如何根据较低和较高的粉红色范围找到 roi_corners，以便它可以在 python 和 opencv 中模糊

Question

I am using hardcoded values in roi_corners to blur number plate of car which is in pink color. I want to find out roi_corners by detecting lower and upper pink color range so that it can be blurred automatically by detecting location of pink color. I am using below code which works fine, only need help in finding roi_corners programmatically based on pink color lower and upper range. Pink color range is provided below for your help. lower_color = np.array([158, 127, 0]) upper_color = np.array([179, 255, 255])

Please find below code which I am using

import cv2 as cv
import numpy as np

# Here I define the list of vertices of an example polygon ROI:
roi_corners = np.array([[(34,188),(30,214),(80,227),(82,200)]],dtype = np.int32)
print ('print roi_corners ')
print (roi_corners)
print (type (roi_corners)) # <class 'numpy.ndarray'>

# Read the original Image:
image = cv.imread('image_new.jpeg')
# create a blurred copy of the entire image:
blurred_image = cv.GaussianBlur(image,(43, 43), 30)

# create a mask for the ROI and fill in the ROI with (255,255,255) color :
mask = np.zeros(image.shape, dtype=np.uint8)
channel_count = image.shape[2]
ignore_mask_color = (255,)*channel_count
cv.fillPoly(mask, roi_corners, ignore_mask_color)

# create a mask for everywhere in the original image except the ROI, (hence mask_inverse) :
mask_inverse = np.ones(mask.shape).astype(np.uint8)*255 - mask

# combine all the masks and above images in the following way :
final_image = cv.bitwise_and(blurred_image, mask) + cv.bitwise_and(image, mask_inverse)

cv.imshow("image", image)
cv.imshow("final_image", final_image)
cv.waitKey()
cv.destroyAllWindows()

Answer 1

Here is one way to get the bounds of the pink on the license plate in Python OpenCV.

 - Read the input
 - Threshold on the pink
 - Apply morphology to clean it up
 - Get the contour
 - Get the rotated rectangle corners from the contour
 - Draw the rotated rectangle on the input image
 - Save the results

Input:

import cv2
import numpy as np

# read image
img = cv2.imread("pink_license.jpg")

# get color bounds of pink
lower =(130,0,220) # lower bound for each channel
upper = (170,255,255) # upper bound for each channel

# create the mask and use it to change the colors
thresh = cv2.inRange(img, lower, upper)

# apply morphology
kernel = np.ones((3,3), np.uint8)
morph = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel)
kernel = np.ones((7,7), np.uint8)
morph = cv2.morphologyEx(morph, cv2.MORPH_DILATE, kernel)

# get contour
contours = cv2.findContours(morph, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
contours = contours[0] if len(contours) == 2 else contours[1]
cntr = contours[0]
# get rotated rectangle from contour
rot_rect = cv2.minAreaRect(cntr)
box = cv2.boxPoints(rot_rect)
box = np.int0(box)
print(box)

# draw rotated rectangle on copy of img
rot_bbox = img.copy()
cv2.drawContours(rot_bbox,[box],0,(0,255,0),1)

# write img with red rotated bounding box to disk
cv2.imwrite("pink_license_thresh.jpg", thresh)
cv2.imwrite("pink_license_morph.jpg", morph)
cv2.imwrite("pink_license_rot_rect.png", rot_bbox)

# display it
cv2.imshow("THRESHOLD", thresh)
cv2.imshow("MORPH", morph)
cv2.imshow("BBOX", rot_bbox)
cv2.waitKey(0)

Threshold image:

Morphology cleaned image:

Green Rotated Rectangle on input:

Corner Coordinates:

[[ 74 212]
 [ 39 209]
 [ 40 197]
 [ 75 200]]