result = img3_2
else:
result = None
if result is not None:
print('이미지 받아왔나? ', result)
print(type(result))
result = cv2.resize(result, dsize=(432, 98), interpolation=cv2.INTER_LINEAR)
cv2.imwrite(outputFileName, result)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
# OCR ########################################################################################################
ocr = out_result(outputFileName)
resultOCR = ocr.output('1024+52font+')
print("OCR 1차 :", resultOCR)
if bool(re.search('\d{2}\D\d{4}', resultOCR)):
resultOCR = re.search('\d{2}\D\d{4}', resultOCR).group()
print("OCR 2차 :", resultOCR)
elif bool(re.search('\d{2}\D', resultOCR)) and bool(re.search('\d{4}', resultOCR)):
resultOCR = re.search('\d{2}\D', resultOCR).group() + re.search('\d{4}', resultOCR).group()
print("OCR 2차_2 :", resultOCR)
txt1 = str(num) + ' : ' + str(resultOCR) + '\n'
print(txt1)
f.write(txt1)
num = num + 1
def make_json(capture_path, car_id, car_capture_path, image_start,
image_detect, date_start, time_start, date_detect, time_detect):
carColor = []
capture_path_tmp = "C:/Users/dlwlg/Desktop/capture_tmp/"
inputCarDIR = capture_path_tmp + car_capture_path
print("자동차사진 : " + str(car_id))
# 자동차 색 판별################################################################################################
print("자동차사진 경로 : ", inputCarDIR)
carIMG = cv2.imread(inputCarDIR)
identificator = colorIdentification(carIMG)
color = identificator.color()
print("차량 색상 :", color)
carColor.append(color)
#print(carIMG)
# 자동차 번호판 자르기##########################################################################################
detect = carplateDetecting(carIMG, 'cfg/obj.names',
'cfg/yolov2-carplate.cfg',
'cfg/yolov2-carplate_2200.weights')
saveIMG = detect.parse()
#cv2.imwrite('img/plate/' + str(car_id) + '.jpg', saveIMG)
image_plate = date_start + time_start + "_" + str(
car_id) + "_plate" + ".jpg"
cv2.imwrite(str(capture_path + image_plate), saveIMG)
print("image_plate : ", capture_path + image_plate)
# 번호판 이미지 전처리#########################################################################################
inputFileName = capture_path + image_plate
outputFileName = capture_path_tmp + "plate_result_id" + str(
car_id) + '.jpg'
opencvIMG = extract_opencv(inputFileName, outputFileName)
# 이미지 로드
img_original = cv2.imread(inputFileName, cv2.IMREAD_COLOR)
# 이미지 높이, 넓이
height, width, channel = img_original.shape
if (height > 10):
# 이미지 확대
img_original = cv2.resize(img_original,
None,
fx=4,
fy=4,
interpolation=cv2.INTER_LINEAR)
height, width, channel = img_original.shape
# 이미지 전처리(img:canny 처리된 이미지, img2:이진화까지만 처리된 이미지)
img, img2 = opencvIMG.img_preprocessing(img_original)
# 직선 검출, 직선 각도 검출
degree = opencvIMG.detect_line(img)
# 전처리(이진화)된 이미지 회전
#cv2.imshow('img2', img2)
#cv2.waitKey(0)
#cv2.destroyAllWindows()
result = opencvIMG.img_rotate(img2, degree, height, width)
#cv2.imshow('rotate', result)
#cv2.waitKey(0)
#cv2.destroyAllWindows()
# 원본 이미지 회전
img_original2 = opencvIMG.img_rotate(img_original, degree, height,
width)
# cv2.imshow('img_original_rotated', img_original)
# 글자영역 찾기 - 1차
high_x = 0
high_y = 0
row_x = 0
row_y = 0
result2, high_y, high_x, row_y, row_x = opencvIMG.find_number(
result, img_original2, high_y, high_x, row_y, row_x, height, width)
# 글자영역 자르기 - 1차
final_result = img_original2[row_y:high_y, row_x:high_x]
# 1차로 자른 사진 다시 노이즈 처리
result = opencvIMG.remove_noise(result[row_y:high_y, row_x:high_x])
# cv2.imshow('result', result)
# 글자영역 찾기 - 2차
result, high_y, high_x, row_y, row_x = opencvIMG.find_number2(
result, final_result, high_y, high_x, row_y, row_x, height, width)
