Skip to content

周伟奇 / bmw-ocr

Go to a project
Toggle navigation
Toggle navigation pinning
e0d31a26 by 周伟奇
Options

Merge branch 'fix/report_ca' into feature/uat-tmp

2 parents 4398d1df e2de024d
Expand all
Inline Side-by-side
Showing 14 changed files with 251 additions and 26 deletions
src/apps/doc/consts.py
......@@ -11,7 +11,7 @@ PAGE_SIZE_DEFAULT = 10
FIXED_APPLICATION_ID_PREFIX = 'CH-S'
DOC_SCHEME_LIST = ['ACCEPTANCE', 'SETTLEMENT', 'CONTRACTMANAGEMENT']
DATA_SOURCE_LIST = ['POS', 'EAPP', 'ECONTRACT']
DATA_SOURCE_LIST = ['POS', 'EAPP', 'ECONTRACT', 'OVP']
COMPARE_DOC_SCHEME_LIST = ['CA', 'SE']
HIL_PREFIX = 'HIL'
......@@ -1057,7 +1057,25 @@ HIL_CONTRACT_2_CLASSIFY = 44
HIL_CONTRACT_3_CN_NAME = '车辆处置协议'
HIL_CONTRACT_3_CLASSIFY = 45
CONTRACT_SET = {CONTRACT_QRS_CLASSIFY, CONTRACT_CLASSIFY, HIL_CONTRACT_1_CLASSIFY, HIL_CONTRACT_2_CLASSIFY, HIL_CONTRACT_3_CLASSIFY}
FSM_CONTRACT_WEP_CN_NAME = '延长保修合同'
FSM_CONTRACT_WEP_CLASSIFY = 51
FSM_CONTRACT_MSI_CN_NAME = '长悦保养合同'
FSM_CONTRACT_MSI_CLASSIFY = 52
FSM_CONTRACT_SC_CN_NAME = '汽车销售合同'
FSM_CONTRACT_SC_CLASSIFY = 53
CONTRACT_SET = {
CONTRACT_QRS_CLASSIFY,
CONTRACT_CLASSIFY,
HIL_CONTRACT_1_CLASSIFY,
HIL_CONTRACT_2_CLASSIFY,
HIL_CONTRACT_3_CLASSIFY,
FSM_CONTRACT_WEP_CLASSIFY,
FSM_CONTRACT_MSI_CLASSIFY,
FSM_CONTRACT_SC_CLASSIFY,
}
CONTRACT_MAP = {
HIL_CONTRACT_1_CLASSIFY: HIL_CONTRACT_1_CN_NAME,
......@@ -1065,8 +1083,13 @@ CONTRACT_MAP = {
HIL_CONTRACT_3_CLASSIFY: HIL_CONTRACT_3_CN_NAME,
CONTRACT_CLASSIFY: CONTRACT_CN_NAME,
CONTRACT_QRS_CLASSIFY: CONTRACT_QRS_CN_NAME,
FSM_CONTRACT_WEP_CLASSIFY: FSM_CONTRACT_WEP_CN_NAME,
FSM_CONTRACT_MSI_CLASSIFY: FSM_CONTRACT_MSI_CN_NAME,
FSM_CONTRACT_SC_CLASSIFY: FSM_CONTRACT_SC_CN_NAME,
}
FSM_CONTRACT_CLASSIFY_SET = {FSM_CONTRACT_WEP_CLASSIFY, FSM_CONTRACT_MSI_CLASSIFY, FSM_CONTRACT_SC_CLASSIFY}
# 保单
INSURANCE_CN_NAME = '保单'
INSURANCE_CLASSIFY = 42
......@@ -1214,6 +1237,11 @@ BS_FIELD = 'bss_ocr'
HIL_CONTRACT_1_FIELD = 'hil_contract_1_ocr'
HIL_CONTRACT_2_FIELD = 'hil_contract_2_ocr'
HIL_CONTRACT_3_FIELD = 'hil_contract_3_ocr'
FSM_CONTRACT_WEP_FIELD = 'fsm_wep_ocr'
FSM_CONTRACT_MSI_FIELD = 'fsm_msi_ocr'
FSM_CONTRACT_SC_FIELD = 'fsm_sc_ocr'
