def get_grid_img(frame):
image = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
image = cv2.GaussianBlur(image, (11, 11), 0)
image = cv2.adaptiveThreshold(image, 255, cv2.ADAPTIVE_THRESH_MEAN_C | cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV,11,4)
# cv2.imshow("original", image)
# display_image(image)
contours = get_sorted_contours(image)
contour, corners = get_grid_contour(contours)
x, y, width, height = cv2.boundingRect(contour)
average_grid_wh = int((int(width) + int(height)) / 2)
# show outline of selected grid on full image
# grid_outline = draw_grid_outline(frame.copy(), corners)
# cv2.imshow("grid_outline", grid_outline)
if contour is None:
return None, None
warped_grid, transform_matrix_inv = warp(image, corners, average_grid_wh)
return warped_grid, transform_matrix_inv
def large_image_demo(image):
print(image.shape)
cw = 256
ch = 256
h,w = image.shape[:2]
gray = cv.cvtColor(image,cv.COLOR_BGR2GRAY)
for row in range(0,h,ch):
for col in range(0,w,cw):
roi = gray[row:row+ch,col:col+cw]
dst = cv.adaptiveThreshold(roi,255,cv.ADAPTIVE_THRESH_MEAN_C,cv.THRESH_BINARY,25,20)
gray[row:row+ch,col:col+cw] = dst
print(np.std(dst),np.mean(dst))
"""
全局二值化结合方差过滤实现类似自适应二值化的效果
dev = np.std(roi)
if dev<15:
gray[row:row+ch,col:col+cw] = 255
else:
ret,dst = cv.threshold(roi,0,255,cv.THRESH_BINARY|cv.THRESH_OTSU)
gray[row:row+ch,col:col+cw] = dst
"""
cv.imshow("result",gray)
def ImageToBWEdgeDetect(img):
# Load image:
img_class = img.__class__.__name__
if img_class == "str":
img = Image.open(img)
elif img_class == None:
print("Invalid image format")
return None
# Adding white background to trasnaprent images:
working_size = GetImageResize(img, 512, max_ratio=1.5)
img = img.resize(working_size)
img = img.convert("RGBA")
white_bg = Image.new("RGB", img.size, "WHITE")
white_bg.paste(img, (0, 0), img)
img = white_bg.convert("RGB")
# Usin OpenCV to find edges:
# img = cv2.imread(img, 0)
img = cv2.cvtColor(numpy.array(img), cv2.COLOR_RGB2GRAY)
img = cv2.resize(img, working_size, interpolation=cv2.INTER_LINEAR)
img = cv2.medianBlur(img, 5)
img = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 2)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img = Image.fromarray(img)
# img = img.convert("RGBA")
# white_bg = Image.new("RGB", img.size, "WHITE")
# white_bg.paste(img, (0, 0), img)
# img = white_bg.convert("RGB")
# img = img.quantize(colors = 2, method = Image.MAXCOVERAGE)
# img = img.convert(mode = "1", dither = Image.NONE)
#return img.quantize(colors = 2)
return img
# img = ImageToBWEdgeDetect("RobloxImages/5918162200.jpg")
# img.save("grayscale_test.png")
def cartoonizer(img, num_down=2, num_bi=5):
#Params
#num_down = 2 #DOWNSAMPLE STEPS
#num_bi = 5 # BILATERAL FILTERING STEPS
img_c = img
for ix in range(num_down):
img_c = cv2.pyrDown(img) # Pyramid Down : Downsampling
# print(img_c.shape)
for iy in range(num_bi):
img_c = cv2.bilateralFilter(img_c, d=9, sigmaColor=9,
sigmaSpace=7) #Filtering
# print(img_c.shape)
#UPSAMPLING
for ix in range(num_down):
img_c = cv2.pyrUp(img_c) # Pyramid Down : Downsampling
# print(img_c.shape)
#BLUR and Threshold
img_gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY) # GRAY SCALE
img_blur = cv2.medianBlur(img_gray, 7) #MEDIAN BLUR
img_edge = cv2.adaptiveThreshold(img_blur,
255,
cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY,
blockSize=9,
C=2)
img_c = cv2.resize(img_c, (800, 800))
#RGB CONVERSION + BITWISE &
img_edge = cv2.cvtColor(img_edge, cv2.COLOR_GRAY2RGB)
# print(img_c.shape)
# print(img_edge.shape)
img_cartoon = cv2.bitwise_and(img_c, img_edge)
stack = np.hstack([img, img_cartoon])
return stack
def get_text(PDF_file):
# Path of the pdf
# Store all the pages of the PDF in a variable
pages = convert_from_path(PDF_file, 500)
# Counter to store images of each page of PDF to image
image_counter = 1
# Iterate through all the pages stored above
t = ""
for page in pages:
# Declaring filename for each page of PDF as JPG
# For each page, filename will be:
# PDF page 1 -> page_1.jpg
# PDF page 2 -> page_2.jpg
# PDF page 3 -> page_3.jpg
# ....
# PDF page n -> page_n.jpg
filename = "page_" + str(image_counter) + ".jpg"
page.save(filename, 'JPEG')
img = cv2.imread(filename)
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# Apply dilation and erosion to remove some noise
kernel = np.ones((1, 1), np.uint8)
img = cv2.dilate(img, kernel, iterations=1)
img = cv2.erode(img, kernel, iterations=1)
img = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 31, 2)
text = str(
((pytesseract.image_to_string(img, config=custom_oem_psm_config))))
t += text + "\n"
os.remove(filename)
return t
def try_wrap():
img = cv2.imread("datas/images/namecard_01.jpg")
pts = np.float32([[400, 464], [934, 544], [96, 1174], [960, 1328]])
img_wrap = wrapping(img, pts)
img_gray, _, _ = preprocessing(img_wrap)
img_gray2 = cv2.resize(img_gray,
None,
fx=0.5,
fy=0.5,
interpolation=cv2.INTER_CUBIC)
img_th = cv2.adaptiveThreshold(img_gray2, 255,
cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, -10)
# ret,img_th = cv2.threshold(img_sharp,127,255,cv2.THRESH_BINARY)
img_ocr, words = test_pytesseract2.get_OCR2(255 - img_th)
img_gray = cv2.cvtColor(img_gray, cv2.COLOR_GRAY2BGR)
# img_blur = cv2.cvtColor(img_blur, cv2.COLOR_GRAY2BGR)
# img_sharp = cv2.cvtColor(img_sharp, cv2.COLOR_GRAY2BGR)
img_th = cv2.cvtColor(img_th, cv2.COLOR_GRAY2BGR)
final1 = np.hstack((img_wrap, img_gray))
final2 = np.hstack((img_th, img_ocr))
final1 = cv2.resize(final1,
None,
fx=0.5,
fy=0.5,
interpolation=cv2.INTER_CUBIC)
final = np.vstack((final1, final2))
cv2.imshow('img1', final1)
cv2.imshow('img2', final2)
cv2.waitKey(0)
cv2.destroyAllWindows()
def try_wrap2():
img = cv2.imread("datas/images/licenseplate_04.jpg")
