Insurance policy identification method and system

Abstract

The invention relates to an insurance policy identification method and system. The method comprises the steps of insurance policy text identification: obtaining full-amount insurance policy data in an insurance policy image to be identified by adopting a trained text detection and identification model; determining an insurance policy type: matching characters in the insurance policy data with preset insurance policy type key characters, and determining the insurance policy type to which the to-be-identified insurance policy image belongs according to a matching result; and insurance policy content matching: according to multiple preset matching fields corresponding to the insurance policy type, matching the total amount of insurance policy data by using multiple matching rules to obtain an insurance policy identification result, and storing the insurance policy identification result. Compared with the prior art, the method has the advantages that the method is suitable for various different types of insurance policy layouts, and the identification accuracy is high.

Description

Technical Field

[0001] The present invention relates to the field of image recognition, and in particular to an insurance policy recognition method and system. Background Art

[0002] With the continuous development of image processing technology, text recognition technology has been widely used in the recognition of various simple documents, such as identity cards, bank cards, etc.

[0003] In the insurance field, existing technologies mostly extract text based on the layout of the insurance policy and after identifying and presetting the position. However, due to the large amount of information and complex layout of the insurance policy, it is difficult to find a versatile optical character recognition product on the market with high recognition accuracy that can be applied to most insurance policy templates. Summary of the invention

[0004] The purpose of the present invention is to provide an insurance policy identification method in order to overcome the defect of the above-mentioned prior art that the insurance policy identification accuracy is low for various types of insurance policy pages.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] According to a first aspect of the present invention, there is provided an insurance policy identification method, the method comprising:

[0007] Insurance policy text recognition: Use the trained text detection and recognition model to obtain the full amount of insurance policy data in the insurance policy image to be recognized;

[0008] Insurance policy type determination: matching characters in the insurance policy data with preset insurance policy category key characters, and determining the insurance policy type to which the insurance policy image to be identified belongs based on the matching result;

[0009] Insurance policy content matching: According to the multiple preset matching fields corresponding to the insurance policy type, multiple matching rules are used to match the full amount of insurance policy data to obtain and store the insurance policy recognition results.

[0010] Preferably, the characters in the insurance policy data are matched with preset insurance policy category key characters, and the insurance policy type to which the insurance policy image to be identified belongs is determined according to the matching result, specifically:

[0011] According to the preset character rules, search for text content that meets the rules in the first M characters of the full policy data, and return the recognition results to the type parameter;

[0012] According to the secondary preset character rule, determine whether the corresponding character is within the type parameter to determine the insurance policy type; the insurance policy types include commercial insurance and traffic insurance, the character corresponding to commercial insurance is 'commercial', and the character corresponding to compulsory traffic insurance is any one of 'traffic' and 'compulsory'.

[0013] Preferably, the primary preset character rule is: obtain under the same line of text, 1 to 20 characters + 'electronic insurance policy' or 'insurance policy'.

[0014] Preferably, when the insurance policy type is compulsory traffic insurance, the multiple preset matching fields include vehicle use nature, insurance company, amount, vehicle identification number, name of the insured, compulsory traffic insurance policy number, start date of insurance, end date of insurance, effective date, premium, and vehicle and vessel tax;

[0015] When the insurance policy type is commercial insurance, the multiple preset matching fields include: vehicle use nature, insurance company, vehicle identification number, name of the insured, commercial insurance policy number, start date of insurance, end date of insurance, effective date, premium, name of vehicle damage insurance, insured amount of vehicle damage insurance, deductible of vehicle damage insurance, premium of vehicle damage insurance, name of third-party liability insurance, insured amount of third-party liability insurance, premium of third-party liability insurance, name of driver liability insurance on the vehicle, insured amount of driver liability insurance on the vehicle, premium of driver liability insurance on the vehicle, name of passenger insurance on the vehicle, insured amount of passenger insurance on the vehicle, and premium of passenger insurance on the vehicle;

[0016] Among them, each multiple preset matching field includes at least one expression form and is updated regularly according to the type of insurance policy.

