def process(img):
# this folder is used to save the image
# temp_folder = 'C:\\Users\\PC\\Desktop\\temp\\'
# ap = argparse.ArgumentParser()
# ap.add_argument("-i", "--image", type=str, required=True, help="path to image")
# args = vars(ap.parse_args())
# img = cv2.imread(args["image"])
# cv2.imshow('original', img)
# cv2.imwrite(temp_folder + '1 - original.png', img)
# hsv transform - value = gray image
# img = cv2.imread(img_location)
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
hue, saturation, value = cv2.split(hsv)
# cv2.imshow('HSV',hsv)
# cv2.imwrite('HSV.jpg',hsv)
# cv2.imshow('gray', value)
# cv2.imwrite(temp_folder + '2 - gray.png', value)
# kernel to use for morphological operations
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
# applying topHat/blackHat operations
topHat = cv2.morphologyEx(value, cv2.MORPH_TOPHAT, kernel)
blackHat = cv2.morphologyEx(value, cv2.MORPH_BLACKHAT, kernel)
#cv2.imshow('topHat', topHat)
#cv2.imshow('blackHat', blackHat)
# cv2.imwrite(temp_folder + '3 - topHat.png', topHat)
# cv2.imwrite(temp_folder + '4 - blackHat.png', blackHat)
# add and subtract between morphological operations
add = cv2.add(value, topHat)
subtract = cv2.subtract(add, blackHat)
# cv2.imshow('subtract', subtract)
# cv2.imwrite(temp_folder + '5 - subtract.png', subtract)
# applying gaussian blur on subtract image
blur = cv2.GaussianBlur(subtract, (5, 5), 0)
# cv2.imshow('blur', blur)
# cv2.imwrite(temp_folder + '6 - blur.png', blur)
# thresholding
thresh = cv2.adaptiveThreshold(blur, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY_INV, 19, 9)
# cv2.imshow('thresh', thresh)
# cv2.imwrite(temp_folder + '7 - thresh.png', thresh)
# check for contours on thresh
imageContours, contours, hierarchy = cv2.findContours(
thresh, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
# get height and width
height, width = thresh.shape
# create a numpy array with shape given by threshed image value dimensions
imageContours = np.zeros((height, width, 3), dtype=np.uint8)
# list and counter of possible chars
possibleChars = []
countOfPossibleChars = 0
# loop to check if any (possible) char is found
for i in range(0, len(contours)):
# draw contours based on actual found contours of thresh image
cv2.drawContours(imageContours, contours, i, (255, 255, 255))
# retrieve a possible char by the result ifChar class give us
possibleChar = Functions.ifChar(contours[i])
# by computing some values (area, width, height, aspect ratio) possibleChars list is being populated
if Functions.checkIfChar(possibleChar) is True:
countOfPossibleChars = countOfPossibleChars + 1
possibleChars.append(possibleChar)
# cv2.imshow("contours", imageContours)
# cv2.imwrite(temp_folder + '8 - imageContours.png', imageContours)
imageContours = np.zeros((height, width, 3), np.uint8)
ctrs = []
# populating ctrs list with each char of possibleChars
for char in possibleChars:
ctrs.append(char.contour)
# using values from ctrs to draw new contours
cv2.drawContours(imageContours, ctrs, -1, (255, 255, 255))
# cv2.imshow("contoursPossibleChars", imageContours)
# cv2.imwrite(temp_folder + '9 - contoursPossibleChars.png', imageContours)
plates_list = []
listOfListsOfMatchingChars = []
for possibleC in possibleChars:
# the purpose of this function is, given a possible char and a big list of possible chars,
# find all chars in the big list that are a match for the single possible char, and return those matching chars as a list
def matchingChars(possibleC, possibleChars):
listOfMatchingChars = []
# if the char we attempting to find matches for is the exact same char as the char in the big list we are currently checking
# then we should not include it in the list of matches b/c that would end up double including the current char
# so do not add to list of matches and jump back to top of for loop
for possibleMatchingChar in possibleChars:
if possibleMatchingChar == possibleC:
continue
# compute stuff to see if chars are a match
distanceBetweenChars = Functions.distanceBetweenChars(
possibleC, possibleMatchingChar)
angleBetweenChars = Functions.angleBetweenChars(
possibleC, possibleMatchingChar)
changeInArea = float(
abs(possibleMatchingChar.boundingRectArea -
possibleC.boundingRectArea)) / float(
possibleC.boundingRectArea)
changeInWidth = float(
abs(possibleMatchingChar.boundingRectWidth -
possibleC.boundingRectWidth)) / float(
possibleC.boundingRectWidth)
changeInHeight = float(
abs(possibleMatchingChar.boundingRectHeight -
possibleC.boundingRectHeight)) / float(
possibleC.boundingRectHeight)
# check if chars match
if distanceBetweenChars < (possibleC.diagonalSize * 5) and \
angleBetweenChars < 12.0 and \
changeInArea < 0.5 and \
changeInWidth < 0.8 and \
changeInHeight < 0.2:
listOfMatchingChars.append(possibleMatchingChar)
return listOfMatchingChars
# here we are re-arranging the one big list of chars into a list of lists of matching chars
# the chars that are not found to be in a group of matches do not need to be considered further
listOfMatchingChars = matchingChars(possibleC, possibleChars)
listOfMatchingChars.append(possibleC)
# if current possible list of matching chars is not long enough to constitute a possible plate
# jump back to the top of the for loop and try again with next char
if len(listOfMatchingChars) < 3:
continue
# here the current list passed test as a "group" or "cluster" of matching chars
listOfListsOfMatchingChars.append(listOfMatchingChars)
# remove the current list of matching chars from the big list so we don't use those same chars twice,
# make sure to make a new big list for this since we don't want to change the original big list
listOfPossibleCharsWithCurrentMatchesRemoved = list(
set(possibleChars) - set(listOfMatchingChars))
recursiveListOfListsOfMatchingChars = []
for recursiveListOfMatchingChars in recursiveListOfListsOfMatchingChars:
listOfListsOfMatchingChars.append(recursiveListOfMatchingChars)
break
imageContours = np.zeros((height, width, 3), np.uint8)
for listOfMatchingChars in listOfListsOfMatchingChars:
contoursColor = (255, 0, 255)
contours = []
for matchingChar in listOfMatchingChars:
contours.append(matchingChar.contour)
cv2.drawContours(imageContours, contours, -1, contoursColor)
# cv2.imshow("finalContours", imageContours)
# cv2.imwrite(temp_folder + '10 - finalContours.png', imageContours)
for listOfMatchingChars in listOfListsOfMatchingChars:
possiblePlate = Functions.PossiblePlate()
# sort chars from left to right based on x position
listOfMatchingChars.sort(key=lambda matchingChar: matchingChar.centerX)
# calculate the center point of the plate
plateCenterX = (
listOfMatchingChars[0].centerX +
listOfMatchingChars[len(listOfMatchingChars) - 1].centerX) / 2.0
plateCenterY = (
listOfMatchingChars[0].centerY +
listOfMatchingChars[len(listOfMatchingChars) - 1].centerY) / 2.0
plateCenter = plateCenterX, plateCenterY
# calculate plate width and height
plateWidth = int(
(listOfMatchingChars[len(listOfMatchingChars) - 1].boundingRectX +
listOfMatchingChars[len(listOfMatchingChars) -
1].boundingRectWidth -
listOfMatchingChars[0].boundingRectX) * 1.3)
totalOfCharHeights = 0
for matchingChar in listOfMatchingChars:
totalOfCharHeights = totalOfCharHeights + matchingChar.boundingRectHeight
averageCharHeight = totalOfCharHeights / len(listOfMatchingChars)
plateHeight = int(averageCharHeight * 1.5)
# calculate correction angle of plate region
opposite = listOfMatchingChars[len(
listOfMatchingChars) - 1].centerY - listOfMatchingChars[0].centerY
hypotenuse = Functions.distanceBetweenChars(
listOfMatchingChars[0],
listOfMatchingChars[len(listOfMatchingChars) - 1])
correctionAngleInRad = math.asin(opposite / hypotenuse)
