Transportation license information identification method and system

Abstract

The invention discloses a transportation certificate information identification method and system. The method comprises the following steps: a graphical user interface module is used for receiving a certificate image to be identified and encoding the certificate image to obtain first image data; the data communication module is used for receiving the first image data from the graphical user interface module and decoding the first image data to obtain an image with a preset size; the character OCR module is used for receiving the image with the preset size from the data communication module and identifying character information on the image with the preset size; the keyword extraction module is used for receiving the character information from the character OCR module, extracting keywords from the character information and transmitting the keywords to the data communication module; the data communication module is further used for receiving the keywords from the keyword extraction module and transmitting the keywords to the graphical user interface. According to the invention, automation of certificate review work in large piece transportation review can be realized, the review process is accelerated, the review efficiency is improved, and the labor cost is reduced.

Description

Technical Field

[0001] This application belongs to the technical field of image recognition, and particularly relates to a method and system for identifying transportation license information. Background Art

[0002] With the rapid progress of the industrial level and the country's increased investment and construction in fields such as transportation infrastructure and logistics hubs, the efficient connection of the logistics network has been promoted. Industries such as mechanical manufacturing, the energy industry, and aerospace have a strong demand for the transportation of large-sized goods. To ensure the safety and compliance of the entire process of large-sized goods transportation and ensure that the transportation department can promptly carry out route planning, the importance of the review work for relevant qualifications and licenses for large-sized goods transportation is self-evident. The traditional method of manual inspection is inefficient and will cause a waste of a large amount of human resources, making it difficult to adapt to the increasing demand for large-sized goods transportation.

[0003] Regarding the four types of commonly used certificates (road transport business license, large-sized goods transportation power of attorney, driving license, temporary license plate) during the review process of large-sized goods transportation, due to the large number of types, the workload of the review personnel to individually check the key information is relatively large, and omissions are likely to occur. Summary of the Invention

[0004] In order to overcome the deficiencies of the prior art, this application provides a method and system for identifying transportation license information. It can accurately identify the text information of licenses in most scenarios, can accurately associate the text at the specified position with the corresponding label, and is applicable to the structured recognition solution for intelligent review of large-sized goods transportation. It can realize the automation of the review work for the four types of certificates, speed up the review process, improve the review efficiency, and reduce the labor cost.

[0005] To achieve the above object, this application adopts the following technical solutions:

[0006] A transportation license information recognition system, comprising:

[0007] A graphical user interface module, configured to receive the image of the certificate to be recognized and perform encoding to obtain first image data;

[0008] A data communication module, coupled to the graphical user interface module, configured to receive the first image data from the graphical user interface module and decode the first image data to obtain an image of a predetermined size;

[0009] A text OCR module, coupled to the data communication module, configured to receive the image of the predetermined size from the data communication module and recognize the text information on the image of the predetermined size; and

[0010] The keyword extraction module is respectively coupled to the data communication module and the text OCR module, and is used to receive the text information from the text OCR module, extract keywords from the text information and transmit the keywords to the data communication module; the keyword extraction module is further used to judge whether the keywords are complete; if so, the keyword extraction module transmits the keywords to the data communication module; if not, the keyword extraction module generates second rotation information and transmits the second rotation information to the text OCR module; the text OCR module is further used to rotate the image of the predetermined size according to the second rotation information to obtain a rotated image and recognize the text information on the rotated image;

[0011] Wherein, the data communication module is further used to receive the keywords from the keyword extraction module and transmit the keywords to the graphical user interface.

[0012] Further, it further includes:

[0013] The seal fingerprint detection module is respectively coupled to the data communication module and the text OCR module, and is used to receive the image of the predetermined size from the data communication module and detect the seal information, fingerprint information or first rotation information of the image of the predetermined size.

[0014] Further, the text OCR module is further used to receive the first rotation information from the seal fingerprint detection module, and the text OCR module is further used to rotate the image of the predetermined size according to the first rotation information to obtain a rotated image and recognize the text information on the rotated image.

