def sift_test(save = False):
global sift_time_use
# Initiate SIFT detector
sift = cv2.xfeatures2d.SIFT_create()
img1 = get_mask("idcard_mask.jpg") # queryImage
for i in range(get_idcard_count()):
img2, img_name = get_idcard(i) # trainImage
save_name = "result" + img_name
start = time.time()
img1, _ = iu.img_resize(img1, MATCH_PIC_WIDTH)
img2, _ = iu.img_resize(img2, MATCH_PIC_WIDTH)
# find the keypoints and descriptors with SIFT
kp1, des1 = sift.detectAndCompute(img1, None)
kp2, des2 = sift.detectAndCompute(img2, None)
# BFMatcher with default params
bf = cv2.BFMatcher()
matches = bf.knnMatch(des1, des2, k=2)
# Apply ratio test
good = []
for m, n in matches:
if m.distance < 0.75 * n.distance:
good.append(m)
end = time.time()
sift_time_use += end - start
if save:
folder = "test_sift"
save_match(folder, save_name, img1, kp1, img2, kp2, good)
print("finish pic %s" % save_name)
print("sift pic num=%d time used = %d"%(get_idcard_count(),int(sift_time_use*1000)))
def orb_test():
img1 = get_mask("idcard_mask.jpg") # queryImage
for i in range(get_idcard_count()):
img2, img_name = get_idcard(i) # trainImage
save_name = "result" + img_name
img1, _ = iu.img_resize(img1, MATCH_PIC_WIDTH)
img2, _ = iu.img_resize(img2, MATCH_PIC_WIDTH)
# Initiate SIFT detector
orb = cv2.ORB_create()
# find the keypoints and descriptors with SIFT
kp1, des1 = orb.detectAndCompute(img1, None)
kp2, des2 = orb.detectAndCompute(img2, None)
# create BFMatcher object
bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
# Match descriptors.
matches = bf.match(des1, des2)
# Sort them in the order of their distance.
matches = sorted(matches, key=lambda x: x.distance)
img3 = np.zeros((1000, 1000, 3), dtype=np.uint8)
# Draw first 10 matches.
# img3 = cv2.drawMatches(img1,kp1,img2,kp2,matches[:10], outImg=img3,flags=cv2.DRAW_MATCHES_FLAGS_DEFAULT)
#
# plt.imshow(img3),plt.show()
folder = "test_orb"
save_match(folder, save_name, img1, kp1, img2, kp2, matches[:10])
print("finish pic %s" % save_name)
def idcard_ocr(request):
start = time.time()
ori_time = start
# 初始化返回对象
response_data = dict(code=20001, message="", result=None)
if request.method == 'POST':
# 0-身份证正面 1-反面
card_side = request.POST.get('side')
if not card_side:
# 默认检测正面
card_side = 0
# 是否有伪造风险，如果有则开启风险检测
detect_risk = request.POST.get('risk')
if detect_risk is None:
# 默认无风险检测
detect_risk = False
# 是否检测身份证旋转角度
detect_direction = request.POST.get('direction')
if detect_direction is None:
detect_direction = False
time_used = time.time() - start
start += time_used
logger.info("prehandle timeUsed = %d ms" % (int(time_used * 1000)))
try:
card_file = request.FILES['image']
image = Image.open(card_file, mode="r").convert("RGB")
imgArray = np.asarray(image)
logger.info("get a image shape = %s" % str(imgArray.shape))
img_mat = cv2.cvtColor(imgArray, cv2.COLOR_RGB2BGR)
img_mat, scale = iu.img_resize(img_mat, 1920)
time_used = time.time() - start
start += time_used
logger.info("file load timeUsed = %d ms" % (int(time_used * 1000)))
img_full, _ = locate_card.locate(img_mat)
time_used = time.time() - start
start += time_used
logger.info("card location timeUsed = %d ms" %
(int(time_used * 1000)))
result_dict = idcardocr.idcardocr(img_full)
time_used = time.time() - ori_time
logger.info("total procession timeUsed = %d ms" %
(int(time_used * 1000)))
response_data = dict(code=0, message="ok", result=result_dict)
except (MultiValueDictKeyError, OSError) as error:
logger.error("图片参数错误" + traceback.format_exc())
response_data["message"] = "图片参数错误"
except ServiceException as error:
logger.error(str(error))
response_data["message"] = str(error)
else:
response_data["message"] = "请求服务方式错误"
ret_str = json.dumps(response_data)
logger.info("respond = %s" % response_data)
return HttpResponse(ret_str, content_type='application/json')
def locate(self, img_target):
"""
寻找图像中的身份证并进行图像矫正（透视变换）
:param target:
:return:
"""
global second_homo_time
start = time.time()
img_target_gray = cv2.cvtColor(
