Detecting soccer field

I think one of the best course I have ever had was the computer vision. Not just the fascinating content of the lesson, but the challenging weekly exercises made me spend much time every week to practice and study more and more. Today I want to share one of my first computer vision homework, which is a necessary and straightforward exercise for the first weeks, so there is no need for vast knowledge and skill in computer vision and can be done with simple techniques. In this exercise, you will learn one of the required techniques for your future serious projects.

most of computer vision projects need a preprocess step, that has many different types. Our practice today is just the same. It is detecting specific areas. It may not be used alone, but it is implicit in many of the most crucial tasks of computer vision. The characteristics of each area help us as identifiers, such as color, shape, and design ( borders ) .for example, in the rat tracker project, we filter frames by color to identify different areas.

In this exercise, we want to detect soccer field grass using simple techniques. in first step we import opencv and numpy and then read the image and split its channels:

import cv2
import numpy as np
#read image
image = cv2.imread('stadium1/Stadium1.jpg')
#split channels
b, g, r = cv2.split(image)

To solve this problem, we first need to convert the image into a black and white image, working in the following way to get the green areas, white content, and other areas as black.

1. For each RGB pixel, we compute the image ratio g to sum r and g and b(g/(r+b+g)).
2. Then, by test and error, we determine the threshold rate of 0.4, and lower values from the threshold get black and higher values white.

for i in range(len(b)):
    for j in range(len(b[i])):
        #find greenness
        if g[i][j] > 0:
            a = float(g[i][j])/(float(r[i][j]) + float(b[i][j]) + float(g[i][j]))
        else:
            a = 0
        if a > 0.40:
            final[i][j] = 255
        else:
            final[i][j] = 0

Then, to remove all scattered white spots outside the grass, we apply the open operator once :

#open op
opening = cv2.morphologyEx(final, cv2.MORPH_OPEN, kernel)
cv2.imwrite("stadium1/opening_Stadium1.jpg", opening)

Now, we use the closed operator to uniform the white areas :

#clos op
closing = cv2.morphologyEx(opening, cv2.MORPH_CLOSE, kernel, iterations=5)
cv2.imwrite("stadium1/closing_Stadium1.jpg", closing)

Finally, we sum the original image with the result :

img = cv2.merge((closing, closing, closing))
final_image = image
# merge with original image
for i in range(len(img)):
    for j in range(len(img[i])):
        if img[i][j][0] == 0 and img[i][j][1] == 0 and img[i][j][2] == 0 :
            final_image[i][j] = image[i][j]
        else:
            final_image[i][j] = img[i][j]

cv2.imwrite("stadium1/Stadium1_result.jpg", final_image)
cv2.waitKey(0)