[0017] Preferably, the matching process of the policy number includes:

[0018] First rule of the policy number: obtain under the same line of text, 'insurance' + 0 to 2 arbitrary characters + 'policy' + 0 to 2 arbitrary characters + 'number' + 0 or 1 ':' + 1 to 40 characters of arbitrary numbers or English letter characters + line break

[0019] Second rule of the policy number: obtain under the same line of text, 'insurance' + 0 or 1 arbitrary character + 'number' + 0 or 1 arbitrary character + 'code' + 0 or 1 ':' + 1 to 40 characters of arbitrary numbers or English letter characters + line break

[0020] Among them, '*' represents the character **;

[0021] When the content obtained by the first rule of the policy number is equal to the null value, identify the content obtained by the second rule of the policy number, otherwise skip the content obtained by the second rule of the policy number.

[0022] Preferably, the matching process of the premium extracts data in the way of extracting Chinese capitalized amount, including:

[0023] The first - stage cleaning of the premium is used to obtain the premium field containing the amount in Chinese capital numbers, and the characters after the first - stage cleaning are obtained.

[0024] The second - stage cleaning of the premium is used to remove the redundant fields in the first - stage cleaning of the premium, and for the remaining amount in Chinese capital numbers, by defining a dictionary to parse Arabic numbers and defining a dictionary to parse monetary units, the Arabic - form numerical characters are obtained after conversion.

[0025] Preferably, the process of the first - stage cleaning of the premium is as follows:

[0026] The first rule of the first stage of the premium: 'Total insurance premium'+ 0 - ∞ Chinese numerical characters + 'whole'

[0027] The second rule of the first stage of the premium: 'Total'+ 0 - 4 arbitrary characters + 'Renminbi in capital letters'+ 0 - ∞ Chinese numerical characters + 'whole'

[0028] The third rule of the first stage of the premium: 'Total'+ 0 - 4 arbitrary characters + 'Renminbi in capital letters'+ 0 - ∞ Chinese numerical characters + 'jiao'+ 0 - ∞ Chinese characters + 0 or 1 'fen'

[0029] When the first rule of the first stage of the premium is not satisfied, the second rule of the first stage of the premium is run. When the second rule of the first stage of the premium is not satisfied, the third rule of the first stage of the premium is run, and the characters after the first - stage cleaning are obtained.

[0030] Preferably, the matching process of the start - insurance time and the end - insurance time includes three cleaning processes, specifically:

[0031] 1) The first - stage cleaning:

[0032] Rule: 'Insurance'+ 0 - 2 arbitrary characters + 'period'+ 0 - 2 arbitrary characters + 'interval'+ 0 - 70 arbitrary characters + line - break character

[0033] 2) The second - stage cleaning:

[0034] The first rule of time: [2 - 4 digits from 0 - 9+0 - 2 arbitrary characters + 'year'+0 - 2 arbitrary characters + 1 - 2 digits from 0 - 9+0 - 2 arbitrary characters +'month'+0 - 2 arbitrary characters + 1 - 2 digits from 0 - 9+0 - 2 arbitrary characters + 'day'+0 - 2 arbitrary characters + 1 - 2 digits from 0 - 9+0 - 2 arbitrary characters + 1 - 2 digits from 0 - 9+0 - 2 arbitrary characters + 1 - 2 digits from 0 - 9+0 - 2 arbitrary characters]

[0035] Second rule for time: 2-4 0-9 numbers + 0-2 arbitrary characters + 'year' + 0-2 arbitrary characters + 1-2 0-9 numbers + 0-2 arbitrary characters + 'month' + 0-2 arbitrary characters + 1-2 0-9 numbers + 0-2 arbitrary characters] + 'day' + 0-2 arbitrary characters + [1-2 0-9 numbers] + 'hour'

[0036] When the first time rule is not satisfied, the second time rule is run, only numbers are obtained without text, and characters are obtained through secondary cleaning;

[0037] 3) The third cleaning process makes a rational judgment on the characters obtained by the second cleaning process.

[0038] Preferably, the third cleaning makes a rationality judgment on the characters obtained by the second cleaning, including:

[0039] If there is no string to obtain minutes and seconds, append '00 minutes 00 seconds';

[0040] If the obtained time is in Chinese, convert it into the corresponding Arabic numerals through the dictionary list;

[0041] If the obtained time is not reasonable, the time '24:00:00' is processed back one second to '23:59:59';

[0042] If the number of characters in the obtained string is insufficient, fill it up.