img = cv2.resize(img, None, fx=0.5, fy=0.5, interpolation=cv2.INTER_CUBIC)
pts = np.float32([[200, 232], [467, 272], [48, 587], [480, 664]])
img_wrap = wrapping(img, pts)
img_gray, img_blur, img_sharp = preprocessing(img_wrap)
img_th = cv2.adaptiveThreshold(img_sharp, 255,
cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 4)
img_ocr, words = test_pytesseract2.get_OCR2(img_th)
img_gray = cv2.cvtColor(img_gray, cv2.COLOR_GRAY2BGR)
img_blur = cv2.cvtColor(img_blur, cv2.COLOR_GRAY2BGR)
img_sharp = cv2.cvtColor(img_sharp, cv2.COLOR_GRAY2BGR)
img_th = cv2.cvtColor(img_th, cv2.COLOR_GRAY2BGR)
final1 = np.hstack((img_wrap, img_gray, img_blur))
final2 = np.hstack((img_sharp, img_th, img_ocr))
final = np.vstack((final1, final2))
cv2.imshow('img1', img)
# cv2.imshow('img2', final2)
cv2.waitKey(0)
cv2.destroyAllWindows()
def get_binary_thresholded_image(cv2_image):
img = convert_to_gray_image(cv2_image)
img = cv2.medianBlur(img, 5)
img = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 2)
return img
def predict_folder(self, pather):
dirs = os.listdir(pather)
corrects = []
wrongs = []
every = []
for folders in dirs:
fold = os.fsdecode(folders)
if fold.endswith((".DS_Store")):
pass
else:
for filename in os.listdir(pather + "/" + fold):
if filename.endswith((".DS_Store")):
pass
else:
gray = cv2.imread(
pather + "/%s" % fold + "/%s" % filename, 0)
dilated_gray = cv2.dilate(gray,
np.ones((7, 7), np.uint8))
bg_gray = cv2.medianBlur(dilated_gray, 21)
diff_gray = 255 - cv2.absdiff(gray, bg_gray)
norm_gray = diff_gray.copy()
cv2.normalize(diff_gray,
norm_gray,
alpha=0,
beta=255,
norm_type=cv2.NORM_MINMAX,
dtype=cv2.CV_8UC1)
_, thr_gray = cv2.threshold(norm_gray, 253, 0,
cv2.THRESH_TRUNC)
cv2.normalize(thr_gray,
thr_gray,
alpha=0,
beta=255,
norm_type=cv2.NORM_MINMAX,
dtype=cv2.CV_8UC1)
gray1 = cv2.resize(255 - thr_gray, (112, 112),
interpolation=cv2.INTER_AREA)
gray = cv2.adaptiveThreshold(
gray1, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY_INV, 5, 3)
(thresh, gray2) = cv2.threshold(
gray1, 128, 255,
cv2.THRESH_BINARY | cv2.THRESH_OTSU)
img = self.image_manip(gray)
img2 = self.image_manip(gray2)
img = image.img_to_array(img)
img2 = image.img_to_array(img2)
test_img = img.reshape((1, 112, 112, 1))
test_img2 = img2.reshape((1, 112, 112, 1))
test_img = img.astype('float32')
test_img2 = img2.astype('float32')
test_img /= 255
test_img2 /= 255
test_img = np.expand_dims(test_img, axis=0)
test_img2 = np.expand_dims(test_img2, axis=0)
prediction = self.model.predict(test_img)
prediction2 = self.model.predict(test_img2)
x1 = np.argmax(prediction)
x2 = np.argmax(prediction2)
c = prediction[0][x1]
v = prediction2[0][x2]
b = prediction[0][x2]
n = prediction2[0][x1]
k = (c + b) / 2
j = (v + n) / 2
if k > j:
classname = x1
if j > k:
classname = x2
print(filename + "prediction: ", classname)
# This list is to keep count of how many images are being classified
every.append(filename)
# This is for benchmark tests during development and demonstrating larger sets of images,
# This requires the images to be labeled to see how many predictions were right or wrong,
# THIS IS OPTIONAL, if you dont label your images you wont get a list of prediction accuracy,
# But the predictions will be printed as outputs next to each image filename.
if filename.startswith(str(classname)):
corrects.append(filename)
else:
wrongs.append(filename)
co = 0
w = 0
e = 0
for i in every:
e += 1
for i in corrects:
co += 1
for i in wrongs:
w += 1
print("all numbers")
print(e)
print("right guess")
print(co)
print("wrong guess")
print(w)
print(wrongs)
print("Running: " + "ALnet-2.0")
def DetectColorGrids(Color_checker_image):
img_gray = cv.cvtColor(Color_checker_image,cv.COLOR_BGR2GRAY)
img_denoise = cv.fastNlMeansDenoising(img_gray,10,7,21)
img_threshold = cv.adaptiveThreshold(img_denoise,255,cv.ADAPTIVE_THRESH_MEAN_C,cv.THRESH_BINARY,21,3)
kernel = np.ones((3,3), np.uint8)
img_eroded = cv.erode(img_threshold,kernel)
small_grid_size_range = GetSmallGridSizeRange(Color_checker_image)
contours, _hierarchy = cv.findContours(img_eroded, cv.RETR_TREE, cv.CHAIN_APPROX_NONE)
color_grid_contours = FilterColorGrid(contours, small_grid_size_range)
centroid_positions = []
centroid_positions = FindContourCenter(color_grid_contours)
centroids_no_overlap = []
centroids_no_overlap = FilterOutOverlapPoint(color_grid_contours,centroid_positions)
#small to big, base on y axis
centroids_no_overlap = sorted(centroids_no_overlap,key=lambda centroid: centroid[1])
rad_top = math.atan((centroids_no_overlap[1][1] - centroids_no_overlap[0][1]) /
(centroids_no_overlap[1][0] - centroids_no_overlap[0][0]))
angle_top = rad_top * 180 / PI
centroids_no_overlap_size = len(centroids_no_overlap)
rad_bottom = math.atan((centroids_no_overlap[centroids_no_overlap_size - 1][1] - centroids_no_overlap[centroids_no_overlap_size - 2][1]) /
(centroids_no_overlap[centroids_no_overlap_size - 1][0] - centroids_no_overlap[centroids_no_overlap_size - 2][0]))
angle_bottom = rad_bottom * 180 / PI
angle_subtract = abs(angle_top - angle_bottom)
if angle_subtract >= ANGLE_SUBTRACT_MAX_ENDURANCE:
orientation_left_top = centroids_no_overlap[0]
orientation_right_bottom=[0.0,0.0]
for centroid in centroids_no_overlap:
if centroid[0] < orientation_left_top[0]:
orientation_left_top[0] = centroid[0]
if centroid[1] < orientation_left_top[1]:
orientation_left_top[1] = centroid[1]
if centroid[0] > orientation_right_bottom[0]:
orientation_right_bottom[0] = centroid[0]
if centroid[1] > orientation_right_bottom[1]:
orientation_right_bottom[1] = centroid[1]
translation=[0.0,0.0]
translation[0] = (orientation_right_bottom[0] - orientation_left_top[0]) / (TOTAL_COLUMNS - 1)
translation[1] = (orientation_right_bottom[1] - orientation_left_top[1]) / (TOTAL_ROWS - 1)