# cv2.imshow('result2', result)
# 글자영역 자르기 - 2차
final_result = final_result[row_y:high_y, row_x:high_x]
# 최종 이미지 수축
kernel2 = np.ones((3, 3), np.uint8)
result = cv2.dilate(result, kernel2, iterations=1)
# 전처리(이진화)된 이미지 최종 결과
# cv2.imshow('result', result)
# 원본 이미지 최종 결과
# cv2.imshow('final_result', final_result)
# 원본이미지 최종결과 사이즈 조정
final_result = cv2.resize(final_result,
dsize=(432, 98),
interpolation=cv2.INTER_LINEAR)
############### 광학문자 추출 ################
# 이미지 로드
# Color Change BGR to RGB
img_original = cv2.cvtColor(final_result, cv2.COLOR_BGR2RGB)
# 선명화를 위한 필터링1
filter_img = opencvIMG.filterIMG_for_char(img_original)
# Color Change GRAY to BGR
filter_img = cv2.cvtColor(filter_img, cv2.COLOR_GRAY2BGR)
# 이미지 합성을 통한 선명화
final_fil_img = opencvIMG.addIMG2_for_char(img_original, filter_img)
# 이미지 합성하면서 원본이미지 수정됨, 다시 불러옴
img_original = cv2.imread(inputFileName, cv2.IMREAD_COLOR)
# height, width, channel = img_original.shape
# 주영쓰 필터가 추가된 이미지 전처리
#cv2.imshow('show filtering IMG', final_fil_img)
# 전처리
img_2 = opencvIMG.img_preprocessing_for_char(final_fil_img)
# Color Change GRAY to BGR
copy_img_2 = cv2.cvtColor(img_2, cv2.COLOR_GRAY2BGR)
height, width, channel = copy_img_2.shape
# 이미지 합성으로 위한 하얀바탕 만들기
blank_img_2 = np.zeros((height + 6, width + 6, 3), np.uint8)
cv2.rectangle(blank_img_2, (0, 0), (width + 6, height + 6),
(255, 255, 255), -1)
# 이미지 합성
img2_2 = opencvIMG.addIMG_for_char(blank_img_2, copy_img_2)
# 노이즈 처리
img3_2 = opencvIMG.remove_noise_for_char(img2_2)
#cv2.imshow("noise_filterImg", img3_2)
result = img3_2
else:
result = None
if result is not None:
print('Image Load Success!!!')
print("전체사진 :", car_capture_path)
print("자동차 사진 :", inputCarDIR)
print("번호판 사진 :", inputFileName)
print("광학문자 :", outputFileName)
result = cv2.resize(result,
dsize=(1296, 294),
interpolation=cv2.INTER_LINEAR)
cv2.imwrite(outputFileName, result)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
# OCR #####################################################################################################
ocr = out_result(outputFileName)
resultOCR = ocr.output('1499lstm2+52font')
print("OCR 1차 :", resultOCR)
if bool(re.search('\d{2}\D\d{4}', resultOCR)):
resultOCR = re.search('\d{2}\D\d{4}', resultOCR).group()
print("OCR 2차 :", resultOCR)
elif bool(re.search('\d{2}\D', resultOCR)) and bool(
re.search('\d{4}', resultOCR)):
resultOCR = re.search('\d{2}\D', resultOCR).group() + re.search(
'\d{4}', resultOCR).group()
print("OCR 2차_2 :", resultOCR)
txt1 = str(car_id) + ' : ' + str(resultOCR) + '\n'
print(txt1)
print('------------------------------------------')
#########################################################################################################
# response
data = {
"capture_path": capture_path,
"saveEvidenceCrop": image_detect,
"saveTrackerCrop": car_capture_path,
"final_plate": inputFileName,
"final_save": outputFileName,
"txt": resultOCR
}
real_data = {
"date_start": date_start,
"time_start": time_start,
"date_detect": date_detect,
"time_detect": time_detect,
"color": color,
"plate_num": resultOCR,
"image_start": image_start,
"image_detect": image_detect,
"image_plate": image_plate
}
print("진짜 저장되는 데이터 : ", real_data)
path_data = json.dumps(real_data, ensure_ascii=False)
path_data2 = json.dumps(data, ensure_ascii=False)
print(path_data)
#response = make_response(path_data)
#response.headers['Content-Type'] = 'application/json;charset=utf-8'
#print(str(response))
print("path_data :", str(path_data))
return path_data.encode('utf-8'), path_data2.encode('utf-8')
else:
print('Image Load Fail!!!')