correctionAngleInDeg = correctionAngleInRad * (180.0 / math.pi)
# pack plate region center point, width and height, and correction angle into rotated rect member variable of plate
possiblePlate.rrLocationOfPlateInScene = (tuple(plateCenter),
(plateWidth, plateHeight),
correctionAngleInDeg)
# get the rotation matrix for our calculated correction angle
rotationMatrix = cv2.getRotationMatrix2D(tuple(plateCenter),
correctionAngleInDeg, 1.0)
height, width, numChannels = img.shape
# rotate the entire image
imgRotated = cv2.warpAffine(img, rotationMatrix, (width, height))
# crop the image/plate detected
imgCropped = cv2.getRectSubPix(imgRotated, (plateWidth, plateHeight),
tuple(plateCenter))
# copy the cropped plate image into the applicable member variable of the possible plate
possiblePlate.Plate = imgCropped
# cv2.imshow('Plate',imgCropped)
cv2.imwrite('./static/outputs/NumberPlate.jpg', imgCropped)
imgT = imgCropped
# populate plates_list with the detected plate
if possiblePlate.Plate is not None:
plates_list.append(possiblePlate)
# draw a ROI on the original image
for i in range(0, len(plates_list)):
# finds the four vertices of a rotated rect - it is useful to draw the rectangle.
p2fRectPoints = cv2.boxPoints(
plates_list[i].rrLocationOfPlateInScene)
# roi rectangle colour
rectColour = (0, 255, 0)
cv2.line(imageContours, tuple(p2fRectPoints[0]),
tuple(p2fRectPoints[1]), rectColour, 2)
cv2.line(imageContours, tuple(p2fRectPoints[1]),
tuple(p2fRectPoints[2]), rectColour, 2)
cv2.line(imageContours, tuple(p2fRectPoints[2]),
tuple(p2fRectPoints[3]), rectColour, 2)
cv2.line(imageContours, tuple(p2fRectPoints[3]),
tuple(p2fRectPoints[0]), rectColour, 2)
cv2.line(img, tuple(p2fRectPoints[0]), tuple(p2fRectPoints[1]),
rectColour, 2)
cv2.line(img, tuple(p2fRectPoints[1]), tuple(p2fRectPoints[2]),
rectColour, 2)
cv2.line(img, tuple(p2fRectPoints[2]), tuple(p2fRectPoints[3]),
rectColour, 2)
cv2.line(img, tuple(p2fRectPoints[3]), tuple(p2fRectPoints[0]),
rectColour, 2)
imageContours = cv2.cvtColor(imageContours, cv2.COLOR_BGR2GRAY)
sobel = cv2.Sobel(imageContours, cv2.CV_8U, 1, 0, ksize=3)
element1 = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 1))
element2 = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5))
dilation = cv2.dilate(sobel, element2, iterations=1)
erosion = cv2.erode(dilation, element1, iterations=1)
dilation2 = cv2.dilate(erosion, element2, iterations=10)
#cv2.imshow("detected", dilation2)
# cv2.imwrite(temp_folder + '11 - detected.png', imageContours)
#cv2.imshow("detectedOriginal", img)
# cv2.imwrite(temp_folder + '12 - detectedOriginal.png', img)
# cv2.imshow("plate", plates_list[i].Plate)
# cv2.imwrite(temp_folder + '13 - plate.png', plates_list[i].Plate)
return dilation2
def read(img):
#args = vars(ap.parse_args())
#img = cv2.imread(args["image"])
retorno = ""
# cv2.imshow('original', img)
# cv2.imwrite(temp_folder + '1 - original.png', img)
# hsv transform - value = gray image
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
hue, saturation, value = cv2.split(hsv)
# cv2.imshow('gray', value)
# cv2.imwrite(temp_folder + '2 - gray.png', value)
# kernel a ser usado para operações morfológicas
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
# aplicação de operações topHat / blackHat
topHat = cv2.morphologyEx(value, cv2.MORPH_TOPHAT, kernel)
blackHat = cv2.morphologyEx(value, cv2.MORPH_BLACKHAT, kernel)
# cv2.imshow('topHat', topHat)
# cv2.imshow('blackHat', blackHat)
# cv2.imwrite(temp_folder + '3 - topHat.png', topHat)
# cv2.imwrite(temp_folder + '4 - blackHat.png', blackHat)
# adicionar e subtrair entre operações morfológicas
add = cv2.add(value, topHat)
subtract = cv2.subtract(add, blackHat)
# cv2.imshow('subtract', subtract)
# cv2.imwrite(temp_folder + '5 - subtract.png', subtract)
# aplicação de desfoque gaussiano na subtração da imagem
blur = cv2.GaussianBlur(subtract, (5, 5), 0)