[0015] Further, the text OCR module uses the det_r50_db++_icdar15 pre-trained model of PaddleOCR and conducts secondary training, and introduces the differentiable binarization DB++ detection algorithm to detect text information;

[0016] Alternatively, the text OCR module uses the ch_PP-OCRv3_rec pre-trained model of PaddleOCR and conducts secondary training, uses the SVTR algorithm for feature enhancement and uses the knowledge distillation method to compress the model to improve the recognition speed.

[0017] Further, the keyword extraction module is further used to extract the keywords in the form of regular expressions according to the text information, and replace and correct the incorrect words in the keywords.

[0018] In addition, the present application also provides a method for identifying transportation license information using the transportation license information identification system as described above, including:

[0019] The graphical user interface module receives the document image to be recognized and encodes it to obtain the first image data;

[0020] The data communication module receives the first image data from the graphical user interface module and decodes the first image data to obtain an image of a predetermined size;

[0021] The text OCR module receives the image of the predetermined size from the data communication module and recognizes the text information on the image of the predetermined size;

[0022] The keyword extraction module receives the text information from the text OCR module, extracts keywords from the text information and transmits the keywords to the data communication module; the keyword extraction module determines whether the keywords are complete; if so, the keyword extraction module transmits the keywords to the data communication module; if not, the keyword extraction module generates second rotation information and transmits the second rotation information to the text OCR module; the text OCR module is further configured to rotate the image of the predetermined size according to the second rotation information to obtain a rotated image and recognize the text information on the rotated image; and

[0023] The data communication module receives the keywords from the keyword extraction module and transmits the keywords to the graphical user interface.

[0024] Further, the transportation license information recognition system further includes a seal fingerprint detection module;

[0025] The seal fingerprint detection module receives the image of the predetermined size from the data communication module and detects the seal information, fingerprint information or first rotation information of the image of the predetermined size.

[0026] Further, the text OCR module receives the first rotation information from the seal fingerprint detection module, and the text OCR module is further configured to rotate the image of the predetermined size according to the first rotation information to obtain a rotated image and recognize the text information on the rotated image;

[0027] The keyword extraction module extracts the keywords in the form of a regular expression according to the text information and replaces and corrects the incorrect words in the keywords.

[0028] Further, the text OCR module uses the det_r50_db++_icdar15 pre-trained model of PaddleOCR and conducts secondary training, introducing the differentiable binarized DB++ detection algorithm for text information detection;

[0029] Alternatively, the text OCR module uses the ch_PP-OCRv3_rec pre-trained model of PaddleOCR and conducts secondary training, uses the SVTR algorithm for feature enhancement and the knowledge distillation method to compress the model to improve the recognition speed.

[0030] This application is specifically targeted at transportation licenses and mainly includes the structured recognition functions for four types of certificates, including the Road Transport Business License, the Consignment Note for Special Cargo Transport, the Vehicle Registration Certificate, and the Temporary License Plate. Compared with the existing recognition technologies, this application can achieve a higher recognition accuracy for the specific four types of certificates, with an overall recognition accuracy reaching 90%; compared with the existing recognition technologies, this application can extract the required key information and assign corresponding labels to all information while recognizing text information; compared with the existing recognition technologies, this application also provides the function of detecting whether there are seals and fingerprints in the certificates; at the same time, this application also optimizes the recognition accuracy in the case of overlap between the template and printed text due to printing problems; optimizes the recognition accuracy under complex backgrounds and anti-counterfeiting marks. This application can also automate the review work of the four types of certificates used in the review of special cargo transport, speed up the review process, improve the review efficiency, and reduce the labor cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, but do not constitute an improper limitation to the present application. In the drawings:

[0032] Figure 1 is a schematic diagram of the system of the present application;

[0033] Figure 2 is another schematic diagram of the system of the present application;

[0034] Figure 3 is another system example diagram of the present application;

[0035] Figure 4 is the flow chart of the consignment note recognition of the present application;

[0036] Figure 5 is the flow chart of the license recognition of the present application;

