使用 OpenCV 检测垂直曲线

Question

I want to detect the vertical curved line which is in the middle of the image. Do you have any idea how to do it? I want to separate the left and right side by the line. What feature or filter should I use? Thank you very much for your ideas how to detect the line.

Here is link to the image:

BTW: I use OpenCV library in Python.

Answer 1

Sorry, I do not know OpenCV that well. So here is an outline of how to do it. I also present code using Imagemagick and also Python Wand code. Note, that I save the intermediate images to show the steps.

Read the image

Blur it some

Threshold is to binary

Do connected components processing to remove all small regions (see contours or blobs in OpenCV)

Use morphology open and close to smooth the edges

Extract an edge outline of the transition between white and black (there are many edge operators: laplacian, gradients, canny, etc)

Input:

convert img.jpg -blur 0x1 -threshold 9% -type bilevel +write threshold.png \
-define connected-components:mean-color=true \
-define connected-components:area-threshold=10000 \
-connected-components 4 +write ccl.png \
-morphology open disk:5 \
-morphology close disk:5 +write smooth.png \
-morphology edge diamond:1 \
result.png

Threshold:

Connected Components:

Smooth:

Result:

Here is the equivalent code using Python Wand 0.5.6 (currently under development) with Imagemagick 7.0.8.56

#!/bin/python3.7

from wand.image import Image
from wand.display import display

with Image(filename='curve.jpg') as img:
    img.blur(radius=0, sigma=1)
    img.threshold(threshold=0.09)
    img.connected_components(connectivity=4, area_threshold=1000, mean_color=True)
    img.morphology(method='open', kernel='disk:5')
    img.morphology(method='close', kernel='disk:5')
    img.morphology(method='edge', kernel='diamond:1')
    img.save(filename='0_wand_trim.png')
    display(img)

Answer 2

Here's a simple approach

• Convert image to grayscale and median blur
• Threshold to get binary image
• Perform morphological transformations to smooth image
• Perform canny edge detection

---

Median blur → Threshold → Opening → Close → Canny

import cv2

image = cv2.imread('1.jpg', 0)
blur = cv2.medianBlur(image, 9)
thresh = cv2.threshold(blur, 25, 255, cv2.THRESH_BINARY)[1]
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (9,9))
opening = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel, iterations=3)
close = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel, iterations=3)
canny = cv2.Canny(close, 120, 255, 1)

cv2.imshow('canny', canny)
cv2.waitKey()