# cv2.imshow('blur', blur)
# cv2.imwrite(temp_folder + '6 - blur.png', blur)
# thresholding
thresh = cv2.adaptiveThreshold(blur, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY_INV, 19, 9)
# cv2.imshow('thresh', thresh)
# cv2.imwrite(temp_folder + '7 - thresh.png', thresh)
# cv2.findCountours() function changed from OpenCV3 to OpenCV4: now it have only two parameters instead of 3
cv2MajorVersion = cv2.__version__.split(".")[0]
# verifique se há contornos na imagem
if int(cv2MajorVersion) >= 4:
contours, hierarchy = cv2.findContours(thresh, cv2.RETR_LIST,
cv2.CHAIN_APPROX_SIMPLE)
else:
imageContours, contours, hierarchy = cv2.findContours(
thresh, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
# obter altura e largura
height, width = thresh.shape
# crie uma matriz numpy com a forma dada pelas dimensões do valor da imagem combinada
imageContours = np.zeros((height, width, 3), dtype=np.uint8)
# lista e contador de caracteres possíveis
possibleChars = []
countOfPossibleChars = 0
# loop para verificar se algum (possível) caractere é encontrado
for i in range(0, len(contours)):
# desenhe contornos com base nos contornos encontrados reais da imagem thresh
cv2.drawContours(imageContours, contours, i, (255, 255, 255))
# recuperar um possível caractere pelo resultado da classe ifChar nos fornecer
possibleChar = Functions.ifChar(contours[i])
#calculando alguns valores (área, largura, altura, proporção) possíveis
if Functions.checkIfChar(possibleChar) is True:
countOfPossibleChars = countOfPossibleChars + 1
possibleChars.append(possibleChar)
# cv2.imshow("contours", imageContours)
# cv2.imwrite(temp_folder + '8 - imageContours.png', imageContours)
# usando valores de ctrs para desenhar novos contornos
imageContours = np.zeros((height, width, 3), np.uint8)
ctrs = []
# populating ctrs list with each char of possibleChars
for char in possibleChars:
ctrs.append(char.contour)
# using values from ctrs to draw new contours
cv2.drawContours(imageContours, ctrs, -1, (255, 255, 255))
# cv2.imshow("contoursPossibleChars", imageContours)
# cv2.imwrite(temp_folder + '9 - contoursPossibleChars.png', imageContours)
plates_list = []
listOfListsOfMatchingChars = []
for possibleC in possibleChars:
# o objetivo desta função é, dado um possível caractere e uma grande lista de possíveis caracteres,
# encontre todos os caracteres na grande lista que correspondem ao único
#caractere possível e retorne esses caracteres como uma lista
def matchingChars(possibleC, possibleChars):
listOfMatchingChars = []
# se o caractere para o qual estamos tentando encontrar correspondências for exatamente o mesmo caractere que o caractere na grande lista que estamos verificando
# então não devemos incluí-lo na lista de correspondências b / c que acabariam em dobro, incluindo o caractere atual
# então não adicione à lista de correspondências e volte ao topo do loop for
for possibleMatchingChar in possibleChars:
if possibleMatchingChar == possibleC:
continue
# calcular coisas para ver se os caracteres são uma correspondência
distanceBetweenChars = Functions.distanceBetweenChars(
possibleC, possibleMatchingChar)
angleBetweenChars = Functions.angleBetweenChars(
possibleC, possibleMatchingChar)
changeInArea = float(
abs(possibleMatchingChar.boundingRectArea -
possibleC.boundingRectArea)) / float(
possibleC.boundingRectArea)
changeInWidth = float(
abs(possibleMatchingChar.boundingRectWidth -
possibleC.boundingRectWidth)) / float(
possibleC.boundingRectWidth)
changeInHeight = float(
abs(possibleMatchingChar.boundingRectHeight -
possibleC.boundingRectHeight)) / float(
possibleC.boundingRectHeight)
# verifique se os caracteres correspondem
if distanceBetweenChars < (possibleC.diagonalSize * 5) and \
angleBetweenChars < 12.0 and \
changeInArea < 0.5 and \