grids_position = np.zeros((TOTAL_ROWS, TOTAL_COLUMNS, 2), dtype=np.int)
row_count = 0
col_count = 0
while(row_count < TOTAL_ROWS):
col_count = 0
while(col_count < TOTAL_COLUMNS):
grids_position[row_count][col_count][0] = orientation_left_top[0] + translation[0]*col_count
grids_position[row_count][col_count][1] = orientation_left_top[1] + translation[1]*row_count
col_count += 1
row_count += 1
return grids_position
else:
angle_avg = (angle_top + angle_bottom) / 2
img_rotation = np.zeros(img_gray.shape, dtype=np.uint8)
rows, cols = img_gray.shape
center = [cols / 2, rows / 2]
rotation_mat = cv.getRotationMatrix2D((center[0],center[1]),angle_avg,1.0)
img_rotation = cv.warpAffine(img_eroded,rotation_mat,(cols,rows))
contours, _hierarchy = cv.findContours(img_rotation, cv.RETR_TREE, cv.CHAIN_APPROX_NONE)
color_grid_contours = FilterColorGrid(contours, small_grid_size_range)
centroid_positions = []
centroid_positions = FindContourCenter(color_grid_contours)
orientation_left_top = centroid_positions[0]
orientation_right_bottom=[0.0,0.0]
# contour_arr_all = np.zeros(img_gray.shape, dtype=np.uint8)
i=0
for centroid in centroid_positions:
if centroid[0] < orientation_left_top[0]:
orientation_left_top[0] = centroid[0]
if centroid[1] < orientation_left_top[1]:
orientation_left_top[1] = centroid[1]
if centroid[0] > orientation_right_bottom[0]:
orientation_right_bottom[0] = centroid[0]
if centroid[1] > orientation_right_bottom[1]:
orientation_right_bottom[1] = centroid[1]
i+=1
translation=[0.0,0.0]
translation[0] = (orientation_right_bottom[0] - orientation_left_top[0]) / (TOTAL_COLUMNS - 1)
translation[1] = (orientation_right_bottom[1] - orientation_left_top[1]) / (TOTAL_ROWS - 1)
grid_coordinate = np.zeros((TOTAL_ROWS, TOTAL_COLUMNS, 2), dtype=np.int)
row_count = 0
col_count = 0
while(row_count < TOTAL_ROWS):
col_count = 0
while(col_count < TOTAL_COLUMNS):
grid_coordinate[row_count][col_count][0] = orientation_left_top[0] + translation[0]*col_count
grid_coordinate[row_count][col_count][1] = orientation_left_top[1] + translation[1]*row_count
col_count += 1
row_count +=1
grids_position = np.zeros((TOTAL_ROWS, TOTAL_COLUMNS, 2), dtype=np.int)
temp_coordinate = [0,0]
row_count = 0
col_count = 0
while(row_count < TOTAL_ROWS):
col_count = 0
while(col_count < TOTAL_COLUMNS):
temp_coordinate = GetRotationPoint((grid_coordinate[row_count][col_count][0],grid_coordinate[row_count][col_count][1]), cols, rows, -1 * angle_avg)
grids_position[row_count][col_count][0] = temp_coordinate[0]
grids_position[row_count][col_count][1] = temp_coordinate[1]
col_count += 1
row_count += 1
return grids_position
from cv2 import cv2
'''
Применить на практике бинаризацию всех типов для изображений.
Подобрано по 2 примера изображений на каждый тип, иллюстрирующие разницу результатов использованных методов.
'''
"""Для дорог лучше подходит метод TRESH_TOZERO, т.к. картинка монотонна, а полученное изображение четчё при ручном подборе трешхолда"""
img = cv2.imread('Doroga1.png', cv2.IMREAD_GRAYSCALE)
treshold, img1 = cv2.threshold(img,
thresh=200,
maxval=255,
type=cv2.THRESH_TOZERO)
treshold, img2 = cv2.threshold(img, 200, 255, cv2.THRESH_OTSU)
img3 = cv2.adaptiveThreshold(img,
maxValue=255,
adaptiveMethod=cv2.ADAPTIVE_THRESH_MEAN_C,
thresholdType=cv2.THRESH_BINARY,
blockSize=11,
C=-4)
img_2 = cv2.imread('Doroga2.png', cv2.IMREAD_GRAYSCALE)
treshold, img4 = cv2.threshold(img_2, 200, 255, cv2.THRESH_TOZERO)
treshold, img5 = cv2.threshold(img_2, 200, 255, cv2.THRESH_OTSU)
img6 = cv2.adaptiveThreshold(img_2, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY, 3, -4)
"""Для сканов лучше подходит как адаптивный метод, так и TRESH_OTSU, т.к. скан документа не монотонный"""
# img = cv2.imread('Scan1.png', cv2.IMREAD_GRAYSCALE)
# treshold, img1 = cv2.threshold(img, 128, 255, cv2.THRESH_TOZERO)
# treshold, img2 = cv2.threshold(img, 128, 255, cv2.THRESH_OTSU)
# img3 = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 31, 30)
# from googletrans import Translator
# translator = Translator()
# hindi_sent = "नमस्कार, धृति!"
# translation = translator.translate(hindi_sent,dest="en")
# translated_sentence = (f"{translation.text}")
# print(translated_sentence)
from cv2 import cv2
import numpy as np
import pytesseract
# from google.colab import cv2_imshow
img = cv2.imread('test2.png')
img = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
reduced_noise = cv2.adaptiveThreshold(img,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,cv2.THRESH_BINARY,11,12)
# cv2.imshow("",reduced_noise)
# cv2.waitKey(0)
print("\nImage is loaded succesfully")
text=pytesseract.image_to_string(reduced_noise,lang='eng')
print("The text is :\n",text)
def decide(filename):
# reading image to grayscale
gray = cv2.imread(f"/content/mnistcnn/digits/{filename}", 0)
# dilate img
dilated_gray = cv2.dilate(gray, np.ones((7, 7), np.uint8))
# median blur the dilated img to further suppress detail
bg_gray = cv2.medianBlur(dilated_gray, 21)
# calculate difference between gray(original img) and bg_gray.
# identical pixels will be black(close to 0 difference), the digits details will be white(large difference).
diff_gray = 255 - cv2.absdiff(gray, bg_gray)
# normalize the image, so that we use the full dynamic range.
norm_gray = diff_gray.copy()
cv2.normalize(diff_gray,
norm_gray,
alpha=0,
beta=255,
norm_type=cv2.NORM_MINMAX,
dtype=cv2.CV_8UC1)
# now the image is still gray in some areas, so we truncate that away and re-normalize the image.
_, thr_gray = cv2.threshold(norm_gray, 253, 0, cv2.THRESH_TRUNC)
cv2.normalize(thr_gray,
thr_gray,
alpha=0,
beta=255,
norm_type=cv2.NORM_MINMAX,
dtype=cv2.CV_8UC1)
# inverting color and rezising image to 112x112
gray1 = cv2.resize(255 - thr_gray, (112, 112),
interpolation=cv2.INTER_AREA)