[0037] Figure 6 is the flow chart of the vehicle registration certificate recognition of the present application;

[0038] Figure 7Flow chart of temporary license plate recognition for this application;

[0039] Figure 8 Flow chart of the method for this application;

[0040] Figure 9 Keyword field table of the large-scale transportation entrustment letter for this application;

[0041] Figure 10 Keyword field table of the road transport business license for this application;

[0042] Figure 11 Keyword field table of the vehicle license for this application;

[0043] Figure 12 Keyword field table of the temporary license plate for this application. Detailed implementation manners

[0044] In order to enable those skilled in the art of this technology to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application.

[0045] In the description of this application, it should be understood that the orientation or positional relationship indicated by terms etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing this application and simplifying the description, rather than indicating or implying that the device, element, module, system, platform or device referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation to this application. The following description of this application is only understood as a description of individual embodiments of the technical solution of this application. Other embodiments are not reflected in the following description, but this does not mean that this application excludes these other embodiments. The technical solution of this application is not limited to the specific implementation manners described below, and the protection scope of this application is not limited to only the specific implementation manners described below. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.

[0046] It should be noted that if terms such as "first", "second", etc. appear in the description, claims and the above-mentioned drawings of this application, such descriptions are only used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of this application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0047] In some embodiments, as Figure 1 shown, a transportation license information recognition system 1 includes:

[0048] A graphical user interface module 11, configured to receive a document image to be recognized and perform encoding to obtain first image data; this graphical user interface module 11 can also be referred to as a front-end page module, and is used to receive the document image and display the recognition result.

[0049] A data communication module 12, coupled to the graphical user interface module 11, configured to receive the first image data from the graphical user interface module 11 and decode the first image data to obtain an image of a predetermined size;

[0050] A text OCR module 13, coupled to the data communication module 12, configured to receive the image of the predetermined size from the data communication module 12 and recognize the text information on the image of the predetermined size; and

[0051] A keyword extraction module 14, coupled to the data communication module 12 and the text OCR module 13 respectively, configured to receive the text information from the text OCR module 13, extract keywords from the text information and transmit the keywords to the data communication module 12;

[0052] Wherein, the data communication module 12 is further configured to receive the keywords from the keyword extraction module 14 and transmit the keywords to the graphical user interface 11.

[0053] In some embodiments, as Figure 2 shown, it further includes:

[0054] The seal and fingerprint detection module 15 is respectively coupled to the data communication module 12 and the text OCR module 13, and is configured to receive the image of the predetermined size from the data communication module 12 and detect the seal information, fingerprint information or first rotation information of the image of the predetermined size.

[0055] Specifically, as Figure 3 shown in an embodiment, the relationship between the specific modules in the exemplary specification is shown.

[0056] Specifically, the specific recognition processes of four types of commonly used certificates in the large-piece transportation review process are as Figures 4 - 7 shown, which are the large-piece transportation power of attorney, road transport business license, vehicle license, and temporary license plate respectively.

[0057] In some embodiments, the text OCR module 13 is further configured to receive the first rotation information from the seal and fingerprint detection module 15, and the text OCR module 13 is further configured to rotate the image of the predetermined size according to the first rotation information to obtain a rotated image and recognize the text information on the rotated image.

[0058] In some embodiments, the keyword extraction module 14 is further configured to determine whether the keyword is complete;

[0059] If so, the keyword extraction module 14 transmits the keyword to the data communication module 12;

[0060] If not, the keyword extraction module 14 generates second rotation information and transmits the second rotation information to the text OCR module 13; the text OCR module 13 is further configured to rotate the image of the predetermined size according to the second rotation information to obtain a rotated image and recognize the text information on the rotated image.

[0061] In some embodiments, the text OCR module 13 uses the det_r50_db++_icdar15 pre-trained model of PaddleOCR and performs secondary training, and introduces the differentiable binarization DB++ detection algorithm to detect text information;

[0062] Alternatively, the text OCR module 13 uses the ch_PP-OCRv3_rec pre-trained model of PaddleOCR and performs secondary training, uses the SVTR algorithm for feature enhancement and uses the knowledge distillation method to compress the model to improve the recognition speed.

