Number Plate Detection in OpenCV - Python

 Script:

import cv2
# Importing the Opencv Library
import numpy as np
# Importing NumPy,which is the fundamental package for scientific computing with Python

# Reading Image
img = cv2.imread("/home/whoami/Pictures/number1.jpg")
cv2.namedWindow("Original Image",cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Original Image",img)
# Display image

# RGB to Gray scale conversion
img_gray = cv2.cvtColor(img,cv2.COLOR_RGB2GRAY)
cv2.namedWindow("Gray Converted Image",cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Gray Converted Image",img_gray)
# Display Image

# Noise removal with iterative bilateral filter(removes noise while preserving edges)
noise_removal = cv2.bilateralFilter(img_gray,9,75,75)
cv2.namedWindow("Noise Removed Image",cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Noise Removed Image",noise_removal)
# Display Image

# Histogram equalisation for better results
equal_histogram = cv2.equalizeHist(noise_removal)
cv2.namedWindow("After Histogram equalisation",cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("After Histogram equalisation",equal_histogram)
# Display Image

# Morphological opening with a rectangular structure element
kernel = cv2.getStructuringElement(cv2.MORPH_RECT,(5,5))
morph_image = cv2.morphologyEx(equal_histogram,cv2.MORPH_OPEN,kernel,iterations=15)
cv2.namedWindow("Morphological opening",cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Morphological opening",morph_image)
# Display Image

# Image subtraction(Subtracting the Morphed image from the histogram equalised Image)
sub_morp_image = cv2.subtract(equal_histogram,morph_image)
cv2.namedWindow("Subtraction image", cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Subtraction image", sub_morp_image)
# Display Image

# Thresholding the image
ret,thresh_image = cv2.threshold(sub_morp_image,0,255,cv2.THRESH_OTSU)
cv2.namedWindow("Image after Thresholding",cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Image after Thresholding",thresh_image)
# Display Image

# Applying Canny Edge detection
canny_image = cv2.Canny(thresh_image,250,255)
cv2.namedWindow("Image after applying Canny",cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Image after applying Canny",canny_image)
# Display Image
canny_image = cv2.convertScaleAbs(canny_image)

# dilation to strengthen the edges
kernel = np.ones((3,3), np.uint8)
# Creating the kernel for dilation
dilated_image = cv2.dilate(canny_image,kernel,iterations=1)
cv2.namedWindow("Dilation", cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Dilation", dilated_image)
# Displaying Image

# Finding Contours in the image based on edges
new,contours, hierarchy = cv2.findContours(dilated_image, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
contours= sorted(contours, key = cv2.contourArea, reverse = True)[:10]
# Sort the contours based on area ,so that the number plate will be in top 10 contours
screenCnt = None
# loop over our contours
for c in contours:
# approximate the contour
peri = cv2.arcLength(c, True)
approx = cv2.approxPolyDP(c, 0.06 * peri, True)  # Approximating with 6% error
# if our approximated contour has four points, then
# we can assume that we have found our screen
if len(approx) == 4:  # Select the contour with 4 corners
screenCnt = approx
break
final = cv2.drawContours(img, [screenCnt], -1, (0, 255, 0), 3)
# Drawing the selected contour on the original image
cv2.namedWindow("Image with Selected Contour",cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Image with Selected Contour",final)

# Masking the part other than the number plate
mask = np.zeros(img_gray.shape,np.uint8)
new_image = cv2.drawContours(mask,[screenCnt],0,255,-1,)
new_image = cv2.bitwise_and(img,img,mask=mask)
cv2.namedWindow("Final_image",cv2.WINDOW_NORMAL)
cv2.imshow("Final_image",new_image)

# Histogram equal for enhancing the number plate for further processing
y,cr,cb = cv2.split(cv2.cvtColor(new_image,cv2.COLOR_RGB2YCrCb))
# Converting the image to YCrCb model and splitting the 3 channels
y = cv2.equalizeHist(y)
# Applying histogram equalisation
final_image = cv2.cvtColor(cv2.merge([y,cr,cb]),cv2.COLOR_YCrCb2RGB)
# Merging the 3 channels
cv2.namedWindow("Enhanced Number Plate",cv2.WINDOW_NORMAL)
# Creating a Named window to display image
cv2.imshow("Enhanced Number Plate",final_image)
# Display image
cv2.waitKey() # Wait for a keystroke from the user


The test cases and the result obtained from the above script are uploaded in the below pdf file

The code which is written above is also uploaded in my github account here.