# we have variables gray and gray2 to differentiate between two ways of thresholding.
# Alnet-2.0 will make predictions on both of these, and combine the two to get our resulting classification.
gray = cv2.adaptiveThreshold(gray1, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY_INV, 7, 3)
(thresh, gray2) = cv2.threshold(gray1, 128, 255,
cv2.THRESH_BINARY | cv2.THRESH_OTSU)
# converting PIL image to numpy array
img = image.img_to_array(gray)
img2 = image.img_to_array(gray2)
# reshaping to 1 channel (its only 1 image), 112x112 shape, and 1 color channel (grayscale)
test_img = img.reshape((1, 112, 112, 1))
test_img2 = img2.reshape((1, 112, 112, 1))
# change datatype from uint8 to float32
test_img = img.astype('float32')
test_img2 = img2.astype('float32')
# normalizing pixels in the decimal range 0 to 1
test_img /= 255
test_img2 /= 255
# extend axis at index[0] to ready for prediction
test_img = np.expand_dims(test_img, axis=0)
test_img2 = np.expand_dims(test_img2, axis=0)
# get network output (10 nodes)
# prediction_list is prediction of gray
# prediction2_list is prediction of gray2
prediction_list = model.predict(test_img)
prediction2_list = model.predict(test_img2)
return prediction_list, prediction2_list, test_img, test_img2
from cv2 import cv2 as cv
import numpy as np
img = cv.imread('cafe.jpg', 0)
img2 = cv.imread('olho.PNG', 0)
metodo = cv.THRESH_BINARY_INV # objeto de interesse na cor branca
#metodo = cv.THRESH_BINARY # objeto de interesse na cor preta
ret, imgBin = cv.threshold(img, 200, 255, metodo) # pixels com valores entre o limiar sao considerado como objeto de interesse
img_gaussian = cv.medianBlur(img2, 7)
#diferentes valores de limiar para cada regiao da imagem
#obtem melhor resultado considerando constraste
mean = cv.adaptiveThreshold(img_gaussian, 255, cv.ADAPTIVE_THRESH_MEAN_C, metodo, 11, 5)
gaussian = cv.adaptiveThreshold(img_gaussian, 255, cv.ADAPTIVE_THRESH_GAUSSIAN_C, metodo, 11, 2)
cv.imshow('original', img)
cv.imshow('imgBin', imgBin)
cv.imshow('mean', mean)
cv.imshow('gaussian', gaussian)
cv.waitKey(0)
cv.destroyAllWindows()
from cv2 import cv2 as cv
def rescaleFrame(frame, scale=0.75):
width = int(frame.shape[1] * scale)
height = int(frame.shape[0] * scale)
dimensions = (width,height)
return cv.resize(frame, dimensions, interpolation=cv.INTER_AREA)
#getting image
init = cv.imread('imagen/wppstalkerlarge.jpg')
img = rescaleFrame(init, 0.4)
cv.imshow('stalker', img)
gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
cv.imshow('GRay', gray)
#Simple Thresholding
threshold, thresh = cv.threshold(gray, 100, 255, cv.THRESH_BINARY)
cv.imshow('threshImaged', thresh)
threshold, threshInv = cv.threshold(gray, 100, 255, cv.THRESH_BINARY_INV)
cv.imshow('threshImagedInverse', threshInv)
#Adaptive Thresholding
adaptiveThresh = cv.adaptiveThreshold(gray, 255, cv.ADAPTIVE_THRESH_MEAN_C, cv.THRESH_BINARY, 11, 5)
cv.imshow('adaptive thresholding', adaptiveThresh)
cv.waitKey(0)
def videoCap():
# pytesseract.pytesseract.tesseract_cmd = "C:/Program Files/Tesseract-OCR/tesseract"
cap = cv2.VideoCapture(
'C:/Users/jeawa/Desktop/project/LPR_Project/project/video/test_car.mp4'
)
count = 0
before_chars = ""
while True:
el = cv2.getTickCount()
fps = cap.get(5)
ret, img_ori = cap.read()
height, width, channel = img_ori.shape # 높이, 너비, 채널 확보
gray = cv2.cvtColor(img_ori, cv2.COLOR_BGR2GRAY)
# img_ori = cv2.imread('LPR_Project/project/image/3.jpg') # 이미지 불러오기
# gray = cv2.cvtColor(img_ori, cv2.COLOR_BGR2GRAY)
img_blurred = cv2.GaussianBlur(gray, ksize=(3, 3), sigmaX=0) # 노이즈 제거
img_thresh = cv2.adaptiveThreshold(
img_blurred,
maxValue=255.0,
adaptiveMethod=cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
thresholdType=cv2.THRESH_BINARY_INV,
blockSize=19, # 12 통과 15,20
C=9)
contours, _ = cv2.findContours( # 윤곽선을 찾기
img_thresh,
mode=cv2.RETR_LIST, #
method=cv2.CHAIN_APPROX_TC89_KCOS #
)
temp_result = np.zeros((height, width, channel), dtype=np.uint8)
cv2.drawContours(
temp_result, # 원본이미지
contours=contours, # contours 정보
contourIdx=-1, # -1 : 전체
color=(255, 255, 255),
thickness=1)
contours_dict = []
for contour in contours:
x, y, w, h = cv2.boundingRect(contour) # 컴투어를 감싸는 사각형
# cv2.rectangle(
# temp_result,
# pt1=(x, y),
# pt2=(x+w, y+h),
# color=(255, 255, 255),
# thickness=1
# )
contours_dict.append({
'contour': contour,
'x': x,
'y': y,
'w': w,
'h': h,
'cx': x + (w / 2), # 중심 좌표
'cy': y + (h / 2)
})
MIN_AREA, MAX_AREA = 100, 1000 # boundingRect(사각형)의 최소넓이
MIN_WIDTH, MIN_HEIGHT = 2, 8 # boundingRect의 최소 넓이, 높이 2, 8
MIN_RATIO, MAX_RATIO = 0.3, 0.9 # boundingRect의 가로 세로 비율
possible_contours = [] # 위의 조건의 만족하는 사각형
cnt = 0
for d in contours_dict:
area = d['w'] * d['h'] # 가로 * 세로 = 면적
ratio = d['w'] / d['h'] # 가로 / 세로 = 비율
if MIN_AREA < area < MAX_AREA and d['w'] > MIN_WIDTH and d[
'h'] > MIN_HEIGHT and MIN_RATIO < ratio < MAX_RATIO:
d['idx'] = cnt # 조건의 맞는 값을 idx에 저장한다.
cnt += 1
possible_contours.append(d) # possible_contour를 업데이트한다.
# temp_result = np.zeros((height, width, channel), dtype=np.uint8)
for d in possible_contours:
# cv2.drawContours(temp_result, d['contour'], -1, (255, 255, 255))
cv2.rectangle(temp_result,
pt1=(d['x'], d['y']),
pt2=(d['x'] + d['w'], d['y'] + d['h']),
color=(0, 0, 255),
thickness=1)
MAX_DIAG_MULTIPLYER = 4 # 대각선의 5배 안에 있어야함
MAX_ANGLE_DIFF = 12.0 # 12.0 #세타의 최대값 12, 0.7, 0.5, 0.6, 3
MAX_AREA_DIFF = 0.3 # 0.5 #면적의 차이
MAX_WIDTH_DIFF = 0.5 # 너비차이
MAX_HEIGHT_DIFF = 0.6 # 높이차이
MIN_N_MATCHED = 4 # 3 #위의 조건이 3개 미만이면 뺀다
def find_chars(contour_list): #
matched_result_idx = [] # idx 값 저장
for d1 in contour_list:
matched_contours_idx = []
for d2 in contour_list:
if d1['idx'] == d2['idx']: # d1 과 d2가 같으면 컨티뉴
continue
dx = abs(d1['cx'] - d2['cx'])
dy = abs(d1['cy'] - d2['cy'])
diagonal_length1 = np.sqrt(d1['w']**2 + d1['h']**2)
distance = np.linalg.norm(
np.array([d1['cx'], d1['cy']]) -
np.array([d2['cx'], d2['cy']])) # 대각선 길이
if dx == 0:
angle_diff = 90 # 0일 때 각도 90도 (예외처리)
else:
angle_diff = np.degrees(np.arctan(dy / dx)) # 세타 구하기
