def single_test_detect():
path = os.path.join(RESOURCE_PATH, 'test3')
files = iu.get_imgs_from_path(path)
image_array = []
text_array = []
for file_name in files:
print("识别图片 %s" % file_name)
file_path = os.path.join(path, file_name)
if not os.path.isfile(file_path):
continue
img = iu.read_img(file_path, width=1000)
try:
contour_img, imgs, txts = td.detect(img,
debug=True,
save_folder=os.path.join(
path, "middle_result"))
except ServiceException as exp:
print("识别出错 %s" % exp.message)
image_array += imgs
text_array += txts
# 保存绿框结果
iu.write_middle_result(contour_img, file_name, "final_result")
for img in image_array:
cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
su.auto_subplt(image_array, 5, [plt.imshow], title=text_array)
def preprocess(gray, debug, save_folder):
# 1. Sobel算子，x方向求梯度
sobel = cv2.Sobel(gray, cv2.CV_8U, 1, 0, ksize=3)
# 2. 二值化
# TODO 关于二值化阈值的选择，或者二值化阈值算法的选择THRESH_TRIANGLE/THRESH_OTSU留待之后进行试验。目前来看THRESH_OTSU算法更适合当前情况下的图片
ret, binary = cv2.threshold(sobel, 0, 255,
cv2.THRESH_OTSU + cv2.THRESH_BINARY)
# binary = img_binary(gray)
# 3. 膨胀和腐蚀操作的核函数
element1 = cv2.getStructuringElement(cv2.MORPH_RECT, (30, 9))
element2 = cv2.getStructuringElement(cv2.MORPH_RECT, (24, 6))
# 4. 膨胀一次，让轮廓突出
dilation = cv2.dilate(binary, element2, iterations=1)
# 5. 腐蚀一次，去掉细节，如表格线等。注意这里去掉的是竖直的线
erosion = cv2.erode(dilation, element1, iterations=1)
# 6. 再次膨胀，让轮廓明显一些
dilation2 = cv2.dilate(erosion, element2, iterations=3)
if debug:
# 7. 存储中间图片
iu.write_middle_result(binary, "1_binary.png", save_folder)
iu.write_middle_result(dilation, "2_dilation.png", save_folder)
iu.write_middle_result(erosion, "3_erosion.png", save_folder)
iu.write_middle_result(dilation2, "4_dilation2.png", save_folder)
return dilation2
def detect(img, debug=False, save_folder="result"):
# 1. 转化成灰度图
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# 提取人脸区域 face_window = [x,y,w,h]
face_window = fd.find_face(img)
if face_window is None:
raise ServiceException("未识别到人脸")
# TODO 暂时松弛边界写为固定，之后可以先扣取身份证区域，根据身份证大小决定边界大小，或者将身份证区域处理成统一大小
shape = gray.shape
gray = fd.remove_face(gray,
face_window,
relax_border=np.intp([shape[0] / 15, shape[1] / 12]))
fd.draw_face_box(img, face_window)
# 2. 形态学变换的预处理，得到可以查找矩形的图片
dilation = preprocess(gray, debug, save_folder)
# 3. 查找和筛选文字区域
region = findTextRegion(dilation, debug, save_folder)
# 识别文字
result_text = ""
image_array = []
text_array = []
for box in region:
# 识别框内文字
x_min, y_min = np.min(box, 0)
x_max, y_max = np.max(box, 0)
text_window = [x_min, y_min, x_max - x_min, y_max - y_min]
if not region_filter(img, text_window, face_window):
continue
image_part, text = tr.ocr_result(img, text_window)
result_text = text + "\n" + result_text
image_array.append(image_part)
text_array.append(text)
print(result_text)
print("-" * 10)
# 4. 用绿线画出这些找到的轮廓
for box in region:
cv2.drawContours(img, [box], 0, (0, 255, 0), 2)
if debug:
# 带轮廓的图片
iu.write_middle_result(img, "contours.png", save_folder)
cv2.waitKey(0)
return img, image_array, text_array
def group_test_detect():
path = os.path.join(RESOURCE_PATH, 'idcard_img')
files = iu.get_imgs_from_path(path)
image_array = []
text_array = []
for file_name in files:
print("识别图片 %s" % file_name)
file_path = os.path.join(path, file_name)
if not os.path.isfile(file_path):
continue
img = iu.read_img(file_path, width=1000)
try:
contour_img, imgs, txts = td.detect(img)
image_array += imgs
text_array += txts
# 保存绿框结果
iu.write_middle_result(contour_img, file_name, "final_result")
except ServiceException as exp:
print("识别出错 %s" % str(exp))
def get_card_area(img, debug=False, save_folder="result"):
"""
剪裁出身份证文字区域
:param img:
:param debug:
:param save_folder:
:return:
"""
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
sobel = cv2.Sobel(gray, cv2.CV_8U, 1, 1, ksize=3)
ret, binary = cv2.threshold(sobel, 0, 255,
cv2.THRESH_OTSU + cv2.THRESH_BINARY)
# 3. 膨胀和腐蚀操作的核函数
element1 = cv2.getStructuringElement(cv2.MORPH_RECT, (30, 9))
element2 = cv2.getStructuringElement(cv2.MORPH_RECT, (24, 6))
# 4. 膨胀一次，让轮廓突出
dilation = cv2.dilate(binary, element2, iterations=1)
# 5. 腐蚀一次，去掉细节，如表格线等。注意这里去掉的是竖直的线
erosion = cv2.erode(dilation, element1, iterations=1)
# 6. 再次膨胀，让轮廓明显一些
dilation2 = cv2.dilate(erosion, element2, iterations=3)
# 查找和筛选文字区域
box = find_max_region(sobel)
# 用绿线画出这些找到的轮廓
cv2.drawContours(img, [box], 0, (0, 255, 0), 2)
if debug:
# 存储中间图片
iu.write_middle_result(binary, "1_binary.png", save_folder)
iu.write_middle_result(dilation, "2_dilation.png", save_folder)
iu.write_middle_result(erosion, "3_erosion.png", save_folder)
iu.write_middle_result(dilation2, "4_dilation2.png", save_folder)
iu.write_middle_result(img, "contour.jpg", save_folder)
cv2.waitKey()