修正heic图片保存问题

This reverts commit a9f172fdb0.

.dockerignore

@@ -238,8 +238,11 @@ cython_debug/
 #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
 #.idea/
 
-# 模型通过卷绑定挂载到容器中
+# 通过卷绑定挂载到容器中
-/model
+/log
+/services/paddle_services/log
+/services/paddle_services/model
+/tmp_img
 # docker
 Dockerfile
 docker-compose*.yml

.gitignore

@@ -142,7 +142,11 @@ cython_debug/
 .idea
 
 ### Model
-model
+services/paddle_services/model
 
 ### Log Backups
-*.log.*-*-*
+*.log.*-*-*
+
+### Tmp Files
+/tmp_img
+/test_img

Dockerfile

@@ -1,5 +1,5 @@
-# 使用官方的paddle镜像作为基础
+# 使用官方的python镜像作为基础
-FROM registry.baidubce.com/paddlepaddle/paddle:2.6.1-gpu-cuda12.0-cudnn8.9-trt8.6
+FROM python:3.10.15-bookworm
 
 # 设置工作目录
 WORKDIR /app
@@ -13,11 +13,10 @@ ENV PYTHONUNBUFFERED=1 \
 
 # 安装依赖
 COPY requirements.txt /app/requirements.txt
-COPY packages /app/packages
 RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo '$TZ' > /etc/timezone \
-    && pip install --no-cache-dir -r requirements.txt \
+    && sed -i 's|deb.debian.org|mirrors.aliyun.com|g' /etc/apt/sources.list.d/debian.sources \
-    && pip uninstall -y onnxruntime onnxruntime-gpu \
+    && apt-get update && apt-get install libgl1 -y \
-    && pip install onnxruntime-gpu==1.18.0 --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/
+    && pip install --no-cache-dir -r requirements.txt
 
 # 将当前目录内容复制到容器的/app内
 COPY . /app

README.md

@@ -6,7 +6,7 @@
 
 1. 从Git远程仓库克隆项目到本地。
 
-2. 将深度学习模型复制到./model目录下，具体请看[模型更新](#模型更新)部分。
+2. 将深度学习模型复制到./services/paddle_services/model目录下，具体请看[模型更新](#模型更新)部分。
 
 3. 安装docker和docker-compose。
 
@@ -125,4 +125,6 @@ bash update.sh
     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
+    1. 新增二维码识别替换高清图片功能
+22. 版本号：2.0.0
+    1. 项目架构调整，模型全部采用接口调用

api_test.py

@@ -0,0 +1,4 @@
+import time
+
+if __name__ == '__main__':
+    time.sleep(3600)

auto_generator.py

@@ -1,15 +0,0 @@
-# 自动生成数据库表和sqlalchemy对应的Model
-import subprocess
-
-from db import DB_URL
-
-if __name__ == '__main__':
-    table = input("请输入表名：")
-    out_file = f"db/{table}.py"
-    command = f"sqlacodegen {DB_URL} --outfile={out_file} --tables={table}"
-
-    try:
-        subprocess.run(command, shell=True, check=True)
-        print(f"{table}.py文件生成成功！请检查并复制到合适的文件中！")
-    except Exception as e:
-        print(f"生成{table}.py文件时发生错误: {e}")

check_ie_result/__init__.py

@@ -1 +0,0 @@
-

check_ie_result/check_ie_result.py

@@ -1,102 +0,0 @@
-import datetime
-import json
-import os
-from decimal import Decimal
-from io import BytesIO
-from itertools import groupby
-
-import requests
-from PIL import ImageDraw, Image, ImageFont
-
-from db import MysqlSession
-from db.mysql import ZxIeCost, ZxIeDischarge, ZxIeSettlement, ZxPhhd, ZxIeResult, ZxPhrec
-from ucloud import ufile
-from util import image_util
-
-
-def check_ie_result(pk_phhd):
-    os.makedirs(f"./check_result/{pk_phhd}", exist_ok=True)
-    json_result = {"pk_phhd": pk_phhd}
-    session = MysqlSession()
-    phhd = session.query(ZxPhhd.cXm).filter(ZxPhhd.pk_phhd == pk_phhd).one()
-    json_result["cXm"] = phhd.cXm
-    settlement = (session.query(ZxIeSettlement.pk_ie_settlement, ZxIeSettlement.name, ZxIeSettlement.admission_date,
-                                ZxIeSettlement.discharge_date, ZxIeSettlement.medical_expenses,
-                                ZxIeSettlement.personal_cash_payment, ZxIeSettlement.personal_account_payment,
-                                ZxIeSettlement.personal_funded_amount, ZxIeSettlement.medical_insurance_type,
-                                ZxIeSettlement.admission_id, ZxIeSettlement.settlement_id)
-                  .filter(ZxIeSettlement.pk_phhd == pk_phhd).one())
-    settlement_result = settlement._asdict()
-    json_result["settlement"] = settlement_result
-
-    discharge = (session.query(ZxIeDischarge.pk_ie_discharge, ZxIeDischarge.hospital, ZxIeDischarge.pk_yljg,
-                               ZxIeDischarge.department, ZxIeDischarge.pk_ylks, ZxIeDischarge.name, ZxIeDischarge.age,
-                               ZxIeDischarge.admission_date, ZxIeDischarge.discharge_date, ZxIeDischarge.doctor,
-                               ZxIeDischarge.admission_id)
-                 .filter(ZxIeDischarge.pk_phhd == pk_phhd).one())
-    discharge_result = discharge._asdict()
-    json_result["discharge"] = discharge_result
-
-    cost = session.query(ZxIeCost.pk_ie_cost, ZxIeCost.name, ZxIeCost.admission_date, ZxIeCost.discharge_date,
-                         ZxIeCost.medical_expenses).filter(ZxIeCost.pk_phhd == pk_phhd).one()
-    cost_result = cost._asdict()
-    json_result["cost"] = cost_result
-
-    phrecs = session.query(ZxPhrec.pk_phrec, ZxPhrec.pk_phhd, ZxPhrec.cRectype, ZxPhrec.cfjaddress).filter(
-        ZxPhrec.pk_phhd == pk_phhd).all()
-    for phrec in phrecs:
-        img_name = phrec.cfjaddress
-        img_path = ufile.get_private_url(img_name)
-
-        response = requests.get(img_path)
-        image = Image.open(BytesIO(response.content)).convert("RGB")
-        font_size = image.width * image.height / 200000
-        font = ImageFont.truetype("./font/simfang.ttf", size=font_size)
-
-        ocr = session.query(ZxIeResult.id, ZxIeResult.content, ZxIeResult.rotation_angle, ZxIeResult.x_offset,
-                            ZxIeResult.y_offset).filter(ZxIeResult.pk_phrec == phrec.pk_phrec).all()
-        if not ocr:
-            os.makedirs(f"./check_result/{pk_phhd}/0", exist_ok=True)
-            image.save(f"./check_result/{pk_phhd}/0/{img_name}")
-
-        for _, group_results in groupby(ocr, key=lambda x: x.id):
-            draw = ImageDraw.Draw(image)
-            for ocr_item in group_results:
-                result = json.loads(ocr_item.content)
-                rotation_angle = ocr_item.rotation_angle
-                x_offset = ocr_item.x_offset
-                y_offset = ocr_item.y_offset
-                for key in result:
-                    for value in result[key]:
-                        box = value["bbox"][0]
-
-                        if rotation_angle:
-                            box = image_util.invert_rotate_rectangle(box, (image.width / 2, image.height / 2),
-                                                                     rotation_angle)
-                        if x_offset:
-                            box[0] += x_offset
-                            box[2] += x_offset
-                        if y_offset:
-                            box[1] += y_offset
-                            box[3] += y_offset
-
-                        draw.rectangle(box, outline="red", width=2)  # 绘制矩形
-                        draw.text((box[0], box[1] - font_size), key, fill="blue", font=font)  # 在矩形上方绘制文本
-                        draw.text((box[0], box[3]), value["text"], fill="blue", font=font)  # 在矩形下方绘制文本
-                os.makedirs(f"./check_result/{pk_phhd}/{ocr_item.id}", exist_ok=True)
-                image.save(f"./check_result/{pk_phhd}/{ocr_item.id}/{img_name}")
-    session.close()
-
-    # 自定义JSON处理器
-    def default(obj):
-        if isinstance(obj, Decimal):
-            return float(obj)
-        if isinstance(obj, datetime.date):
-            return obj.strftime("%Y-%m-%d")
-
-    with open(f"./check_result/{pk_phhd}/result.json", "w", encoding="utf-8") as json_file:
-        json.dump(json_result, json_file, indent=4, ensure_ascii=False, default=default)
-
-
-if __name__ == '__main__':
-    check_ie_result(0)

db/mysql.py

@@ -63,6 +63,8 @@ class ZxIeCost(Base):
     discharge_date = Column(Date, comment='出院日期')
     medical_expenses_str = Column(String(255), comment='费用总额字符串')
     medical_expenses = Column(DECIMAL(18, 2), comment='费用总额')
+    page_nums = Column(String(255), comment='页码')
+    page_count = Column(TINYINT(4), comment='页数')
     create_time = Column(DateTime, server_default=text("CURRENT_TIMESTAMP"), comment='创建时间')
     creator = Column(String(255), comment='创建人')
     update_time = Column(DateTime, server_default=text("CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP"),
@@ -120,7 +122,7 @@ class ZxIeResult(Base):
     pk_ocr = Column(INTEGER(11), primary_key=True, comment='图片OCR识别主键')
     pk_phhd = Column(INTEGER(11), nullable=False, comment='报销单主键')
     pk_phrec = Column(INTEGER(11), nullable=False, comment='图片主键')
-    id = Column(INTEGER(11), nullable=False, comment='识别批次')
+    id = Column(CHAR(32), nullable=False, comment='识别批次')
     cfjaddress = Column(String(200), nullable=False, comment='云存储文件名')
     content = Column(String(5000), comment='OCR识别内容')
     rotation_angle = Column(INTEGER(11), comment='旋转角度')
@@ -412,17 +414,19 @@ class ZxIeReview(Base):
     pk_ie_review = Column(INTEGER(11), primary_key=True, comment='自动审核主键')
     pk_phhd = Column(INTEGER(11), nullable=False, comment='报销案子主键')
     success = Column(BIT(1))
+    integrity = Column(BIT(1))
     has_settlement = Column(BIT(1))
     has_discharge = Column(BIT(1))
     has_cost = Column(BIT(1))
     full_page = Column(BIT(1))
     page_description = Column(String(255), comment='具体缺页描述')
+    consistency = Column(BIT(1), comment='三项资料一致性。0：不一致；1：一致')
     name_match = Column(CHAR(1), server_default=text("'0'"),
-                        comment='三项资料姓名是否一致。0：不一致；1：一致；2：结算单不一致；3：出院记录不一致；4：费用清单不一致；5：与报销申请对象不一致')
+                        comment='三项资料姓名是否一致。0：都不一致；1：一致；2：结算单不一致；3：出院记录不一致；4：费用清单不一致；5：与报销申请对象不一致')
     admission_date_match = Column(CHAR(1), server_default=text("'0'"),
-                                  comment='三项资料入院日期是否一致。0：不一致；1：一致；2：结算单不一致；3：出院记录不一致；4：费用清单不一致')
+                                  comment='三项资料入院日期是否一致。0：都不一致；1：一致；2：结算单不一致；3：出院记录不一致；4：费用清单不一致')
     discharge_date_match = Column(CHAR(1), server_default=text("'0'"),
-                                  comment='三项资料出院日期是否一致。0：不一致；1：一致；2：结算单不一致；3：出院记录不一致；4：费用清单不一致')
+                                  comment='三项资料出院日期是否一致。0：都不一致；1：一致；2：结算单不一致；3：出院记录不一致；4：费用清单不一致')
     create_time = Column(DateTime, server_default=text("CURRENT_TIMESTAMP"), comment='创建时间')
     creator = Column(String(255), comment='创建人')
     update_time = Column(DateTime, server_default=text("CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP"),

det_api.py

@@ -1,32 +0,0 @@
-import base64
-
-import cv2
-import numpy as np
-from flask import Flask, request, jsonify
-
-from paddle_detection import detector
-
-app = Flask(__name__)
-
-
-@app.route("/det/detect_books", methods=['POST'])
-def detect_books():
-    try:
-        file = request.files['image']
-        image_data = file.read()
-        nparr = np.frombuffer(image_data, np.uint8)
-        image = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
-        result = detector.get_book_areas(image)
-        encoded_images = []
-        for i in result:
-            _, encoded_image = cv2.imencode('.jpg', i)
-            byte_stream = encoded_image.tobytes()
-            img_str = base64.b64encode(byte_stream).decode('utf-8')
-            encoded_images.append(img_str)
-        return jsonify(encoded_images), 200
-    except Exception as e:
-        return jsonify({'error': str(e)}), 500
-
-
-if __name__ == '__main__':
-    app.run("0.0.0.0")

doc_dewarp/__init__.py

@@ -1,4 +0,0 @@
-from onnxruntime import InferenceSession
-
-DOC_TR = InferenceSession("model/dewarp_model/doc_tr_pp.onnx",
-                          providers=["CUDAExecutionProvider"], provider_options=[{"device_id": 0}])

docker-compose.yml

@@ -1,46 +1,32 @@
-x-env:
+x-base:
-  &template
+  &base_template
-  image: fcb_photo_review:1.14.6
   restart: always
 