[0043] According to a second aspect of the present invention, there is provided an insurance policy identification system, the system comprising:

[0044] An insurance policy text recognition module is used to obtain the full amount of insurance policy data in the insurance policy image to be recognized based on the insurance policy electronic document through a trained text detection and recognition model;

[0045] The insurance policy type determination module is used to match the text content of the header through preset character rules, determine key characters, and determine the insurance policy type;

[0046] Insurance policy content matching module: Based on multiple preset fields corresponding to the insurance policy type, match the full amount of insurance policy data to obtain preliminary recognition results, and then preset fields and judgment logic again to obtain and store the final insurance policy recognition results;

[0047] The storage module is used to store the identified insurance policy data.

[0048] Compared with the prior art, the present invention has the following advantages:

[0049] 1) The present invention recognizes the characters of the entire insurance policy by using multiple preset character rules, avoiding the use of positioning coordinates to extract insurance policy information. It can perform more targeted and accurate insurance policy recognition for different types of insurance policies and can cope with a variety of different types of layouts, alleviating the technical problem of difficulty in recognizing insurance policies for a variety of different types of insurance policy layouts.

[0050] 2) It uses multiple preset matching fields and can be updated regularly according to the type of insurance policy, which can effectively cope with various types of layouts;

[0051] 3) Multiple matching rules are used for content matching, which makes insurance policy identification more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a flow chart of the method of the present invention;

[0053] Figure 2 This is a schematic diagram of data flow of the insurance policy identification system in the embodiment;

[0054] Figure 3 This is a diagram showing the insurance policy type identification result in the embodiment. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0056] Example

[0057] like Figure 1 As shown, this embodiment provides an insurance policy identification method, which includes:

[0058] Insurance policy text recognition: Based on the insurance policy image to be recognized, the full amount of insurance policy data in the insurance policy image to be recognized is obtained through the trained text detection and recognition model;

[0059] Insurance policy type determination: matching characters in the insurance policy data with preset insurance policy category key characters, and determining the insurance policy type to which the insurance policy image to be identified belongs based on the matching result;

[0060] Insurance policy content matching: According to the multiple preset matching fields corresponding to the insurance policy type, multiple matching rules are used to match the full amount of insurance policy data to obtain and store the insurance policy recognition results; specifically, the full amount of insurance policy data is matched to obtain preliminary recognition results, and the fields and judgment logic are preset again to finally obtain accurate recognition results.

[0061] Next, the method of this embodiment is introduced in detail.

[0062] 1. Insurance Policy Text Recognition

[0063] The insurance policy electronic file obtains the full amount of insurance policy data in the insurance policy image to be identified through the trained text detection and recognition model;

[0064] The text detection and recognition model in this embodiment can adopt any existing text detection and recognition model, for example, the .extract_text method related to the PDF plumber.PDF library of Python can be used to obtain a string of the entire policy text, or the page.extractText() method related to the PyPDF2 library can be used to obtain it.

[0065] 2. Determination of Insurance Policy Type

[0066] 1. According to a preset character rule: search for matching text content in the first M characters of the full policy data, and return the recognition result to the ins_type parameter; in this embodiment, M can be set to 200;

[0067] Rule 1: [Number of characters (any from 1 to 20) + 'electronic policy or insurance policy']

[0068] Rule 2: On the same line of text

[0069] 2. According to the secondary preset character rules, determine whether the corresponding character is within the ins_type type parameter to determine the insurance policy type; the insurance policy types include commercial insurance and traffic insurance. The character corresponding to commercial insurance is "commercial", and the character corresponding to compulsory traffic insurance is any one of "traffic" and "compulsory traffic insurance".

