新增二维码识别替换高清图片功能

README.md

@@ -118,3 +118,5 @@
 20. 版本号：1.13.0
     1. 新增文档检测功能
     2. 新增扭曲矫正功能
+21. 版本号：1.14.0
+    1. 新增二维码识别替换高清图片功能

docker-compose.yml

@@ -1,8 +1,35 @@
-x-env:
+template:
   &template
-  image: fcb_photo_review:1.13.10
+  image: fcb_photo_review:1.14.1
   restart: always
 
+review_template:
+  &review_template
+  <<: *template
+  volumes:
+    - ./log:/app/log
+    - ./model:/app/model
+  deploy:
+    resources:
+      reservations:
+        devices:
+          - device_ids: [ '0', '1' ]
+            capabilities: [ 'gpu' ]
+            driver: 'nvidia'
+
+mask_template:
+  &mask_template
+  <<: *template
+  volumes:
+    - ./log:/app/log
+  deploy:
+    resources:
+      reservations:
+        devices:
+          - device_ids: [ '1' ]
+            capabilities: [ 'gpu' ]
+            driver: 'nvidia'
+
 services:
   det_api:
     <<: *template
@@ -13,153 +40,67 @@ services:
     volumes:
       - ./log:/app/log
       - ./model:/app/model
-    command: [ "det_api.py" ]
+#    command: [ 'det_api.py' ]
     deploy:
       resources:
         reservations:
           devices:
-            - device_ids: [ "0" ]
+            - device_ids: [ '0' ]
-              capabilities: [ "gpu" ]
+              capabilities: [ 'gpu' ]
-              driver: "nvidia"
+              driver: 'nvidia'
 
   photo_review_1:
-    <<: *template
+    <<: *review_template
     container_name: photo_review_1
     hostname: photo_review_1
-    volumes:
-      - ./log:/app/log
-      - ./model:/app/model
     depends_on:
       - det_api
-    command: [ "photo_review.py", "--clean", "True" ]
+    command: [ 'photo_review.py', '--clean', 'True' ]
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - device_ids: [ "0", "1" ]
-              capabilities: [ "gpu" ]
-              driver: "nvidia"
 
   photo_review_2:
-    <<: *template
+    <<: *review_template
     container_name: photo_review_2
     hostname: photo_review_2
-    volumes:
-      - ./log:/app/log
-      - ./model:/app/model
     depends_on:
       - photo_review_1
-    command: [ "photo_review.py" ]
+    command: [ 'photo_review.py' ]
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - device_ids: [ "0", "1" ]
-              capabilities: [ "gpu" ]
-              driver: "nvidia"
 
   photo_review_3:
-    <<: *template
+    <<: *review_template
     container_name: photo_review_3
     hostname: photo_review_3
-    volumes:
-      - ./log:/app/log
-      - ./model:/app/model
     depends_on:
       - photo_review_2
-    command: [ "photo_review.py" ]
+    command: [ 'photo_review.py' ]
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - device_ids: [ "0", "1" ]
-              capabilities: [ "gpu" ]
-              driver: "nvidia"
 
   photo_review_4:
-    <<: *template
+    <<: *review_template
     container_name: photo_review_4
     hostname: photo_review_4
-    volumes:
-      - ./log:/app/log
-      - ./model:/app/model
     depends_on:
       - photo_review_3
-    command: [ "photo_review.py" ]
+    command: [ 'photo_review.py' ]
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - device_ids: [ "0", "1" ]
-              capabilities: [ "gpu" ]
-              driver: "nvidia"
 
   photo_review_5:
-    <<: *template
+    <<: *review_template
     container_name: photo_review_5
     hostname: photo_review_5
-    volumes:
-      - ./log:/app/log
-      - ./model:/app/model
     depends_on:
       - photo_review_4
-    command: [ "photo_review.py" ]
+    command: [ 'photo_review.py' ]
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - device_ids: [ "0", "1" ]
-              capabilities: [ "gpu" ]
-              driver: "nvidia"
 