-x-review:
+x-project:
-  &review_template
+  &project_template
-  <<: *template
+  <<: *base_template
+  image: fcb_photo_review:2.0.0
   volumes:
     - ./log:/app/log
-    - ./model:/app/model
+    - ./tmp_img:/app/tmp_img
-  deploy:
-    resources:
-      reservations:
-        devices:
-          - device_ids: [ '0', '1' ]
-            capabilities: [ 'gpu' ]
-            driver: 'nvidia'
 
-x-mask:
+x-paddle:
-  &mask_template
+  &paddle_template
-  <<: *template
+  <<: *base_template
+  image: fcb_paddle:0.0.1
   volumes:
-    - ./log:/app/log
+    - ./services/paddle_services/log:/app/log
-  deploy:
+    - ./services/paddle_services/model:/app/model
-    resources:
+    - ./tmp_img:/app/tmp_img
-      reservations:
-        devices:
-          - device_ids: [ '1' ]
-            capabilities: [ 'gpu' ]
-            driver: 'nvidia'
 
 services:
-  det_api:
+  ocr:
-    <<: *template
+    <<: *paddle_template
     build:
-      context: .
+      context: ./services/paddle_services
-    container_name: det_api
+    container_name: ocr
-    hostname: det_api
+    hostname: ocr
-    volumes:
+    command: [ '-w', '4', 'ocr:app', '--bind', '0.0.0.0:5001' ]
-      - ./log:/app/log
-      - ./model:/app/model
-#    command: [ 'det_api.py' ]
     deploy:
       resources:
         reservations:
@@ -49,16 +35,116 @@ services:
               capabilities: [ 'gpu' ]
               driver: 'nvidia'
 
+  ie_settlement:
+    <<: *paddle_template
+    container_name: ie_settlement
+    hostname: ie_settlement
+    command: [ '-w', '5', 'ie_settlement:app', '--bind', '0.0.0.0:5002' ]
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - device_ids: [ '0' ]
+              capabilities: [ 'gpu' ]
+              driver: 'nvidia'
+
+  ie_discharge:
+    <<: *paddle_template
+    container_name: ie_discharge
+    hostname: ie_discharge
+    command: [ '-w', '5', 'ie_discharge:app', '--bind', '0.0.0.0:5003' ]
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - device_ids: [ '1' ]
+              capabilities: [ 'gpu' ]
+              driver: 'nvidia'
+
+  ie_cost:
+    <<: *paddle_template
+    container_name: ie_cost
+    hostname: ie_cost
+    command: [ '-w', '5', 'ie_cost:app', '--bind', '0.0.0.0:5004' ]
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - device_ids: [ '1' ]
+              capabilities: [ 'gpu' ]
+              driver: 'nvidia'
+
+  clas_orientation:
+    <<: *paddle_template
+    container_name: clas_orientation
+    hostname: clas_orientation
+    command: [ '-w', '3', 'clas_orientation:app', '--bind', '0.0.0.0:5005' ]
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - device_ids: [ '0' ]
+              capabilities: [ 'gpu' ]
+              driver: 'nvidia'
+
+  det_book:
+    <<: *paddle_template
+    container_name: det_book
+    hostname: det_book
+    command: [ '-w', '4', 'det_book:app', '--bind', '0.0.0.0:5006' ]
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - device_ids: [ '1' ]
+              capabilities: [ 'gpu' ]
+              driver: 'nvidia'
+
+  dewarp:
+    <<: *paddle_template
+    container_name: dewarp
+    hostname: dewarp
+    command: [ '-w', '4', 'dewarp:app', '--bind', '0.0.0.0:5007' ]
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - device_ids: [ '0' ]
+              capabilities: [ 'gpu' ]
+              driver: 'nvidia'
+
+  #  clas_text:
+  #    <<: *paddle_template
+  #    container_name: clas_text
+  #    hostname: clas_text
+  #    command: [ '-w', '3', 'clas_text:app', '--bind', '0.0.0.0:5008' ]
+  #    deploy:
+  #      resources:
+  #        reservations:
+  #          devices:
+  #            - device_ids: [ '1' ]
+  #              capabilities: [ 'gpu' ]
+  #              driver: 'nvidia'
+
   photo_review_1:
-    <<: *review_template
+    <<: *project_template
+    build:
+      context: .
     container_name: photo_review_1
     hostname: photo_review_1
     depends_on:
-      - det_api
+      - ocr
+      - ie_settlement
+      - ie_discharge
+      - ie_cost
+      - clas_orientation
+      - det_book
+      - dewarp
+    #      - clas_text
     command: [ 'photo_review.py', '--clean', 'True' ]
 
   photo_review_2:
-    <<: *review_template
+    <<: *project_template
     container_name: photo_review_2
     hostname: photo_review_2
     depends_on:
@@ -66,57 +152,41 @@ services:
     command: [ 'photo_review.py' ]
 
   photo_review_3:
-    <<: *review_template
+    <<: *project_template
     container_name: photo_review_3
     hostname: photo_review_3
     depends_on:
-      - photo_review_2
+      - photo_review_1
     command: [ 'photo_review.py' ]
 
   photo_review_4:
-    <<: *review_template
+    <<: *project_template
     container_name: photo_review_4
     hostname: photo_review_4
     depends_on:
-      - photo_review_3
+      - photo_review_1
     command: [ 'photo_review.py' ]
 
   photo_review_5:
-    <<: *review_template
+    <<: *project_template
     container_name: photo_review_5
     hostname: photo_review_5
     depends_on:
-      - photo_review_4
+      - photo_review_1
     command: [ 'photo_review.py' ]
 
   photo_mask_1:
-    <<: *mask_template
+    <<: *project_template
     container_name: photo_mask_1
     hostname: photo_mask_1
     depends_on:
-      - photo_review_5
+      - photo_review_1
     command: [ 'photo_mask.py', '--clean', 'True' ]
 
   photo_mask_2:
-    <<: *mask_template
+    <<: *project_template
     container_name: photo_mask_2
     hostname: photo_mask_2
     depends_on:
       - photo_mask_1
-    command: [ 'photo_mask.py' ]
+    command: [ 'photo_mask.py' ]
-#
-#  photo_review_6:
-#    <<: *review_template
-#    container_name: photo_review_6
-#    hostname: photo_review_6
-#    depends_on:
-#      - photo_mask_2
-#    command: [ 'photo_review.py' ]
-#
-#  photo_review_7:
-#    <<: *review_template
-#    container_name: photo_review_7
-#    hostname: photo_review_7
-#    depends_on:
-#      - photo_review_6
-#    command: [ 'photo_review.py' ]

log/__init__.py

@@ -1,13 +1,15 @@
 import os
 import socket
 
+# 项目根目录
+PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
 # 获取主机名，方便区分容器
 HOSTNAME = socket.gethostname()
 # 检测日志文件的路径是否存在，不存在则创建
 LOG_PATHS = [
-    f"log/{HOSTNAME}/ucloud",
+    os.path.join(PROJECT_ROOT, 'log', HOSTNAME, 'ucloud'),
-    f"log/{HOSTNAME}/error",
+    os.path.join(PROJECT_ROOT, 'log', HOSTNAME, 'error'),
-    f"log/{HOSTNAME}/qr",
+    os.path.join(PROJECT_ROOT, 'log', HOSTNAME, 'qr'),
 ]
 for path in LOG_PATHS:
     if not os.path.exists(path):

my_email/__init__.py

@@ -8,15 +8,13 @@ MAX_WAIT_TIME = 3
 # 程序异常短信配置
 ERROR_EMAIL_CONFIG = {
     # SMTP服务器地址
-    "smtp_server": "smtp.163.com",
+    'smtp_server': 'smtp.163.com',
     # 连接SMTP的端口
-    "port": 994,
+    'port': 994,
     # 发件人邮箱地址，请确保开启了SMTP邮件服务！
-    "sender": "EchoLiu618@163.com",
+    'sender': 'EchoLiu618@163.com',
     # 授权码--用于登录第三方邮件客户端的专用密码，不是邮箱密码
-    "authorization_code": "OKPQLIIVLVGRZYVH",
+    'authorization_code': 'OKPQLIIVLVGRZYVH',
     # 收件人邮箱地址
-    "receivers": ["1515783401@qq.com"],
+    'receivers': ['1515783401@qq.com'],
-    # 尝试次数
-    "retry_times": 3,
 }

my_email/error_email.py

@@ -5,18 +5,18 @@ from email.mime.text import MIMEText
 
 from tenacity import retry, stop_after_attempt, wait_random
 
-from auto_email import ERROR_EMAIL_CONFIG, TRY_TIMES, MIN_WAIT_TIME, MAX_WAIT_TIME
 from log import HOSTNAME
+from my_email import ERROR_EMAIL_CONFIG, TRY_TIMES, MIN_WAIT_TIME, MAX_WAIT_TIME
 
 
 @retry(stop=stop_after_attempt(TRY_TIMES), wait=wait_random(MIN_WAIT_TIME, MAX_WAIT_TIME), reraise=True,
-       after=lambda x: logging.warning("发送邮件失败！"))
+       after=lambda x: logging.warning('发送邮件失败！'))
 def send_email(email_config, massage):
-    smtp_server = email_config["smtp_server"]
+    smtp_server = email_config['smtp_server']
-    port = email_config["port"]
+    port = email_config['port']
-    sender = email_config["sender"]
+    sender = email_config['sender']
-    authorization_code = email_config["authorization_code"]
+    authorization_code = email_config['authorization_code']
-    receivers = email_config["receivers"]
+    receivers = email_config['receivers']
     mail = smtplib.SMTP_SSL(smtp_server, port)  # 连接SMTP服务
     mail.login(sender, authorization_code)  # 登录到SMTP服务
     mail.sendmail(sender, receivers, massage.as_string())  # 发送邮件
@@ -34,13 +34,13 @@ def send_error_email(program_name, error_name, error_detail):
     """
 
     # SMTP 服务器配置
-    sender = ERROR_EMAIL_CONFIG["sender"]
+    sender = ERROR_EMAIL_CONFIG['sender']
-    receivers = ERROR_EMAIL_CONFIG["receivers"]
+    receivers = ERROR_EMAIL_CONFIG['receivers']
 
     # 获取程序出错的时间
-    error_time = datetime.datetime.strftime(datetime.datetime.today(), "%Y-%m-%d %H:%M:%S:%f")
+    error_time = datetime.datetime.strftime(datetime.datetime.today(), '%Y-%m-%d %H:%M:%S:%f')
     # 邮件内容
-    subject = f"【程序异常提醒】{program_name}({HOSTNAME}) {error_time}"  # 邮件的标题
+    subject = f'【程序异常提醒】{program_name}({HOSTNAME}) {error_time}'  # 邮件的标题
     content = f'''<div class="emailcontent" style="width:100%;max-width:720px;text-align:left;margin:0 auto;padding-top:80px;padding-bottom:20px">
         <div class="emailtitle">
             <h1 style="color:#fff;background:#51a0e3;line-height:70px;font-size:24px;font-weight:400;padding-left:40px;margin:0">程序运行异常通知</h1>

paddle_detection/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/__init__.py

@@ -1,4 +0,0 @@
-from onnxruntime import InferenceSession
-
-PADDLE_DET = InferenceSession("model/object_det_model/ppyoloe_plus_crn_l_80e_coco_w_nms.onnx",
-                              providers=["CPUExecutionProvider"], provider_options=[{"device_id": 0}])

paddle_detection/configs/mot/bytetrack/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/configs/mot/bytetrack/detector/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/configs/mot/centertrack/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/configs/mot/deepsort/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/configs/mot/headtracking21/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/configs/mot/mtmct/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/configs/mot/ocsort/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/configs/mot/pedestrian/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/configs/mot/vehicle/README.md