changeInWidth < 0.8 and \
changeInHeight < 0.2:
listOfMatchingChars.append(possibleMatchingChar)
return listOfMatchingChars
# aqui estamos reorganizando a grande lista de caracteres em uma lista de listas de caracteres correspondentes
# os caracteres que não foram encontrados em um grupo de correspondências não precisam ser considerados mais
listOfMatchingChars = matchingChars(possibleC, possibleChars)
listOfMatchingChars.append(possibleC)
if len(listOfMatchingChars) < 6:
continue
listOfListsOfMatchingChars.append(listOfMatchingChars)
# remova a lista atual de caracteres correspondentes da grande lista para não usarmos os mesmos caracteres duas vezes,
# certifique-se de criar uma nova grande lista para isso, pois não queremos alterar a grande lista original
#listOfPossibleCharsWithCurrentMatchesRemoved = list(set(possibleChars) - set(listOfMatchingChars))
recursiveListOfListsOfMatchingChars = []
for recursiveListOfMatchingChars in recursiveListOfListsOfMatchingChars:
listOfListsOfMatchingChars.append(recursiveListOfMatchingChars)
break
imageContours = np.zeros((height, width, 3), np.uint8)
for listOfMatchingChars in listOfListsOfMatchingChars:
contoursColor = (255, 0, 255)
contours = []
for matchingChar in listOfMatchingChars:
contours.append(matchingChar.contour)
cv2.drawContours(imageContours, contours, -1, contoursColor)
# cv2.imshow("finalContours", imageContours)
# cv2.imwrite(temp_folder + '10 - finalContours.png', imageContours)
for listOfMatchingChars in listOfListsOfMatchingChars:
possiblePlate = Functions.PossiblePlate()
# sort chars from left to right based on x position
listOfMatchingChars.sort(key=lambda matchingChar: matchingChar.centerX)
# calcular o ponto central da placa
plateCenterX = (
listOfMatchingChars[0].centerX +
listOfMatchingChars[len(listOfMatchingChars) - 1].centerX) / 2.0
plateCenterY = (
listOfMatchingChars[0].centerY +
listOfMatchingChars[len(listOfMatchingChars) - 1].centerY) / 2.0
plateCenter = plateCenterX, plateCenterY
# calcular largura e altura da placa
plateWidth = int(
(listOfMatchingChars[len(listOfMatchingChars) - 1].boundingRectX +
listOfMatchingChars[len(listOfMatchingChars) -
1].boundingRectWidth -
listOfMatchingChars[0].boundingRectX) * 1.3)
totalOfCharHeights = 0
for matchingChar in listOfMatchingChars:
totalOfCharHeights = totalOfCharHeights + matchingChar.boundingRectHeight
averageCharHeight = totalOfCharHeights / len(listOfMatchingChars)
plateHeight = int(averageCharHeight * 1.5)
# calcular o ângulo de correção da região da placa
opposite = listOfMatchingChars[len(
listOfMatchingChars) - 1].centerY - listOfMatchingChars[0].centerY
hypotenuse = Functions.distanceBetweenChars(
listOfMatchingChars[0],
listOfMatchingChars[len(listOfMatchingChars) - 1])
correctionAngleInRad = math.asin(opposite / hypotenuse)
correctionAngleInDeg = correctionAngleInRad * (180.0 / math.pi)
# empacote o ponto central, a largura e a altura da região da placa e o ângulo de correção na variável de membro retira da placa
possiblePlate.rrLocationOfPlateInScene = (tuple(plateCenter),
(plateWidth, plateHeight),
correctionAngleInDeg)
# obtenha a matriz de rotação para o ângulo de correção calculado
rotationMatrix = cv2.getRotationMatrix2D(tuple(plateCenter),
correctionAngleInDeg, 1.0)
height, width, numChannels = img.shape
# girar a imagem inteira
imgRotated = cv2.warpAffine(img, rotationMatrix, (width, height))
# cortar a imagem / placa detecta
imgCropped = cv2.getRectSubPix(imgRotated, (plateWidth, plateHeight),
tuple(plateCenter))
# copie a imagem da placa cortada na variável de membro aplicável da placa possível
possiblePlate.Plate = imgCropped
# preenche a plates_list com as placas detectadas
if possiblePlate.Plate is not None:
plates_list.append(possiblePlate)
# Desenha um ROI na imagem original
for i in range(0, len(plates_list)):