[0070] 3. Insurance policy content matching:

[0071] When the insurance policy type is compulsory traffic insurance, the corresponding multiple preset matching fields include vehicle usage, insurance company, amount, frame number, insured person name, compulsory traffic insurance policy number, insurance start time, insurance end time, effective time, premium and vehicle and vessel tax;

[0072] When the insurance policy type is commercial insurance, the corresponding multiple preset matching fields include: vehicle usage nature, insurance company, frame number, insured person name, commercial insurance policy number, insurance start time, insurance end time, effective time, premium, vehicle damage insurance name, vehicle damage insurance coverage, vehicle damage insurance deductible, vehicle damage insurance premium, third party liability insurance name, third party liability insurance coverage, third party liability insurance premium, driver liability insurance name, driver liability insurance coverage, driver liability insurance premium, passenger insurance name, passenger insurance coverage, and passenger insurance premium;

[0073] Among them, each multiple preset matching field includes at least one expression form and is updated regularly according to the type of insurance policy.

[0074] 1. The matching process of the policy number is as follows:

[0075] The first rule of the policy number: Obtained in the same line of text, ['保' + 0 to 2 arbitrary characters + '险' + 0 to 2 arbitrary characters + '单' + 0 to 2 arbitrary characters + '号' + 0 or 1 ':' + 1 to 40 characters of arbitrary numbers or English letter characters + line break]

[0076] The second rule of the policy number: Obtained in the same line of text, ['保' + 0 or 1 arbitrary character + '险' + 0 or 1 arbitrary character + '号' + 0 or 1 arbitrary character + '码' + 0 or 1 ':' + 1 to 40 characters of arbitrary numbers or English letter characters + line break]

[0077] Among them, '*' represents a character **;

[0078] When the content obtained by the first rule of the policy number is equal to the null value, identify the content obtained by the second rule of the policy number, otherwise skip identifying the content obtained by the second rule of the policy number.

[0079] 2. The matching process of the premium extracts data in the way of extracting Chinese capitalized amounts, including:

[0080] The first cleaning of the premium is used to obtain the premium field containing capitalized numerical amounts, and the first-cleaned characters are obtained, including:

[0081] The second cleaning of the premium is used to remove the redundant fields in the first cleaning of the premium, and for the remaining capitalized numerical amounts, define a dictionary to parse Arabic numerals and define a dictionary to parse monetary units, and convert them to obtain Arabic-form numerical characters.

[0082] Among them, the process of the first cleaning of the premium is as follows:

[0083] The first rule of the first stage of the premium: ['保险费合计' + 0 - ∞ Chinese numerical characters + '整']

[0084] The second rule of the first stage of the premium: ['合计' + 0 - 4 arbitrary characters + '人民币大写' + 0 - ∞ Chinese numerical characters + '整']

[0085] The third rule of the first stage of the premium: ['合计' + 0 - 4 arbitrary characters + '人民币大写' + 0 - ∞ Chinese numerical characters + '角' + 0 - ∞ Chinese characters + 0 or 1 '分']

[0086] When the first rule of the first stage of the premium is not satisfied, the second rule of the first stage of the premium is run. When the second rule of the first stage of the premium is not satisfied, the third rule of the first stage of the premium is run to obtain the first-stage cleaned characters.

[0087] Further, the rule: person + [any character from 0 to 1]min + [any character from 0 to 1]currency

Explanation of the left-side characters: They must be included but not used in the returned result

Explanation of the characters: The content within "" is the actual result to be returned

[0088] Through secondary cleaning, characters such as "Five thousand six hundred and forty-two yuan and seventy-six fen" are obtained. Then, by defining a dictionary to parse Arabic numerals ['zero': 0, 'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5,'six': 6,'seven': 7, 'eight': 8, 'nine': 9], and defining a dictionary to parse monetary units ['fen': 0.01, 'jiao': 0.1, 'yuan': 1,'shi': 10, 'bai': 100, 'qian': 1000, 'yuan': 1], the premium amount after conversion is 5642.76 yuan.

