Python OpenCV line detection to detect `X` symbol in image

Question

I have an image where I need to detect a X symbol inside the line.

The image:

As you can see in the image above there is a X symbol inside a line. I want to know the X & Y coordinates of the symbol. Is there a way to find this symbol within this picture or is it to small?

import cv2
import numpy as np


def calculateCenterSpot(results):
    startX, endX = results[0][0], results[0][2]
    startY, endY = results[0][1], results[0][3]
    centerSpotX = (endX - startX) / 2 + startX
    centerSpotY = (endY - startY) / 2 + startY
    return [centerSpotX, centerSpotY]

img = cv2.imread('crop_1.png')
res2 = img.copy()

cords = [[1278, 704, 1760, 1090]]
center = calculateCenterSpot(cords)
cv2.circle(img, (int(center[0]), int(center[1])), 1, (0,0,255), 30)
cv2.line(img, (int(center[0]), 0), (int(center[0]), img.shape[0]), (0,255,0), 10)
cv2.line(img, (0, int(center[1])), (img.shape[1], int(center[1])), (255,0,0), 10)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# You can either use threshold or Canny edge for HoughLines().
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
#edges = cv2.Canny(gray, 50, 150, apertureSize=3)

# Perform HoughLines tranform.
lines = cv2.HoughLines(thresh,0.5,np.pi/180,1000)
for line in lines:
    for rho,theta in line:
            a = np.cos(theta)
            b = np.sin(theta)
            x0 = a*rho
            y0 = b*rho
            x1 = int(x0 + 5000*(-b))
            y1 = int(y0 + 5000*(a))
            x2 = int(x0 - 5000*(-b))
            y2 = int(y0 - 5000*(a))
            if x2 == int(center[0]):
                cv2.circle(img, (x2,y1), 100, (0,0,255), 30)

            if y2 == int(center[1]):
                print('hell2o')
                # cv2.line(res2,(x1,y1),(x2,y2),(0,0,255),2)

#Display the result.
cv2.imwrite('h_res1.png', img)
cv2.imwrite('h_res3.png', res2)

cv2.imwrite('image.png', img)

I already tried doing it with HoughLines, but it wasn't a success.

Answer 1

Instead of using cv2.HoughLines(), an alternative approach is to use template matching. The idea is to search and find the location of a template image in a larger image. To perform this method, the template slides over the input image (similar to 2D convolution) where comparison methods are performed to determine pixel similarity. This is the basic idea behind template matching. Unfortunately, this basic method has flaws since it only works if the template image size is the same as the desired item to find in the input image. So if your template image was smaller than the desired region to find in the input image, this method would not work.

To get around this limitation, we can dynamically rescale the image for better template matching using np.linspace(). With each iteration, we resize the input image and keep track of the ratio. We continue resizing until the template image size is larger than the resized image while keeping track of the highest correlation value. A higher correlation value means a better match. Once we iterate through various scales, we find the ratio with the largest match and then compute the coordinates of the bounding box to determine the ROI.

---

Using this screenshotted template image

Here's the result

import cv2
import numpy as np

# Resizes a image and maintains aspect ratio
def maintain_aspect_ratio_resize(image, width=None, height=None, inter=cv2.INTER_AREA):
    # Grab the image size and initialize dimensions
    dim = None
    (h, w) = image.shape[:2]

    # Return original image if no need to resize
    if width is None and height is None:
        return image

    # We are resizing height if width is none
    if width is None:
        # Calculate the ratio of the height and construct the dimensions
        r = height / float(h)
        dim = (int(w * r), height)
    # We are resizing width if height is none
    else:
        # Calculate the ratio of the 0idth and construct the dimensions
        r = width / float(w)
        dim = (width, int(h * r))

    # Return the resized image
    return cv2.resize(image, dim, interpolation=inter)

# Load template, convert to grayscale, perform canny edge detection
template = cv2.imread('template.png')
template = cv2.cvtColor(template, cv2.COLOR_BGR2GRAY)
template = cv2.Canny(template, 50, 200)
(tH, tW) = template.shape[:2]
cv2.imshow("template", template)

# Load original image, convert to grayscale
original_image = cv2.imread('1.png')
gray = cv2.cvtColor(original_image, cv2.COLOR_BGR2GRAY)
found = None

# Dynamically rescale image for better template matching
for scale in np.linspace(0.1, 3.0, 20)[::-1]:

    # Resize image to scale and keep track of ratio
    resized = maintain_aspect_ratio_resize(gray, width=int(gray.shape[1] * scale))
    r = gray.shape[1] / float(resized.shape[1])

    # Stop if template image size is larger than resized image
    if resized.shape[0] < tH or resized.shape[1] < tW:
        break

    # Detect edges in resized image and apply template matching
    canny = cv2.Canny(resized, 50, 200)
    detected = cv2.matchTemplate(canny, template, cv2.TM_CCOEFF)
    (_, max_val, _, max_loc) = cv2.minMaxLoc(detected)

    # Uncomment this section for visualization
    '''
    clone = np.dstack([canny, canny, canny])
    cv2.rectangle(clone, (max_loc[0], max_loc[1]), (max_loc[0] + tW, max_loc[1] + tH), (0,255,0), 2)
    cv2.imshow('visualize', clone)
    cv2.waitKey(0)
    '''

    # Keep track of correlation value
    # Higher correlation means better match
    if found is None or max_val > found[0]:
        found = (max_val, max_loc, r)

# Compute coordinates of bounding box
(_, max_loc, r) = found
(start_x, start_y) = (int(max_loc[0] * r), int(max_loc[1] * r))
(end_x, end_y) = (int((max_loc[0] + tW) * r), int((max_loc[1] + tH) * r))

# Draw bounding box on ROI
cv2.rectangle(original_image, (start_x, start_y), (end_x, end_y), (0,255,0), 2)
cv2.imshow('detected', original_image)
cv2.imwrite('detected.png', original_image)
cv2.waitKey(0)

Answer 2

For multiple template images you can use a for loop with the amount of different template images you have and then using the threshold to method find multiple template matches.

for i in range(templateAmount):
    template = cv2.imread('template{}.png'.format(i),0)
    w, h = template.shape[::-1]
    res = cv2.matchTemplate(img_gray,template,cv2.TM_CCOEFF_NORMED)
    threshold = 0.8
    loc = np.where( res >= threshold)
    for pt in zip(*loc[::-1]):
        cv2.rectangle(img_rgb, pt, (pt[0] + w, pt[1] + h), (0,0,255), 2)

Answer 3

If you have multiple images where you need to detect this X symbol, and if this X symbol is always the same and has the same dimension, you could run a two-dimensional convolution over each image, where the kernel that you are convoluting is the X symbol, isolated, that you are trying to detect. You could then check the output of this two-dimensional convolution for the pixel of maximum intensity, whose normalize coordinate (x/w,y/h) would correspond with high probability to the normalize coordinate of the X symbol in the input image. Here is the mathematical expression for a two-dimensional convolution:

In opencv you can define your own kernel (make sure to only keep the cross and nothing else in the background), and then apply it to your image.