add contract 8.5

周伟奇
Showing 3 changed files with 105 additions and 29 deletions
src/celery_compare/tasks.py
src/common/electronic_afc_contract/afc_contract_ocr.py
src/common/electronic_afc_contract/get_char.py
--- a/src/celery_compare/tasks.py
View file @fd55633
+++ b/src/celery_compare/tasks.py
View file @fd55633
@@ -2970,6 +2970,8 @@ def compare(application_id, application_entity, uniq_seq, ocr_res_id, is_ca=True
        auto_obj = auto_class.objects.filter(application_id=application_id, on_off=True).first()
        if auto_obj is not None:
            auto_result = se_compare_auto(application_id, application_entity, ocr_res_id, last_obj, ocr_res_dict, auto_obj)
+        else:
+            auto_result = None
        full_result = se_compare(application_id, application_entity, ocr_res_id, last_obj, ocr_res_dict, is_cms, auto_result)
--- a/src/common/electronic_afc_contract/afc_contract_ocr.py
View file @fd55633
+++ b/src/common/electronic_afc_contract/afc_contract_ocr.py
View file @fd55633
@@ -6,28 +6,35 @@
 # @Description   :
 from .get_char import Finder
+import numpy as np
 def predict(pdf_info):
    ocr_results = {}
    for pno in pdf_info:
        ocr_results[pno] = {}
+        ocr_result = []
        for key, block in enumerate(pdf_info[pno]['blocks']):
            if block['type'] != 0:
                continue
            for line in block['lines']:
                for span in line['spans']:
                    bbox, text = span['bbox'], span['text']
+                    if len(text) == 0:
+                        continue
                    # print(text)
                    xmin, ymin, xmax, ymax = bbox
                    polygon = [xmin, ymin, xmax, ymin, xmax, ymax, xmin, ymax]
+                    polygon = np.array(polygon, dtype=np.int32).tolist()
                    text = text.replace("：", ":").replace("　", "")
-                    ocr_results[pno][key] = [polygon, text]
+                    ocr_result.append([polygon, text])
+        ocr_result = sorted(ocr_result, key=lambda x: x[0][1], reverse=False)  # 按 y0 从小到大排
+        keys = list(range(len(ocr_result)))
+        ocr_result = dict(zip(keys, ocr_result))
+        ocr_results[pno] = ocr_result
    # 输入是整个 PDF 中的信息
    f = Finder(pdf_info, ocr_results=ocr_results)
    results = f.get_info()
    return results
--- a/src/common/electronic_afc_contract/get_char.py
View file @fd55633
+++ b/src/common/electronic_afc_contract/get_char.py
View file @fd55633
@@ -8,6 +8,7 @@
 import re
 import numpy as np
 from fuzzywuzzy import fuzz
+from shapely.geometry import Polygon
 class Finder:
@@ -111,6 +112,28 @@ class Finder:
                                                    },
                                          }
+    def get_top_iou(self, poly, ocr_result):
+        """传入一个多边形, 找到与之最匹配的多边形
+        Args:
+            poly (TYPE): Description
+        """
+        iou_list = []
+        for key in ocr_result:
+            bbox, text = ocr_result[key]
+            g = Polygon(np.array(bbox).reshape((-1, 2)))
+            p = Polygon(np.array(poly).reshape((-1, 2)))
+            if not g.is_valid or not p.is_valid:
+                continue
+            inter = Polygon(g).intersection(Polygon(p)).area
+            union = g.area + p.area - inter
+            iou = inter / union
+            iou_list.append([iou, key])
+        if len(iou_list) == 0:
+            return -1, -1
+        top_iou = sorted(iou_list, key=lambda x: x[0])[-1]
+        return top_iou
    def poly_to_rectangle(self, poly):
        xmin, ymin, xmax, ymin, xmax, ymax, xmin, ymax = poly
        bbox = [xmin, ymin, xmax, ymax]
@@ -253,38 +276,67 @@ class Finder:
        if mergeindex == -1:
            return text_list
        else:
-            new_text_list = text_list[:mergeindex] + [text_list[mergeindex] + text_list[mergeindex + 1]] + \
+            new_text_list = text_list[:mergeindex] + [text_list[mergeindex] + text_list[mergeindex + 1]] + text_list[
-                            text_list[mergeindex + 2:]
+                                                                                                           mergeindex + 2:]
            return self.mergelist(new_text_list)
    def get_asp_details(self, page_num):