print("전체사진 경로:", image_start)
print("자동차 사진 경로:", inputCarDIR)
txt2 = str(car_id) + ' : X\n'
print('-----------------------------------------')
# response
data = {
"capture_path": capture_path,
"saveEvidenceCrop": image_start,
"saveTrackerCrop": car_capture_path,
"final_plate": inputFileName,
"final_save": outputFileName,
"txt": "X"
}
real_data = {
"date_start": date_start,
"time_start": time_start,
"date_detect": date_detect,
"time_detect": time_detect,
"color": color,
"plate_num": "X",
"image_start": image_start,
"image_detect": image_detect,
"image_plate": image_plate
}
print("진짜 저장되는 데이터 : ", real_data)
path_data = json.dumps(real_data, ensure_ascii=False)
path_data2 = json.dumps(data, ensure_ascii=False)
print(path_data)
# response = make_response(path_data)
# response.headers['Content-Type'] = 'application/json;charset=utf-8'
# print(str(response))
print("path_data :", str(path_data))
return path_data.encode('utf-8'), path_data2.encode('utf-8')
def getPath():
print("Success connected")
if request.method == 'POST':
capture_path = "C:/Users/dlwlg/Desktop/capture/"
json_path = request.get_json()
path = json_path['serverPath']
saveFileName = json_path['saveFileName']
print(json_path['serverPath'])
print(json_path['saveFileName'])
###################################################################################################################################
videotrack = traker()
cap = cv2.VideoCapture(path + saveFileName)
tracker = Sort()
totalFrame = 0
# 재생할 파일의 넓이 얻기
width = cap.get(cv2.CAP_PROP_FRAME_WIDTH)
# 재생할 파일의 높이 얻기
height = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
# 재생할 파일의 프레임 레이트 얻기
fps = cap.get(cv2.CAP_PROP_FPS)
# 저장할 비디오 코덱
fourcc = cv2.VideoWriter_fourcc(*'DIVX')
# 저장할 파일 이름
output_avi = 'output.avi'
# 파일 stream 생성
out = cv2.VideoWriter(capture_path + output_avi, fourcc, fps,
(int(width), int(height)))
# filename : 파일 이름
# fourcc : 코덱
# fps : 초당 프레임 수
# width : 넓이
# height : 높이
#################################
IDcount = {}
XYbox = {}
time = {}
print("IDcount : ")
print(IDcount)
print("XYbox : ")
print(XYbox)
################################
while True:
# Capture frame-by-frame
ret, frame = cap.read()
if ret == True:
(H, W) = frame.shape[:2]
frame = np.asarray(frame)
results = videotrack.tfnet.return_predict(frame) # JSON
dets = videotrack.getPoint(results)
trackers_id = tracker.update(dets)
tracking = videotrack.id_box(frame, trackers_id)
# ID 별로 frame 수 카운트
num = len(trackers_id)
for i in range(num):
print(trackers_id[i][4])
strID = str(trackers_id[i][4])
if strID in IDcount.keys():
IDcount[strID] += 1
else:
IDcount[strID] = 1
# 90 프레임 (3초) 넘어가는 애들은 첫 사진 촬영 (이미지 저장)
if IDcount[strID] == 150:
XYbox[strID + "3s"] = [
1, trackers_id[i][0], trackers_id[i][1],
trackers_id[i][2], trackers_id[i][3]
]
print("XYbox 90프레임 때 캡쳐 된 좌표 추가 : ")
print(XYbox)
xy = XYbox[strID + "3s"]
topx = xy[1]
topy = xy[2]
bottomx = xy[3]
bottomy = xy[4]
cv2.imwrite(
capture_path + "img_id" + strID + "_3s" + ".jpg",
tracking)
cropped = frame[int(topy):int(bottomy) + 1,
int(topx):int(bottomx) + 1]
cv2.imwrite(
capture_path + "img_id" + strID + "_3s" + ".jpg",
cropped)
# 150 프레임(5초) 넘어가는 애들 증거 사진 촬영 (이미지 저장)
if IDcount[strID] == 150:
time1 = datetime.datetime.now()
time[strID] = str(time1)
print(json.dumps(time, indent="\t"))
with open('img/time.json', 'w') as make_file:
json.dump(time, make_file, indent='\t')
XYbox[strID + "5s"] = [
2, trackers_id[i][0], trackers_id[i][1],
trackers_id[i][2], trackers_id[i][3]
]
print("XYbox 150프레임(5초) 때 캡쳐 된 좌표 추가 : ")
print(strID + "차량 불법주정차 시작 시각 : " + str(time1))
print(XYbox)
xy = XYbox[strID + "5s"]
topx = xy[1]
topy = xy[2]
bottomx = xy[3]
bottomy = xy[4]
saveEvidenceCrop = "img_id" + strID + "_5s" + ".jpg"
saveTrackerCrop = "box_id" + strID + "_5s" + ".jpg"
cv2.imwrite(
capture_path + "img_id" + strID + "_5s" + ".jpg",
tracking)
cropped = frame[int(topy):int(bottomy) + 1,
int(topx):int(bottomx) + 1]
cv2.imwrite(
capture_path + "box_id" + strID + "_5s" + ".jpg",
cropped)
print(capture_path + "box_id" + strID + "_5s" + ".jpg")
totalFrame = totalFrame + 1
# Display the resulting frame
cv2.imshow('frame', tracking)
# print(totalFrame)
out.write(tracking)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
else:
break
total = 0
idlist = []
# 불법 주정차 차량 판단하기#####################################
for id, fps in IDcount.items():
if fps >= 150: # 150 (5초 이상 일때)
print("ID : " + id + "는 불법 주정차 차량입니다.")