@@ -1 +0,0 @@
-README_cn.md

paddle_detection/deploy/pptracking/README.md

@@ -1 +0,0 @@
-README_en.md

paddle_detection/detector.py

@@ -1,76 +0,0 @@
-import base64
-import logging
-import tempfile
-from collections import defaultdict
-
-import cv2
-import numpy as np
-import requests
-from tenacity import retry, stop_after_attempt, wait_random
-
-from paddle_detection import PADDLE_DET
-from paddle_detection.deploy.third_engine.onnx.infer import PredictConfig
-from paddle_detection.deploy.third_engine.onnx.preprocess import Compose
-from util import image_util, util
-
-
-def predict_image(infer_config, predictor, img_path):
-    # load preprocess transforms
-    transforms = Compose(infer_config.preprocess_infos)
-    # predict image
-    inputs = transforms(img_path)
-    inputs["image"] = np.array(inputs["image"]).astype('float32')
-    inputs_name = [var.name for var in predictor.get_inputs()]
-    inputs = {k: inputs[k][None,] for k in inputs_name}
-
-    outputs = predictor.run(output_names=None, input_feed=inputs)
-
-    bboxes = np.array(outputs[0])
-    result = defaultdict(list)
-    for bbox in bboxes:
-        if bbox[0] > -1 and bbox[1] > infer_config.draw_threshold:
-            result[bbox[0]].append({"score": bbox[1], "box": bbox[2:]})
-    return result
-
-
-def detect_image(img_path):
-    infer_cfg = "model/object_det_model/infer_cfg.yml"
-    # load infer config
-    infer_config = PredictConfig(infer_cfg)
-
-    return predict_image(infer_config, PADDLE_DET, img_path)
-
-
-def get_book_areas(image):
-    with tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") as temp_file:
-        cv2.imwrite(temp_file.name, image)
-    detect_result = detect_image(temp_file.name)
-    util.delete_temp_file(temp_file.name)
-    book_areas = detect_result[73]
-    result = []
-    for book_area in book_areas:
-        result.append(image_util.capture(image, book_area["box"]))
-    return result
-
-
-@retry(stop=stop_after_attempt(3), wait=wait_random(1, 3), reraise=True,
-       after=lambda x: logging.warning("获取文档区域失败！"))
-def request_book_areas(image):
-    url = "http://det_api:5000/det/detect_books"
-    _, encoded_image = cv2.imencode('.jpg', image)
-    byte_stream = encoded_image.tobytes()
-    files = {"image": ("image.jpg", byte_stream)}
-    response = requests.post(url, files=files)
-    if response.status_code == 200:
-        img_str_list = response.json()
-        result = []
-        for img_str in img_str_list:
-            img_data = base64.b64decode(img_str)
-            np_array = np.frombuffer(img_data, np.uint8)
-            img = cv2.imdecode(np_array, cv2.IMREAD_COLOR)
-            height, width = img.shape[:2]
-            if max(height, width) / min(height, width) <= 6.5:
-                result.append(img)  # 过滤异常结果
-        return result
-    else:
-        return []

photo_mask.py

@@ -5,35 +5,36 @@ from time import sleep
 
 from sqlalchemy import update
 
-from auto_email.error_email import send_error_email
 from db import MysqlSession
 from db.mysql import ZxPhhd
 from log import LOGGING_CONFIG
+from my_email.error_email import send_error_email
 from photo_mask import auto_photo_mask, SEND_ERROR_EMAIL
 
 if __name__ == '__main__':
-    program_name = "照片审核自动涂抹脚本"
+    program_name = '照片审核自动涂抹脚本'
     logging.config.dictConfig(LOGGING_CONFIG)
 
+    logging.info('等待接口服务启动...')
+    sleep(60)
+
     parser = argparse.ArgumentParser()
-    parser.add_argument("--clean", default=False, type=bool, help="是否将涂抹中的案子改为待涂抹状态")
+    parser.add_argument('--clean', default=False, type=bool, help='是否将涂抹中的案子改为待涂抹状态')
     args = parser.parse_args()
     if args.clean:
         # 主要用于启动时，清除仍在涂抹中的案子
         session = MysqlSession()
-        update_flag = (update(ZxPhhd).where(ZxPhhd.paint_flag == "2").values(paint_flag="1"))
+        update_flag = (update(ZxPhhd).where(ZxPhhd.paint_flag == '2').values(paint_flag='1'))
         session.execute(update_flag)
         session.commit()
         session.close()
-        logging.info("已释放残余的涂抹案子！")
+        logging.info('已释放残余的涂抹案子！')
-    else:
-        sleep(5)
 
     try:
-        logging.info(f"【{program_name}】开始运行")
+        logging.info(f'【{program_name}】开始运行')
         auto_photo_mask.main()
     except Exception as e:
-        error_logger = logging.getLogger("error")
+        error_logger = logging.getLogger('error')
         error_logger.error(traceback.format_exc())
         if SEND_ERROR_EMAIL:
             send_error_email(program_name, repr(e), traceback.format_exc())

photo_mask/__init__.py

@@ -1,5 +1,3 @@
-from paddleocr import PaddleOCR
-
 """
 项目配置
 """
@@ -40,13 +38,3 @@ SIMILAR_CHAR = {
     "侯": ["候"],
     "宇": ["字"],
 }
-
-# 如果不希望识别出空格，可以设置use_space_char=False。做此项设置一定要测试，2.7.3版本此项设置有bug，会导致识别失败
-OCR = PaddleOCR(
-    gpu_id=0,
-    show_log=False,
-    det_db_thresh=0.1,
-    det_db_box_thresh=0.3,
-    det_limit_side_len=1248,
-    drop_score=0.3
-)

photo_mask/auto_photo_mask.py

@@ -1,7 +1,9 @@
 import logging.config
+import os
 import re
-import tempfile
+import shutil
 import time
+import uuid
 from time import sleep
 
 import cv2
@@ -10,9 +12,10 @@ from sqlalchemy import update, and_
 from db import MysqlSession
 from db.mysql import ZxPhrec, ZxPhhd
 from log import HOSTNAME
-from photo_mask import OCR, PHHD_BATCH_SIZE, SLEEP_MINUTES, NAME_KEYS, ID_CARD_NUM_KEYS, SIMILAR_CHAR
+from photo_mask import PHHD_BATCH_SIZE, SLEEP_MINUTES, NAME_KEYS, ID_CARD_NUM_KEYS, SIMILAR_CHAR
+from photo_review import set_batch_id
 from ucloud import BUCKET, ufile
-from util import image_util, util
+from util import image_util, common_util, model_util
 
 
 def find_boxes(content, layout, offset=0, length=None, improve=False, image_path=None, extra_content=None):
@@ -55,14 +58,15 @@ def find_boxes(content, layout, offset=0, length=None, improve=False, image_path
     if improve:
         # 再次识别，提高精度
         image = cv2.imread(image_path)
+        img_name, img_ext = common_util.parse_save_path(image_path)
         # 截图时偏大一点
-        capture_box = util.zoom_rectangle(box, 0.2)
+        capture_box = common_util.zoom_rectangle(box, 0.2)
         captured_image = image_util.capture(image, capture_box)
-        with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as temp_file:
+        captured_image_path = common_util.get_processed_img_path(f'{img_name}.capture.{img_ext}')
-            captured_image, offset_x, offset_y = image_util.expand_to_a4_size(captured_image)
+        cv2.imwrite(captured_image_path, captured_image)
-            cv2.imwrite(temp_file.name, captured_image)
+        captured_a4_img_path, offset_x, offset_y = image_util.expand_to_a4_size(captured_image_path)
         try:
-            layouts = util.get_ocr_layout(OCR, temp_file.name)
+            layouts = common_util.ocr_result_to_layout(model_util.ocr(captured_a4_img_path))
         except TypeError:
             # 如果是类型错误，大概率是没识别到文字
             layouts = []
@@ -86,22 +90,17 @@ def find_boxes(content, layout, offset=0, length=None, improve=False, image_path
                             temp_box[3] + capture_box[1] - offset_y,
                         ])
                         break
-        util.delete_temp_file(temp_file.name)
 
     if not boxes:
         boxes.append(box)
     return boxes
 
 
-def get_mask_layout(image, name, id_card_num):
+def get_mask_layout(img_path, name, id_card_num):
-    with tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") as temp_file:
-        cv2.imwrite(temp_file.name, image)
-
     result = []
     try:
         try:
-            layouts = util.get_ocr_layout(OCR, temp_file.name)
+            layouts = common_util.ocr_result_to_layout(model_util.ocr(img_path))
-            # layouts = OCR.parse({"doc": temp_file.name})["layout"]
         except TypeError:
             # 如果是类型错误，大概率是没识别到文字
             layouts = []
@@ -135,12 +134,12 @@ def get_mask_layout(image, name, id_card_num):
                 find_id_card_num_by_key = True
                 matches = re.findall(r, layout[1])
                 for match in matches:
-                    result += find_boxes(match, layout, improve=True, image_path=temp_file.name, extra_content=r)
+                    result += find_boxes(match, layout, improve=True, image_path=img_path, extra_content=r)
                     find_name_by_key = False
                     break
 
                 if id_card_num in layout[1]:
-                    result += find_boxes(id_card_num, layout, improve=True, image_path=temp_file.name)
+                    result += find_boxes(id_card_num, layout, improve=True, image_path=img_path)
                     find_id_card_num_by_key = False
 
                 def _find_boxes_by_keys(keys):
@@ -163,8 +162,6 @@ def get_mask_layout(image, name, id_card_num):
     except Exception as e:
         logging.error("涂抹时出错！", exc_info=e)
         return result
-    finally:
-        util.delete_temp_file(temp_file.name)
 
 
 def handle_image_for_mask(split_result):
@@ -174,11 +171,15 @@ def handle_image_for_mask(split_result):
     return expand_img, split_result["x_offset"], split_result["y_offset"]
 
 
-def mask_photo(img_url, name, id_card_num, color=(255, 255, 255)):
+def mask_photo(img_path, name, id_card_num, color=(255, 255, 255)):
-    def _mask(i, n, icn, c):
+    def _mask(ip, n, icn, c):
+        i = cv2.imread(ip)
+        img_name, img_ext = common_util.parse_save_path(ip)
         do_mask = False
-        split_results = image_util.split(i)
+        split_results = image_util.split(ip)
         for split_result in split_results:
+            if not split_result['img']:
+                continue
             to_mask_img, x_offset, y_offset = handle_image_for_mask(split_result)
             results = get_mask_layout(to_mask_img, n, icn)
 
@@ -193,27 +194,27 @@ def mask_photo(img_url, name, id_card_num, color=(255, 255, 255)):
                     result[3] + y_offset,
                 )
                 cv2.rectangle(i, (int(result[0]), int(result[1])), (int(result[2]), int(result[3])), c, -1, 0)
-        return do_mask, i
+        masked_path = common_util.get_processed_img_path(f'{img_name}.mask.{img_ext}')
+        cv2.imwrite(masked_path, i)
+        return do_mask, masked_path
 
-    # 打开图片
+    original_image = img_path
-    image = image_util.read(img_url)
+    is_masked, img_path = _mask(img_path, name, id_card_num, color)
-    original_image = image
-    is_masked, image = _mask(image, name, id_card_num, color)
     if not is_masked:
         # 如果没有涂抹，可能是图片方向不对
-        angles = image_util.parse_rotation_angles(image)
+        angles = model_util.clas_orientation(img_path)
         angle = angles[0]
         if angle != "0":
-            image = image_util.rotate(image, int(angle))
+            img_path = image_util.rotate(img_path, int(angle))
-            is_masked, image = _mask(image, name, id_card_num, color)
+            is_masked, img_path = _mask(img_path, name, id_card_num, color)
         if not is_masked:
             # 如果旋转后也没有涂抹，恢复原来的方向
-            image = original_image
+            img_path = original_image
         else:
             # 如果旋转有效果，打一个日志
             logging.info(f"图片旋转了{angle}°")
 
-    return is_masked, image
+    return is_masked, img_path
 
 
 def photo_mask(pk_phhd, name, id_card_num):
@@ -223,32 +224,37 @@ def photo_mask(pk_phhd, name, id_card_num):
         ZxPhrec.cRectype.in_(["3", "4"])
     )).all()
     session.close()
+    # 同一批图的标识
+    set_batch_id(uuid.uuid4().hex)
+    processed_img_dir = common_util.get_processed_img_path('')
+    os.makedirs(processed_img_dir, exist_ok=True)
     for phrec in phrecs:
         img_url = ufile.get_private_url(phrec.cfjaddress)
         if not img_url:
             continue
-
+        original_img_path = common_util.save_to_local(img_url)
-        is_masked, image = mask_photo(img_url, name, id_card_num)
+        img_path = common_util.get_processed_img_path(phrec.cfjaddress)
+        shutil.copy2(original_img_path, img_path)
+        is_masked, image = mask_photo(img_path, name, id_card_num)
 