[0063] In some embodiments, the keyword extraction module 14 is further configured to extract the keyword in the form of a regular expression according to the text information, and replace and correct the incorrect text in the keyword.

[0064] In some embodiments, as Figure 8 shown, the present application further provides a method for identifying transportation license information using the transportation license information identification system as described above, including:

[0065] S1: The graphical user interface module receives the image of the document to be identified and encodes it to obtain first image data;

[0066] Specifically, the graphical user interface module (or the front-end page module) receives the original image, converts the picture data in formats such as jpg and png into base64 encoding, and transmits it to the data communication module.

[0067] S2: The data communication module receives the first image data from the graphical user interface module and decodes the first image data to obtain an image of a predetermined size;

[0068] S3: The text OCR module receives the image of the predetermined size from the data communication module and identifies the text information on the image of the predetermined size;

[0069] Specifically, the text OCR module includes an orientation correction model, a text detection model, and a text recognition model. The orientation correction model first rotates and corrects the initial image to the correct orientation, then the text detection model detects the positions of all texts row by row, and finally the text recognition model recognizes the texts within the range detected by the text detection model.

[0070] S4: The keyword extraction module receives the text information from the text OCR module, extracts keywords from the text information, and transmits the keywords to the data communication module;

[0071] Specifically, the keyword extraction module extracts keyword fields in the way of regular expressions according to the texts recognized by the text OCR module, and replaces and corrects the misrecognized texts in information such as license plates and dates. The keyword fields extracted from the four types of documents are as Figures 9 - 12 shown. The specific logic of the regular expressions and corrections for the four types of documents is as follows:

[0072] 1. Logic for extracting text from the power of attorney.

[0073] ① Delete the spaces in the recognition result.

[0074] ② Divide the text into the main text part and the ID card part.

[0075] (1) Replace the text at the end of the main text with Complete transportation\n.

[0076] (2) If the body text cannot be located, find 2 to 4 characters among the 2 to 4 characters that are not in the process of transportation and are wrapped by two line breaks in the recognition result, and consider them as the name. Use this as the basis.

[0077] (3) If both of the above two schemes fail, it is considered that all the recognition results are the body text, and the extraction of the ID card text is abandoned.

[0078] ③ Delete the redundant characters in the body text, including ['[', ']', '{', '}', '、', '.', '°', ',', '—', '-', '_', ':', ':', '\n', ”, ''].

[0079] ④ Delete all redundant characters in the ID card text, and only keep Chinese characters, numbers and line breaks.

[0080] ⑤ Judge the power of attorney template.

[0081] If the seal detection model only detects one national emblem, and the words 'legal', 'legal person' or 'oneself' are found in the body text, it is considered as the legal person template, otherwise it is the entrusted person template.

[0082] ⑥ Obtain the information of the operator.

[0083] Copy the body text as the text to be processed.

[0084] (1) Search for the words 'highway in transit', 'dismantling', 'due to transportation', 'ID card', 'this carrier' in the text in turn. Once one of the words is found, use this word as a node to split, and take the second half. Once split, stop searching for the next word.

[0085] (2) Delete the Chinese and English brackets in the text.

[0086] (3) Find the first ID card number in the text. If not found, find 17 consecutive digits.

[0087] (4) Search for the fields starting with 'by this legal representative' | 'by legal representative' | 'by legal person' | 'current legal representative' | 'current legal person' | 'this legal representative' | 'legal representative' | 'authorized' | 'entrusted' | 'designated' | 'current oneself' | 'by oneself' | 'legal representative' and ending with the ID card number in turn, and take the middle part as the name of the operator.

[0088] (5) Capitalize the 'x' in the ID card number as the ID card number of the operator.

[0089] (6) If the extraction of the operator information fails, add 'error in recognizing the name and ID card in the body text' to the error message and make the operator information empty.

