Python cv2 HoughLines grid line detection

Question

I have a simple grid in an image, I am trying to determine the grid size, e.g. 6x6, 12x12, etc. Using Python and cv2.

I am testing it with the above 3x3 grid, I was planning to determine the grid size by counting how many vertical / horizontal lines there are by detecting them in the image:

import cv2
import numpy as np

im = cv2.imread('photo2.JPG')
gray = cv2.cvtColor(im,cv2.COLOR_BGR2GRAY)

imgSplit = cv2.split(im)
flag,b = cv2.threshold(imgSplit[2],0,255,cv2.THRESH_OTSU) 

element = cv2.getStructuringElement(cv2.MORPH_CROSS,(1,1))
cv2.erode(b,element)

edges = cv2.Canny(b,150,200,3,5)

while(True):

    img = im.copy()

    lines = cv2.HoughLinesP(edges,1,np.pi/2,2, minLineLength = 620, maxLineGap = 100)[0]

    for x1,y1,x2,y2 in lines:        
        cv2.line(img,(x1,y1),(x2,y2),(0,255,0),1)

    cv2.imshow('houghlines',img)

    if k == 27:
        break

cv2.destroyAllWindows()

My code detects the lines, as can be seen below, however there are multiple lines detected for each line in my image:

(there are two 1px green lines drawn for every line in the image)

I cannot simply divide the number of lines by two because (depending on the grid size) sometimes just the one line will be drawn.

How can I more accurately detect and draw a single line for every line detected in the original image?

I have tweaked threshold settings, reducing the image to black and white, yet I still get multiple lines. I assume this is because of the canny edge detection?

Answer 1

I ended up iterating through the lines and removing lines that were within 10px of one another:

lines = cv2.HoughLinesP(edges,1,np.pi/180,275, minLineLength = 600, maxLineGap = 100)[0].tolist()

for x1,y1,x2,y2 in lines:
    for index, (x3,y3,x4,y4) in enumerate(lines):

        if y1==y2 and y3==y4: # Horizontal Lines
            diff = abs(y1-y3)
        elif x1==x2 and x3==x4: # Vertical Lines
            diff = abs(x1-x3)
        else:
            diff = 0

        if diff < 10 and diff is not 0:
            del lines[index]

gridsize = (len(lines) - 2) / 2

Answer 2

you can dilate the image with kernel = cv2.getStructuringElement(cv2.MORPH_CROSS, (2, 2)) dilated = cv2.dilate(edges, kernel, iterations=5) then apply cv2.HoughLinesP

Answer 3

Doesn't the Hough function have a parameter that does exactly this? MaxLineGap? So if your lines were 2px thick, you set that parameter to 3? Does it not work?