BS_CLASSIFY = 10089
RESULT_MAPPING = {
......@@ -1238,6 +1266,9 @@ RESULT_MAPPING = {
HIL_CONTRACT_1_CLASSIFY: HIL_CONTRACT_1_FIELD,
HIL_CONTRACT_2_CLASSIFY: HIL_CONTRACT_2_FIELD,
HIL_CONTRACT_3_CLASSIFY: HIL_CONTRACT_3_FIELD,
FSM_CONTRACT_WEP_CLASSIFY: FSM_CONTRACT_WEP_FIELD,
FSM_CONTRACT_MSI_CLASSIFY: FSM_CONTRACT_MSI_FIELD,
FSM_CONTRACT_SC_CLASSIFY: FSM_CONTRACT_SC_FIELD,
}
CA_ADD_COMPARE_FIELDS = (IC_OCR_FIELD, BL_OCR_FIELD, BS_FIELD)
......@@ -2313,29 +2344,42 @@ APPLICANT_TYPE_MAP = {
APPLICANT_TYPE_ORDER = ['Borrower', 'Co-Borrower', 'Guarantor', 'Mortgager']
FILE_NAME_PREFIX_MAP = {
AFC_PREFIX: [
((CONTRACT_CLASSIFY, 0), '{0}_电子签署-汽车抵押贷款合同'),
((HMH_CLASSIFY, 0), '{0}_电子签署-抵押登记豁免函'),
],
HIL_PREFIX: [
((HIL_CONTRACT_1_CLASSIFY, HIL_CONTRACT_3_CLASSIFY), '{0}_电子签署-售后回租合同'),
((HIL_CONTRACT_2_CLASSIFY, 0), '{0}_电子签署-汽车租赁抵押合同'),
((HMH_CLASSIFY, 0), '{0}_电子签署-抵押登记豁免函'),
]
}
# FILE_NAME_PREFIX_MAP = {
# AFC_PREFIX: [
# ((CONTRACT_CLASSIFY, 0), '{0}_电子签署-汽车抵押贷款合同'),
# ((HMH_CLASSIFY, 0), '{0}_电子签署-抵押登记豁免函'),
# ],
# HIL_PREFIX: [
# ((HIL_CONTRACT_1_CLASSIFY, HIL_CONTRACT_3_CLASSIFY), '{0}_电子签署-售后回租合同'),
# ((HIL_CONTRACT_2_CLASSIFY, 0), '{0}_电子签署-汽车租赁抵押合同'),
# ((HMH_CLASSIFY, 0), '{0}_电子签署-抵押登记豁免函'),
# ]
# }
ECONTRACT_KEYWORDS_MAP = {
AFC_PREFIX: [
('抵押贷款合同', CONTRACT_CLASSIFY),
('送达地址确认书', CONTRACT_QRS_CLASSIFY),
# ('电子签署-抵押登记豁免函', HMH_CLASSIFY, 0),
('抵押登记豁免函', HMH_CLASSIFY),
],
HIL_PREFIX: [
('售后回租合同', HIL_CONTRACT_1_CLASSIFY),
('租赁抵押合同', HIL_CONTRACT_2_CLASSIFY),
('车辆处置协议', HIL_CONTRACT_3_CLASSIFY),
# ('电子签署-抵押登记豁免函', HMH_CLASSIFY, 0),
('抵押登记豁免函', HMH_CLASSIFY),
]
}
FSM_ECONTRACT_KEYWORDS_MAP = {
AFC_PREFIX: [
('延长保修条款与条件', FSM_CONTRACT_WEP_CLASSIFY),
('长悦保养套餐服务合约', FSM_CONTRACT_MSI_CLASSIFY),
('汽车销售合同', FSM_CONTRACT_SC_CLASSIFY),
],
HIL_PREFIX: [
('延长保修条款与条件', FSM_CONTRACT_WEP_CLASSIFY),
('长悦保养套餐服务合同', FSM_CONTRACT_MSI_CLASSIFY),
('汽车销售合同', FSM_CONTRACT_SC_CLASSIFY),
]
}
......@@ -2345,6 +2389,12 @@ HIL_CONTRACT_TYPE_MAP = {
str(HIL_CONTRACT_3_CLASSIFY): 1,
}
FSM_CONTRACT_TYPE_MAP = {
str(FSM_CONTRACT_WEP_CLASSIFY): 0,
str(FSM_CONTRACT_MSI_CLASSIFY): 1,
str(FSM_CONTRACT_SC_CLASSIFY): 2,
}
RESULT_MAP = {
0: None,
1: True,
......