area_diff = abs(d1['w'] * d1['h'] - d2['w'] * d2['h']) / (
d1['w'] * d1['h']) # 면적 비율
width_diff = abs(d1['w'] - d2['w']) / d1['w'] # 너비비율
height_diff = abs(d1['h'] - d2['h']) / d1['h'] # 높이비율
# 조건들
if distance < diagonal_length1 * MAX_DIAG_MULTIPLYER \
and angle_diff < MAX_ANGLE_DIFF and area_diff < MAX_AREA_DIFF \
and width_diff < MAX_WIDTH_DIFF and height_diff < MAX_HEIGHT_DIFF:
# d2만 넣었기 때문에 마지막으로 d1을 넣은다
matched_contours_idx.append(d2['idx'])
# append this contour
matched_contours_idx.append(d1['idx'])
if len(matched_contours_idx) < MIN_N_MATCHED: # 3개 이하이면 번호판 x
continue
matched_result_idx.append(matched_contours_idx) # 최종 후보군
unmatched_contour_idx = [] # 최종후보군이 아닌 애들
for d4 in contour_list:
if d4['idx'] not in matched_contours_idx:
unmatched_contour_idx.append(d4['idx'])
unmatched_contour = np.take(possible_contours,
unmatched_contour_idx)
# recursive
recursive_contour_list = find_chars(unmatched_contour)
for idx in recursive_contour_list:
matched_result_idx.append(idx) # 번호판 이외의 값을 재정의
break
return matched_result_idx
result_idx = find_chars(possible_contours)
matched_result = []
for idx_list in result_idx:
matched_result.append(np.take(possible_contours, idx_list))
# visualize possible contours
temp_result = np.zeros((height, width, channel), dtype=np.uint8)
for r in matched_result:
for d in r:
cv2.rectangle(temp_result,
pt1=(d['x'], d['y']),
pt2=(d['x'] + d['w'], d['y'] + d['h']),
color=(0, 0, 255),
thickness=1)
PLATE_WIDTH_PADDING = 1.2 # 1.3
PLATE_HEIGHT_PADDING = 1.5 # 1.5
MIN_PLATE_RATIO = 3
MAX_PLATE_RATIO = 7 #10
plate_imgs = []
plate_infos = []
for i, matched_chars in enumerate(matched_result):
sorted_chars = sorted(matched_chars,
key=lambda x: x['cx']) # x방향으로 순차적으로 정렬
plate_cx = (sorted_chars[0]['cx'] +
sorted_chars[-1]['cx']) / 2 # 센터 좌표 구하기
plate_cy = (sorted_chars[0]['cy'] + sorted_chars[-1]['cy']) / 2
plate_width = (sorted_chars[-1]['x'] + sorted_chars[-1]['w'] -
sorted_chars[0]['x']) * PLATE_WIDTH_PADDING # 너비
# sum_height = 0
# for d in sorted_chars:
# sum_height += d['h']
# plate_height = int(sum_height / len(sorted_chars)
# * PLATE_HEIGHT_PADDING) # 높이
plate_height = (sorted_chars[-1]['y'] - sorted_chars[0]['y'] +
sorted_chars[-1]['h']) * PLATE_HEIGHT_PADDING
#
triangle_height = sorted_chars[-1]['cy'] - sorted_chars[0]['cy']
triangle_hypotenus = np.linalg.norm(
np.array([sorted_chars[0]['cx'], sorted_chars[0]['cy']]) -
np.array([sorted_chars[-1]['cx'], sorted_chars[-1]['cy']]))
angle = np.degrees(np.arcsin(triangle_height / triangle_hypotenus))
rotation_matrix = cv2.getRotationMatrix2D(center=(plate_cx,
plate_cy),
angle=angle,
scale=1.0)
# 회전
img_rotated = cv2.warpAffine(img_thresh,
M=rotation_matrix,
dsize=(width, height))
img_cropped = cv2.getRectSubPix(img_rotated,
patchSize=(int(plate_width),
int(plate_height)),
center=(int(plate_cx),
int(plate_cy)))
if img_cropped.shape[1] / img_cropped.shape[
0] < MIN_PLATE_RATIO or img_cropped.shape[
1] / img_cropped.shape[
0] < MIN_PLATE_RATIO > MAX_PLATE_RATIO:
continue
plate_imgs.append(img_cropped)
plate_infos.append({
'x': int(plate_cx - plate_width / 2),
'y': int(plate_cy - plate_height / 2),
'w': int(plate_width),
'h': int(plate_height)
})
for i, plate_img in enumerate(plate_imgs):
plate_img = cv2.resize(plate_img, dsize=(0, 0), fx=1.6, fy=1.6)
_, plate_img = cv2.threshold(plate_img,
thresh=0.0,
maxval=255.0,
type=cv2.THRESH_BINARY
| cv2.THRESH_OTSU)
# plt.imshow(plate_img, cmap='gray')
# plt.show()
# find contours again (same as above)
contours, _ = cv2.findContours(plate_img,
mode=cv2.RETR_LIST,
method=cv2.CHAIN_APPROX_SIMPLE)
plate_min_x, plate_min_y = plate_img.shape[1], plate_img.shape[
0]
plate_max_x, plate_max_y = 0, 0
for contour in contours:
x, y, w, h = cv2.boundingRect(contour)
area = w * h
ratio = w / h
if area > MIN_AREA \
and w > MIN_WIDTH and h > MIN_HEIGHT \
and MIN_RATIO < ratio < MAX_RATIO:
if x < plate_min_x:
plate_min_x = x
if y < plate_min_y:
plate_min_y = y
if x + w > plate_max_x:
plate_max_x = x + w
if y + h > plate_max_y:
plate_max_y = y + h
img_result = plate_img[plate_min_y:plate_max_y,
plate_min_x:plate_max_x]
img_result = cv2.GaussianBlur(img_result, ksize=(3, 3), sigmaX=0)
_, img_result = cv2.threshold(img_result,
thresh=0.0,
maxval=255.0,
type=cv2.THRESH_BINARY
| cv2.THRESH_OTSU)
img_result = cv2.copyMakeBorder(img_result,
top=10,
bottom=10,
left=20,
right=10,
borderType=cv2.BORDER_CONSTANT,
value=(0, 0, 0))
chars = pytesseract.image_to_string(img_result,
lang='kor',
config='--psm 7 --oem 0')
result_chars = ''
has_digit = False
# <= ord('힣')
for c in chars:
if \
+ ord('가') == ord(c) or \
+ ord('나') == ord(c) or \
+ ord('다') == ord(c) or \
+ ord('라') == ord(c) or \
+ ord('마') == ord(c) or \
+ ord('거') == ord(c) or \
+ ord('너') == ord(c) or \
+ ord('더') == ord(c) or \
+ ord('러') == ord(c) or \
+ ord('머') == ord(c) or \
+ ord('버') == ord(c) or \
+ ord('서') == ord(c) or \
+ ord('어') == ord(c) or \
+ ord('저') == ord(c) or \
+ ord('고') == ord(c) or \
+ ord('노') == ord(c) or \
+ ord('도') == ord(c) or \
+ ord('로') == ord(c) or \
+ ord('모') == ord(c) or \
+ ord('보') == ord(c) or \
+ ord('소') == ord(c) or \
+ ord('오') == ord(c) or \
+ ord('조') == ord(c) or \
+ ord('구') == ord(c) or \
+ ord('누') == ord(c) or \
+ ord('두') == ord(c) or \
+ ord('루') == ord(c) or \
+ ord('무') == ord(c) or \
+ ord('부') == ord(c) or \
+ ord('수') == ord(c) or \
+ ord('우') == ord(c) or \
+ ord('주') == ord(c) or \
+ ord('아') == ord(c) or \
+ ord('바') == ord(c) or \
+ ord('사') == ord(c) or \
+ ord('자') == ord(c) or \
+ ord('배') == ord(c) or \
+ ord('하') == ord(c) or \
+ ord('허') == ord(c) or \
+ ord('호') == ord(c) or c.isdigit():
if c.isdigit():
has_digit = True
result_chars += c
# print(result_chars)
# print(len(result_chars))
if result_chars == "":
pass
else:
if before_chars == result_chars and 6 < len(result_chars) < 9:
count += 1
else:
before_chars = result_chars
count = 0
if count == 1:
print(result_chars)
count = 0
before_chars = ""
