CODEWITHSUNDEEP
pythonimage-processinghistogram
SA1
SA1

Reputation: 11

How to find total quantity of green crosses in images

Working with an image processing project to find all green (128) crosses from multiple images.

1

Should achieve the following:

  1. total amount of image files
  2. Count the ones with less/most amount of crosses in image file including names of the files with it.
  3. The mean of the crosses and total amount of the crosses.
  4. Histogram over the distribution of the number of plus.

The crosses should at least look like this, meaning two crosses can be next to each other. Should count all crosses (inculding copies). The white boxes outside the cross can have various colors

2

In the left picture below are 2 crosses and in the right you will find four (there are other variants than the two below):

3

Quite new to all this so thanks in advance, this is what I have to go on:

from PIL import Image

import numpy as np

im = Image.open("testbild.bmp")

M = np.asarray(im)


def antalkors(M):

    kors = 0
    for i in range(2,M.shape[0]-2):
        for j in range(2,M.shape[1]-2):
            if M[i][j] == 128 and M[i][j-1] == 128 and M[i][j-2]==128 and M[i][j+1] == 128 and M[i][j+2] == 128 and M[i-1][j] == 128 and M[i-2][j] == 128 and M[i+1][j] == 128 and M[i+2][j] == 128:
                  kors += 1
        return kors

Upvotes: 1

Views: 106

Answers (1)

Mark Setchell
Mark Setchell

Reputation: 207698

As a first step, I would load the image and make a mask of where the green pixels are so that I just had one channel rather than three to deal with, and to speed up processing:

import cv2
import numpy as np

# Load image as BGR
im = cv2.imread('iejDj.png', cv2.IMREAD_COLOR)

# Now make a single channel image that is 255 where green and black elsewhere:
green = np.all(im==[0,255,0], axis=2) * 255

That looks like this:

enter image description here

Next, make a template along these lines:

template = np.zeros((5,5), np.uint8)
template[2,:] = 255
template[:,2] = 255

That looks like this:

enter image description here

Then follow Christoph's advice and use OpenCV Template Matching.

Upvotes: 1

Related Questions