[0089] 3. The matching process of the start insurance time and the end insurance time includes three cleaning processes, specifically:

[0090] 1) First cleaning:

[0091] Rule: 'insurance' + 0 - 2 arbitrary characters + 'period' + 0 - 2 arbitrary characters + 'time' + 0 - 70 arbitrary characters + line break

[0092] 2) Second cleaning:

[0093] First time rule: [2 - 4 digits from 0 to 9 + 0 - 2 arbitrary characters + 'year' + 0 - 2 arbitrary characters + 1 - 2 digits from 0 to 9 + 0 - 2 arbitrary characters +'month' + 0 - 2 arbitrary characters + 1 - 2 digits from 0 to 9 + 0 - 2 arbitrary characters + 'day' + 0 - 2 arbitrary characters + 1 - 2 digits from 0 to 9 + 0 - 2 arbitrary characters + 1 - 2 digits from 0 to 9 + 0 - 2 arbitrary characters + 1 - 2 digits from 0 to 9 + 0 - 2 arbitrary characters]

[0094] Second time rule: 2 - 4 digits from 0 to 9 + 0 - 2 arbitrary characters + 'year' + 0 - 2 arbitrary characters + 1 - 2 digits from 0 to 9 + 0 - 2 arbitrary characters +'month' + 0 - 2 arbitrary characters + 1 - 2 digits from 0 to 9 + 0 - 2 arbitrary characters] + 'day' + 0 - 2 arbitrary characters + [1 - 2 digits from 0 to 9] + 'hour'

[0095] When the first time rule is not met, the second time rule is run to obtain only numbers but not text, and obtain characters through secondary cleaning, such as: ['2022','06','13','00','00'], ['2023','06','12','23','59'];

[0096] 3) The third cleaning process makes a reasonable judgment on the characters obtained by the second cleaning process, including:

[0097] If there is no string to obtain minutes and seconds, such as ['2022','06','13','20','',''], then append '00 minutes 00 seconds';

[0098] If the obtained time is in Chinese, such as 'zero hour', convert it into the corresponding Arabic numerals through the dictionary list ["21","22","23","24","0","11","12","13","14","15","16","17","18","19","20","10","1","2","3","4","5","6","7","8","9"];

[0099] If the obtained time is not reasonable, the time '24:00:00' is processed back one second to '23:59:59';

[0100] If the number of characters in the obtained string is insufficient, it will be padded, such as ['2022','6','3','23','59','59'], and the padded string will become ['2022','06','03','23','59','59']

[0101] Next, a specific example of recognition results is given.

[0102] like Figure 3 The following is a schematic diagram of the results of insurance policy type identification.

[0103] For insurance policy content matching, the result example is as follows:

[0104] Insurance policy number: ['Insurance policy number: PDAA202231010000170293\n']

[0105] Secondary cleaning: ['PDAA202231010000170293']

[0106] vin:['Vehicle identification code LSJA24031HS201680 Brand model CSA6456BEV1\nVehicle\n']

[0107] Secondary cleaning: ['LSJA24031HS201680']

[0108] Company Name: ['Company Name: xxx Company xxx Branch INDIGO (Insurer's Signature)\nInsurance Company Address']

[0109] Secondary cleaning: ['xxx company xxx branch']

[0110] Insured: ['Insured by Name xxx Company\n\n']

[0111] Secondary cleaning: ['xxx company']

[0112] Insurance premium: ['New energy vehicle third party liability insurance 500000.003383.15\nNew energy vehicle on-board liability insurance (driver) 50000.00 / seat*1seat 115.09\nTotal insurance premium (RMB in capital letters): 5,642.76 yuan (￥: 5,642.76 yuan)']

[0113] Second cleaning: ['5642.76']

[0114] Insurance period: ['Insurance period: from 00:00 on June 13, 2022 to 23:59 on June 12, 2023\n']

[0115] Second cleaning: [['2022','06','13','00','00'], ['2023','06','12','23','59']]

[0116] Three cleanings ['2022-06-13 00:00:00']['2023-06-12 23:59:00']

[0117] Nature of insurance: ['Nature of use: rental, leasing\nlabel\n']

[0118] Secondary cleaning: ['rental, lease']

[0119] Third party liability insurance: ['Third party liability insurance 500000.00 3383.15\n']

[0120] Second cleaning: ['Motor vehicle third party liability insurance'], ['500000'], ['3383.15']

[0121] Driver on the bus: ['Driver) 50000.00 / seat*115.09\n']

[0122] Second cleaning: ['Driver liability insurance'], ['50000.00'], ['115.09']

[0123] Vehicle damage insurance: ['Vehicle loss insurance 151782.00 2144.52\nMotor vehicle loss insurance absolute deductible per accident 2000.00\n']