         # 如果涂抹了要备份以及更新
         if is_masked:
-            ufile.copy_file(BUCKET, phrec.cfjaddress, "drg2015", phrec.cfjaddress)
-
-            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)
+                ufile.copy_file(BUCKET, phrec.cfjaddress, "drg2015", phrec.cfjaddress)
+                ufile.upload_file(phrec.cfjaddress, image)
                 session = MysqlSession()
                 update_flag = (update(ZxPhrec).where(ZxPhrec.pk_phrec == phrec.pk_phrec).values(
                     paint_user=HOSTNAME,
-                    paint_date=util.get_default_datetime()))
+                    paint_date=common_util.get_default_datetime()))
                 session.execute(update_flag)
                 session.commit()
                 session.close()
             except Exception as e:
                 logging.error("上传图片出错", exc_info=e)
-            finally:
+
-                util.delete_temp_file(temp_file.name)
+    # 删除多余图片
+    if os.path.exists(processed_img_dir) and os.path.isdir(processed_img_dir):
+        shutil.rmtree(processed_img_dir)
 
 
 def main():
@@ -278,7 +284,7 @@ def main():
                 update_flag = (update(ZxPhhd).where(ZxPhhd.pk_phhd == pk_phhd).values(
                     paint_flag="8",
                     paint_user=HOSTNAME,
-                    paint_date=util.get_default_datetime(),
+                    paint_date=common_util.get_default_datetime(),
                     fZcfwfy=time.time() - start_time))
                 session.execute(update_flag)
                 session.commit()

photo_mask/photo_mask_error_check.py

@@ -8,7 +8,7 @@ from db import MysqlSession
 from db.mysql import ZxIeOcrerror, ZxPhrec
 from photo_mask.auto_photo_mask import mask_photo
 from ucloud import ufile
-from util import image_util, util
+from util import image_util, common_util
 
 
 def check_error(error_ocr):
@@ -91,7 +91,7 @@ if __name__ == '__main__':
 
     session = MysqlSession()
     update_error = (update(ZxIeOcrerror).where(ZxIeOcrerror.pk_phrec == ocr_error.pk_phrec).values(
-        checktime=util.get_default_datetime(), cfjaddress2=error_descript))
+        checktime=common_util.get_default_datetime(), cfjaddress2=error_descript))
     session.execute(update_error)
     session.commit()
     session.close()

photo_mask/photo_mask_error_report.py

@@ -7,7 +7,7 @@ from sqlalchemy import update, and_
 from db import MysqlSession
 from db.mysql import ZxIeOcrerror
 from photo_mask.photo_mask_error_check import auto_check_error
-from util import util
+from util import common_util
 
 if __name__ == '__main__':
     today = date.today()
@@ -29,7 +29,7 @@ if __name__ == '__main__':
         if error_descript == "未知错误":
             check_time = None
         else:
-            check_time = util.get_default_datetime()
+            check_time = common_util.get_default_datetime()
 
         session = MysqlSession()
         update_error = (update(ZxIeOcrerror).where(ZxIeOcrerror.pk_phrec == ocr_error.pk_phrec).values(
@@ -41,5 +41,5 @@ if __name__ == '__main__':
     print(result)
     with open("photo_mask_error_report.txt", 'w', encoding='utf-8') as file:
         file.write(json.dumps(result, indent=4, ensure_ascii=False))
-        file.write(util.get_default_datetime())
+        file.write(common_util.get_default_datetime())
     print("结果已保存。")

photo_review.py

@@ -5,36 +5,36 @@ from time import sleep
 
 from sqlalchemy import update
 
-from auto_email.error_email import send_error_email
 from db import MysqlSession
 from db.mysql import ZxPhhd
 from log import LOGGING_CONFIG
+from my_email.error_email import send_error_email
 from photo_review import auto_photo_review, SEND_ERROR_EMAIL
 
-# 项目必须从此处启动，否则代码中的相对路径可能导致错误的发生
+# 照片审核自动识别脚本入口
 if __name__ == '__main__':
     program_name = '照片审核自动识别脚本'
     logging.config.dictConfig(LOGGING_CONFIG)
 
+    logging.info('等待接口服务启动...')
+    sleep(60)
+
     parser = argparse.ArgumentParser()
-    parser.add_argument("--clean", default=False, type=bool, help="是否将识别中的案子改为待识别状态")
+    parser.add_argument('--clean', default=False, type=bool, help='是否将识别中的案子改为待识别状态')
     args = parser.parse_args()
     if args.clean:
-        # 主要用于启动时，清除仍在涂抹中的案子
+        # 启动时清除仍在识别中的案子
         session = MysqlSession()
-        update_flag = (update(ZxPhhd).where(ZxPhhd.exsuccess_flag == "2").values(exsuccess_flag="1"))
+        update_flag = (update(ZxPhhd).where(ZxPhhd.exsuccess_flag == '2').values(exsuccess_flag='1'))
         session.execute(update_flag)
         session.commit()
         session.close()
-        logging.info("已释放残余的识别案子！")
+        logging.info('已释放残余的识别案子！')
-    else:
-        sleep(5)
 
     try:
-        logging.info(f"【{program_name}】开始运行")
+        logging.info(f'【{program_name}】开始运行')
         auto_photo_review.main()
     except Exception as e:
-        error_logger = logging.getLogger('error')
+        logging.getLogger('error').error(traceback.format_exc())
-        error_logger.error(traceback.format_exc())
         if SEND_ERROR_EMAIL:
             send_error_email(program_name, repr(e), traceback.format_exc())

photo_review/__init__.py

@@ -1,6 +1,4 @@
 import jieba
-from paddlenlp import Taskflow
-from paddleocr import PaddleOCR
 
 '''
 项目配置
@@ -11,52 +9,8 @@ PHHD_BATCH_SIZE = 10
 SLEEP_MINUTES = 5
 # 是否发送报错邮件
 SEND_ERROR_EMAIL = True
-# 是否开启布局分析
+# 处理批号（这里仅起声明作用）
-LAYOUT_ANALYSIS = False
+BATCH_ID = ''
-
-"""
-信息抽取关键词配置
-"""
-# 患者姓名
-PATIENT_NAME = ['患者姓名']
-# 入院日期
-ADMISSION_DATE = ['入院日期']
-# 出院日期
-DISCHARGE_DATE = ['出院日期']
-# 发生医疗费
-MEDICAL_EXPENSES = ['费用总额']
-# 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']
-# 个人自费金额
-PERSONAL_FUNDED_AMOUNT = ['自费金额', '个人自费']
-# 医保类别
-MEDICAL_INSURANCE_TYPE = ['医保类型']
-# 就诊医院
-HOSPITAL = ['医院']
-# 就诊科室
-DEPARTMENT = ['科室']
-# 主治医生
-DOCTOR = ['主治医生']
-# 住院号
-ADMISSION_ID = ['住院号']
-# 医保结算单号码
-SETTLEMENT_ID = ['医保结算单号码']
-# 年龄
-AGE = ['年龄']
-# 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
-     + PERSONAL_ACCOUNT_PAYMENT + PERSONAL_FUNDED_AMOUNT + MEDICAL_INSURANCE_TYPE + ADMISSION_ID + SETTLEMENT_ID
-     + UPPERCASE_MEDICAL_EXPENSES)
-
-DISCHARGE_RECORD_SCHEMA = \
-    HOSPITAL + DEPARTMENT + PATIENT_NAME + ADMISSION_DATE + DISCHARGE_DATE + DOCTOR + ADMISSION_ID + AGE
-
-COST_LIST_SCHEMA = PATIENT_NAME + ADMISSION_DATE + DISCHARGE_DATE + MEDICAL_EXPENSES
 
 '''
 别名配置
@@ -92,13 +46,32 @@ jieba.suggest_freq(('胆', '道'), True)
 jieba.suggest_freq(('脾', '胃'), True)
 
 '''
-模型配置
+出院记录缺页判断关键词配置
 '''
-SETTLEMENT_IE = Taskflow('information_extraction', schema=SETTLEMENT_LIST_SCHEMA, model='uie-x-base',
+DISCHARGE_KEY = {
-                         task_path='model/settlement_list_model', layout_analysis=LAYOUT_ANALYSIS, precision='fp16')
+    '入院诊断': ['入院诊断'],
-DISCHARGE_IE = Taskflow('information_extraction', schema=DISCHARGE_RECORD_SCHEMA, model='uie-x-base',
+    '入院情况': ['入院情况', '入院时情况', '入院时主要症状'],
-                        task_path='model/discharge_record_model', layout_analysis=LAYOUT_ANALYSIS, precision='fp16')
+    '入院日期': ['入院日期', '入院时间'],
-COST_IE = Taskflow('information_extraction', schema=COST_LIST_SCHEMA, model='uie-x-base', device_id=1,
+    '诊疗经过': ['诊疗经过', '住院经过', '治疗经过'],
-                   task_path='model/cost_list_model', layout_analysis=LAYOUT_ANALYSIS, precision='fp16')
+    '出院诊断': ['出院诊断'],
+    '出院情况': ['出院情况', '出院时情况'],
+    '出院日期': ['出院日期', '出院时间'],
+    '出院医嘱': ['出院医嘱', '出院医瞩']
+}
 
-OCR = PaddleOCR(use_angle_cls=False, show_log=False, gpu_id=1, det_db_box_thresh=0.3)
+
+def get_batch_id():
+    """
+    获取处理批号
+    :return: 处理批号
+    """
+    return BATCH_ID
+
+
+def set_batch_id(batch_id):
+    """
+    修改处理批号哦
+    :param batch_id: 新批号
+    """
+    global BATCH_ID
+    BATCH_ID = batch_id

photo_review/auto_photo_review.py

@@ -1,113 +1,76 @@
+import copy
 import json
 import logging
 import os
-import tempfile
+import re
+import shutil
 import time
+import uuid
 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 db.mysql import BdYljg, BdYlks, ZxIeCost, ZxIeDischarge, ZxIeSettlement, ZxPhhd, ZxPhrec, ZxIeReview, ZxIeResult
 from log import HOSTNAME
-from photo_review import PATIENT_NAME, ADMISSION_DATE, DISCHARGE_DATE, MEDICAL_EXPENSES, PERSONAL_CASH_PAYMENT, \
+from photo_review import PHHD_BATCH_SIZE, SLEEP_MINUTES, HOSPITAL_ALIAS, HOSPITAL_FILTER, DEPARTMENT_ALIAS, \
-    PERSONAL_ACCOUNT_PAYMENT, PERSONAL_FUNDED_AMOUNT, MEDICAL_INSURANCE_TYPE, HOSPITAL, DEPARTMENT, DOCTOR, \
+    DEPARTMENT_FILTER, DISCHARGE_KEY, set_batch_id, get_batch_id
-    ADMISSION_ID, SETTLEMENT_ID, AGE, OCR, SETTLEMENT_IE, DISCHARGE_IE, COST_IE, PHHD_BATCH_SIZE, SLEEP_MINUTES, \
+from services.paddle_services import IE_KEY
-    UPPERCASE_MEDICAL_EXPENSES, HOSPITAL_ALIAS, HOSPITAL_FILTER, DEPARTMENT_ALIAS, DEPARTMENT_FILTER
+from ucloud import ufile, BUCKET
-from ucloud import ufile
+from util import image_util, common_util, html_util, model_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, \
-from util.data_util import handle_date, handle_decimal, parse_department, handle_name, \
+    handle_original_data, handle_hospital, handle_department, handle_id, handle_age, parse_money, parse_hospital, \
-    handle_insurance_type, handle_original_data, handle_hospital, handle_department, handle_id, handle_age, parse_money, \
+    parse_page_num, handle_tiny_int
-    parse_hospital
 
 
-# 合并信息抽取结果
+def parse_qrcode(img_path, image_id):
-def merge_result(result1, result2):
+    """
-    for key in result2:
+    解析二维码，尝试从中获取高清图片
-        result1[key] = result1.get(key, []) + result2[key]
+    :param img_path: 待解析图片
-    return result1
+    :param image_id: 图片id
+    :return: 解析结果
+    """
 
-
-def ie_temp_image(ie, ocr, image):
-    with tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") as temp_file:
-        cv2.imwrite(temp_file.name, image)
-
-    ie_result = []
-    try:
-        layout = util.get_ocr_layout(ocr, temp_file.name)
-        if not layout:
-            # 无识别结果
-            ie_result = []
-        else:
-            ie_result = ie({"doc": temp_file.name, "layout": layout})[0]
-    except Exception as e:
-        logging.error("信息抽取时出错", exc_info=e)
-    finally:
-        try:
-            os.remove(temp_file.name)
-        except Exception as e:
-            logging.info(f"删除临时文件 {temp_file.name} 时出错", exc_info=e)
-    return ie_result
-
-
-# 关键信息提取
-def request_ie_result(task_enum, phrecs):
-    url = task_enum.request_url()
-    identity = int(time.time())
-    images = []
-    for phrec in phrecs:
-        images.append({"name": phrec.cfjaddress, "pk": phrec.pk_phrec})
-    payload = {"images": images, "schema": task_enum.schema(), "pk_phhd": phrecs[0].pk_phhd, "identity": identity}
-    response = requests.post(url, json=payload)
-
-    if response.status_code == 200:
-        return response.json()["data"]
-    else:
-        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, image_id, dpi=150):
     def _parse_pdf_url(pdf_url_to_parse):
-        pdf_file = None
         local_pdf_path = None
+        img_name, img_ext = common_util.parse_save_path(img_path)
         try:
             local_pdf_path = html_util.download_pdf(pdf_url_to_parse)
-            # 打开PDF文件
+            pdf_imgs = image_util.pdf_to_imgs(local_pdf_path)
-            pdf_file = fitz.open(local_pdf_path)
+            # 结算单部分
-            # 选择第一页
+            better_settlement_path = common_util.get_processed_img_path(f'{img_name}.better_settlement.jpg')
-            page = pdf_file[0]
+            cv2.imwrite(better_settlement_path, pdf_imgs[0][0])
-            # 定义缩放系数（DPI）
+            # 费用清单部分
-            default_dpi = 72
+            better_cost_path = common_util.get_processed_img_path(f'{img_name}.better_cost.jpg')
-            zoom = dpi / default_dpi
+            total_height = sum([p[0].shape[0] for p in pdf_imgs[1:]])
-            # 设置矩阵变换参数
+            common_width = pdf_imgs[1][0].shape[1]
-            mat = fitz.Matrix(zoom, zoom)
+            better_cost_img = np.zeros((total_height, common_width, 3), dtype=np.uint8)
-            # 渲染页面
+            current_y = 0
-            pix = page.get_pixmap(matrix=mat)
+            for pdf in pdf_imgs[1:]:
-            # 将渲染结果转换为OpenCV兼容的格式
+                height = pdf[0].shape[0]
-            img = np.frombuffer(pix.samples, dtype=np.uint8).reshape((pix.height, pix.width, -1))
+                better_cost_img[current_y:current_y + height, :, :] = pdf[0]
-            img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+                current_y += height
-            return img, page.get_text()
+                # cost_text += pdf[1]  # 费用清单文本暂时没用到
+            cv2.imwrite(better_cost_path, better_cost_img)
+
+            return better_settlement_path, pdf_imgs[0][1], better_cost_path
         except Exception as ex:
             logging.getLogger('error').error('解析pdf失败！', exc_info=ex)
-            return None, None
+            return None, None, None
         finally:
-            if pdf_file:
-                pdf_file.close()
             if local_pdf_path:
-                util.delete_temp_file(local_pdf_path)
+                common_util.delete_temp_file(local_pdf_path)
 