print("사각형 좌표 : ")
total = total + 1
idlist.append(id)
###########################################################
print("IDcount : ")
print(IDcount)
print("XYbox : ")
print(XYbox)
print("idlist 불법주정차 : ", idlist)
# When everything done, release the capture
cap.release()
cv2.destroyAllWindows()
###################################################################################################################################
num = 0
endnum = total - 1
print("불법주정차 차량 숫자 : ", total)
#idlist[num]
carColor = []
f = open("img/ocr_result.txt", 'w')
while num <= endnum:
inputCarDIR = capture_path + "box_id" + str(
idlist[num]) + '_5s.jpg'
print("자동차사진 : " + str(idlist[num]))
# 자동차 색 판별################################################################################################
print("자동차사진 경로 : ", inputCarDIR)
carIMG = cv2.imread(inputCarDIR)
identificator = colorIdentification(carIMG)
color = identificator.color()
print("차량 색상 :", color)
carColor.append(color)
#print(carIMG)
# 자동차 번호판 자르기##########################################################################################
detect = carplateDetecting(carIMG, 'cfg/obj.names',
'cfg/yolov2-carplate.cfg',
'cfg/yolov2-carplate_2200.weights')
saveIMG = detect.parse()
#cv2.imwrite('img/plate/' + str(idlist[num]) + '.jpg', saveIMG)
cv2.imwrite(capture_path + "plate_id" + str(idlist[num]) + '.jpg',
saveIMG)
# 번호판 이미지 전처리##########################################################################################
inputFileName = capture_path + "plate_id" + str(
idlist[num]) + '.jpg'
outputFileName = capture_path + "plate_result_id" + str(
idlist[num]) + '.jpg'
opencvIMG = extract_opencv(inputFileName, outputFileName)
# 이미지 로드
img_original = cv2.imread(inputFileName, cv2.IMREAD_COLOR)
# 이미지 높이, 넓이
height, width, channel = img_original.shape
if (height > 10):
# 이미지 확대
img_original = cv2.resize(img_original,
None,
fx=4,
fy=4,
interpolation=cv2.INTER_LINEAR)
height, width, channel = img_original.shape
# 이미지 전처리(img:canny 처리된 이미지, img2:이진화까지만 처리된 이미지)
img, img2 = opencvIMG.img_preprocessing(img_original)
# 직선 검출, 직선 각도 검출
degree = opencvIMG.detect_line(img)
# 전처리(이진화)된 이미지 회전
#cv2.imshow('img2', img2)
#cv2.waitKey(0)
#cv2.destroyAllWindows()
result = opencvIMG.img_rotate(img2, degree, height, width)
#cv2.imshow('rotate', result)
#cv2.waitKey(0)
#cv2.destroyAllWindows()
# 원본 이미지 회전
img_original2 = opencvIMG.img_rotate(img_original, degree,
height, width)
# cv2.imshow('img_original_rotated', img_original)
# 글자영역 찾기 - 1차
high_x = 0
high_y = 0
row_x = 0
row_y = 0
result2, high_y, high_x, row_y, row_x = opencvIMG.find_number(
result, img_original2, high_y, high_x, row_y, row_x,
height, width)
# 글자영역 자르기 - 1차
final_result = img_original2[row_y:high_y, row_x:high_x]
# 1차로 자른 사진 다시 노이즈 처리
result = opencvIMG.remove_noise(result[row_y:high_y,
row_x:high_x])
# cv2.imshow('result', result)
# 글자영역 찾기 - 2차
result, high_y, high_x, row_y, row_x = opencvIMG.find_number2(
result, final_result, high_y, high_x, row_y, row_x, height,
width)
# cv2.imshow('result2', result)
# 글자영역 자르기 - 2차
final_result = final_result[row_y:high_y, row_x:high_x]
# 최종 이미지 수축
kernel2 = np.ones((3, 3), np.uint8)
result = cv2.dilate(result, kernel2, iterations=1)
# 전처리(이진화)된 이미지 최종 결과
# cv2.imshow('result', result)
# 원본 이미지 최종 결과
# cv2.imshow('final_result', final_result)
# 원본이미지 최종결과 사이즈 조정
final_result = cv2.resize(final_result,
