Canny(img, low_threshold, high_threshold)
def gaussian_blur(img, kernel_size):
"""Applies a Gaussian Noise kernel"""
return cv2.GaussianBlur(img, (kernel_size, kernel_size), 0)
def region_of_interest(img, vertices):
"""Applies an image mask.
Only keeps the region of the image defined by the polygon
formed from `vertices`. The rest of the image is set to black.
vertices shall have the shape (num_of_polygon, num_of_vertices, 2)
eg: vertices = np.array([[(wd*.45, ht*.53),(wd*.05, ht), (wd*.98, ht), (wd*.55, ht*.53)]], dtype=np.int32)
"""
#defining a blank mask to start with
mask = np.zeros_like(img)
#defining a 3 channel or 1 channel color to fill the mask with depending on the input image
if len(img.shape) > 2:
channel_count = img.shape[2] # i.e. 3 or 4 depending on your image
mask_color = (255,) * channel_count
else:
mask_color = 255
#filling pixels inside the polygon defined by "vertices" with the fill color
cv2.fillPoly(mask, vertices, mask_color)
#returning the image only where mask pixels are nonzero
masked_image = cv2.bitwise_and(img, mask)
return masked_image
def hough_lines(img, rho, theta, threshold, min_line_len, max_line_gap, line_thickness=2, line_color=[0, 255, 0],):
"""Returns an image with hough lines drawn.
`img` should be the output of a Canny transform.
"""
lines = cv2.HoughLinesP(img, rho, theta, threshold, np.array([]), minLineLength=min_line_len, maxLineGap=max_line_gap)
line_img = np.zeros((*img.shape, 3), dtype=np.uint8)
draw_lines(line_img, lines, thickness=line_thickness, color=line_color)
return line_img
# Python 3 has support for cool math symbols.
def weighted_img(img, initial_img, alpha=0.8, beta=1., gamma=0.):
"""Return weighted sum of two images
`img` is the output of the hough_lines(), An image with lines drawn on it.
Should be a blank image (all black) with lines drawn on it.
`initial_img` should be the image before any processing.
The result image is computed as follows:
initial_img * α + img * β + λ
NOTE: initial_img and img must be the same shape!
"""
return cv2.addWeighted(initial_img, alpha, img, beta, gamma)
def draw_lines(img, lines, thickness=2, color=[255, 0, 0]):
""" Draw interpolated lanes on img
"""
lane_1st, lane_2nd = [], []
height, width, _ = img.shape
# separate the line segments based on slope and their position in the image
for line in lines:
for x1,y1,x2,y2 in line:
if ((x2-x1) != 0) and ((y2-y1)/(x2-x1) < 0) and ((x1+x2)/2/width < 0.55):
lane_1st.append(line)
elif ((x2-x1) != 0) and ((y2-y1)/(x2-x1) > 0) and ((x1+x2)/2/width > 0.55):
lane_2nd.append(line)
# fit the left and right lane separately with ElasticNet
x_pred = np.arange(img.shape[1]).reshape(-1,1)
for lane in [np.array(lane_1st), np.array(lane_2nd)]:
lane = lane.reshape(lane.shape[0]*2, 2)
X, y = lane[:,0], lane[:,1]
reg = ElasticNet().fit(X.reshape(-1,1),y)
y_pred = np.hstack((x_pred, reg.predict(x_pred).reshape(-1,1)))
cv2.polylines(img,np.int32([y_pred]),False,color,thickness)
def select_color(img, colors):
''' Return img with specified color selected
colors is a list of (color_lower, color_upper) tuples
'''
img_color_select = np.zeros_like(img)
for color_lower, color_upper in colors:
color_mask = cv2.inRange(img, color_lower, color_upper)
img_color_select += cv2.bitwise_and(img, img, mask=color_mask)
return img_color_select
In [4]:
# pipline for lanes detection on a single image
def lane_detection(img,
colors,
ROI_vertex_scales=[(0.48, 0.60), (0.55, 0.60), (0.95, 1), (0.05, 1)],
GB_kernel_size = 3,
Canny_low_th = 40,
Canny_high_th = 120,
Hough_rho = 1.5,
Hough_theta = 1,
Hough_threshold = 5,
Hough_min_line_length = 7,
Hough_max_line_gap = 75,
Hough_line_thickness = 10,
Hough_line_color=[255, 0, 0],
overlay_alpha = 1,
overlay_beta = 0.85,
overlay_gamma = 0,
debug = 0):
'''Return an image with detected lanes overlaid on the original image
======================
Lane detction pipeline
original image -> color selection -> grayscale -> ROI selection -> Gaussian blur
-> Canny edge detection -> Hough line detection -> lane separation
-> lane interpolation -> overlay of detected lanes and original image
===================================
Explaination of selected parameters
colors:
a list of (color_lower, color_upper) tuples
ROI_vertex_scales:
a list of (x_scale, y_scale) tuples in terms of percentages of (width, heigh),
specifiying a polygonal region of interest
debug:
when debug is set to 1, the detected lane image would also be returned
'''
img = np.array(img).astype('uint8')
img_color_select = select_color(img, colors)
img_gray = grayscale(img_color_select)