     jsczt_base_url = 'http://einvoice.jsczt.cn'
     try:
-        results = zxingcpp.read_barcodes(image)
+        img = cv2.imread(img_path)
+        results = zxingcpp.read_barcodes(img, text_mode=zxingcpp.TextMode.HRI)
     except Exception as e:
         logging.getLogger('error').info('二维码识别失败', exc_info=e)
         results = []
@@ -132,138 +95,122 @@ def get_better_image_from_qrcode(image, image_id, dpi=150):
                 if not pdf_url:
                     continue
                 return _parse_pdf_url(pdf_url)
-            elif url.startswith('http://weixin.qq.com'):
+            elif (url.startswith('http://weixin.qq.com')
+                  or url == 'https://ybj.jszwfw.gov.cn/hsa-app-panel/index.html'):
                 # 无效地址
                 continue
+            elif url.startswith('http://dzpj.ntzyy.com'):
+                # 南通市中医院
+                return _parse_pdf_url(url)
+            # elif url.startswith('https://apph5.ztejsapp.cn/nj/view/elecInvoiceForOther/QRCode2Invoice'):
+            #     pdf_url = html_util.get_dtsrmyy_pdf_url(url)
+            #     if not pdf_url:
+            #         continue
+            #     return _parse_pdf_url(pdf_url)
             else:
                 logging.getLogger('qr').info(f'[{image_id}]中有未知二维码内容：{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
 
-    return None, None
+    return None, None, None
 
 
 # 关键信息提取
-def information_extraction(ie, phrecs, identity):
+def information_extraction(phrec, pk_phhd):
-    result = {}
+    """
-    for phrec in phrecs:
+    处理单张图片
-        img_path = ufile.get_private_url(phrec.cfjaddress)
+    :param phrec:图片信息
-        if not img_path:
+    :param pk_phhd:案子主键
-            continue
+    :return:记录类型，信息抽取结果
-
+    """
-        image = image_util.read(img_path)
+    img_path = common_util.get_processed_img_path(phrec.cfjaddress)
-
+    if not os.path.exists(img_path):
-        # 尝试从二维码中获取高清图片
+        original_img_path = common_util.get_img_path(phrec.cfjaddress)
-        better_image, text = get_better_image_from_qrcode(image, phrec.cfjaddress)
+        if not original_img_path:
-        if phrec.cRectype != '1':
+            img_url = ufile.get_private_url(phrec.cfjaddress)
-            better_image = None  # 非结算单暂时不进行替换
+            if not img_url:
-        zx_ie_results = []
+                return None, None, None
-        if better_image is not None:
+            original_img_path = common_util.save_to_local(img_url)
-            img_angle = '0'
+        shutil.copy2(original_img_path, img_path)
-            image = better_image
+    if image_util.is_photo(img_path):
-            if text:
+        book_img_path = model_util.det_book(img_path)  # 识别文档区域并裁剪
-                info_extract = ie(text)[0]
+        dewarped_img_path = model_util.dewarp(book_img_path)  # 去扭曲
-            else:
+    else:  # todo:也可能是图片，后续添加细分逻辑
-                info_extract = ie_temp_image(ie, OCR, image)
+        dewarped_img_path = img_path
-            ie_result = {'result': info_extract, 'angle': '0'}
+    angles = model_util.clas_orientation(dewarped_img_path)
-
+    ocr_text = ''
-            now = util.get_default_datetime()
+    info_extract = []
-            if not ie_result['result']:
+    rec_type = None
+    for angle in angles:
+        ocr_result = []
+        rotated_img = image_util.rotate(dewarped_img_path, int(angle))
+        split_results = image_util.split(rotated_img)
+        for split_result in split_results:
+            if split_result['img'] is None:
                 continue
+            a4_img = image_util.expand_to_a4_size(split_result['img'])
+            tmp_ocr_result = model_util.ocr(a4_img)
+            if tmp_ocr_result:
+                ocr_result += tmp_ocr_result
+        tmp_ocr_text = common_util.ocr_result_to_text(ocr_result)
 
-            result_json = json.dumps(ie_result['result'], ensure_ascii=False)
+        # if any(key in tmp_ocr_text for key in ['出院记录', '出院小结', '死亡记录']):
-            if len(result_json) > 5000:
+        #     tmp_rec_type = '出院记录'
-                result_json = result_json[:5000]
+        # elif any(key in tmp_ocr_text for key in ['费用汇总清单', '费用清单', '费用明细', '结账清单', '费用小项统计']):
-            zx_ie_results.append(ZxIeResult(pk_phhd=phrec.pk_phhd, pk_phrec=phrec.pk_phrec, id=identity,
+        #     tmp_rec_type = '费用清单'
-                                            cfjaddress=phrec.cfjaddress, content=result_json,
+        # elif any(key in tmp_ocr_text for key in ['住院收费票据', '结算单', '财政部监制', '结算凭证']):
-                                            rotation_angle=int(ie_result['angle']),
+        #     tmp_rec_type = '基本医保结算单'
-                                            x_offset=0, y_offset=0, create_time=now,
+        # else:
-                                            creator=HOSTNAME, update_time=now, updater=HOSTNAME))
+        #     tmp_rec_type = model_util.clas_text(tmp_ocr_text) if tmp_ocr_text else None
-
+        # if not tmp_rec_type:
-            result = merge_result(result, ie_result['result'])
+        rec_dict = {
+            '1': '基本医保结算单',
+            '3': '出院记录',
+            '4': '费用清单',
+        }
+        tmp_rec_type = rec_dict.get(phrec.cRectype)
+        if tmp_rec_type == '基本医保结算单':
+            tmp_info_extract = model_util.ie_settlement(rotated_img, common_util.ocr_result_to_layout(ocr_result))
+        elif tmp_rec_type == '出院记录':
+            tmp_info_extract = model_util.ie_discharge(rotated_img, common_util.ocr_result_to_layout(ocr_result))
+        elif tmp_rec_type == '费用清单':
+            tmp_info_extract = model_util.ie_cost(rotated_img, common_util.ocr_result_to_layout(ocr_result))
         else:
-            target_images = []
+            tmp_info_extract = []
-            # 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)
 
-                split_results = image_util.split(dewarped_image)
+        if tmp_info_extract and len(tmp_info_extract) > len(info_extract):
-                for split_result in split_results:
+            info_extract = tmp_info_extract
-                    if split_result['img'] is None or split_result['img'].size == 0:
+            ocr_text = tmp_ocr_text
-                        continue
+            rec_type = tmp_rec_type
-                    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 info_extract:
-                        if len(result_json) > 5000:
+        result_json = json.dumps(info_extract, ensure_ascii=False)
-                            result_json = result_json[:5000]
+        if len(result_json) > 5000:
-                        zx_ie_results.append(ZxIeResult(pk_phhd=phrec.pk_phhd, pk_phrec=phrec.pk_phrec, id=identity,
+            result_json = result_json[:5000]
-                                                        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' or better_image is not None:
-            image = image_util.rotate(image, int(img_angle))
-            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':
-                    logging.info(f'旋转图片[{phrec.cfjaddress}]替换成功，已旋转{img_angle}度。')
-                    # 修正旋转角度
-                    for zx_ie_result in zx_ie_results:
-                        zx_ie_result.rotation_angle -= int(img_angle)
-                else:
-                    logging.info(f'高清图片[{phrec.cfjaddress}]替换成功！')
-            except Exception as e:
-                logging.error(f'上传图片({phrec.cfjaddress})失败', exc_info=e)
-            finally:
-                util.delete_temp_file(temp_file.name)
 
+        now = common_util.get_default_datetime()
         session = MysqlSession()
-        session.add_all(zx_ie_results)
+        session.add(ZxIeResult(pk_phhd=pk_phhd, pk_phrec=phrec.pk_phrec, id=get_batch_id(),
+                               cfjaddress=phrec.cfjaddress, content=result_json, create_time=now,
+                               creator=HOSTNAME, update_time=now, updater=HOSTNAME))
         session.commit()
         session.close()
-
+    return rec_type, info_extract, ocr_text
-    return result
 
 
 # 从keys中获取准确率最高的value
-def get_best_value_in_keys(source, keys):
+def get_best_value_of_key(source, key):
     # 最终结果
     result = None
     # 最大可能性
     best_probability = 0
-    for key in keys:
+    values = source.get(key)
-        values = source.get(key)
+    if values:
-        if values:
+        for value in values:
-            for value in values:
+            for v in value:
-                text = value.get("text")
+                text = v.get("text")
-                probability = value.get("probability")
+                probability = v.get("probability")
                 if text and probability > best_probability:
                     result = text
                     best_probability = probability
@@ -271,11 +218,11 @@ def get_best_value_in_keys(source, keys):
 
 
 # 从keys中获取所有value组成list
-def get_values_of_keys(source, keys):
+def get_values_of_key(source, key):
     result = []
-    for key in keys:
+    values = source.get(key)
-        value = source.get(key)
+    if values:
-        if value:
+        for value in values:
             for v in value:
                 v = v.get("text")
                 if v:
@@ -289,7 +236,7 @@ def save_or_update_ie(table, pk_phhd, data):
     obj = table(**data)
     session = MysqlSession()
     db_data = session.query(table).filter_by(pk_phhd=pk_phhd).one_or_none()
-    now = util.get_default_datetime()
+    now = common_util.get_default_datetime()
     if db_data:
         # 更新
         db_data.update_time = now
@@ -364,23 +311,24 @@ def search_department(department):
     return best_match
 
 
-def settlement_task(pk_phhd, settlement_list, identity):
+def settlement_task(pk_phhd, settlement_list_ie_result):
-    settlement_list_ie_result = information_extraction(SETTLEMENT_IE, settlement_list, identity)
     settlement_data = {
         "pk_phhd": pk_phhd,
-        "name": handle_name(get_best_value_in_keys(settlement_list_ie_result, PATIENT_NAME)),
+        "name": handle_name(get_best_value_of_key(settlement_list_ie_result, IE_KEY['name'])),
-        "admission_date_str": handle_original_data(get_best_value_in_keys(settlement_list_ie_result, ADMISSION_DATE)),
+        "admission_date_str": handle_original_data(
-        "discharge_date_str": handle_original_data(get_best_value_in_keys(settlement_list_ie_result, DISCHARGE_DATE)),
+            get_best_value_of_key(settlement_list_ie_result, IE_KEY['admission_date'])),
+        "discharge_date_str": handle_original_data(
+            get_best_value_of_key(settlement_list_ie_result, IE_KEY['discharge_date'])),
         "personal_cash_payment_str": handle_original_data(
-            get_best_value_in_keys(settlement_list_ie_result, PERSONAL_CASH_PAYMENT)),
+            get_best_value_of_key(settlement_list_ie_result, IE_KEY['personal_cash_payment'])),
         "personal_account_payment_str": handle_original_data(
-            get_best_value_in_keys(settlement_list_ie_result, PERSONAL_ACCOUNT_PAYMENT)),
+            get_best_value_of_key(settlement_list_ie_result, IE_KEY['personal_account_payment'])),
         "personal_funded_amount_str": handle_original_data(
-            get_best_value_in_keys(settlement_list_ie_result, PERSONAL_FUNDED_AMOUNT)),
+            get_best_value_of_key(settlement_list_ie_result, IE_KEY['personal_funded_amount'])),
         "medical_insurance_type_str": handle_original_data(
-            get_best_value_in_keys(settlement_list_ie_result, MEDICAL_INSURANCE_TYPE)),
+            get_best_value_of_key(settlement_list_ie_result, IE_KEY['medical_insurance_type'])),
-        "admission_id": handle_id(get_best_value_in_keys(settlement_list_ie_result, ADMISSION_ID)),
+        "admission_id": handle_id(get_best_value_of_key(settlement_list_ie_result, IE_KEY['admission_id'])),
-        "settlement_id": handle_id(get_best_value_in_keys(settlement_list_ie_result, SETTLEMENT_ID)),
+        "settlement_id": handle_id(get_best_value_of_key(settlement_list_ie_result, IE_KEY['settlement_id'])),
     }
     settlement_data["admission_date"] = handle_date(settlement_data["admission_date_str"])
     settlement_data["admission_date"] = handle_date(settlement_data["admission_date_str"])
@@ -390,27 +338,30 @@ def settlement_task(pk_phhd, settlement_list, identity):
     settlement_data["personal_funded_amount"] = handle_decimal(settlement_data["personal_funded_amount_str"])
     settlement_data["medical_insurance_type"] = handle_insurance_type(settlement_data["medical_insurance_type_str"])
 