[0124] Second cleaning: ['Motor vehicle loss insurance'], ['151782.00'], ['2144.52'], ['2000']

[0125] This embodiment also provides a system embodiment, such as Figure 2 As shown, an insurance policy identification system, the system comprising:

[0126] An insurance policy text recognition module is used to obtain the full amount of insurance policy data in the insurance policy image to be recognized based on the insurance policy electronic document through a trained text detection and recognition model;

[0127] The insurance policy type determination module is used to match the text content of the header through preset character rules, determine key characters, and determine the insurance policy type;

[0128] Insurance policy content matching module: Based on the multiple preset fields corresponding to the insurance policy type, match the full amount of insurance policy data to obtain preliminary recognition results, preset the fields and judgment logic again, and finally obtain an accurate recognition result storage module for data storage and result output of insurance policy data.

[0129] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for identifying an insurance policy, characterized in that: The method includes: Insurance policy text recognition: obtaining all the insurance policy data in the insurance policy image to be recognized by using the trained text detection and recognition model; Insurance policy type determination: matching the characters in the insurance policy data with the preset key characters of the insurance policy categories, and determining the insurance policy type to which the insurance policy image to be recognized belongs according to the matching result; Insurance policy content matching: matching all the insurance policy data by using multiple matching rules according to the multiple preset matching fields corresponding to the insurance policy type, and obtaining and storing the insurance policy recognition result.

2. The insurance policy identification method according to claim 1, characterized in that: The matching of the characters in the insurance policy data with the preset key characters of the insurance policy categories and determining the insurance policy type to which the insurance policy image to be recognized belongs is specifically as follows: According to the first preset character rule, search for the text content that conforms to the rule among the first M characters of all the insurance policy data, and return the recognition result to the type parameter; According to the second preset character rule, determine whether the corresponding characters are within the type parameter to determine the insurance policy type; the insurance policy types include commercial insurance and traffic insurance, the character corresponding to commercial insurance is 'commercial', and the characters corresponding to compulsory traffic insurance are any one of 'traffic' and 'compulsory'.

3. The insurance policy identification method according to claim 2, characterized in that: The first preset character rule is: obtained under the same line of text, 1 to 20 characters + 'electronic insurance policy' or 'insurance policy'.

4. The insurance policy identification method according to claim 1, characterized in that: The multiple preset matching fields corresponding to the insurance policy type of compulsory traffic insurance include vehicle use nature, insurance company, amount, vehicle identification number, name of the insured, compulsory traffic insurance policy number, start date of insurance, end date of insurance, effective date, premium, and vehicle and vessel tax; The multiple preset matching fields corresponding to the insurance policy type of commercial insurance include: vehicle use nature, insurance company, vehicle identification number, name of the insured, commercial insurance policy number, start date of insurance, end date of insurance, effective date, premium, name of vehicle damage insurance, insured amount of vehicle damage insurance, deductible of vehicle damage insurance, premium of vehicle damage insurance, name of third-party liability insurance, insured amount of third-party liability insurance, premium of third-party liability insurance, name of on-board driver liability insurance, insured amount of on-board driver liability insurance, premium of on-board driver liability insurance, name of on-board passenger insurance, insured amount of on-board passenger insurance, and premium of on-board passenger insurance; Among them, each multiple preset matching field includes at least one expression form and is updated regularly according to the insurance policy type.

5. An insurance policy identification method according to claim 4, characterized in that: The matching process of the policy number includes: First rule of the policy number: obtained under the same line of text, 'insurance' + 0 to 2 arbitrary characters + 'policy' + 0 to 2 arbitrary characters + 'number' + 0 or 1 ':' + 1 to 40 characters of arbitrary numbers or English letter characters + line break Second rule of the policy number: obtained under the same line of text, 'insurance' + 0 or 1 arbitrary character + 'number' + 0 or 1 arbitrary character + 'code' + 0 or 1 ':' + 1 to 40 characters of arbitrary numbers or English letter characters + line break Among them, '*' represents a character **; When the content obtained by the first rule of the policy number is equal to the null value, recognize the content obtained by the second rule of the policy number, otherwise skip the content obtained by the second rule of the policy number.