   photo_mask_1:
-    <<: *template
+    <<: *mask_template
     container_name: photo_mask_1
     hostname: photo_mask_1
-    volumes:
-      - ./log:/app/log
     depends_on:
       - photo_review_5
-    command: [ "photo_mask.py", "--clean", "True" ]
+    command: [ 'photo_mask.py', '--clean', 'True' ]
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - device_ids: [ "1" ]
-              capabilities: [ "gpu" ]
-              driver: "nvidia"
 
   photo_mask_2:
-    <<: *template
+    <<: *mask_template
     container_name: photo_mask_2
     hostname: photo_mask_2
-    volumes:
-      - ./log:/app/log
     depends_on:
       - photo_mask_1
-    command: [ "photo_mask.py" ]
+    command: [ 'photo_mask.py' ]
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - device_ids: [ "1" ]
-              capabilities: [ "gpu" ]
-              driver: "nvidia"
-
-#  photo_review_6:
-#    <<: *template
-#    container_name: photo_review_6
-#    hostname: photo_review_6
-#    volumes:
-#      - ./log:/app/log
-#      - ./model:/app/model
-#    depends_on:
-#      - photo_mask_2
-#    command: [ "photo_review.py" ]
-#    deploy:
-#      resources:
-#        reservations:
-#          devices:
-#            - device_ids: [ "0", "1" ]
-#              capabilities: [ "gpu" ]
-#              driver: "nvidia"

photo_review/__init__.py

@@ -2,9 +2,9 @@ import jieba
 from paddlenlp import Taskflow
 from paddleocr import PaddleOCR
 
-"""
+'''
 项目配置
-"""
+'''
 # 每次从数据库获取的案子数量
 PHHD_BATCH_SIZE = 10
 # 没有查询到案子的等待时间(分钟)
@@ -18,35 +18,35 @@ LAYOUT_ANALYSIS = False
 信息抽取关键词配置
 """
 # 患者姓名
-PATIENT_NAME = ["患者姓名"]
+PATIENT_NAME = ['患者姓名']
 # 入院日期
-ADMISSION_DATE = ["入院日期"]
+ADMISSION_DATE = ['入院日期']
 # 出院日期
-DISCHARGE_DATE = ["出院日期"]
+DISCHARGE_DATE = ['出院日期']
 # 发生医疗费
-MEDICAL_EXPENSES = ["费用总额"]
+MEDICAL_EXPENSES = ['费用总额']
 # 个人现金支付
-PERSONAL_CASH_PAYMENT = ["个人现金支付"]
+PERSONAL_CASH_PAYMENT = ['个人现金支付']
 # 个人账户支付
-PERSONAL_ACCOUNT_PAYMENT = ["个人账户支付"]
+PERSONAL_ACCOUNT_PAYMENT = ['个人账户支付']
 # 个人自费金额
-PERSONAL_FUNDED_AMOUNT = ["自费金额"]
+PERSONAL_FUNDED_AMOUNT = ['自费金额', '个人自费']
 # 医保类别
-MEDICAL_INSURANCE_TYPE = ["医保类型"]
+MEDICAL_INSURANCE_TYPE = ['医保类型']
 # 就诊医院
-HOSPITAL = ["医院"]
+HOSPITAL = ['医院']
 # 就诊科室
-DEPARTMENT = ["科室"]
+DEPARTMENT = ['科室']
 # 主治医生
-DOCTOR = ["主治医生"]
+DOCTOR = ['主治医生']
 # 住院号
-ADMISSION_ID = ["住院号"]
+ADMISSION_ID = ['住院号']
 # 医保结算单号码
-SETTLEMENT_ID = ["医保结算单号码"]
+SETTLEMENT_ID = ['医保结算单号码']
 # 年龄
-AGE = ["年龄"]
+AGE = ['年龄']
 # 大写总额
-UPPERCASE_MEDICAL_EXPENSES = ["大写总额"]
+UPPERCASE_MEDICAL_EXPENSES = ['大写总额']
 