-    parse_money_result = parse_money(get_best_value_in_keys(settlement_list_ie_result, UPPERCASE_MEDICAL_EXPENSES),
+    parse_money_result = parse_money(
-                                     get_best_value_in_keys(settlement_list_ie_result, MEDICAL_EXPENSES))
+        get_best_value_of_key(settlement_list_ie_result, IE_KEY['uppercase_medical_expenses']),
+        get_best_value_of_key(settlement_list_ie_result, IE_KEY['medical_expenses']))
     settlement_data["medical_expenses_str"] = handle_original_data(parse_money_result[0])
     settlement_data["medical_expenses"] = parse_money_result[1]
     save_or_update_ie(ZxIeSettlement, pk_phhd, settlement_data)
+    return settlement_data
 
 
-def discharge_task(pk_phhd, discharge_record, identity):
+def discharge_task(pk_phhd, discharge_record_ie_result):
-    discharge_record_ie_result = information_extraction(DISCHARGE_IE, discharge_record, identity)
+    hospitals = get_values_of_key(discharge_record_ie_result, IE_KEY['hospital'])
-    hospitals = get_values_of_keys(discharge_record_ie_result, HOSPITAL)
+    departments = get_values_of_key(discharge_record_ie_result, IE_KEY['department'])
-    departments = get_values_of_keys(discharge_record_ie_result, DEPARTMENT)
     discharge_data = {
         "pk_phhd": pk_phhd,
         "hospital": handle_hospital(",".join(hospitals)),
         "department": handle_department(",".join(departments)),
-        "name": handle_name(get_best_value_in_keys(discharge_record_ie_result, PATIENT_NAME)),
+        "name": handle_name(get_best_value_of_key(discharge_record_ie_result, IE_KEY['name'])),
-        "admission_date_str": handle_original_data(get_best_value_in_keys(discharge_record_ie_result, ADMISSION_DATE)),
+        "admission_date_str": handle_original_data(
-        "discharge_date_str": handle_original_data(get_best_value_in_keys(discharge_record_ie_result, DISCHARGE_DATE)),
+            get_best_value_of_key(discharge_record_ie_result, IE_KEY['admission_date'])),
-        "doctor": handle_name(get_best_value_in_keys(discharge_record_ie_result, DOCTOR)),
+        "discharge_date_str": handle_original_data(
-        "admission_id": handle_id(get_best_value_in_keys(discharge_record_ie_result, ADMISSION_ID)),
+            get_best_value_of_key(discharge_record_ie_result, IE_KEY['discharge_date'])),
-        "age": handle_age(get_best_value_in_keys(discharge_record_ie_result, AGE)),
+        "doctor": handle_name(get_best_value_of_key(discharge_record_ie_result, IE_KEY['doctor'])),
+        "admission_id": handle_id(get_best_value_of_key(discharge_record_ie_result, IE_KEY['admission_id'])),
+        "age": handle_age(get_best_value_of_key(discharge_record_ie_result, IE_KEY['age'])),
     }
     discharge_data["admission_date"] = handle_date(discharge_data["admission_date_str"])
     discharge_data["discharge_date"] = handle_date(discharge_data["discharge_date_str"])
@@ -467,52 +418,270 @@ def discharge_task(pk_phhd, discharge_record, identity):
         if best_match:
             discharge_data["pk_ylks"] = best_match[2]
     save_or_update_ie(ZxIeDischarge, pk_phhd, discharge_data)
+    return discharge_data
 
 
-def cost_task(pk_phhd, cost_list, identity):
+def cost_task(pk_phhd, cost_list_ie_result):
-    cost_list_ie_result = information_extraction(COST_IE, cost_list, identity)
     cost_data = {
         "pk_phhd": pk_phhd,
-        "name": handle_name(get_best_value_in_keys(cost_list_ie_result, PATIENT_NAME)),
+        "name": handle_name(get_best_value_of_key(cost_list_ie_result, IE_KEY['name'])),
-        "admission_date_str": handle_original_data(get_best_value_in_keys(cost_list_ie_result, ADMISSION_DATE)),
+        "admission_date_str": handle_original_data(
-        "discharge_date_str": handle_original_data(get_best_value_in_keys(cost_list_ie_result, DISCHARGE_DATE)),
+            get_best_value_of_key(cost_list_ie_result, IE_KEY['admission_date'])),
-        "medical_expenses_str": handle_original_data(get_best_value_in_keys(cost_list_ie_result, MEDICAL_EXPENSES))
+        "discharge_date_str": handle_original_data(
+            get_best_value_of_key(cost_list_ie_result, IE_KEY['discharge_date'])),
+        "medical_expenses_str": handle_original_data(
+            get_best_value_of_key(cost_list_ie_result, IE_KEY['medical_expenses']))
     }
     cost_data["admission_date"] = handle_date(cost_data["admission_date_str"])
     cost_data["discharge_date"] = handle_date(cost_data["discharge_date_str"])
     cost_data["medical_expenses"] = handle_decimal(cost_data["medical_expenses_str"])
+    if cost_list_ie_result.get(IE_KEY['page']):
+        page_nums, page_count = parse_page_num(cost_list_ie_result[IE_KEY['page']])
+        if page_nums:
+            page_nums_str = [str(num) for num in page_nums]
+            cost_data['page_nums'] = handle_original_data(','.join(page_nums_str))
+            cost_data['page_count'] = handle_tiny_int(page_count)
     save_or_update_ie(ZxIeCost, pk_phhd, cost_data)
+    return cost_data
 
 
-def photo_review(pk_phhd):
+def parse_pdf_text(settlement_text):
-    settlement_list = []
+    pattern = (r'(?:交款人：(.*?)\n|住院时间：(.*?)至(.*?)\n|\(小写\)(.*?)\n|个人现金支付：(.*?)\n|个人账户支付：(.*?)\n'
-    discharge_record = []
+               r'|个人自费：(.*?)\n|医保类型：(.*?)\n|住院科别：(.*?)\n|住院号：(.*?)\n|票据号码：(.*?)\n|)')
-    cost_list = []
+    # 查找所有匹配项
+    matches = re.findall(pattern, settlement_text)
+    results = {}
+    keys = ['患者姓名', '入院日期', '出院日期', '费用总额', '个人现金支付', '个人账户支付', '个人自费', '医保类型',
+            '科室', '住院号', '医保结算单号码']
+
+    for match in matches:
+        for key, value in zip(keys, match):
+            if value:
+                results[key] = [[{'text': value, 'probability': 1}]]
+    settlement_key = ['患者姓名', '入院日期', '出院日期', '费用总额', '个人现金支付', '个人账户支付', '个人自费',
+                      '医保类型', '住院号', '医保结算单号码']
+    discharge_key = ['科室', '患者姓名', '入院日期', '出院日期', '住院号']
+    cost_key = ['患者姓名', '入院日期', '出院日期', '费用总额']
+    settlement_result = {key: copy.copy(results[key]) for key in settlement_key if key in results}
+    discharge_result = {key: copy.copy(results[key]) for key in discharge_key if key in results}
+    cost_result = {key: copy.copy(results[key]) for key in cost_key if key in results}
+    return settlement_result, discharge_result, cost_result
+
+
+def photo_review(pk_phhd, name):
+    """
+    处理单个报销案子
+    :param pk_phhd: 报销单主键
+    :param name: 报销人姓名
+    """
+    settlement_result = defaultdict(list)
+    discharge_result = defaultdict(list)
+    cost_result = defaultdict(list)
 
     session = MysqlSession()
-    phrecs = session.query(ZxPhrec.pk_phrec, ZxPhrec.pk_phhd, ZxPhrec.cRectype, ZxPhrec.cfjaddress).filter(
+    phrecs = session.query(ZxPhrec.pk_phrec, ZxPhrec.cRectype, ZxPhrec.cfjaddress).filter(
         ZxPhrec.pk_phhd == pk_phhd
-    ).all()
+    ).order_by(ZxPhrec.cRectype, ZxPhrec.rowno).all()
     session.close()
-    for phrec in phrecs:
-        if phrec.cRectype == "1":
-            settlement_list.append(phrec)
-        elif phrec.cRectype == "3":
-            discharge_record.append(phrec)
-        elif phrec.cRectype == "4":
-            cost_list.append(phrec)
 
