Opencv|Road Lane Line Detection

Opencv|Road Lane Line Detection

Today, I want to share a program of detection for road lane line in an image.
In a similar way, we can also detect road lane line in a video.(Seen in the Extra Expansion part)

step 1. import the image and show it in pyplot.

import matplotlib.pyplot as plt
import cv2
import numpy as np
image= cv2.imread('road.jpeg')#375*500
image = cv2.cvtColor(image,cv2.COLOR_BGR2RGB)
#the order of colors is different in pyplot and opencv.
#So we need to transform the color of the original image.
print(image.shape)
height = image.shape[0]
width=image.shape[1]
plt.imshow(image)
plt.show()

result:

(375, 500, 3)

Step2. Determine the detection area.
I created a function called region_of_interest to get the area to remove the extra part of the image, leaving only the part to be detected.
Besides, I created an array named region_of_interest_vertices to save the vertices of the area I wanted to keep.
Details can be seen in the notes of the code below.

def region_of_interest(img,vertices):
    mask =np.zeros_like(img)
    #create an array as large as the original image.
    match_mask_color = 255 
    # the color filled in the area.
    cv2.fillPoly(mask,vertices,match_mask_color)
    # fill mask with the color match_mask_color in detection region in mask.
    masked_image=cv2.bitwise_and(img,mask)
    #combine img and mask to get rid of the unneccessary part for detection in image.
    return masked_image
    
region_of_interest_vertices = [
    (100,height),
    (280,275),
    (width,height)]
   #To define the vertices of the region we need to detect.

Step 3. Perform preliminary processing on the image

gray_image=cv2.cvtColor(image,cv2.COLOR_RGB2GRAY)
#turn the image to gray which is easier to detect.
canny_image=cv2.Canny(gray_image,100,200)
#Use the function Canny to find outlines of objects in the detected area.
cropped_image = region_of_interest(canny_image,np.array([region_of_interest_vertices],np.int32))
#to get the region picture we need.(shown as the picture below)

result:

Step 4.Use HoughLinesP to detect the lines in cropped_image.

Parameters in HoughLinesP:
rho :the resolution of the parameter polar diameter r in pixels.
theta: the polar angle \ theta resolution in radians
threshold: the minimum number of curve intersections needed to detect a straight line.
lines :a container storing the parameter pairs (x_ {start}, y_ {start}, x_ {end}, y_ {end}) of the detected straight lines.( the coordinates of the two end points of the line segment)
minLineLength: the minimum number of points that can form a straight line. A straight line with insufficient number of points will be discarded.
maxLineGap:the maximum distance of bright points that can be considered on a straight line.

lines=cv2.HoughLinesP(cropped_image,rho=6,theta=np.pi/180,threshold=140,lines=np.array([]), minLineLength=60,maxLineGap=30)

Step 5. Draw the detected straight lines in the original image.
Details can be seen in the notes of the code below.

def draw_the_lines(img,lines):#to draw the lines in img
    img=np.copy(img)
    blank_image= np.zeros((img.shape[0],img.shape[1],3),dtype=np.uint8)
    #create an blank_image as large as img
    for line in lines:
        for x1,y1,x2,y2 in line:
            cv2.line(blank_image,(x1,y1),(x2,y2),(0,255,0),
            thickness=3)
            #draw lines in blank_image
    img = cv2.addWeighted(img,0.8,blank_image,1,0.0)
    #combine the img and blank_image
    return img

img_with_lines=draw_the_lines(image,lines)

Step 6. Print the image we got in the end.

plt.title('img_with_lines')
plt.imshow(img_with_lines)
plt.show()

result:

Extra Expansion

With minor changes, the program can be applied to videos to detect lane lines.

import matplotlib.pylab as plt
import cv2
import numpy as np


def region_of_interest(img, vertices):
    mask = np.zeros_like(img)
    match_mask_color = 255
    cv2.fillPoly(mask, vertices, match_mask_color)
    masked_image = cv2.bitwise_and(img, mask)
    return masked_image


def drow_the_lines(img, lines):
    img = np.copy(img)
    blank_image = np.zeros((img.shape[0], img.shape[1], 3), dtype=np.uint8)
    for line in lines:
        for x1, y1, x2, y2 in line:
            cv2.line(img, (x1, y1), (x2, y2), (0, 255, 255), thickness=3)
    img = cv2.addWeighted(img, 0.8, blank_image, 1, 0.0)
    return img


def process(image):

    print(image.shape)
    height=image.shape[0]
    width=image.shape[1]
    region_of_interest_vertices =[
        (0,height),
        (width/2,height/1.75),
        (width,height)
         ]


    gray_image= cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
    canny_image=cv2.Canny(gray_image,100,200)
    cropped_image = region_of_interest(canny_image,np.array([region_of_interest_vertices],np.int32))
    lines =cv2.HoughLinesP(cropped_image,
                           rho=6,theta=np.pi/180,
                           threshold=160,
                           lines=np.array([]),
                           minLineLength=0,
                           maxLineGap=25
                           )
    image_with_lines=drow_the_lines(image,lines)
    return image_with_lines

cap = cv2.VideoCapture('solidWhiteRight.avi')
while(cap.isOpened()):
    ret,frame =cap.read()
    frame= process(frame)
    cv2.imshow('frame',frame)
    if cv2.waitKey(1)&0xFF ==ord('q'):
        break
cap.release()
cv2.destroyAllWindows()

result:
优快云 does not support uploading video so it is displayed in screenshot form.

---

Gains:
At first, the initial setting of those parameters bothered me a lot. It is difficult to find the most suitable parameter values. After a long groping with several tries, finally I found the parameters that were sufficient to achieve the purpose. But the accuracy of recognition is still not enough, which can be further optimized.
Thank you for reading!

---

Resources:
https://www.bilibili.com/video/av77215007?
https://blog.youkuaiyun.com/e01528/article/details/82749816
test video footage from:
https://github.com/naokishibuya/car-finding-lane-lines