 SETTLEMENT_LIST_SCHEMA = \
     (PATIENT_NAME + ADMISSION_DATE + DISCHARGE_DATE + MEDICAL_EXPENSES + PERSONAL_CASH_PAYMENT
@@ -58,47 +58,47 @@ DISCHARGE_RECORD_SCHEMA = \
 
 COST_LIST_SCHEMA = PATIENT_NAME + ADMISSION_DATE + DISCHARGE_DATE + MEDICAL_EXPENSES
 
-"""
+'''
 别名配置
-"""
+'''
 # 使用别名中的value替换key。考虑到效率问题，只会替换第一个匹配到的key。
 HOSPITAL_ALIAS = {
-    "沐阳": ["沭阳"],
+    '沐阳': ['沭阳'],
-    "连水": ["涟水"],
+    '连水': ['涟水'],
-    "唯宁": ["睢宁"],
+    '唯宁': ['睢宁'],  # 雕宁
-    "九〇四": ["904"],
+    '九〇四': ['904'],
-    "漂水": ["溧水"],
+    '漂水': ['溧水'],
 }
 DEPARTMENT_ALIAS = {
-    "耳鼻喉": ["耳鼻咽喉"],
+    '耳鼻喉': ['耳鼻咽喉'],
-    "急症": ["急诊"],
+    '急症': ['急诊'],
 }
 
-"""
+'''
 搜索过滤配置
-"""
+'''
 # 默认会过滤单字
-HOSPITAL_FILTER = ["医院", "人民", "第一", "第二", "第三", "大学", "附属"]
+HOSPITAL_FILTER = ['医院', '人民', '第一', '第二', '第三', '大学', '附属']
 
-DEPARTMENT_FILTER = ["医", "伤", "西", "新"]
+DEPARTMENT_FILTER = ['医', '伤', '西', '新']
 
-"""
+'''
 分词配置
-"""
+'''
-jieba.suggest_freq(("肿瘤", "医院"), True)
+jieba.suggest_freq(('肿瘤', '医院'), True)
-jieba.suggest_freq(("骨", "伤"), True)
+jieba.suggest_freq(('骨', '伤'), True)
-jieba.suggest_freq(("感染", "性"), True)
+jieba.suggest_freq(('感染', '性'), True)
-jieba.suggest_freq(("胆", "道"), True)
+jieba.suggest_freq(('胆', '道'), True)
-jieba.suggest_freq(("脾", "胃"), True)
+jieba.suggest_freq(('脾', '胃'), True)
 
-"""
+'''
 模型配置
-"""
+'''
-SETTLEMENT_IE = Taskflow("information_extraction", schema=SETTLEMENT_LIST_SCHEMA, model="uie-x-base",
+SETTLEMENT_IE = Taskflow('information_extraction', schema=SETTLEMENT_LIST_SCHEMA, model='uie-x-base',
-                         task_path="model/settlement_list_model", layout_analysis=LAYOUT_ANALYSIS, precision='fp16')
+                         task_path='model/settlement_list_model', layout_analysis=LAYOUT_ANALYSIS, precision='fp16')
-DISCHARGE_IE = Taskflow("information_extraction", schema=DISCHARGE_RECORD_SCHEMA, model="uie-x-base",
+DISCHARGE_IE = Taskflow('information_extraction', schema=DISCHARGE_RECORD_SCHEMA, model='uie-x-base',
-                        task_path="model/discharge_record_model", layout_analysis=LAYOUT_ANALYSIS, precision='fp16')
+                        task_path='model/discharge_record_model', layout_analysis=LAYOUT_ANALYSIS, precision='fp16')
-COST_IE = Taskflow("information_extraction", schema=COST_LIST_SCHEMA, model="uie-x-base", device_id=1,
+COST_IE = Taskflow('information_extraction', schema=COST_LIST_SCHEMA, model='uie-x-base', device_id=1,
-                   task_path="model/cost_list_model", layout_analysis=LAYOUT_ANALYSIS, precision='fp16')
+                   task_path='model/cost_list_model', layout_analysis=LAYOUT_ANALYSIS, precision='fp16')
 