src/apps/doc/management/commands/ocr_process.py
......@@ -20,6 +20,8 @@ from common.tools.file_tools import get_pwd_list_from_str, extract_zip_or_rar, g
from common.tools.pdf_to_img import PDFHandler
from common.electronic_afc_contract.afc_contract_ocr import predict as afc_predict
from common.electronic_hil_contract.hil_contract_ocr import predict as hil_predict
from common.fsm_econtract.fsm_contract_ocr import predict as fsm_predict
from common.fsm_econtract.hmh_ocr import predict as hmh_predict
from apps.doc import consts
# from apps.doc.ocr.edms import EDMS, rh
from apps.doc.ocr.ecm import ECM, rh
......@@ -996,7 +998,7 @@ class Command(BaseCommand, LoggerMixin):
res.setdefault(consts.ALL_POSITION_KEY, dict())[key] = page_info_dict.get(str(pno), {}).get(
consts.ALL_POSITION_KEY, {}).get(key1, [])
license_summary[classify] = [res]
else:
elif classify in consts.SE_HIL_CON_MAP: # TODO FSM新合同写入数据库用于比对
res = {}
for key, (pno1, pno2, end_idx, key1, key2) in consts.SE_HIL_CON_MAP[classify].items():
if pno1 is None:
......@@ -1442,7 +1444,7 @@ class Command(BaseCommand, LoggerMixin):
self.log_base, traceback.format_exc()))
error_list.append(1)
return
else: # e-contract
else: # e-contract or or e-fsm-contract or e-hmh
try:
# pdf下载 处理 图片存储 识别
for times in range(consts.RETRY_TIMES):
......@@ -1472,8 +1474,10 @@ class Command(BaseCommand, LoggerMixin):
self.online_log.error('{0} [process error (db save)] [error={1}]'.format(
self.log_base, traceback.format_exc()))
# AFC合同
if classify_1_str == str(consts.CONTRACT_CLASSIFY):
ocr_result = afc_predict(pdf_handler.pdf_info)
is_fsm = doc.data_source == consts.DATA_SOURCE_LIST[3]
ocr_result = afc_predict(pdf_handler.pdf_info, is_fsm=is_fsm)
page_res = {}
for page_num, page_info in ocr_result.get('page_info', {}).items():
if isinstance(page_num, str) and page_num.startswith('page_'):
......@@ -1483,6 +1487,7 @@ class Command(BaseCommand, LoggerMixin):
'page_num': page_num,
'page_info': page_info
}
# 送达地址确认书
elif classify_1_str == str(consts.CONTRACT_QRS_CLASSIFY):
ocr_result = afc_predict(pdf_handler.pdf_info, is_qrs=True)
page_num = 'page_1'
......@@ -1493,9 +1498,11 @@ class Command(BaseCommand, LoggerMixin):
'page_info': ocr_result.pop(page_num, {})
}
}
else:
# HIL合同
elif classify_1_str in consts.HIL_CONTRACT_TYPE_MAP:
is_fsm = doc.data_source == consts.DATA_SOURCE_LIST[3]
file_type_1 = consts.HIL_CONTRACT_TYPE_MAP.get(classify_1_str)
ocr_result_1 = hil_predict(pdf_handler.pdf_info, file_type_1)
ocr_result_1 = hil_predict(pdf_handler.pdf_info, file_type_1, is_fsm=is_fsm)
rebuild_res_1 = {}
page_res = {}
for field_name, field_info in ocr_result_1.items():
......@@ -1508,9 +1515,36 @@ class Command(BaseCommand, LoggerMixin):
'page_num': page_num,
'page_info': page_info
}
# FSM合同 WEP MSI SC
elif classify_1_str in consts.FSM_CONTRACT_TYPE_MAP:
file_type = consts.FSM_CONTRACT_TYPE_MAP.get(classify_1_str)
ocr_result = fsm_predict(pdf_handler.pdf_info, file_type)
page_res = {}
for page_num, page_info in ocr_result.items():
if isinstance(page_num, str) and page_num.startswith('page_'):
page_res[page_num] = {
'classify': int(classify_1_str),
'page_num': page_num,
'page_info': page_info
}
# hmh
# else:
# pass
contract_res = {}
for img_path_tmp, page_key in pdf_handler.img_path_pno_list:
if classify_1_str == str(consts.HMH_CLASSIFY):
img_contract_res = {
'code': 1,
'data': [
{
'classify': consts.HMH_CLASSIFY,
'data': hmh_predict(pdf_handler.pdf_info)
}
]
}
else:
if page_key in page_res:
img_contract_res = {
'code': 1,
......
src/apps/doc/named_enum.py
......@@ -36,6 +36,7 @@ class RequestTrigger(NamedEnum):
DOCUPLOAD = (3, 'Document Upload')
SUBMITING = (4, 'Submiting')
UPLOADING = (5, 'Uploading')
OVP = (6, 'OVP')
class FailureReason(NamedEnum):
......
src/apps/doc/ocr/wb.py
......@@ -780,10 +780,12 @@ class BSWorkbook(Workbook):
if field_str is not None:
count_list.append((field_str, count))
def contract_rebuild(self, contract_result_dict):
def contract_rebuild(self, contract_result_dict, is_ca=False):
for classify, contract_result in contract_result_dict.items():
if len(contract_result) == 0:
continue
if is_ca and classify not in consts.FSM_CONTRACT_CLASSIFY_SET:
continue
ws = self.create_sheet(consts.CONTRACT_MAP.get(classify))
for i in range(30):
if str(i) in contract_result:
......@@ -906,6 +908,7 @@ class BSWorkbook(Workbook):
else:
self.bs_rebuild(bs_summary, res_count_tuple, metadata)
self.license_rebuild(license_summary, document_scheme, count_list)
self.contract_rebuild(contract_result, True)
self.move_res_sheet()
self.remove_base_sheet()
return count_list
......
src/apps/doc/views.py
......@@ -602,13 +602,22 @@ class UploadDocView(GenericView, DocHandler):
is_zip = False
classify_1 = 0
# 电子合同
if data_source == consts.DATA_SOURCE_LIST[-1] and document_scheme == consts.DOC_SCHEME_LIST[1]:
# 电子合同 Econtract or OVP(FSM)
if data_source == consts.DATA_SOURCE_LIST[2] or data_source == consts.DATA_SOURCE_LIST[3]:
if document_scheme == consts.DOC_SCHEME_LIST[1]:
for keyword, classify_1_tmp in consts.ECONTRACT_KEYWORDS_MAP.get(prefix):
if keyword in document_name:
classify_1 = classify_1_tmp
break
elif document_name.endswith('.zip') or document_name.endswith('.rar') or document_name.endswith('.ZIP') \
# FSM合同：WEP/MSI/SC
elif data_source == consts.DATA_SOURCE_LIST[0] and document_scheme == consts.DOC_SCHEME_LIST[0]:
for keyword, classify_1_tmp in consts.FSM_ECONTRACT_KEYWORDS_MAP.get(prefix):
if keyword in document_name:
classify_1 = classify_1_tmp
break
if document_name.endswith('.zip') or document_name.endswith('.rar') or document_name.endswith('.ZIP') \
or document_name.endswith('.RAR'):
is_zip = True
......