        asp_details_table_term = self.item.copy()
-        asp_details_table = []
+        asp_details_table = [['附加产品融资贷款本金总金额及贷款利率明细'], ['项目1', '用途总金额2', '贷款本金3']]
-        asp_details_text_list = []
+        bbox_xm = None
-        table = False
+        bbox_ytzje = None
-        for block in self.pdf_info[page_num]['blocks']:
+        bbox_dkbj = None
-            if block['type'] != 0:
+        bbox_total = None
+        for key in self.ocr_results[page_num]:
+            bbox, text = self.ocr_results[page_num][key]
+            if text == '项目1':
+                bbox_xm = bbox
+            if text == '用途总金额2':
+                bbox_ytzje = bbox
+            if text == '贷款本金3':
+                bbox_dkbj = bbox
+            if text == '附加产品融资贷款本':
+                bbox_total = bbox
+        # print(bbox_xm, bbox_ytzje, bbox_dkbj, bbox_total)
+        if bbox_xm:
+            for i in range(10):
+                rh = abs(bbox_xm[1] - bbox_xm[-1])
+                anchor = np.array(bbox_xm).reshape((-1, 2))
+                anchor[:, 1] += int(rh * 1.4)
+                _iou, _key = self.get_top_iou(poly=anchor, ocr_result=self.ocr_results[page_num])
+                if _iou > 0:
+                    bbox, xm_text = self.ocr_results[page_num][_key]
+                    bbox_xm = bbox
+                    # 解决项目内容是两行的问题
+                    if not '所购' in xm_text:
+                        line = asp_details_table[-1]
+                        line[0] += xm_text
+                        asp_details_table[-1] = line
                        continue
-            for line in block['lines']:
+                    # print(xm_text)
-                for span in line['spans']:
+                    anchor_1 = [bbox_ytzje[0], bbox[1], bbox_ytzje[2], bbox[3],
-                    bbox, text = span['bbox'], span['text']
+                                bbox_ytzje[4], bbox[5], bbox_ytzje[6], bbox[7]]
-                    if '附加产品融资贷款本金总金额明细' == text:
+                    _iou, _key = self.get_top_iou(poly=anchor_1, ocr_result=self.ocr_results[page_num])
-                        table = True
+                    bbox, ytzje_text = self.ocr_results[page_num][_key]
-                    if '第二条' in text or '征信管理' in text:
+                    # print(ytzje_text)
-                        table = False
+                    anchor_2 = [bbox_dkbj[0], bbox[1], bbox_dkbj[2], bbox[3],
-                    if table == True:
+                                bbox_dkbj[4], bbox[5], bbox_dkbj[6], bbox[7]]
-                        # print(text)
+                    _iou, _key = self.get_top_iou(poly=anchor_2, ocr_result=self.ocr_results[page_num])
-                        asp_details_text_list.append(text)
+                    bbox, dkbj_text = self.ocr_results[page_num][_key]
-        asp_details_text_list = self.mergelist(asp_details_text_list)
+                    # print(dkbj_text)
-        for i in range((len(asp_details_text_list) + 2) // 3):
+                    if xm_text == ytzje_text:
-            line = []
+                        xm_text, ytzje_text = xm_text.split(' ')
-            if i == 0:
+                    line = [xm_text, ytzje_text, dkbj_text]
-                line = [asp_details_text_list[0]]
-            else:
-                for j in range(3):
-                    line.append(asp_details_text_list[i * 3 - 2 + j])
                    asp_details_table.append(line)
-        if len(asp_details_table) > 0:
+                else:
+                    break
+        if bbox_total:
+            anchor = [bbox_total[0], bbox[1], bbox_total[2], bbox[3],
+                      bbox_total[4], bbox[5], bbox_total[6], bbox[7]]
+            _iou, _key = self.get_top_iou(poly=anchor_2, ocr_result=self.ocr_results[page_num])
+            bbox, total_text = self.ocr_results[page_num][_key]
+            asp_details_table.append(['附加产品融资贷款本金总金额:', '', total_text])
            asp_details_table_term['words'] = asp_details_table
        return asp_details_table_term
@@ -678,38 +730,48 @@ class Finder:
            if '附加产品融资贷款本金总金额' in text:
                self.is_asp = True
        self.gen_init_result(self.is_asp)
+        if len(list(self.ocr_results.keys())) <= 8:  # 8.5 版本客户提供的样本出现串页的情况，暂时无法识别
            # Page 1