6. An insurance policy identification method according to claim 4, characterized in that: The matching process of the premium extracts data in the way of extracting Chinese capitalized amounts, including: The first cleaning of the premium is used to obtain the premium field containing the amount in capital numbers, and the characters after the first cleaning are obtained. The second cleaning of the premium is used to remove the redundant fields in the first cleaning of the premium, and for the remaining amount in capital numbers, the Arabic numerals are parsed through a defined dictionary, and the monetary unit is parsed through a defined dictionary, and the characters in Arabic numeral form are obtained after conversion.

7. An insurance policy identification method according to claim 6, characterized in that: The process of the first cleaning of the premium is as follows: The first rule in the first stage of the premium: 'Total insurance premium' + 0 to ∞ Chinese numeral characters + 'whole' The second rule in the first stage of the premium: 'Total' + 0 to 4 arbitrary characters + 'Renminbi in capital letters' + 0 to ∞ Chinese numeral characters + 'whole' The third rule in the first stage of the premium: 'Total' + 0 to 4 arbitrary characters + 'Renminbi in capital letters' + 0 to ∞ Chinese numeral characters + 'jiao' + 0 to ∞ Chinese characters + 0 or 1 'fen' When the first rule in the first stage of the premium is not satisfied, the second rule in the first stage of the premium is run. When the second rule in the first stage of the premium is not satisfied, the third rule in the first stage of the premium is run to obtain the characters after the first cleaning.

8. An insurance policy identification method according to claim 4, characterized in that: The matching process of the start insurance time and the end insurance time includes three cleaning processes, specifically: 1) The first cleaning: Rule: 'Insurance' + 0 to 2 arbitrary characters + 'period' + 0 to 2 arbitrary characters + 'between' + 0 to 70 arbitrary characters + line break 2) The second cleaning: The first time rule: [2 to 4 digits from 0 to 9 + 0 to 2 arbitrary characters + 'year' + 0 to 2 arbitrary characters + 1 to 2 digits from 0 to 9 + 0 to 2 arbitrary characters +'month' + 0 to 2 arbitrary characters + 1 to 2 digits from 0 to 9 + 0 to 2 arbitrary characters + 'day' + 0 to 2 arbitrary characters + 1 to 2 digits from 0 to 9 + 0 to 2 arbitrary characters + 1 to 2 digits from 0 to 9 + 0 to 2 arbitrary characters + 1 to 2 digits from 0 to 9 + 0 to 2 arbitrary characters] The second time rule: 2 to 4 digits from 0 to 9 + 0 to 2 arbitrary characters + 'year' + 0 to 2 arbitrary characters + 1 to 2 digits from 0 to 9 + 0 to 2 arbitrary characters +'month' + 0 to 2 arbitrary characters + 1 to 2 digits from 0 to 9 + 0 to 2 arbitrary characters] + 'day' + 0 to 2 arbitrary characters + [1 to 2 digits from 0 to 9] + 'hour' When the first time rule is not satisfied, the second time rule is run, and only the numbers are obtained without obtaining the words, and the characters are obtained through the second cleaning. 3) The third cleaning makes a rationality judgment on the characters obtained by the second cleaning.

9. An insurance policy identification method according to claim 8, characterized in that: The third cleaning makes a rationality judgment on the characters obtained by the second cleaning, including: If the strings of minutes and seconds are not obtained, '00 minutes 00 seconds' is appended. If the obtained is the Chinese time, it is converted into the corresponding Arabic numerals through a dictionary list. When the obtained time lacks rationality, including the operation of retreating one second for the time '24:00:00' to be processed as '23:59:59'. If the number of characters in the obtained string is insufficient, it is supplemented.

10. An insurance policy identification system, characterized in that: Using the method described in any one of claims 1 to 9, the system includes: An insurance policy text recognition module is used to obtain the full amount of insurance policy data in the insurance policy image to be recognized based on the insurance policy electronic document through a trained text detection and recognition model; The insurance policy type determination module is used to match the text content of the header through preset character rules, determine key characters, and determine the insurance policy type; Insurance policy content matching module: matches the full amount of insurance policy data using multiple matching rules according to multiple preset matching fields corresponding to the insurance policy type to obtain insurance policy recognition results; The storage module is used to store the identified insurance policy data.

Images