[0090] ⑦ Obtain the tractor-trailer information.

[0091] (1) Preprocess the body text for the following situations:

[0092] I Process the problem of missing city names.

[0093] II For some repeated characters, such as license plate numbers 'GanGanA00001', 'Limited CompanyCompany', 'JingA0001ChaoChao', etc., merge the repeated characters.

[0094] III Replace heavy flatbed trucks, prime movers, truck carriers, company vehicles, trucks, and ordinary trucks with tractors.

[0095] IV For forms like 'Jing(A00001)', delete the parentheses.

[0096] V Compensate for misspelled license plate numbers, model errors caused by self - errors or similar characters in the certificates {'Yun': 'Liao', 'Shou': 'Lu', 'Yi': 'Ji', 'Si': 'Liao', 'Mo': 'Hei', 'Cai': 'Ji', 'Yu': 'Yu', 'Zhi': 'Jin', 'Zhi': 'Ji', 'Kong': 'Ning', 'Xiang': 'Xiang', 'Hu': 'Xiang', 'Wan': 'Wan'}.

[0097] VI The fields of 'tractor' and 'trailer' are incomplete, with only 'gua' without 'che', only 'qianyin' without 'che', only 'qianche' without 'yin', only 'yinche' without 'qian', replace them with the correct fields.

[0098] VII In this correction of freight vehicles, correct 'freight axles' to 'freight vehicles', 'vehicle out' to'vehicle from', 'rented by' to 'rented', 'goods' to 'cargo', 'transportation' to 'transport', and the situation where the freight vehicle is directly followed by the license plate number without 'from' in the middle.

[0099] (2) Search for the number of license plate numbers in the body text. If it is greater than 2, output the license plate numbers normally. If it is less than or equal to 2, adopt them in turn.

[0100] Scenario 1: Only the case of tractors.

[0101] In this freight vehicle [vehicle transportation], it is from (.+?) tractor (.+?) (? : complete transportation | engage in this over - limit transportation | transportation | complete transportation).

[0102] Scenario 2: General situation.

[0103] In this freight vehicle [vehicle transportation], it is from (.+?) tractor (.+?) and (.+?) trailer (.+?) (? : complete transportation | engage in this over - limit transportation | transportation | complete transportation).

[0104] Scenario 3: The situation where the company to which the trailer belongs is omitted. If the length of the trailer license plate number is greater than 8, the extraction fails.

[0105] This freight vehicle transportation is completed by a tractor (.+?) and a trailer (.+?) (?: for transportation | engaging in this over-limit transportation | for this transportation | completing the transportation).

[0106] If any of the three extraction methods is successful, delete the word "Company" from the names of the tractor and trailer owners that are less than 6 characters long and end with "Company", and then remove the parentheses on both sides of the owners.

[0107] If all three extraction methods are incorrect, then use Method 4.

[0108] Method 4 has multiple tractor-trailer pairs.

[0109] If it is linked with "and", replace it with "by".

[0110] If it is linked with "by", find all the license plate numbers of the tractors and all the license plate numbers of the trailers.

[0111] Use the following regular expression to locate tractor information:

[0112] r'by.+?[A-Z0-9][))]? and'

[0113] Match a string that starts with "by", followed by any characters (as few as possible), then a capital letter or a number, and finally optionally followed by a right parenthesis (including full-width and half-width).

[0114] Use the following regular expression to locate trailer information:

[0115] r'[A-Z0-9][))]? and.+?[0-9].+?[completed by]'

[0116] Match a string that contains a capital letter or a number, followed by an optional right parenthesis (including full-width and half-width). Match the Chinese character "and", followed by any characters (as few as possible), then a number, and then any characters (as few as possible), and finally match the Chinese characters "completed" or "by".

[0117] The above two regular expressions also consider the cases where the names of the tractor and trailer owners contain "by" and "and".

[0118] Link all the tractor information found with '&' that is more than 8 characters long, and link all the trailer information found with '&' that is more than 8 characters long.