# print(fps + " : 프레임 영상입니다.")
print(fps)
a = np.hstack((img_ori, temp_result))
cv2.imshow("Go", a)
if cv2.waitKey(42) == ord('q'):
break
e2 = cv2.getTickCount()
time = (e2 - el) / cv2.getTickFrequency()
print(time)
# print("1프레임 : 1초당 처리속도는 " + time + "입니다.")
cap.release()
cv2.destroyAllWindows()
from cv2 import cv2
import numpy as np
from matplotlib import pyplot as plt
img = cv2.imread('./images/sudoku.png', 0)
img = cv2.medianBlur(img, 5)
ret, th1 = cv2.threshold(img, 127, 255, cv2.THRESH_BINARY)
th2 = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY, 11, 2)
th3 = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 2)
titles = [
'Original Image', 'Global Thresholding (v = 127)',
'Adaptive Mean Thresholding', 'Adaptive Gaussian Thresholding'
]
images = [img, th1, th2, th3]
for i in range(4):
plt.subplot(2, 2, i + 1), plt.imshow(images[i], 'gray')
plt.title(titles[i])
plt.xticks([]), plt.yticks([])
plt.show()
def capture_mrz(window: sg.Window,
camera_id: int) -> Tuple[List[str], Image.Image]:
"""
Capture the MRZ by using OCR and the camera footage.
:returns: MRZ lines in a list
"""
cap = cv2.VideoCapture(camera_id)
tess_api = PyTessBaseAPI(init=False, psm=PSM.SINGLE_BLOCK_VERT_TEXT)
tess_api.InitFull(
# https://github.com/DoubangoTelecom/ultimateMRZ-SDK/tree/master/assets/models
path="text_detection",
lang="mrz",
variables={
"load_system_dawg": "false",
"load_freq_dawg": "false",
"tessedit_char_whitelist": "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ<",
},
)
# mrz_list: List[List[str]] = []
pool = ThreadPool(processes=1)
ocr_running = False
while True:
_, frame = cap.read()
mrz = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
mrz = cv2.adaptiveThreshold(mrz, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY, 21, 10)
# mrz = cv2.adaptiveThreshold(mrz, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY,3,2)
# mrz = cv2.GaussianBlur(mrz, (5,5), 0)
# _, mrz = cv2.threshold(mrz, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
# mrz = cv2.GaussianBlur(mrz, (5,5), 0)
# mrz = cv2.medianBlur(mrz, 3)
frame_shown = copy.deepcopy(mrz)
width = 320
height = int(frame_shown.shape[0] * (320 / frame_shown.shape[1]))
frame_shown = cv2.resize(frame_shown, (width, height))
alpha = 0.8
frame_overlay = add_mrz_overlay(copy.deepcopy(frame_shown),
"<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<", 3,
0.9, False)
frame_overlay = add_mrz_overlay(
frame_overlay, "<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<", 2,
0.9, True)
cv2.addWeighted(frame_shown, alpha, frame_overlay, 1 - alpha, 0,
frame_shown)
imgbytes = cv2.imencode(".png", frame_shown)[1].tobytes()
window.write_event_value("-SHOW MRZ-", [imgbytes])
mrz = Image.fromarray(mrz)
if not ocr_running:
checked_frame = Image.fromarray(frame[:, :, ::-1])
tess_api.SetImage(mrz)
async_result = pool.apply_async(tess_api.GetUTF8Text)
ocr_running = True
if async_result.ready():
ocr_running = False
mrz_text = async_result.get()
result = parse_mrz_ocr(mrz_text)
if result is not None:
break
# if result and len(mrz_list) < 3:
# mrz_list.append(result)
# elif not result:
# mrz_list = []
# else:
# if all(x == mrz_list[0] for x in mrz_list):
# break
# When everything done, release the capture
cap.release()
# cv2.destroyAllWindows()
tess_api.End()
# return mrz_list[0]
window.write_event_value("-HIDE MRZ-", "")
return (result, checked_frame)
from cv2 import cv2
import numpy as np
img = cv2.imread('sudoku.jpg')
# NOTE: If not converted to grayscale the following error is coming:
# error: (-215:Assertion failed) src.type() == CV_8UC1 in function 'cv::adaptiveThreshold'
# The problem is that you are trying to use adaptive thresholding to an image that is not in greyscale.
# And the function only works with a greyscale images.
# So you have to convert your image to a greyscale format as it is described in documentation.
# They read the image in a greyscale format with: img = cv2.imread('dave.jpg',0). You can also convert it
# to greyscale
img_grey = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# SYNTAX: adaptiveThreshold(src, maxValue, adaptiveMethod, thresholdType, blockSize, C)
# NOTE: More the value of 'C' the more white the image becomes
adap_img = cv2.adaptiveThreshold(img_grey, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY, 11, 1) # MEAN
adap_img1 = cv2.adaptiveThreshold(img_grey, 255,
cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 2) # GAUSSIAN
cv2.imshow('Original Image', img)
cv2.imshow('Adaptive_Mean Image', adap_img)
cv2.imshow('Adaptive_Gaussian Image', adap_img1)
if cv2.waitKey(0) == 27:
cv2.destroyAllWindows()
def predict_chosen(self):
print("")
# selecting image to predict
filename = askopenfilename()
# reading image to grayscale
gray = cv2.imread(filename, 0)
# dilate img
dilated_gray = cv2.dilate(gray, np.ones((7, 7), np.uint8))
# median blur the dilated img to further suppress detail
bg_gray = cv2.medianBlur(dilated_gray, 21)
# calculate difference between gray(original img) and bg_gray.
# identical pixels will be black(close to 0 difference), the digits details will be white(large difference).
diff_gray = 255 - cv2.absdiff(gray, bg_gray)
# normalize the image, so that we use the full dynamic range.
norm_gray = diff_gray.copy()
cv2.normalize(diff_gray,
norm_gray,
alpha=0,
beta=255,
norm_type=cv2.NORM_MINMAX,
dtype=cv2.CV_8UC1)
# now the image is still gray in some areas, so we truncate that away and re-normalize the image.
_, thr_gray = cv2.threshold(norm_gray, 253, 0, cv2.THRESH_TRUNC)
cv2.normalize(thr_gray,
thr_gray,
alpha=0,
beta=255,
norm_type=cv2.NORM_MINMAX,
dtype=cv2.CV_8UC1)
# inverting color and rezising image to 112x112
gray1 = cv2.resize(255 - thr_gray, (112, 112),
interpolation=cv2.INTER_AREA)
