Calculating the center of mass of a binary image (openCV Python)

Question

I study mechanical engineering and this was one of the first programming tasks that I had to do so my coding experience especially in python and open cv was almost zero when I started this.

We were given a picture and should write a program to calculate the center of mass of the shape.

Here is the code that I came up with,

import cv2
import numpy as np
from matplotlib import pyplot as plt

img = cv2.imread('C:/Amoebe/Amoebebesser.png',0)
img = cv2.medianBlur(img,5)
ret,th1 = cv2.threshold(img,100,255,cv2.THRESH_BINARY)
plt.imshow(th1, cmap = 'gray', interpolation = 'none')
plt.show()

momentx = 0
momenty = 0
count = 0

for i in range(th1.shape[0]):
    for j in range(th1.shape[1]):
        if th1[i, j] >= 255:
            momentx = momentx + j
            momenty = momenty + i
            count = count + 1

centx = int(momentx / count)
centy = int(momenty / count)
print(centx)
print(centy)
cv2.circle(th1, (centy, centx), 1, (0, 0, 255), 2)
cv2.imshow('image', th1)
k = cv2.waitKey(0)
if k == 27:         # wait for ESC key to exit
    cv2.destroyAllWindows()
elif k == ord('s'): # wait for 's' key to save and exit
    cv2.imwrite('centerofmasstest.png', th1)
    cv2.destroyAllWindows()

From what I have read there is a lot of optimizing that could be done using numpy but I couldn't figure out how to do it. Also, I am not sure if I am actually getting the center of mass or if there are errors in my method.

Thank you very much in advance for your time and best regards.

Answer 1

We can use np.where and np.average to do this easily.

# Find indices where we have mass
mass_x, mass_y = np.where(th1 >= 255)
# mass_x and mass_y are the list of x indices and y indices of mass pixels

cent_x = np.average(mass_x)
cent_y = np.average(mass_y)

EDIT: Oneliner
center = [ np.average(indices) for indices in np.where(th1 >= 255) ]

Answer 2

your code looks correct. I might leave the int(...) off to get a subpixel resolution centroid value. the coordinate (0.0, 0.0) would be the center of the top left pixel.

apart from the numpy solution, there is cv2.moments , which can be run on contours and mask images directly, and it gives you various moment values (m00, m10, m01, ...) you can use to calculate the centroid (m10/m00, m01/m00)

m00 is count, m10 is momentx, m01 is momenty.