[0119] If the tractor information is not empty, delete the license plate number of the tractor and the useless characters in the tractor information, including:

[0120] ['tractor', '()', '()', '、', ',', ',', ';', ';', ':', ':']

[0121] Otherwise, it is regarded as a failure to extract the tractor information.

[0122] If the trailer information is not empty, delete the trailer license plate number and useless characters in the trailer information, including:

[0123] ['Trailer', '()', '()', '、', ',', ',', ';', ';', ':', ':'];

[0124] Otherwise, it is regarded as a failure to extract the trailer information.

[0125] Use 、 to link all the tractor license plate numbers found and all the trailer license plate numbers found.

[0126] Split the tractor information according to &, remove 'by' and 'and' at both ends of each field in the split, then split each split again according to 'company', remove the parentheses at both ends of all splits, and add 'company' after all splits with a length greater than 4 and not containing 'individual', and finally link these splits with 、.

[0127] Split the trailer information according to &, remove 'completed' and 'by' at both ends of each field in the split, split each split according to 'and' (at most once), then split each split again according to 'company', remove the parentheses at both ends of all splits, and add 'company' after all splits with a length greater than 4 and not containing 'individual', and finally link these splits with 、.

[0128] If there is a tractor owner, a tractor license plate number, and a trailer license plate number, but there is no trailer owner, then regard the tractor owner as the trailer owner as well.

[0129] (3) If the extraction of the operator information fails, then add 'Error in identifying tractor and trailer information' to the error message and set the tractor and trailer information to be empty.

[0130] ⑧ Extract the ID card information and initialize an empty list id_info_list.

[0131] (1) Split the ID card text by line to obtain the ID card information.

[0132] (2) For each item in the ID card information:

[0133] If there is more than one digit in it, then delete all non-numeric and 'long' characters in it.

[0134] If there is the character 'long' in it, then add an item ['date', 'long term'] to id_info_list.

[0135] If its length is less than 17 and greater than 8, add an item ['date','2'+i[-7:]] to id_info_list, where '2'+i[-7:] represents the character '2' plus the 7 characters following the item.

[0136] If its length is greater than 16, add an item ['id',i] to id_info_list where i is the item.

[0137] If the number is equal to or less than 1, and the length is greater than 1, and it is not a field in ['front', 'back', 'backside', 'Public Security Bureau', 'Identity'], then add an item ['name', i.strip('name')] to id_info_list, where i.strip('name') represents the result after removing the first and last characters 'surname' and 'name'.

[0138] (3) Handle the situation where the validity period and ID number are in reverse order.

[0139] If one item in id_info_list is the expiration date and the next item is the ID number, swap the two items.

[0140] (4) Extract the first 6 items of id_info_list (if there are more than 6 items).

[0141] (5) Initialize a list, which is the ID card extraction result, corresponding to the three information of the first person and the three information of the second person in sequence.

[0142] (6) If the ID card number is 17 digits, add X. If it is a lowercase x, replace it with an uppercase X.

[0143] (7) If the ID card information extraction fails, add 'Error in identifying ID card photo' to the error message and make the ID card information empty.

[0144] 2License text extraction logic.

[0145] ① The text containing 1-4 numbers in the recognition results is classified as date text, and the rest is classified as body text.

[0146] ②Remove all line breaks in the text.

[0147] ③ Divide the main text into business name and business scope.

[0148] (1) Delete the remaining line breaks and correct the error-prone characters:

[0149] (2) If the length of a text line is less than 6, then that line is the business name and the remaining lines are linked as the business scope.

[0150] (3) If found in the main text.

[0151] Company|Transportation team|Service department|Center|Engineering team|Fleet.

[0152] If any one of them is found, the business name will be divided into business name and business scope according to this field, and this field is the end of the business name.

[0153] (4) If found in the main text.

[0154] Large object transportation|General freight|General freight transportation by road|Special freight transportation|Freight;|Any one of the inter-county ones, it will be divided into business name and business scope according to this field, and this field is the beginning of the business scope.

