在 python (OpenCV) 中查找 object 的大小？

Question

How to get height and width of bright part using python (opencv)? Having 2000 of these picture and end goal is to make table with length and width values

Answer 1

It is primitive method. Convert to grayscale and check which points have value bigger then some "bright" color ie. 21 and it gives array True/False - using .any(axis=0) you can reduce every row to single value, using .any(axis=1) you can reduce every column to single value and then using sum() you can count how many True was in any row or column (because True/False is converted to 1/0 )

import cv2

img = cv2.imread('image.png')
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
#print(img)

print('height, width, color:', img.shape)

#cv2.imshow('image', img)
#cv2.waitKey(0)
#cv2.destroyAllWindows()

print('width:', sum((img > 21).any(axis=0)))
print('height:', sum((img > 21).any(axis=1)))

For your image it gives me

width: 19
height: 27

For my image (below) it gives me

width: 23
height: 128

---

EDIT: Version with small change.

I set mask = (img > 21) to

• calculate size

• create Black&White image which better shows which points are used to calculate size.

BTW: code ~mask inverts mask (convert True to False and False to True ). It can be used also to invert image - ~img - to create negative for RGB, Grayscale or B&W.

Code:

import cv2

for filename in ['image-1.png', 'image-2.png']:

    img = cv2.imread(filename)
    img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

    print('height, width, color:', img.shape)

    mask = (img > 21)

    # display size 
    print(' width:', sum( mask.any(axis=0) ))
    print('height:', sum( mask.any(axis=1) ))

    # create Black&White version 
    img[  mask ] = 255 # set white
    img[ ~mask ] = 0   # set black

    # display Black&White version
    cv2.imshow('image', img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

    # write Black&White version
    cv2.imwrite('BW-' + filename, img)

--->

--->

---

EDIT: The same result using cv2.boundingRect() instead of sum(mask.any()) - but it still needs img[ mask ] = 255 to create Black&White image.

import cv2

for filename in ['image-1.png', 'image-2.png']:
    print('filename:', filename)

    img = cv2.imread(filename)
    #print('height, width, color:', img.shape)

    img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    #print('height, width, color:', img.shape)

    mask = (img > 21)

    # create Black&White version 
    img[  mask ] = 255 # set white
    img[ ~mask ] = 0   # set black

    x, y, width, height = cv2.boundingRect(img)

    # display size 
    print(' width:', width)
    print('height:', height)

    # display Black&White version
    #cv2.imshow('image', img)
    #cv2.waitKey(0)
    #cv2.destroyAllWindows()

    # write Black&White version
    #cv2.imwrite('BW-' + filename, img)

https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_contours/py_contour_features/py_contour_features.html#bounding-rectangle

---

EDIT: The same result using cv2.boundingRect() and cv2.threshold() - so it doesn't need mask

import cv2

for filename in ['image-1.png', 'image-2.png']:
    print('filename:', filename)

    img = cv2.imread(filename)
    #print('height, width, color:', img.shape)

    img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    #print('height, width, color:', img.shape)

    ret, img = cv2.threshold(img, 21, 255, cv2.THRESH_BINARY)  # the same 21 as in `mask = (img > 21)`
    x, y, width, height = cv2.boundingRect(img)

    # display size 
    print(' width:', width)
    print('height:', height)

    # display Black&White version
    cv2.imshow('image', img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

    # write Black&White version
    #cv2.imwrite('BW-' + filename, img)

https://docs.opencv.org/3.4/d7/d4d/tutorial_py_thresholding.html

https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_thresholding/py_thresholding.html

https://www.learnopencv.com/opencv-threshold-python-cpp/

https://www.geeksforgeeks.org/python-thresholding-techniques-using-opencv-set-1-simple-thresholding/