# we have variables gray and gray2 to differentiate between two ways of thresholding.
# Alnet-2.0 will make predictions on both of these, and combine the two to get our resulting classification.
gray = cv2.adaptiveThreshold(gray1, 255,
cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY_INV, 5, 3)
(thresh, gray2) = cv2.threshold(gray1, 128, 255,
cv2.THRESH_BINARY | cv2.THRESH_OTSU)
img = self.image_manip(gray)
img2 = self.image_manip(gray2)
# converting PIL image to numpy array
img = image.img_to_array(img)
img2 = image.img_to_array(img2)
# reshaping to 1 channel (its only 1 image), 112x112 shape, and 1 color channel (grayscale)
test_img = img.reshape((1, 112, 112, 1))
test_img2 = img2.reshape((1, 112, 112, 1))
# change datatype from uint8 to float32
test_img = img.astype('float32')
test_img2 = img2.astype('float32')
# normalizing pixels in the decimal range 0 to 1
test_img /= 255
test_img2 /= 255
# extend axis at index[0] to ready for prediction
test_img = np.expand_dims(test_img, axis=0)
test_img2 = np.expand_dims(test_img2, axis=0)
# get network output (10 nodes)
# prediction_list is prediction of gray
# prediction2_list is prediction of gray2
prediction_list = self.model.predict(test_img)
prediction2_list = self.model.predict(test_img2)
# x1 and x2 are indexes for elements (what digit is guessed) of highest value for images gray and gray2
x1 = np.argmax(prediction_list)
x2 = np.argmax(prediction2_list)
# c is the highest value in prediction_list
# v is the highest value in prediction2_list
c = prediction_list[0][x1]
v = prediction2_list[0][x2]
# b is the value of prediction_list at index of prediction2_list highest value
# n is the value of prediction2_list at index of predition_list highest value
b = prediction_list[0][x2]
n = prediction2_list[0][x1]
# this helps us decide what prediction is most likely to be correct, by comparing the values below.
k = (c + b) / 2
j = (v + n) / 2
if k > j:
classname = x1
if j > k:
classname = x2
# reshaping again for showing image to user
test_img = test_img.reshape((112, 112))
test_img2 = test_img2.reshape((112, 112))
# lines 181-195 displays to the user gray and gray2, values of k and j, and the estimated classname for the original image chosen
fig = plt.figure(figsize=(8, 8))
columns = 2
rows = 1
plt.gray()
fig.add_subplot(rows, columns,
1).set_title("adaptive_Threshold\n k = {}".format(k))
plt.imshow(test_img)
fig.add_subplot(rows, columns,
2).set_title("Threshold\n j = {}".format(j))
plt.imshow(test_img2)
if k > j:
fig.suptitle("Image shows a {}\n k > j".format(classname))
if j > k:
fig.suptitle("Image shows a {}\n j > k".format(classname))
plt.show()
print("\nadaptive_Threshold prediction:\n", prediction_list, "\n")
print("Threshold prediction:\n", prediction2_list)
print("\nRunning: " + "Alnet-2.0.")
# # cap.set(cv2.CAP_PROP_BUFFERSIZE, 3);
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, height)
ret, original_frame = cap.read()
out.write(original_frame)
#we will use this later
gray = cv2.cvtColor(original_frame, cv2.COLOR_BGR2GRAY)
# Blurring the unwanted stuff
blur = cv2.GaussianBlur(gray, (9, 9), 0)
# Adaptive threshold for to obtain a Binary Image from the frame
# followed by Hough lines probablistic transform to highlight the straight line proposals
# https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_thresholding/py_thresholding.html
adaptive = cv2.adaptiveThreshold(blur, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 2)
processed = cv2.bitwise_not(adaptive, adaptive)
# processed_not_dilated = processed.cop
# Dilate the image to increase the size of the grid lines.
kernel = np.array([[0., 1., 0.], [1., 1., 1.], [0., 1., 0.]],
dtype=np.uint8)
processed = cv2.dilate(processed, kernel)
# Find the proposals sudoku contour- Contours are all posssible points. We will create the grid using contours
contours, hier = cv2.findContours(processed.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
external_contours = cv2.drawContours(processed.copy(), contours, -1,
(255, 0, 0), 2)
current_time = (time.time() - starttime)
def prepareData(self,cvl=False,iam=False,cvl_iam=False):
input_len = []
text_list = []
original_text = []
label_len = []
max_len = 0
image = []
if iam:
print('LOADING IAM DATASET')
image_name_path,text,len_text = self.read_data()
elif cvl:
print('LOADING CVL DATASET')
path = r'D:\cvl-database-1-1\testset\words'
image_name_path = [ ]
text = [ ]
len_text = [ ]
for root,dirs,files in os.walk(path):
for image_name in files:
image_name_path.append(os.path.join(root,image_name))
val = image_name.split('-')[-1].split('.')[0]
text.append(val.split('.')[0])
len_text.append(len(val))
else:
## IN DEVELOPMENT DONT CALL THIS LOOP
image_name_path,text,len_text = self.read_data()
for root,dirs,files in os.walk(path):
for image_name in files:
image_name_path.append(os.path.join(root,image_name))
val = image_name.split('-')[-1].split('.')[0]
text.append(val.split('.')[0])
len_text.append(len(val))
for image_name,text_word,text_len in tqdm(zip(image_name_path,text,len_text)):
load_image = cv2.imread(image_name)
try:
img = cv2.cvtColor(load_image,cv2.COLOR_BGR2GRAY)
img = cv2.adaptiveThreshold(img,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,cv2.THRESH_BINARY,blockSize=91,C=11)
except:
continue
# Dialiate and eroding of the image to make the image look much clearer
kernal = np.ones((5,5),np.uint8)
#img = cv2.erode(img,kernal,1)
#img = cv2.dilate(img,kernal,1)
# Image size adjustment to make all the image of shape (128,32)(h*w)
height,width = img.shape
if ((height > 32) or (width > 128)):
img = cv2.resize(img,(128,32),interpolation = cv2.INTER_AREA)
else:
if height < 32:
add_ones = np.ones((32-height,width)) * 255
try:
img = np.concatenate((img,add_ones))
except:
continue
if width < 128:
add_ones = np.ones((height,128-width)) * 255
try:
img = np.concatenate((img,add_ones),axis=1)
except:
continue
img = np.expand_dims(img,axis=2)
# Encode text
encode_text = self.__encode_string_to_numbers(text_word)
# Len of text
if text_len > max_len:
max_len = text_len
image.append(img)
text_list.append(encode_text)
original_text.append(text_word)
input_len.append(len(encode_text))
label_len.append(len(text_word))
return image,text_list,input_len,label_len,original_text,max_len
while True:
ret, frame = cam.read()
if not ret:
break
cv2.imshow("test", frame)
if cv2.waitKey(1) & 0xFF == 27:
break
elif cv2.waitKey(1) & 0xFF == 32:
img_name = "opencv_frame_{}.png".format(img_counter)
cv2.imwrite(img_name, frame)
img_counter += 1
cam.release()
cv2.destroyAllWindows()
img = cv2.imread('//project//project//opencv_frame_0.png')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
gray = cv2.medianBlur(gray, 5)
edges = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY, 21, 8)
color = cv2.medianBlur(img, 19)
cartoon = cv2.bitwise_and(color, color, mask=edges)
cv2.imwrite("//project//project//media//temp.png", cartoon)
print("/media/temp.png")
import numpy as np
from cv2 import cv2
import pytesseract
from PIL import Image
# Load image, grayscale, adaptive threshold
image = cv2.imread('captchas/10.png')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
thresh = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 3)