            # 找合同编号
            contract_no = self.get_contract_no(page_num='0')
+            # print(contract_no)
            self.init_result['page_1']['合同编号'] = contract_no
            # 所购车辆价格
            vehicle_price = self.get_vehicle_price()
+            # print(vehicle_price)
            self.init_result['page_1']['所购车辆价格'] = vehicle_price
            # 车架号
            vin = self.get_vin()
+            # print(vin)
            self.init_result['page_1']['车架号'] = vin
            # 贷款本金金额(如果是 ASP产品)则'贷款本金金额'项目中包含'车辆贷款本金金额'和'附加产品融资贷款本金总金额'两个项目
            upper, lower, asp_1, asp_2 = self.get_loan_principal()
+            # print(upper, lower, asp_1, asp_2)
            self.init_result['page_1']['贷款本金金额']['大写'] = upper
            self.init_result['page_1']['贷款本金金额']['小写'] = lower
            self.init_result['page_1']['贷款本金金额']['车辆贷款本金金额'] = asp_1
            self.init_result['page_1']['贷款本金金额']['附加产品融资贷款本金总金额'] = asp_2
            # 贷款期限
            loan_term = self.get_loan_term()
+            # print(loan_term)
            self.init_result['page_1']['贷款期限'] = loan_term
            # 附加产品融资贷款本金总金额明细（ASP-表格）
            asp_details_table = self.get_asp_details(page_num='0')
+            # print(asp_details_table)
            self.init_result['page_1']['附加产品融资贷款本金总金额明细'] = asp_details_table
            # 借款人签字及时间
            signature = self.get_signature()
+            # print(signature)
            self.init_result['page_1']['借款人签字及时间'] = signature
            #######################################
            # Page 2
            # 找合同编号
            contract_no = self.get_contract_no(page_num='0')
+            # print(contract_no)
            self.init_result['page_2']['合同编号'] = contract_no
            # 找借款人及抵押人(地址字段原本有空格)
            borrower_name, borrower_id = self.get_somebody(top='借款人及抵押人：', bottom='共同借款人：')
+            # print(borrower_name, borrower_id)
            # 这是为了同时兼容 8.1 版本
            if borrower_name['words'] == None:
                borrower_name, borrower_id = self.get_somebody(top='借款人及抵押人：', bottom='共同借款人及共同抵押人：')
@@ -717,6 +779,7 @@ class Finder:
            self.init_result['page_2']['借款人及抵押人']['id'] = borrower_id
            # 找共同借款人及共同抵押人
            co_borrower_name, co_borrower_id = self.get_somebody(top='共同借款人：', bottom='保证人1：')
+            # print(co_borrower_name, co_borrower_id)
            self.init_result['page_2']['共同借款人及共同抵押人']['name'] = co_borrower_name
            self.init_result['page_2']['共同借款人及共同抵押人']['id'] = co_borrower_id
            # 保证人1
@@ -738,6 +801,7 @@ class Finder:
            self.init_result['page_2']['经销商'] = seller
            # 贷款本金金额(如果是 ASP产品)则'贷款本金金额'项目中包含'车辆贷款本金金额'和'附加产品融资贷款本金总金额'两个项目
            upper, lower, asp_1, asp_2 = self.get_loan_principal(page_num='1')
+            # print(upper, lower, asp_1, asp_2)
            self.init_result['page_2']['贷款本金金额']['大写'] = upper
            self.init_result['page_2']['贷款本金金额']['小写'] = lower
            self.init_result['page_2']['贷款本金金额']['车辆贷款本金金额'] = asp_1
@@ -750,6 +814,7 @@ class Finder:
            self.init_result['page_2']['标准利率'] = standard_rate
            # 还款账户
            account, account_name, account_bank = self.get_payback_account()
+            # print(account, account_name, account_bank)
            self.init_result['page_2']['还款账户']['账号'] = account
            self.init_result['page_2']['还款账户']['户名'] = account_name
            self.init_result['page_2']['还款账户']['开户行'] = account_bank
@@ -760,6 +825,7 @@ class Finder:
            self.init_result['page_3']['合同编号'] = contract_no
            # 还款计划表（表格）
            repayment_schedule_table = self.get_repayment_schedule()
+            # print(repayment_schedule_table)
            self.init_result['page_3']['还款计划表'] = repayment_schedule_table
            #######################################
            # Page 4
@@ -768,6 +834,7 @@ class Finder:
            self.init_result['page_4']['合同编号'] = contract_no
            # 附加产品融资贷款本金总金额明细（ASP-表格）
            asp_details_table = self.get_asp_details(page_num='3')
+            # print(asp_details_table)
            self.init_result['page_4']['附加产品融资贷款本金总金额明细'] = asp_details_table
            #######################################
            # Page 5