     # 同一批图的标识
-    identity = int(time.time())
+    set_batch_id(uuid.uuid4().hex)
-    settlement_task(pk_phhd, settlement_list, identity)
+    processed_img_dir = common_util.get_processed_img_path('')
-    discharge_task(pk_phhd, discharge_record, identity)
+    os.makedirs(processed_img_dir, exist_ok=True)
-    cost_task(pk_phhd, cost_list, identity)
+
+    has_pdf = False  # 是否获取到了pdf，获取到可以直接利用pdf更快的获取信息
+    better_settlement_path = None
+    better_cost_path = None
+    settlement_text = ''
+    qrcode_img_id = None
+    for phrec in phrecs:
+        original_img_path = common_util.get_img_path(phrec.cfjaddress)
+        if not original_img_path:
+            img_url = ufile.get_private_url(phrec.cfjaddress)
+            if not img_url:
+                continue
+            original_img_path = common_util.save_to_local(img_url)
+        img_path = common_util.get_processed_img_path(phrec.cfjaddress)
+        shutil.copy2(original_img_path, img_path)
+        # 尝试从二维码中获取高清图片
+        better_settlement_path, settlement_text, better_cost_path = parse_qrcode(img_path, phrec.cfjaddress)
+        if better_settlement_path:
+            has_pdf = True
+            qrcode_img_id = phrec.cfjaddress
+            break
+
+    discharge_text = ''
+    if has_pdf:
+        settlement_result, discharge_result, cost_result = parse_pdf_text(settlement_text)
+        discharge_ie_result = defaultdict(list)
+
+        is_cost_updated = False
+        for phrec in phrecs:
+            if phrec.cRectype == '1':
+                if phrec.cfjaddress == qrcode_img_id:
+                    try:
+                        ufile.copy_file(BUCKET, phrec.cfjaddress, "drg2015", phrec.cfjaddress)
+                        ufile.upload_file(phrec.cfjaddress, better_settlement_path)
+                    except Exception as e:
+                        logging.error("更新结算单pdf图片出错", exc_info=e)
+            elif phrec.cRectype == '3':
+                rec_type, ie_result, ocr_text = information_extraction(phrec, pk_phhd)
+                if rec_type == '出院记录':
+                    discharge_text += ocr_text
+                    for key, value in ie_result.items():
+                        discharge_ie_result[key].append(value)
+            # 暂不替换费用清单
+            # elif phrec.cRectype == '4':
+            #     if not is_cost_updated:
+            #         try:
+            #             ufile.copy_file(BUCKET, phrec.cfjaddress, "drg2015", phrec.cfjaddress)
+            #             ufile.upload_file(phrec.cfjaddress, better_cost_path)
+            #         except Exception as e:
+            #             logging.error("更新费用清单pdf图片出错", exc_info=e)
+            #         finally:
+            #             is_cost_updated = True
+
+        # 合并出院记录
+        for key, value in discharge_ie_result.items():
+            ie_value = get_best_value_of_key(discharge_ie_result, key)
+            pdf_value = discharge_result.get(key)[0][0]['text'] if discharge_result.get(key) else ''
+            similarity_ratio = fuzz.ratio(ie_value, pdf_value)
+            if similarity_ratio < 60:
+                discharge_result[key] = [[{'text': ie_value, 'probability': 1}]]
+    else:
+        for phrec in phrecs:
+            rec_type, ie_result, ocr_text = information_extraction(phrec, pk_phhd)
+            if rec_type == '基本医保结算单':
+                rec_result = settlement_result
+            elif rec_type == '出院记录':
+                rec_result = discharge_result
+                discharge_text += ocr_text
+            elif rec_type == '费用清单':
+                rec_result = cost_result
+            else:
+                rec_result = None
+            if rec_result is not None:
+                for key, value in ie_result.items():
+                    rec_result[key].append(value)
+
+    # 删除多余图片
+    if os.path.exists(processed_img_dir) and os.path.isdir(processed_img_dir):
+        shutil.rmtree(processed_img_dir)
+
+    settlement_data = settlement_task(pk_phhd, settlement_result)
+    discharge_data = discharge_task(pk_phhd, discharge_result)
+    cost_data = cost_task(pk_phhd, cost_result)
+
+    # 三项资料完整性判断
+    # 三项资料缺项判断
+    review_result = {
+        'pk_phhd': pk_phhd,
+        'has_settlement': bool(settlement_result),
+        'has_discharge': bool(discharge_result),
+        'has_cost': bool(cost_result),
+    }
+    if (review_result['has_settlement'] and settlement_data.get('personal_account_payment')
+            and settlement_data.get('personal_cash_payment') and settlement_data.get('medical_expenses')):
+        review_result['has_settlement'] &= (
+                float(settlement_data['personal_account_payment']) + float(settlement_data['personal_cash_payment'])
+                < float(settlement_data['medical_expenses'])
+        )
+    if has_pdf:
+        review_result['has_discharge'] &= bool(discharge_text)
+
+    # 三项资料缺页判断
+    page_description = []
+    if review_result['has_discharge']:
+        for discharge_item in DISCHARGE_KEY:
+            if not any(key in discharge_text for key in DISCHARGE_KEY[discharge_item]):
+                page_description.append(f"《出院记录》缺页")
+                break
+
+    if review_result['has_cost']:
+        cost_missing_page = {}
+        if cost_data.get('page_nums') and cost_data.get('page_count'):
+            page_nums = cost_data['page_nums'].split(',')
+            required_set = set(range(1, cost_data['page_count'] + 1))
+            page_set = set([int(num) for num in page_nums])
+            cost_missing_page = required_set - page_set
+        if cost_missing_page:
+            cost_missing_page = sorted(cost_missing_page)
+            cost_missing_page = [str(num) for num in cost_missing_page]
+            page_description.append(f"《住院费用清单》，缺第{','.join(cost_missing_page)}页")
+
+    if page_description:
+        review_result['full_page'] = False
+        review_result['page_description'] = ';'.join(page_description)
+    else:
+        review_result['full_page'] = True
+
+    review_result['integrity'] = (review_result['has_settlement'] and review_result['has_discharge']
+                                  and review_result['has_cost'] and review_result['full_page'])
+
+    # 三项资料一致性判断
+    # 姓名一致性
+    name_list = [settlement_data['name'], discharge_data['name'], cost_data['name']]
+    if sum(not bool(n) for n in name_list) > 1:  # 有2个及以上空值直接认为都不一致
+        review_result['name_match'] = '0'
+    else:
+        unique_name = set(name_list)
+        if len(unique_name) == 1:
+            review_result['name_match'] = '1' if name == unique_name.pop() else '5'
+        elif len(unique_name) == 2:
+            if settlement_data['name'] != discharge_data['name'] and settlement_data['name'] != cost_data['name']:
+                review_result['name_match'] = '2'
+            elif discharge_data['name'] != settlement_data['name'] and discharge_data['name'] != cost_data['name']:
+                review_result['name_match'] = '3'
+            else:
+                review_result['name_match'] = '4'
+        else:
+            review_result['name_match'] = '0'
+
+    # 住院日期一致性
+    if (settlement_data['admission_date'] and discharge_data['admission_date']
+            and settlement_data['discharge_date'] and discharge_data['discharge_date']
+            and settlement_data['admission_date'] == discharge_data['admission_date']
+            and settlement_data['discharge_date'] == discharge_data['discharge_date']):
+        review_result['admission_date_match'] = '1'
+    else:
+        review_result['admission_date_match'] = '0'
+
+    # 出院日期一致性
+    discharge_date_list = [settlement_data['discharge_date'], discharge_data['discharge_date'],
+                           cost_data['discharge_date']]
+    if sum(not bool(d) for d in discharge_date_list) > 1:
+        review_result['discharge_date_match'] = '0'
+    else:
+        unique_discharge_date = set(discharge_date_list)
+        if len(unique_discharge_date) == 1:
+            review_result['discharge_date_match'] = '1'
+        elif len(unique_discharge_date) == 2:
+            if (settlement_data['discharge_date'] != discharge_data['discharge_date']
+                    and settlement_data['discharge_date'] != cost_data['discharge_date']):
+                review_result['discharge_date_match'] = '2'
+            elif (discharge_data['discharge_date'] != settlement_data['discharge_date']
+                  and discharge_data['discharge_date'] != cost_data['discharge_date']):
+                review_result['discharge_date_match'] = '3'
+            else:
+                review_result['discharge_date_match'] = '4'
+        else:
+            review_result['discharge_date_match'] = '0'
+
+    review_result['consistency'] = (
+            review_result['name_match'] == '1' and review_result['admission_date_match'] == '1'
+            and review_result['discharge_date_match'] == '1')
+
+    review_result['success'] = review_result['integrity'] and review_result['consistency']
+    save_or_update_ie(ZxIeReview, pk_phhd, review_result)
 
 
 def main():
+    """
+    照片审核批量控制
+    """
     while 1:
         session = MysqlSession()
-        phhds = (session.query(ZxPhhd.pk_phhd)
+        phhds = (session.query(ZxPhhd.pk_phhd, ZxPhhd.cXm)
                  .join(ZxPhrec, ZxPhhd.pk_phhd == ZxPhrec.pk_phhd, isouter=True)
                  .filter(ZxPhhd.exsuccess_flag == "1")
                  .filter(ZxPhrec.pk_phrec.isnot(None))
@@ -529,14 +698,14 @@ def main():
                 pk_phhd = phhd.pk_phhd
                 logging.info(f"开始识别：{pk_phhd}")
                 start_time = time.time()
-                photo_review(pk_phhd)
+                photo_review(pk_phhd, phhd.cXm)
 
                 # 识别完成更新标识
                 session = MysqlSession()
                 update_flag = (update(ZxPhhd).where(ZxPhhd.pk_phhd == pk_phhd).values(
                     exsuccess_flag="8",
                     ref_id1=HOSTNAME,
-                    checktime=util.get_default_datetime(),
+                    checktime=common_util.get_default_datetime(),
                     fFSYLFY=time.time() - start_time))
                 session.execute(update_flag)
                 session.commit()

photo_review/photo_review_error_report.py

@@ -6,7 +6,7 @@ from sqlalchemy.sql.functions import count
 
 from db import MysqlSession
 from db.mysql import ZxPhhd, ViewErrorReview
-from util import util
+from util import common_util
 
 
 def handle_reason(reason):
@@ -95,5 +95,5 @@ if __name__ == '__main__':
     print(result)
     with open("photo_review_error_report.txt", 'w', encoding='utf-8') as file:
         file.write(json.dumps(result, indent=4, ensure_ascii=False))
-        file.write(util.get_default_datetime())
+        file.write(common_util.get_default_datetime())
     print("结果已保存。")

requirements.txt

@@ -1,16 +1,16 @@
+beautifulsoup4==4.12.3  # 网页分析
+jieba==0.42.1  # 中文分词
 numpy==1.26.4
-onnxconverter-common==1.14.0
+OpenCC==1.1.9  # 中文繁简转换
-OpenCC==1.1.6
 opencv-python==4.6.0.66
-paddle2onnx==1.2.3
+opencv-python-headless==4.10.0.84
-paddleclas==2.5.2
-paddlenlp==2.6.1
-paddleocr==2.7.3
 pillow==10.4.0
+PyMuPDF==1.24.9  # pdf处理
 pymysql==1.1.1
+rapidfuzz==3.9.4  #文本相似度
 requests==2.32.3
-sqlacodegen==2.3.0.post1
+sqlacodegen==2.3.0.post1  # 实体类生成
-sqlalchemy==1.4.52
+sqlalchemy==1.4.52  # ORM框架
-tenacity==8.5.0
+tenacity==8.5.0  # 重试
-ufile==3.2.9
+ufile==3.2.9  # 云空间
-zxing-cpp==2.2.0
+zxing-cpp==2.2.0  # 二维码识别

services/paddle_services/.dockerignore

@@ -0,0 +1,245 @@
+### PyCharm template
+# Covers JetBrains IDEs: IntelliJ, RubyMine, PhpStorm, AppCode, PyCharm, CLion, Android Studio, WebStorm and Rider
+# Reference: https://intellij-support.jetbrains.com/hc/en-us/articles/206544839
+
+# User-specific stuff
+.idea/**/workspace.xml
+.idea/**/tasks.xml
+.idea/**/usage.statistics.xml
+.idea/**/dictionaries
+.idea/**/shelf
+
+# AWS User-specific
+.idea/**/aws.xml
+
+# Generated files
+.idea/**/contentModel.xml
+
+# Sensitive or high-churn files
+.idea/**/dataSources/
+.idea/**/dataSources.ids
+.idea/**/dataSources.local.xml
+.idea/**/sqlDataSources.xml
+.idea/**/dynamic.xml
+.idea/**/uiDesigner.xml
+.idea/**/dbnavigator.xml
+
+# Gradle
+.idea/**/gradle.xml
+.idea/**/libraries
+
+# Gradle and Maven with auto-import
+# When using Gradle or Maven with auto-import, you should exclude module files,
+# since they will be recreated, and may cause churn.  Uncomment if using
+# auto-import.
+# .idea/artifacts
+# .idea/compiler.xml
+# .idea/jarRepositories.xml
+# .idea/modules.xml
+# .idea/*.iml
+# .idea/modules
+# *.iml
+# *.ipr
+
+# CMake
+cmake-build-*/
+
+# Mongo Explorer plugin
+.idea/**/mongoSettings.xml
+
+# File-based project format
+*.iws
+
+# IntelliJ
+out/
+
+# mpeltonen/sbt-idea plugin
+.idea_modules/
+
+# JIRA plugin
+atlassian-ide-plugin.xml
+
+# Cursive Clojure plugin
+.idea/replstate.xml
+
+# SonarLint plugin
+.idea/sonarlint/
+
+# Crashlytics plugin (for Android Studio and IntelliJ)
+com_crashlytics_export_strings.xml
+crashlytics.properties
+crashlytics-build.properties
+fabric.properties
+
+# Editor-based Rest Client
+.idea/httpRequests
+
+# Android studio 3.1+ serialized cache file
+.idea/caches/build_file_checksums.ser
+
+### Python template
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# 通过卷绑定挂载到容器中
+/log
+/model
+# docker
+Dockerfile

services/paddle_services/Dockerfile

@@ -0,0 +1,28 @@
+# 使用官方的paddle镜像作为基础
+FROM registry.baidubce.com/paddlepaddle/paddle:2.6.1-gpu-cuda12.0-cudnn8.9-trt8.6
+
+# 设置工作目录
+WORKDIR /app
+
+# 设置环境变量
+ENV PYTHONUNBUFFERED=1 \
+    # 设置时区
+    TZ=Asia/Shanghai \
+    # 设置pip镜像地址，加快安装速度
+    PIP_INDEX_URL=https://pypi.tuna.tsinghua.edu.cn/simple
+
+# 安装依赖
+COPY requirements.txt /app/requirements.txt
+RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo '$TZ' > /etc/timezone \
+    && pip install --no-cache-dir -r requirements.txt \
+    && pip uninstall -y onnxruntime onnxruntime-gpu \
+    && pip install onnxruntime-gpu==1.18.0 --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/
+
+# 将当前目录内容复制到容器的/app内
+COPY . /app
+
+# 暴露端口
+# EXPOSE 8081
+
+# 运行api接口，具体接口在命令行或docker-compose.yml文件中定义
+ENTRYPOINT ["gunicorn"]

services/paddle_services/__init__.py

@@ -0,0 +1,21 @@
+"""
+信息抽取关键词配置
+"""
+IE_KEY = {
+    'name': '患者姓名',
+    'admission_date': '入院日期',
+    'discharge_date': '出院日期',
+    'medical_expenses': '费用总额',
+    'personal_cash_payment': '个人现金支付',
+    'personal_account_payment': '个人账户支付',
+    'personal_funded_amount': '自费金额',
+    'medical_insurance_type': '医保类型',
+    'hospital': '医院',
+    'department': '科室',
+    'doctor': '主治医生',
+    'admission_id': '住院号',
+    'settlement_id': '医保结算单号码',
+    'age': '年龄',
+    'uppercase_medical_expenses': '大写总额',
+    'page': '页码',
+}

services/paddle_services/clas_orientation.py

@@ -0,0 +1,30 @@
+import logging.config
+
+from flask import Flask, request
+from paddleclas import PaddleClas
+
+from log import LOGGING_CONFIG
+from utils import process_request
+
+app = Flask(__name__)
+CLAS = PaddleClas(model_name='text_image_orientation')
+
+
+@app.route(rule='/', methods=['POST'])
+@process_request
+def main():
+    """
+    判断图片旋转角度，逆时针旋转该角度后为正。可能值['0', '90', '180', '270']
+    :return: 最有可能的两个角度
+    """
+    img_path = request.form.get('img_path')
+    clas_result = CLAS.predict(input_data=img_path)
+    clas_result = next(clas_result)[0]
+    if clas_result['scores'][0] < 0.5:
+        return ['0', '90']
+    return clas_result['label_names']
+
+
+if __name__ == '__main__':
+    logging.config.dictConfig(LOGGING_CONFIG)
+    app.run('0.0.0.0', 5005)