# Morph open
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (4, 3))
opening = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel, iterations=1)
# Remove noise by filtering using contour area
cnts = cv2.findContours(opening, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]
for c in cnts:
area = cv2.contourArea(c)
if area < 10:
cv2.drawContours(opening, [c], -1, (0, 0, 0), -1)
# Invert image for result
result = 255 - opening
cv2.imshow('thresh', thresh)
cv2.imshow('opening', opening)
cv2.imshow('result', result)
cv2.imwrite("result2.png", result)
motoren = simulator
# CPU quälen:
while True:
# frame lesen
dieser_frame = kamera.get_frame()
########################################
# verschiedene Varianten von threshold #
########################################
# Kopie von dieser_frame in Graustufen konvertieren
frame_graustufen = cv2.cvtColor(dieser_frame, cv2.COLOR_BGR2GRAY)
_, frame_schwarz_weiss1 = cv2.threshold(frame_graustufen, 110, 255,
cv2.THRESH_BINARY_INV)
frame_schwarz_weiss2 = cv2.adaptiveThreshold(
frame_graustufen, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY_INV, 11, 9)
frame_schwarz_weiss3 = cv2.adaptiveThreshold(
frame_graustufen, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY_INV, 11, 6)
# Kamerabild anzeigen
cv2.imshow("Original", dieser_frame)
cv2.imshow("Global Threshold", frame_schwarz_weiss1)
cv2.imshow("Adaptive Mean", frame_schwarz_weiss2)
cv2.imshow("Adaptive Gaussian", frame_schwarz_weiss3)
# Motoren ansteuern
motoren.set_speed(30, 30)
# Pausieren mit p, bewegen mit w,a,s,d, rotieren mit q,e, beenden mit k
As = [0,30,60,90,120,150]
filtered_img = np.zeros(image_nor.shape, np.float32)
for A in As:
#cv.getGaborKernel(ksize, sigma, theta, lambd, gamma[, psi[, ktype]])
g_kernel = cv2.getGaborKernel((27, 27), 3.4, np.pi*A/180, 8.1, 0.9, 0, ktype=cv2.CV_32F)
_filtered_img = cv2.filter2D(image_nor, cv2.CV_8UC3, g_kernel)
filtered_img += _filtered_img
# normalizacja filtru
filtered_img = filtered_img / filtered_img.max() *255
filtered_img = filtered_img.astype(np.uint8)
# uzyskanie z filtracji kodu dwuwymiarowego i zbinaryzowanego
filtered_img = cv2.cvtColor(filtered_img, cv2.COLOR_BGR2GRAY)
filtered_img = cv2.adaptiveThreshold(filtered_img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 19, 1)
# obsługa zapisu kodów lub wyszukiwania
if Baza.count(file)==True:
cv2.imwrite("Baza/kod"+array2string(np.array(files.index(file)))+".bmp",filtered_img)
print(file)
else:
Search(filtered_img, file)
else:
NotFound.append(file)
#print(file + " ----> błąd: nie wykryto tęczówki")
continue
cv2.waitKey(0)
width_ = w
height_ = h
#Or force it to be A4 format in 300 DPI
width_ = 1240
height_ = 1754
#New coordinates
new_rect = np.array(
[[0, 0], [width_ - 1, 0], [width_ - 1, height_ - 1], [0, height_ - 1]],
dtype="float32")
M = cv2.getPerspectiveTransform(rect, new_rect)
output_gray = cv2.warpPerspective(gray, M, (width_, height_))
output = cv2.adaptiveThreshold(output_gray, 255,
cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY, 5, 2)
except Exception as e:
print(str(e))
sys.exit(0)
print("Print")
# MATPLOTLIB
imgs = [img, output_gray, output]
i = 1
for img in imgs:
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.subplot(1, len(imgs), i)
plt.imshow(img)
im2 = imgWarpColored.copy()
im2 = unsharp_mask(im2)
cv2.imshow("b", im2)
im3 = im1.copy()
im3 = unsharp_mask(im3)
cv2.imshow("a+b", im3)
im4 = imgWarpColored.copy()
im4 = cv2.GaussianBlur(im4, (5, 5), 3)
im4 = cv2.addWeighted(im4, 1.5, im4, -0.5, 0, im4)
cv2.imshow("a+b+c", im4)
# APPLY ADAPTIVE THRESHOLD
imgWarpGray = cv2.cvtColor(imgWarpColored, cv2.COLOR_BGR2GRAY)
imgAdaptiveThre = cv2.adaptiveThreshold(imgWarpGray, 255, 1, 1, 7, 2)
imgAdaptiveThre = cv2.bitwise_not(imgAdaptiveThre)
imgAdaptiveThre = cv2.medianBlur(imgAdaptiveThre, 3)
cv2.imshow("Warped Image With corners after adaptive theshold1",
imgAdaptiveThre)
imgWarpGray1 = cv2.cvtColor(im3, cv2.COLOR_BGR2GRAY)
imgAdaptiveThre1 = cv2.adaptiveThreshold(imgWarpGray1, 255, 1, 1, 7, 2)
imgAdaptiveThre1 = cv2.bitwise_not(imgAdaptiveThre1)
imgAdaptiveThre1 = cv2.medianBlur(imgAdaptiveThre1, 3)
cv2.imshow("Warped Image With corners after adaptive theshold2",
imgAdaptiveThre1)
if cv2.waitKey(0) & 0xFF == ord('s'):
break
import numpy as np
from cv2 import cv2
img = cv2.imread('detect_blob.png',1)
gray = cv2.cvtColor(img,cv2.COLOR_RGB2GRAY)
thresh_img = cv2.adaptiveThreshold(gray,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,cv2.THRESH_BINARY,115,1) #guassion threshold with output binary threshold type
cv2.imshow("Adaptive Thresh Image",thresh_img)
#Contours
contours, hierarchy = cv2.findContours(thresh_img,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
thickness = 2
color = (255,0,255)
object=np.zeros([img.shape[0],img.shape[1],3],'uint8')
for c in contours:
cv2.drawContours(object,[c],-1,color,thickness) #Draw contour of one object at time, -1 means all but c have only 1 object contour
area = cv2.contourArea(c) #Area of one contour
perimeter = cv2.arcLength(c,True) #Perimiter of one contour
print("Perimeter : {}, Area: {}".format(perimeter,area))
M=cv2.moments(c)
cx = int(M['m10']/M['m00'])
cy = int(M['m01']/M['m00'])
cv2.circle(object,(cx,cy),4,color,-1) #draw centroid as a small unit cirlce in the image at centroid pos
cv2.imshow("Centroids",object)
cv2.waitKey(0)
cv2.destroyAllWindows()
def read_grid(image):
sudoku = ""
side = image.shape[0] // 9 # this is a square image - side will be 50
offset = 0.1 * side # for borders
for i in range(9):
for j in range(9):
# coordinates if corners of a small box
top = max(int(round(side * i + offset)), 0)
left = max(int(round(side * j + offset)), 0)
right = min(int(round(side * (j + 1) - offset)), image.shape[0])
bottom = min(int(round(side * (i + 1) - offset)), image.shape[0])
# finding the contour inside the box
crop = image[top:bottom, left:right]
gray = cv2.cvtColor(crop, cv2.COLOR_BGR2GRAY)
blurred = cv2.GaussianBlur(gray, (7, 7), 0)
thresh = cv2.adaptiveThreshold(blurred, 255,
cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY_INV, 11, 2)
contours, _ = cv2.findContours(thresh, cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
# finding the largest valid contour
max_area = 1
largest_contour = None
for cnt in contours:
area = cv2.contourArea(cnt)
if area > side * side * 0.04 and area > max_area:
max_area = area
largest_contour = cnt
# if there is no contour, then the box must be empty
if largest_contour is None:
sudoku += "0"
continue
# crop out the largest contour i.e. the digit
rect = cv2.minAreaRect(largest_contour)
box = cv2.boxPoints(rect)
box = np.int0(box)
minx = max(min(box, key=lambda g: g[0])[0] - 3, 0)
miny = max(min(box, key=lambda g: g[1])[1] - 3, 0)
maxx = min(max(box, key=lambda g: g[0])[0] + 3, int(side))
maxy = min(max(box, key=lambda g: g[1])[1] + 3, int(side))
# gray image of only the number
number_image = gray[miny:maxy, minx:maxx]
# OCR
custom_config = r' --psm 6 -c tessedit_char_whitelist=123456789'
text = pt.image_to_string(number_image, config=custom_config)
# if i == 7 and j == 3:
# cv2.imshow("cropped",number_image)
# print(text)
try:
num = int(text)
sudoku += str(num)
except:
# if there is a big contour but its not a digit,
# this frame is invalid, ignore it
return None
return sudoku