src/common/electronic_afc_contract/afc_contract_ocr.py
......@@ -6,6 +6,7 @@
# @Description :
from .get_char import Finder
from .get_char_fsm import Finder as FSMFinder
import numpy as np
......@@ -23,7 +24,7 @@ def extract_info(ocr_results):
return {'page_1': {'合同编号': contract_no}}
def predict(pdf_info, is_qrs=False):
def predict(pdf_info, is_qrs=False, is_fsm=False):
ocr_results = {}
for pno in pdf_info:
ocr_results[pno] = {}
......@@ -50,6 +51,9 @@ def predict(pdf_info, is_qrs=False):
results = extract_info(ocr_results)
else:
# 输入是整个 PDF 中的信息
if is_fsm:
f = FSMFinder(pdf_info, ocr_results=ocr_results)
else:
f = Finder(pdf_info, ocr_results=ocr_results)
results = f.get_info()
return results
......
src/common/electronic_hil_contract/hil_contract_ocr.py
......@@ -6,9 +6,10 @@
# @Description :
from .get_char import Finder
from .get_char_fsm import Finder as FSMFinder
def predict(pdf_info, file_cls):
def predict(pdf_info, file_cls, is_fsm=False):
"""Summary
Args:
......@@ -58,6 +59,10 @@ def predict(pdf_info, file_cls):
pdf_info = dict()
for pno, page_info in enumerate(pdf_info_1):
pdf_info[str(pno)] = page_info
if is_fsm:
f = FSMFinder(pdf_info)
else:
f = Finder(pdf_info)
if file_cls == 0:
results = f.get_info()
......
src/common/fsm_econtract/const.py 0 → 100644
WEP_FIELD = {
"0": {
'keys': {
'客户姓名': [('客户姓名', (r'^客户姓名.?$', r'^企业名称.?$'), 'top1', {})],
'证件类型': [('证件类型', (r'^证件类型.?$', ), 'top1', {})],
'证件号码': [('证件号码', (r'^证件号码.?$', r'^统一社会信用代码.?$'), 'top1', {})],
'合同价格（小写）': [('人民币', (r'^人民币￥.?$', ), 'top1', {})],
'客户签名': [('客户签名／盖章', (r'^客户签名／盖章.*$', ), 'top1', {})],
'签单日期': [('签单日期', (r'^签单日期.*签单日期.?$', ), 'top1', {})],
},
'value': {
'客户姓名': ('text', 'right', {'offset_tuple': (-1.1, 1, 0.3, 0)}, ''),
'证件类型': ('text', 'right', {'offset_tuple': (-1, 1, 0, 0)}, ''),
'证件号码': ('text', 'right', {'offset_tuple': (-1, 2, 0.3, 0)}, ''),
'合同价格（小写）': ('text', 'right', {'offset_tuple': (-1, 1, 0.3, 0)}, ''),
'客户签名': ('img', 'under', {'offset_tuple': (0, 0, 0, 4), 'rigorous': True}, '无'),
'签单日期': ('img', 'right', {'offset_tuple': (0, 0, 1.1, 0), 'rigorous': True}, '无'),
},
}
}
MSI_FIELD = {
"0": {
'keys': {
'客户姓名': [('客户姓名', (r'^客户姓名.?$', r'^企业名称.?$'), 'top1', {})],
'证件类型': [('证件类型', (r'^证件类型.?$', ), 'top1', {})],
'证件号码': [('证件号码', (r'^证件号码.?$', r'^统一社会信用代码.?$'), 'top1', {})],
'合同价格（小写）': [('人民币', (r'^人民币￥.?$', ), 'top1', {})],
},
'value': {
'客户姓名': ('text', 'right', {'offset_tuple': (-1.2, 1, 0.3, 0)}, ''),
'证件类型': ('text', 'right', {'offset_tuple': (-1, 1, 0, 0)}, ''),
'证件号码': ('text', 'right', {'offset_tuple': (-1, 2, 0.3, 0)}, ''),
'合同价格（小写）': ('text', 'right', {'offset_tuple': (-1, 1, 0.3, 0)}, ''),
},
},
"1": {
'keys': {
'客户签名': [('客户签名／盖章', (r'^客户签名／盖章.*$', ), 'top1', {})],