dsize=(432, 98),
interpolation=cv2.INTER_LINEAR)
############### 광학문자 추출 #################
# 이미지 로드
# Color Change BGR to RGB
img_original = cv2.cvtColor(final_result, cv2.COLOR_BGR2RGB)
# 선명화를 위한 필터링1
filter_img = opencvIMG.filterIMG_for_char(img_original)
# Color Change GRAY to BGR
filter_img = cv2.cvtColor(filter_img, cv2.COLOR_GRAY2BGR)
# 이미지 합성을 통한 선명화
final_fil_img = opencvIMG.addIMG2_for_char(
img_original, filter_img)
# 이미지 합성하면서 원본이미지 수정됨, 다시 불러옴
img_original = cv2.imread(inputFileName, cv2.IMREAD_COLOR)
# height, width, channel = img_original.shape
# 주영쓰 필터가 추가된 이미지 전처리
#cv2.imshow('show filtering IMG', final_fil_img)
# 전처리
img_2 = opencvIMG.img_preprocessing_for_char(final_fil_img)
# Color Change GRAY to BGR
copy_img_2 = cv2.cvtColor(img_2, cv2.COLOR_GRAY2BGR)
height, width, channel = copy_img_2.shape
# 이미지 합성으로 위한 하얀바탕 만들기
blank_img_2 = np.zeros((height + 6, width + 6, 3), np.uint8)
cv2.rectangle(blank_img_2, (0, 0), (width + 6, height + 6),
(255, 255, 255), -1)
# 이미지 합성
img2_2 = opencvIMG.addIMG_for_char(blank_img_2, copy_img_2)
# 노이즈 처리
img3_2 = opencvIMG.remove_noise_for_char(img2_2)
#cv2.imshow("noise_filterImg", img3_2)
result = img3_2
else:
result = None
if result is not None:
print('Image Load Success!!!')
print("전체사진 :", saveEvidenceCrop)
print("자동차 사진 :", inputCarDIR)
print("번호판 사진 :", inputFileName)
print("광학문자 :", outputFileName)
result = cv2.resize(result,
dsize=(432, 98),
interpolation=cv2.INTER_LINEAR)
cv2.imwrite(outputFileName, result)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
# OCR ########################################################################################################
ocr = out_result(outputFileName)
resultOCR = ocr.output('1024+52font')
print("OCR 1차 :", resultOCR)
if bool(re.search('\d{2}\D\d{4}', resultOCR)):
resultOCR = re.search('\d{2}\D\d{4}', resultOCR).group()
print("OCR 2차 :", resultOCR)
elif bool(re.search('\d{2}\D', resultOCR)) and bool(
re.search('\d{4}', resultOCR)):
resultOCR = re.search('\d{2}\D',
resultOCR).group() + re.search(
'\d{4}', resultOCR).group()
print("OCR 2차_2 :", resultOCR)
txt1 = str(num) + ' : ' + str(resultOCR) + '\n'
print(txt1)
f.write(txt1)
num = num + 1
print('------------------------------------------')
############################################################################################################
# response
data = {
"capture_path": capture_path,
"saveEvidenceCrop": saveEvidenceCrop,
"saveTrackerCrop": saveTrackerCrop,
"final_plate": inputFileName,
"final_save": outputFileName,
"txt": resultOCR
}
path_data = json.dumps(data, ensure_ascii=False)
print(path_data)
response = make_response(path_data)
response.headers[
'Content-Type'] = 'application/json;charset=utf-8'
print(str(response))
return response
else:
print('Image Load Fail!!!', num)
print("전체사진 :", saveEvidenceCrop)
print("자동차 사진 :", inputCarDIR)
txt2 = str(num) + ' : X\n'
f.write(txt2)
num = num + 1
print('------------------------------------------')
# response
data = {
"capture_path": capture_path,
"saveEvidenceCrop": saveEvidenceCrop,
"saveTrackerCrop": saveTrackerCrop,
"final_plate": inputFileName,
"final_save": outputFileName,
"txt": "X"
}
path_data = json.dumps(data, ensure_ascii=False)
print(path_data)
response = make_response(path_data)
response.headers[
'Content-Type'] = 'application/json;charset=utf-8'
print(str(response))
return response
f.close()