[0155] (5) If the business scope begins with 'Freight;', replace it with 'Freight:'.

[0156] ④Process date text.

[0157] (1) If the number of date text lines is 3 or 6;

[0158] The last three lines are date text.

[0159] (2) If the number of date text lines is 1 or 2, and the first line is the year ((20 or 0) + two arbitrary digits);

[0160] Then add two lines of xx at the end and take the first three lines.

[0161] (3) If the number of date text lines is greater than 6;

[0162] Then the third to sixth lines are date text.

[0163] (4) If the number of date text lines is 4 or 5;

[0164] Then the first year word from the second line and the following two lines, a total of three lines, are date text.

[0165] (5) Compensation for date errors.

[0166] If the year is three digits, add 2 in front of it. If it is two digits, add 20 in front of it.

[0167] If the month is a single digit, add a leading 0.

[0168] If the month is a 3-digit number and the first digit is 0, take the first two digits; if the first digit is not 0, take the last two digits.

[0169] If the day is a single digit, add a leading 0.

[0170] If the day is a 3-digit number and the first digit is 0, take the first two digits; if the first digit is not 0, take the last two digits.

[0171] (6) Compensation for common numbers mistaken for letters, e.g., 'T2' month {'T':'1','l':'1','I':'1','L':'1','B':'3','D':'0','O':'0','o':'0','Z':'2','z':'2'}

[0172] (7) Compensation for errors in the first digit of a character string, e.g., '62' for month.

[0173] The first digit of 6, 8, 9 is replaced by 0, and the first digit of 7 is replaced by 1.

[0174] (8) The combined year, month and day shall be the final date.

[0175] 3. Driving license text extraction logic.

[0176] ①Text classification.

[0177] Classify the text into: certificate validity period, certificate validity period without dates, annual inspection validity period, license plate, and owner.

[0178] ② Extraction of certificate validity period.

[0179] The annual inspection validity period that includes the mandatory scrapping expiration date shall be in the form of xx-xx-xx. The annual inspection validity period that does not include the mandatory scrapping expiration date shall only have the last two digits as a backup. If the mandatory scrapping expiration date is not found, the backup number shall be used instead.

[0180] ③Extract during the annual inspection validity period.

[0181] Find all the words "xx year xx month" in the annual inspection validity period and change them to yyyymm format.

[0182] ③License plate extraction.

[0183] Typos are handled (same as the letter of authorization). If there are more than 2 different license plate numbers, take the one with the most letters. If there are only two different license plate numbers, take the one with more letters.

[0184] ⑤ Withdrawal by all

[0185] (1) If there is only one alternative owner, that owner is the outcome.

[0186] (2) If there are multiple ones, the name will be "yes", "limited", "limited company", "limited company", "transport team", "fleet", "company", "service department", "center", "engineering team".

[0187] Those ending with a length greater than 4 are the results. If not, those with a length greater than 1 and less than 5 are the results. If the length is greater than 4 and the "Limited Company" at the end is incomplete, then supplement "Limited Company".

[0188] (3) Delete all the letters and numbers therein.

[0189] 4 Temporary license plate text extraction logic.

[0190] ① Classify the text lines.

[0191] If it contains a structure of one letter + four consecutive letters or numbers, it is identified as a license plate number; otherwise, it is identified as a date.

[0192] ② Date processing.

[0193] The processing method is the same as that of the license.

[0194] S5: The data communication module receives the keyword from the keyword extraction module and transmits the keyword to the graphical user interface.

[0195] In some embodiments, the transportation license information recognition system further includes a seal fingerprint detection module;

[0196] The seal fingerprint detection module receives the image of the predetermined size from the data communication module and detects the seal information, fingerprint information or first rotation information of the image of the predetermined size.