ht,wd,_ = img.shape
vertices = np.array([[(wd*wd_scale, ht*ht_scale) for (wd_scale, ht_scale) in ROI_vertex_scales]], dtype=np.int32)
img_gray_masked = region_of_interest(img_gray, vertices)
img_gray_blur = gaussian_blur(img_gray_masked, GB_kernel_size)
img_edges = canny(img_gray_blur, Canny_low_th, Canny_high_th)
img_lines = hough_lines(img_edges, Hough_rho, Hough_theta, Hough_threshold,
Hough_min_line_length, Hough_max_line_gap,
Hough_line_thickness, Hough_line_color)
img_lines = region_of_interest(img_lines,vertices)
img_overlay = weighted_img(img, img_lines, overlay_alpha, overlay_beta, overlay_gamma)
if debug == 1:
return img_overlay, img_lines
else:
return img_overlay
In [5]:
import os
img_dir = "test_images/"
img_files = [img_dir + img_file for img_file in os.listdir(img_dir) if img_file.endswith('.jpg')]
# colors = [(white_lower_bound, white_hight_bound),
# (yellow_lower_bound, yellow_hight_bound)]
colors = [(np.array([210, 210, 210], dtype=np.uint8), np.array([255, 255, 255], dtype=np.uint8)),
(np.array([120, 140, 50], dtype=np.uint8), np.array([255, 200, 120], dtype=np.uint8))]
roi_vertex_scales=[(0.48, 0.60), (0.55, 0.60), (0.95, 1), (0.05, 1)]
plt.figure(figsize=(10,8))
for idx, img_file in enumerate(img_files):
img = mpimg.imread(img_file).astype('uint8')
img_overlay = lane_detection(img, colors=colors, ROI_vertex_scales=roi_vertex_scales)
plt.subplot(len(img_files)/2, 2, idx+1)
plt.imshow(img_overlay)
plt.axis('off')
plt.title(img_file)
plt.show()
run your solution on all test_images and make copies into the test_images directory).
In [6]:
# Import everything needed to edit/save/watch video clips
from moviepy.editor import VideoFileClip
from IPython.display import HTML
def process_image(img):
# NOTE: The output you return should be a color image (3 channel) for processing video below
# TODO: put your pipeline here,
# you should return the final output (image with lines are drawn on lanes)
# the solution is detailed in the function lane_detection above
return lane_detection(img, colors=colors, ROI_vertex_scales=roi_vertex_scales)
In [7]:
# Apply lane detection algorithm on every frame for each video clip
clips_files = ["solidWhiteRight.mp4", "solidYellowLeft.mp4","challenge.mp4", ]
for clip_file in clips_files:
clip = VideoFileClip(clip_file)
clip_out = clip.fl_image(process_image)
%time clip_out.write_videofile("z_sol_"+clip_file, audio=False)
print("======================================================")
Play the video inline, or if you prefer find the video in your filesystem (should be in the same directory) and play it in your video player of choice.
In [8]:
import io
import base64
def vid_html(vid_file, vid_width=960, vid_height=540):
'''return a HTML string for embeding mp4 video
==============
example usage:
import io
import base64
from IPython.display import HTML
HTML(vid_html("test_output.mp4"))
'''
vid = io.open(vid_file, 'r+b').read()
vid_encoded = base64.b64encode(vid)
vid_html_str = """
<video width="{0}" height="{1}" controls>
<source src="data:video/mp4;base64,{2}" type="video/mp4">
</video>""".format(vid_width, vid_height, vid_encoded.decode('ascii'))
return vid_html_str
In [9]:
HTML(vid_html("z_sol_" + clips_files[0]))
Out[9]:
In [10]:
HTML(vid_html("z_sol_" + clips_files[1]))
Out[10]:
In [11]:
HTML(vid_html("z_sol_" + clips_files[2]))
Out[11]:
Congratulations on finding the lane lines! As the final step in this project, we would like you to share your thoughts on your lane finding pipeline... specifically, how could you imagine making your algorithm better / more robust? Where will your current algorithm be likely to fail?
Please add your thoughts below, and if you're up for making your pipeline more robust, be sure to scroll down and check out the optional challenge video below!
Despite the adequate performance in relatively ideal road conditions, the current iteration of lane detection algorithm has several disadvantages.
The lane color is specified and tuned manually. In differnt lighting conditions, the lane color might fall outside the color range specified , thus causing the lane detection algorithm to fail. More road and weather conditions need to tested against to imporve the lane color selection.
The lane interplation is currently handeled by a linear regression method (ElasticNet), thus failing to capture the curvature of the lanes. Non-linear interplation methods could be used, but might incur performace penalities.
It's unclear how the algorithm might perform if there's an object (car, pedestrain) within the ROI selected for lane detection.