services/paddle_services/clas_text.py

@@ -0,0 +1,31 @@
+import logging.config
+
+from flask import Flask, request
+from paddlenlp import Taskflow
+
+from log import LOGGING_CONFIG
+from utils import process_request
+
+app = Flask(__name__)
+schema = ['基本医保结算单', '出院记录', '费用清单']
+CLAS = Taskflow('zero_shot_text_classification', model='utc-xbase', schema=schema,
+                task_path='model/text_classification', precision='fp32')
+
+
+@app.route('/', methods=['POST'])
+@process_request
+def main():
+    text = request.form.get('text')
+    cls_result = CLAS(text)
+    cls_result = cls_result[0].get('predictions')
+    if cls_result:
+        cls_result = cls_result[0]
+    if cls_result['score'] and float(cls_result['score']) < 0.8:
+        logging.info(f"识别结果置信度{cls_result['score']}过低！text: {text}")
+        return None
+    return cls_result['label']
+
+
+if __name__ == '__main__':
+    logging.config.dictConfig(LOGGING_CONFIG)
+    app.run('0.0.0.0', 5008)

services/paddle_services/det_book.py

@@ -0,0 +1,31 @@
+import logging.config
+import os.path
+
+import cv2
+from flask import Flask, request
+
+from log import LOGGING_CONFIG
+from paddle_detection import detector
+from utils import process_request, parse_img_path
+
+app = Flask(__name__)
+
+
+@app.route('/', methods=['POST'])
+@process_request
+def main():
+    img_path = request.form.get('img_path')
+    result = detector.get_book_areas(img_path)
+
+    dirname, img_name, img_ext = parse_img_path(img_path)
+    books_path = []
+    for i in range(len(result)):
+        save_path = os.path.join(dirname, f'{img_name}.book_{i}.{img_ext}')
+        cv2.imwrite(save_path, result[i])
+        books_path.append(save_path)
+    return books_path
+
+
+if __name__ == '__main__':
+    logging.config.dictConfig(LOGGING_CONFIG)
+    app.run('0.0.0.0', 5006)

services/paddle_services/dewarp.py

@@ -0,0 +1,28 @@
+import logging.config
+import os
+
+import cv2
+from flask import Flask, request
+
+from doc_dewarp import dewarper
+from log import LOGGING_CONFIG
+from utils import process_request, parse_img_path
+
+app = Flask(__name__)
+
+
+@app.route('/', methods=['POST'])
+@process_request
+def main():
+    img_path = request.form.get('img_path')
+    img = cv2.imread(img_path)
+    dewarped_img = dewarper.dewarp_image(img)
+    dirname, img_name, img_ext = parse_img_path(img_path)
+    save_path = os.path.join(dirname, f'{img_name}.dewarped.{img_ext}')
+    cv2.imwrite(save_path, dewarped_img)
+    return save_path
+
+
+if __name__ == '__main__':
+    logging.config.dictConfig(LOGGING_CONFIG)
+    app.run('0.0.0.0', 5007)

services/paddle_services/doc_dewarp/__init__.py

@@ -0,0 +1,7 @@
+import os
+
+from onnxruntime import InferenceSession
+
+MODEL_PATH = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))),
+                          'model', 'dewarp_model', 'doc_tr_pp.onnx')
+DOC_TR = InferenceSession(MODEL_PATH, providers=['CUDAExecutionProvider'], provider_options=[{'device_id': 0}])

services/paddle_services/doc_dewarp/dewarper.py

@@ -11,10 +11,10 @@ def dewarp_image(image):
     y = to_tensor(image)
 
     img = np.transpose(img, (2, 0, 1))
-    bm = DOC_TR.run(None, {"image": img[None,]})[0]
+    bm = DOC_TR.run(None, {'image': img[None,]})[0]
     bm = paddle.to_tensor(bm)
     bm = paddle.nn.functional.interpolate(
-        bm, y.shape[2:], mode="bilinear", align_corners=False
+        bm, y.shape[2:], mode='bilinear', align_corners=False
     )
     bm_nhwc = np.transpose(bm, (0, 2, 3, 1))
     out = paddle.nn.functional.grid_sample(y, (bm_nhwc / 288 - 0.5) * 2)

services/paddle_services/ie_cost.py

@@ -0,0 +1,36 @@
+import json
+import logging.config
+
+from flask import Flask, request
+from paddlenlp import Taskflow
+
+from __init__ import IE_KEY
+from log import LOGGING_CONFIG
+from utils import process_request
+
+app = Flask(__name__)
+COST_LIST_SCHEMA = tuple(IE_KEY[key] for key in [
+    'name', 'admission_date', 'discharge_date', 'medical_expenses', 'page'
+])
+COST = Taskflow('information_extraction', schema=COST_LIST_SCHEMA, model='uie-x-base',
+                task_path='model/cost_list_model', layout_analysis=False, precision='fp16')
+
+
+@app.route('/', methods=['POST'], endpoint='cost')
+@process_request
+def main():
+    img_path = request.form.get('img_path')
+    layout = request.form.get('layout')
+    return COST({'doc': img_path, 'layout': json.loads(layout)})
+
+
+@app.route('/text', methods=['POST'])
+@process_request
+def text():
+    t = request.form.get('text')
+    return COST(t)
+
+
+if __name__ == '__main__':
+    logging.config.dictConfig(LOGGING_CONFIG)
+    app.run('0.0.0.0', 5004)

services/paddle_services/ie_discharge.py

@@ -0,0 +1,36 @@
+import json
+import logging.config
+
+from flask import Flask, request
+from paddlenlp import Taskflow
+
+from __init__ import IE_KEY
+from log import LOGGING_CONFIG
+from utils import process_request
+
+app = Flask(__name__)
+DISCHARGE_RECORD_SCHEMA = tuple(IE_KEY[key] for key in [
+    'hospital', 'department', 'name', 'admission_date', 'discharge_date', 'doctor', 'admission_id', 'age'
+])
+DISCHARGE = Taskflow('information_extraction', schema=DISCHARGE_RECORD_SCHEMA, model='uie-x-base',
+                     task_path='model/discharge_record_model', layout_analysis=False, precision='fp16')
+
+
+@app.route('/', methods=['POST'], endpoint='discharge')
+@process_request
+def main():
+    img_path = request.form.get('img_path')
+    layout = request.form.get('layout')
+    return DISCHARGE({'doc': img_path, 'layout': json.loads(layout)})
+
+
+@app.route('/text', methods=['POST'])
+@process_request
+def text():
+    t = request.form.get('text')
+    return DISCHARGE(t)
+
+
+if __name__ == '__main__':
+    logging.config.dictConfig(LOGGING_CONFIG)
+    app.run('0.0.0.0', 5003)

services/paddle_services/ie_settlement.py

@@ -0,0 +1,38 @@
+import json
+import logging.config
+
+from flask import Flask, request
+from paddlenlp import Taskflow
+
+from __init__ import IE_KEY
+from log import LOGGING_CONFIG
+from utils import process_request
+
+app = Flask(__name__)
+SETTLEMENT_LIST_SCHEMA = tuple(IE_KEY[key] for key in [
+    'name', 'admission_date', 'discharge_date', 'medical_expenses', 'personal_cash_payment',
+    'personal_account_payment', 'personal_funded_amount', 'medical_insurance_type', 'admission_id', 'settlement_id',
+    'uppercase_medical_expenses'
+])
+SETTLEMENT_IE = Taskflow('information_extraction', schema=SETTLEMENT_LIST_SCHEMA, model='uie-x-base',
+                         task_path='model/settlement_list_model', layout_analysis=False, precision='fp16')
+
+
+@app.route('/', methods=['POST'], endpoint='settlement')
+@process_request
+def main():
+    img_path = request.form.get('img_path')
+    layout = request.form.get('layout')
+    return SETTLEMENT_IE({'doc': img_path, 'layout': json.loads(layout)})
+
+
+@app.route('/text', methods=['POST'])
+@process_request
+def text():
+    t = request.form.get('text')
+    return SETTLEMENT_IE(t)
+
+
+if __name__ == '__main__':
+    logging.config.dictConfig(LOGGING_CONFIG)
+    app.run('0.0.0.0', 5002)

services/paddle_services/log/__init__.py

@@ -0,0 +1,70 @@
+import os
+import socket
+
+# 获取主机名，方便区分容器
+HOSTNAME = socket.gethostname()
+# 检测日志文件的路径是否存在，不存在则创建
+LOG_PATHS = [
+    f'log/{HOSTNAME}/error',
+]
+for path in LOG_PATHS:
+    if not os.path.exists(path):
+        os.makedirs(path)
+
+# 配置字典
+LOGGING_CONFIG = {
+    'version': 1,  # 必需，指定配置格式的版本
+    'disable_existing_loggers': False,  # 是否禁用已经存在的logger实例
+
+    # formatters定义了不同格式的日志样式
+    'formatters': {
+        'standard': {
+            'format': '%(asctime)s [%(levelname)s] %(name)s: %(message)s',
+            'datefmt': '%Y-%m-%d %H:%M:%S',
+        },
+    },
+
+    # handlers定义了不同类型的日志处理器
+    'handlers': {
+        'console': {
+            'class': 'logging.StreamHandler',  # 控制台处理器
+            'level': 'DEBUG',
+            'formatter': 'standard',
+            'stream': 'ext://sys.stdout',  # 输出到标准输出，默认编码跟随系统，一般为UTF-8
+        },
+        'file': {
+            'class': 'logging.handlers.TimedRotatingFileHandler',  # 文件处理器，支持日志滚动
+            'level': 'INFO',
+            'formatter': 'standard',
+            'filename': f'log/{HOSTNAME}/fcb_photo_review.log',  # 日志文件路径
+            'when': 'midnight',
+            'interval': 1,
+            'backupCount': 14,  # 保留的备份文件数量
+            'encoding': 'utf-8',  # 显式指定文件编码为UTF-8以支持中文
+        },
+        'error': {
+            'class': 'logging.handlers.TimedRotatingFileHandler',
+            'level': 'INFO',
+            'formatter': 'standard',
+            'filename': f'log/{HOSTNAME}/error/fcb_photo_review_error.log',
+            'when': 'midnight',
+            'interval': 1,
+            'backupCount': 14,
+            'encoding': 'utf-8',
+        },
+    },
+
+    # loggers定义了日志记录器
+    'loggers': {
+        '': {  # 根记录器
+            'handlers': ['console', 'file'],  # 关联的处理器
+            'level': 'DEBUG',  # 根记录器的级别
+            'propagate': False,  # 是否向上级传播日志信息
+        },
+        'error': {
+            'handlers': ['console', 'file', 'error'],
+            'level': 'DEBUG',
+            'propagate': False,
+        },
+    },
+}

services/paddle_services/model/cost_list_model/README.md

@@ -0,0 +1 @@
+住院费用清单信息抽取微调模型存放目录

services/paddle_services/model/dewarp_model/README.md

@@ -0,0 +1 @@
+图片扭曲矫正模型存放目录

services/paddle_services/model/discharge_record_model/README.md

@@ -0,0 +1 @@
+出院记录信息抽取微调模型存放目录

services/paddle_services/model/object_det_model/README.md

@@ -0,0 +1 @@
+文档检测模型存放目录

services/paddle_services/model/settlement_list_model/README.md

@@ -0,0 +1 @@
+基本医保结算单信息抽取微调模型存放目录

services/paddle_services/model/text_classification/README.md

@@ -0,0 +1 @@
+文本分类模型存放目录

services/paddle_services/ocr.py

@@ -0,0 +1,24 @@
+import logging.config
+
+from flask import Flask, request
+from paddleocr import PaddleOCR
+
+from log import LOGGING_CONFIG
+from utils import process_request
+
+app = Flask(__name__)
+# 如果不希望识别出空格，可以设置use_space_char=False。做此项设置一定要测试，2.7.3版本此项设置有bug，会导致识别失败
+OCR = PaddleOCR(use_angle_cls=False, show_log=False, det_db_thresh=0.1, det_db_box_thresh=0.3, det_limit_side_len=1248,
+                drop_score=0.3)
+
+
+@app.route('/', methods=['POST'])
+@process_request
+def main():
+    img_path = request.form.get('img_path')
+    return OCR.ocr(img_path, cls=False)
+
+
+if __name__ == '__main__':
+    logging.config.dictConfig(LOGGING_CONFIG)
+    app.run('0.0.0.0', 5001)

services/paddle_services/paddle_detection/README.md

@@ -0,0 +1 @@
+README_cn.md

services/paddle_services/paddle_detection/__init__.py

@@ -0,0 +1,7 @@
+import os
+
+from onnxruntime import InferenceSession
+
+MODEL_DIR = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), 'model', 'object_det_model')
+PADDLE_DET = InferenceSession(os.path.join(MODEL_DIR, 'ppyoloe_plus_crn_l_80e_coco_w_nms.onnx'),
+                              providers=['CUDAExecutionProvider'], provider_options=[{'device_id': 0}])