img_target, cv2.COLOR_BGR2GRAY) # trainImage in Gray
img_target_gray_small, _ = iu.img_resize(img_target_gray,
XFEATURE_IMG_WIDTH)
img_target_gray = cv2.cvtColor(img_target, cv2.COLOR_BGR2GRAY)
pts, dst = self.findMatchWithXFeature(img_target_gray_small,
img_template_small)
hl, wl = img_target_gray.shape
h_temp, w_temp = img_template_small.shape
ratio = wl / XFEATURE_IMG_WIDTH
pts = pts * ratio
dst = dst * ratio
# 参数
img_target_gray = cv2.polylines(img_target_gray, [np.int32(dst)], True,
255, 3, cv2.LINE_AA)
# FIXME
# iu.showimg(img_target_gray,"large")
# 进行图像矫正（透视变换）
h_time = time.time()
M_r, mask_r = cv2.findHomography(dst, pts, 0, 5.0)
second_homo_time += time.time() - h_time
im_r = cv2.warpPerspective(img_target, M_r,
(int(w_temp * ratio), int(h_temp * ratio)))
# FIXME
# iu.showimg(im_r,"after",True)
time_used = time.time() - start
start += time_used
print("match timeUsed = %d ms" % (int(time_used * 1000)))
# draw_params = dict(matchColor = (0,255,0), # draw matches in green color
# singlePointColor = None,
# matchesMask = matchesMask, # draw only inliers
# flags = 2)
# img3 = cv2.drawMatches(img1,kp1,img2,kp2,good,None,**draw_params)
# plt.imshow(img3, 'gray'),plt.show()
# im_r图像矫正结果
return im_r, img_target_gray
import numpy as np
import time
import cv2
import os
from django_web.model import ServiceException
from idcard_ocr.settings import BASE_DIR
from django_web.util import img_util as iu
RESOURCE_PATH = os.path.join(BASE_DIR, "django_web/resource")
MASK = os.path.join(BASE_DIR, "django_web/resource", "mask")
template = os.path.join(MASK, 'idcard_mask.jpg')
img_template = cv2.imread(template, 0) # queryImage in Gray
XFEATURE_IMG_WIDTH = 400
img_template_small, _ = iu.img_resize(img_template, XFEATURE_IMG_WIDTH)
MIN_MATCH_COUNT = 10
class LocateCard:
def __init__(self):
pass
def locate_with_file(self, file_name):
# img2 = idocr.hist_equal(img2)
img_target = cv2.imread(file_name)
return self.locate(img_target)
# target为需要识别的图像
def locate(self, img_target):
"""
def sift_test_with_flann(save = False):
global sift_time_use
# Initiate SIFT detector
sift = cv2.xfeatures2d.SIFT_create()
img1 = get_mask("idcard_mask.jpg") # queryImage
for i in range(get_idcard_count()):
img2, img_name = get_idcard(i) # trainImage
save_name = "result" + img_name
start = time.time()
# 模板
img1, _ = iu.img_resize(img1, 600)
# 目标图片
img2, _ = iu.img_resize(img2, 600)
# find the keypoints and descriptors with SIFT
kp1, des1 = sift.detectAndCompute(img1, None)
kp2, des2 = sift.detectAndCompute(img2, None)
FLANN_INDEX_KDTREE = 0
index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
search_params = dict(checks=10)
flann = cv2.FlannBasedMatcher(index_params, search_params)
matches = flann.knnMatch(des1, des2, k=2)
# Apply ratio test
good = []
for m, n in matches:
if m.distance < 0.70 * n.distance:
good.append(m)
if len(good) > MIN_MATCH_COUNT:
src_pts = np.float32([kp1[m.queryIdx].pt for m in good]).reshape(-1, 1, 2)
dst_pts = np.float32([kp2[m.trainIdx].pt for m in good]).reshape(-1, 1, 2)
# 用HomoGraphy计算图像与图像之间映射关系, M为转换矩阵
M, mask = cv2.findHomography(src_pts, dst_pts, cv2.RANSAC, 5.0)
matchesMask = mask.ravel().tolist()
# 使用转换矩阵M计算出img1在img2的对应形状
h, w = img1.shape
pts = np.float32([[0, 0], [0, h - 1], [w - 1, h - 1], [w - 1, 0]]).reshape(-1, 1, 2)
dst = cv2.perspectiveTransform(pts, M)
# 参数
img_target_gray = cv2.polylines(img2, [np.int32(dst)], True, 255, 3, cv2.LINE_AA)
# iu.showimg(img_target_gray)
# 进行图像矫正（透视变换）
M_r, mask_r = cv2.findHomography(dst, pts, 0, 5.0)
ratio = 1
im_r = cv2.warpPerspective(img2, M_r, (int(w*ratio), int(h*ratio)))
# FIXME
# iu.showimg(im_r)
# save_img("perspective",save_name,im_r)
else:
print("身份证匹配度不足 - %d/%d" % (len(good), MIN_MATCH_COUNT))
matchesMask = None
end = time.time()
sift_time_use += end - start
if save:
folder = "test_sift_with_flann"
save_match(folder, save_name, img1, kp1, img2, kp2, good)
print("finish pic %s" % save_name)
print("sift pic num=%d time used = %d"%(get_idcard_count(),int(sift_time_use*1000)))