 OCR = PaddleOCR(use_angle_cls=False, show_log=False, gpu_id=1, det_db_box_thresh=0.3)

photo_review/auto_photo_review.py

@@ -7,22 +7,23 @@ from collections import defaultdict
 from time import sleep
 
 import cv2
+import fitz
 import jieba
+import numpy as np
 import requests
+import zxingcpp
 from rapidfuzz import process, fuzz
 from sqlalchemy import update
 
 from db import MysqlSession
 from db.mysql import BdYljg, BdYlks, ZxIeResult, ZxIeCost, ZxIeDischarge, ZxIeSettlement, ZxPhhd, ZxPhrec
-from doc_dewarp import dewarp
 from log import HOSTNAME
-from paddle_detection import detector
 from photo_review import PATIENT_NAME, ADMISSION_DATE, DISCHARGE_DATE, MEDICAL_EXPENSES, PERSONAL_CASH_PAYMENT, \
     PERSONAL_ACCOUNT_PAYMENT, PERSONAL_FUNDED_AMOUNT, MEDICAL_INSURANCE_TYPE, HOSPITAL, DEPARTMENT, DOCTOR, \
     ADMISSION_ID, SETTLEMENT_ID, AGE, OCR, SETTLEMENT_IE, DISCHARGE_IE, COST_IE, PHHD_BATCH_SIZE, SLEEP_MINUTES, \
     UPPERCASE_MEDICAL_EXPENSES, HOSPITAL_ALIAS, HOSPITAL_FILTER, DEPARTMENT_ALIAS, DEPARTMENT_FILTER
 from ucloud import ufile
-from util import image_util, util
+from util import image_util, util, html_util
 from util.data_util import handle_date, handle_decimal, parse_department, handle_name, \
     handle_insurance_type, handle_original_data, handle_hospital, handle_department, handle_id, handle_age, parse_money, \
     parse_hospital
@@ -73,6 +74,47 @@ def request_ie_result(task_enum, phrecs):
         raise Exception(f"请求信息抽取结果失败，状态码：{response.status_code}")
 
 
+# 尝试从二维码中获取高清图片
+def get_better_image_from_qrcode(image, dpi=150):
+    js_base_url = 'http://einvoice.jsczt.cn'
+    results = zxingcpp.read_barcodes(image)
+    for result in results:
+        pdf = None
+        pdf_path = None
+        try:
+            url = result.text
+            if url.startswith(js_base_url):
+                id_base = html_util.get_jsczt_id_base(url)
+                pdf_url = f'{js_base_url}/download?idBase={id_base}'
+                pdf_path = html_util.download_pdf(pdf_url)
+                # 打开PDF文件
+                pdf = fitz.open(pdf_path)
+                # 选择第一页
+                page = pdf[0]
+                # 定义缩放系数（DPI）
+                default_dpi = 72
+                zoom = dpi / default_dpi
+                # 设置矩阵变换参数
+                mat = fitz.Matrix(zoom, zoom)
+                # 渲染页面
+                pix = page.get_pixmap(matrix=mat)
+                # 将渲染结果转换为OpenCV兼容的格式
+                img = np.frombuffer(pix.samples, dtype=np.uint8).reshape((pix.height, pix.width, -1))
+                img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+                return img, page.get_text()
+            else:
+                logging.getLogger('qr').info(f'未知二维码内容：{url}')
+        except Exception as e:
+            logging.getLogger('error').error('从二维码中获取高清图片时出错', exc_info=e)
+            continue
+        finally:
+            if pdf:
+                pdf.close()
+            if pdf_path:
+                util.delete_temp_file(pdf_path)
+    return None, None
+
+
 # 关键信息提取
 def information_extraction(ie, phrecs, identity):
     result = {}
@@ -83,60 +125,88 @@ def information_extraction(ie, phrecs, identity):
 
         image = image_util.read(img_path)
 
