如何从手绘电路的扫描图像中检测逻辑门？

Question

我们从逻辑电路的正确绘制图表的扫描图像开始，我们能够将逻辑门与电路的扫描图像分开，但是我们无法检测到以及如何进一步处理，我们使用了python open cv for this ，我们上面的代码是

import cv2
import numpy as np
from matplotlib import pyplot as plt
img = cv2.imread('logic.png',0)

ret,img2 = cv2.threshold(img,127,255,cv2.THRESH_BINARY_INV)  # converting the image into binary image.
kernel = cv2.getStructuringElement(cv2.MORPH_RECT,(100,3))  # kernel to detect vertical lines
vertical = cv2.morphologyEx(img2, cv2.MORPH_OPEN, kernel)  # applying morphological opening operation to detect vertical lines
vertical = cv2.dilate(vertical,kernel,iterations = 1)   #dilate the vertical lines obtained

kernel2 = cv2.getStructuringElement(cv2.MORPH_RECT,(3,100))   # kernel to detect horizontal lines
horizontal = cv2.morphologyEx(img2, cv2.MORPH_OPEN, kernel2)   # applying morphological opening operation to detect horizontal lines
horizontal = cv2.dilate(horizontal,kernel2,iterations = 1)    #dilate the horizontal lines obtained

cv2.imshow('d',vertical)    # show the vertical imag
cv2.imshow('b',horizontal)  # show the horizontal image

img = img2 -horizontal - vertical   # subtracting horizontal and vertical lines from original image

cv2.imwrite('horizontal.png',horizontal)   
cv2.imwrite('vertical.png',vertical)
cv2.imwrite('result.png',img)


cv2.imshow('last',img)     # show the resulted image after subtraction

kerne = np.ones((3,3),np.uint8)             # kernel to remove the noise from the last image
opening = cv2.morphologyEx(img, cv2.MORPH_OPEN, kerne)  # applying opening morphological operation to remove the noise from the image 
cv2.imshow('opening',opening)               # show the resulted image after removing noise
cv2.imwrite('noise_removal.png',opening)
cv2.waitKey(0)

检查下面的结果，并建议如何进一步从手绘电路的扫描图像中检测逻辑门？

代码的结果如下：

1）输入图片：

2）输出图像（代码结果）：

Answer 1

逻辑门都有相同的尺寸。我会这样做：

1. 白色区域的连接组件标记。
2. 独立/分离
3. 按尺寸过滤标签。
4. （可选）所有逻辑门都会触摸右侧的白色小图案/标签。