'签单日期': [('签单日期', (r'^签单日期.*签单日期.?$', ), 'top1', {})],
},
'value': {
'客户签名': ('img', 'under', {'offset_tuple': (0, 0, 0, 4), 'rigorous': True}, '无'),
'签单日期': ('img', 'right', {'offset_tuple': (0, 0, 1.1, 0), 'rigorous': True}, '无'),
},
}
}
SC_FIELD = {
"0": {
'keys': {
'姓名': [('姓名', (r'^姓名.?$', r'^企业名称.?$'), 'top1', {})],
'证件类型': [('证件类型', (r'^证件类型.?$', ), 'top1', {})],
'证件号码': [('证件号码', (r'^证件号码.?$', r'^统一社会信用代码.?$'), 'top1', {})],
'总价': [('总价', (r'^总价.?$', ), 'top1', {})],
},
'value': {
'姓名': ('text', 'right', {'offset_tuple': (-2, 8, 0.5, 0)}, ''),
'证件类型': ('text', 'right', {'offset_tuple': (-2, 6, 0.5, 0)}, ''),
'证件号码': ('text', 'right', {'offset_tuple': (-2, 6, 0.5, 0)}, ''),
'总价': ('text', 'right', {'offset_tuple': (-2, 12, 0.5, 0)}, ''),
},
},
"-1": {
'keys': {
'客户签名': [('客户签名/盖章', (r'^客户签名/盖章.*$', r'^客户签名／盖章.*$'), 'top1', {})],
'签单日期': [('签单日期', (r'^签单日期.*签单日期.?$', ), 'top1', {})],
},
'value': {
'客户签名': ('img', 'under', {'offset_tuple': (1.5, 1, 0, 4), 'rigorous': True}, '无'),
'签单日期': ('img', 'right', {'offset_tuple': (0, 0, 1.1, 0), 'rigorous': True}, '无'),
},
}
}
src/common/fsm_econtract/fsm_contract_ocr.py 0 → 100644
from .retriever import Retriever
from .const import WEP_FIELD, MSI_FIELD, SC_FIELD
from .tools import pdf_info_rebuild
retriever_list = [Retriever(WEP_FIELD), Retriever(MSI_FIELD), Retriever(SC_FIELD)]
def predict(pdf_info, file_type=0):
retriever = retriever_list[file_type]
pdf_text_list, pdf_img_list = pdf_info_rebuild(pdf_info)
return retriever.get_target_fields(pdf_text_list, pdf_img_list)
src/common/fsm_econtract/hmh_ocr.py 0 → 100644
from .retriever import HMHRetriever
from .tools import pdf_info_rebuild
hmh_retriever = HMHRetriever()
def predict(pdf_info):
pdf_text_list, _ = pdf_info_rebuild(pdf_info, fix_bbox=False)
return hmh_retriever.get_target_fields(pdf_text_list)
src/common/fsm_econtract/tools.py 0 → 100644
def pdf_info_rebuild(pdf_info, fix_bbox=True):
pdf_text_info = dict()
pdf_img_info = dict()
for pno_str, page_info in pdf_info.items():
text_set = set()
for block in page_info['blocks']:
if block['type'] == 0:
# text有重复的现象
text_set.clear()
for line in block['lines']:
for span in line['spans']:
bbox, text = span['bbox'], span['text'].strip()
if len(text) != 0 and text not in text_set:
text_set.add(text)
# bbox的高，不准
if fix_bbox and bbox[-1] - bbox[1] < span['size']:
bbox[-1] = bbox[-1] + span['size']
pdf_text_info.setdefault(pno_str, list()).append([bbox, text])
elif block['type'] == 1:
pdf_img_info.setdefault(pno_str, list()).append((block['bbox'], '有'))
return pdf_text_info, pdf_img_info
\ No newline at end of file
Styling with Markdown is supported
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!