-        target_images = []
+        # 尝试从二维码中获取高清图片
-        target_images += detector.request_book_areas(image)  # 识别文档区域并裁剪
+        better_image, text = get_better_image_from_qrcode(image)
-        if not target_images:
-            target_images.append(image)  # 识别失败
-        angle_count = defaultdict(int, {"0": 0})  # 分割后图片的最优角度统计
-        for target_image in target_images:
-            dewarped_image = dewarp.dewarp_image(target_image)  # 去扭曲
-            angles = image_util.parse_rotation_angles(dewarped_image)
         zx_ie_results = []
-            split_results = image_util.split(dewarped_image)
+        if better_image is not None:
-            for split_result in split_results:
+            img_angle = '0'
-                if split_result["img"] is None or split_result["img"].size == 0:
+            image = better_image
-                    continue
+            if text:
-                rotated_img = image_util.rotate(split_result["img"], int(angles[0]))
+                info_extract = ie(text)[0]
-                ie_results = [{"result": ie_temp_image(ie, OCR, rotated_img), "angle": angles[0]}]
+            else:
-                if not ie_results[0]["result"] or len(ie_results[0]["result"]) < len(ie.kwargs.get("schema")):
+                info_extract = ie_temp_image(ie, OCR, image)
-                    rotated_img = image_util.rotate(split_result["img"], int(angles[1]))
+            ie_result = {'result': info_extract, 'angle': '0'}
-                    ie_results.append({"result": ie_temp_image(ie, OCR, rotated_img), "angle": angles[1]})
 
             now = util.get_default_datetime()
-                best_angle = ["0", 0]
+            if not ie_result['result']:
-                for ie_result in ie_results:
-                    if not ie_result["result"]:
                 continue
 
-                    result_json = json.dumps(ie_result["result"], ensure_ascii=False)
+            result_json = json.dumps(ie_result['result'], ensure_ascii=False)
             if len(result_json) > 5000:
                 result_json = result_json[:5000]
             zx_ie_results.append(ZxIeResult(pk_phhd=phrec.pk_phhd, pk_phrec=phrec.pk_phrec, id=identity,
                                             cfjaddress=phrec.cfjaddress, content=result_json,
-                                                    rotation_angle=int(ie_result["angle"]),
+                                            rotation_angle=int(ie_result['angle']),
-                                                    x_offset=split_result["x_offset"],
+                                            x_offset=0, y_offset=0, create_time=now,
-                                                    y_offset=split_result["y_offset"], create_time=now,
                                             creator=HOSTNAME, update_time=now, updater=HOSTNAME))
 
-                    result = merge_result(result, ie_result["result"])
+            result = merge_result(result, ie_result['result'])
+        else:
+            target_images = []
+            # target_images += detector.request_book_areas(image)  # 识别文档区域并裁剪
+            if not target_images:
+                target_images.append(image)  # 识别失败
+            angle_count = defaultdict(int, {'0': 0})  # 分割后图片的最优角度统计
+            for target_image in target_images:
+                # dewarped_image = dewarp.dewarp_image(target_image)  # 去扭曲
+                dewarped_image = target_image
+                angles = image_util.parse_rotation_angles(dewarped_image)
 