# finds the four vertices of a rotated rect - it is useful to draw the rectangle.
p2fRectPoints = cv2.boxPoints(
plates_list[i].rrLocationOfPlateInScene)
# roi rectangle colour
rectColour = (0, 255, 0)
cv2.line(imageContours, tuple(p2fRectPoints[0]),
tuple(p2fRectPoints[1]), rectColour, 2)
cv2.line(imageContours, tuple(p2fRectPoints[1]),
tuple(p2fRectPoints[2]), rectColour, 2)
cv2.line(imageContours, tuple(p2fRectPoints[2]),
tuple(p2fRectPoints[3]), rectColour, 2)
cv2.line(imageContours, tuple(p2fRectPoints[3]),
tuple(p2fRectPoints[0]), rectColour, 2)
cv2.line(img, tuple(p2fRectPoints[0]), tuple(p2fRectPoints[1]),
rectColour, 2)
cv2.line(img, tuple(p2fRectPoints[1]), tuple(p2fRectPoints[2]),
rectColour, 2)
cv2.line(img, tuple(p2fRectPoints[2]), tuple(p2fRectPoints[3]),
rectColour, 2)
cv2.line(img, tuple(p2fRectPoints[3]), tuple(p2fRectPoints[0]),
rectColour, 2)
retorno = getCharactres(plates_list[i].Plate)
cv2.waitKey(0)
return retorno
thresh, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
height, width = thresh.shape
imageContours = np.zeros((height, width, 3), dtype=np.uint8)
possibleChars = []
countOfPossibleChars = 0
for i in range(0, len(contours)):
cv2.drawContours(imageContours, contours, i, (255, 255, 255))
possibleChar = Functions.ifChar(contours[i])
if Functions.checkIfChar(possibleChar) is True:
countOfPossibleChars = countOfPossibleChars + 1
possibleChars.append(possibleChar)
imageContours = np.zeros((height, width, 3), np.uint8)
ctrs = []
for char in possibleChars:
ctrs.append(char.contour)
cv2.drawContours(imageContours, ctrs, -1, (255, 255, 255))
plates_list = []
listOfListsOfMatchingChars = []
height, width = thresh.shape
imageContours = np.zeros((height, width, 3), dtype=np.uint8)
possibleChars = []
countOfPossibleChars = 0
for i in range(0, len(contours)):
cv2.drawContours(
imageContours, contours, i,
(255, 255,
255)) # draw contours based on actual found contours of thresh image
possibleChar = Functions.ifChar(
contours[i]
) # retrieve a possible char by the result ifChar class give us
if Functions.checkIfChar(
possibleChar
) is True: # by computing some values (area, width, height, aspect ratio) possibleChars list is being populated
countOfPossibleChars = countOfPossibleChars + 1
possibleChars.append(possibleChar)
imageContours = np.zeros((height, width, 3), np.uint8)
ctrs = []
for char in possibleChars:
ctrs.append(char.contour)
# using values from ctrs to draw new contours
cv2.drawContours(imageContours, ctrs, -1, (255, 255, 255))
#cv2.imshow("contoursPossibleChars", imageContours)
cv2.imwrite(temp_folder + '9 - contoursPossibleChars.png', imageContours)
plates_list = []