-                    if len(ie_result["result"]) > best_angle[1]:
+                split_results = image_util.split(dewarped_image)
-                        best_angle = [ie_result["angle"], len(ie_result["result"])]
+                for split_result in split_results:
+                    if split_result['img'] is None or split_result['img'].size == 0:
+                        continue
+                    rotated_img = image_util.rotate(split_result['img'], int(angles[0]))
+                    ie_results = [{'result': ie_temp_image(ie, OCR, rotated_img), 'angle': angles[0]}]
+                    if not ie_results[0]['result'] or len(ie_results[0]['result']) < len(ie.kwargs.get('schema')):
+                        rotated_img = image_util.rotate(split_result['img'], int(angles[1]))
+                        ie_results.append({'result': ie_temp_image(ie, OCR, rotated_img), 'angle': angles[1]})
+                    now = util.get_default_datetime()
+                    best_angle = ['0', 0]
+                    for ie_result in ie_results:
+                        if not ie_result['result']:
+                            continue
+
+                        result_json = json.dumps(ie_result['result'], ensure_ascii=False)
+                        if len(result_json) > 5000:
+                            result_json = result_json[:5000]
+                        zx_ie_results.append(ZxIeResult(pk_phhd=phrec.pk_phhd, pk_phrec=phrec.pk_phrec, id=identity,
+                                                        cfjaddress=phrec.cfjaddress, content=result_json,
+                                                        rotation_angle=int(ie_result['angle']),
+                                                        x_offset=split_result['x_offset'],
+                                                        y_offset=split_result['y_offset'], create_time=now,
+                                                        creator=HOSTNAME, update_time=now, updater=HOSTNAME))
+
+                        result = merge_result(result, ie_result['result'])
+
+                        if len(ie_result['result']) > best_angle[1]:
+                            best_angle = [ie_result['angle'], len(ie_result['result'])]
 
                     angle_count[best_angle[0]] += 1
-
             img_angle = max(angle_count, key=angle_count.get)
-        if img_angle != "0":
+
+        if img_angle != '0' or better_image is not None:
             image = image_util.rotate(image, int(img_angle))
-            with tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") as temp_file:
+            with tempfile.NamedTemporaryFile(delete=False, suffix='.jpg') as temp_file:
                 cv2.imwrite(temp_file.name, image)
             try:
                 ufile.upload_file(phrec.cfjaddress, temp_file.name)
+                if img_angle != '0':
                     # 修正旋转角度
                     for zx_ie_result in zx_ie_results:
                         zx_ie_result.rotation_angle -= int(img_angle)
             except Exception as e:
-                logging.error(f"上传图片({phrec.cfjaddress})失败", exc_info=e)
+                logging.error(f'上传图片({phrec.cfjaddress})失败', exc_info=e)
             finally:
                 util.delete_temp_file(temp_file.name)
 

requirements.txt

@@ -13,3 +13,4 @@ sqlacodegen==2.3.0.post1
 sqlalchemy==1.4.52
 tenacity==8.5.0
 ufile==3.2.9
+zxing-cpp==2.2.0

util/html_util.py

@@ -0,0 +1,43 @@
+import logging
+import tempfile
+
+import requests
+from bs4 import BeautifulSoup
+from tenacity import retry, stop_after_attempt, wait_random
+
+
+@retry(stop=stop_after_attempt(3), wait=wait_random(1, 3), reraise=True,
+       after=lambda x: logging.warning('获取江苏省财政票据idBase失败！'))
+def get_jsczt_id_base(url):
+    response = requests.get(url)
+    if response.status_code != 200:
+        raise Exception(f'请求江苏省财政票据失败！状态码: {response.status_code}')
+    soup = BeautifulSoup(response.text, 'html.parser')
+    hidden_input = soup.find('input', {'name': "idBase"})
+    if hidden_input:
+        # 获取隐藏字段的值
+        value = hidden_input.get('value')
+        return value
+    return None
+
+
+@retry(stop=stop_after_attempt(3), wait=wait_random(1, 3), reraise=True,
+       after=lambda x: logging.warning('下载pdf失败！'))
+def download_pdf(url, local_filename=None):
+    # 如果没有提供文件名，则使用URL中的文件名
+    if local_filename is None:
+        with tempfile.NamedTemporaryFile(delete=False, suffix='.pdf') as temp_pdf:
+            local_filename = temp_pdf.name
+    # 发送HTTP GET请求
+    response = requests.get(url, stream=True)
+    # 检查请求是否成功
+    if response.status_code != 200:
+        raise Exception(f'下载pdf失败！状态码: {response.status_code}')
+    else:
+        # 打开一个文件用于写入二进制数据
+        with open(local_filename, 'wb') as file:
+            # 迭代写入文件
+            for chunk in response.iter_content(chunk_size=8192):
+                if chunk:  # filter out keep-alive new chunks
+                    file.write(chunk)
+        return local_filename