[0197] In some embodiments, the text OCR module receives the first rotation information from the seal fingerprint detection module, and the text OCR module is further configured to rotate the image of the predetermined size according to the first rotation information to obtain a rotated image and identify the text information on the rotated image;

[0198] The keyword extraction module extracts the keyword in the form of a regular expression according to the text information, and replaces and corrects the incorrect text in the keyword;

[0199] The keyword extraction module determines whether the keyword is complete; if so, the keyword extraction module transmits the keyword to the data communication module; if not, the keyword extraction module generates second rotation information and transmits the second rotation information to the text OCR module; the text OCR module is further configured to rotate the image of the predetermined size according to the second rotation information to obtain a rotated image and identify the text information on the rotated image;

[0200] In some embodiments, the text OCR module uses the det_r50_db++_icdar15 pre-trained model of PaddleOCR and conducts secondary training, introducing the differentiable binarization DB++ detection algorithm for text information detection;

[0201] Alternatively, the text OCR module uses the ch_PP-OCRv3_rec pre-trained model of PaddleOCR and conducts secondary training, uses the SVTR algorithm for feature enhancement and uses the knowledge distillation method to compress the model to improve the recognition speed.

[0202] Optionally, specific examples in this application can refer to the examples described in the above embodiments and optional implementation manners.

[0203] The above specific embodiments of this application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0204] In the above embodiments of this application, the descriptions of each embodiment have their own emphases. For parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0205] In several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the above-described system embodiments are only illustrative. For example, multiple devices can be combined or integrated into another system, or some features can be ignored or not executed.

[0206] The above is only the preferred implementation manner of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and refinements can still be made, and these improvements and refinements should also be regarded as the protection scope of this application.

Claims

1. A transport license information recognition system, characterized in that: include: A graphical user interface module, used for receiving and encoding the image of the document to be identified to obtain first image data; a data communication module, coupled to the graphical user interface module, configured to receive the first image data from the graphical user interface module and decode the first image data to obtain an image of a predetermined size; A text OCR module, coupled to the data communication module, for receiving the image of the predetermined size from the data communication module and identifying text information on the image of the predetermined size; as well as A keyword extraction module is coupled to the data communication module and the text OCR module, respectively, and is used to receive the text information from the text OCR module, extract keywords from the text information and transmit the keywords to the data communication module; the keyword extraction module is also used to determine whether the keywords are complete; if so, the keyword extraction module transmits the keywords to the data communication module; If not, the keyword extraction module generates second rotation information and transmits the second rotation information to the text OCR module; the text OCR module is further used to rotate the image of the predetermined size according to the second rotation information to obtain a rotated image and recognize text information on the rotated image; The data communication module is further used to receive the keyword from the keyword extraction module and transmit the keyword to the graphical user interface.

2. A transport certificate information recognition system according to claim 1, characterized in that: Also includes: The seal fingerprint detection module is coupled to the data communication module and the text OCR module respectively, and is used to receive the image of the predetermined size from the data communication module and detect the seal information, fingerprint information or first rotation information of the image of the predetermined size.

3. A transport certificate information recognition system according to claim 2, characterized in that: The text OCR module is also used to receive the first rotation information from the seal fingerprint detection module, and the text OCR module is also used to rotate the image of the predetermined size according to the first rotation information to obtain a rotated image and recognize text information on the rotated image.

4. A transport certificate information recognition system according to claim 1, characterized in that: The text OCR module uses the det_r50_db++_icdar15 pre-trained model of PaddleOCR and performs secondary training, and introduces the differentiable binary DB++ detection algorithm to detect text information; Alternatively, the text OCR module uses PaddleOCR's ch_PP-OCRv3_rec pre-trained model and performs secondary training, uses the SVTR algorithm for feature enhancement, and uses the knowledge distillation method to compress the model to improve the recognition speed.

5. A transport certificate information recognition system according to claim 4, characterized in that: The keyword extraction module is also used to extract the keywords in the form of regular expressions according to the text information, and replace and correct erroneous words in the keywords.

6. A method for identifying transport certificate information using the transport certificate information identification system according to any one of claims 1 to 5, characterized in that: include: The graphical user interface module receives and encodes the image of the document to be identified to obtain first image data; The data communication module receives the first image data from the graphical user interface module and decodes the first image data to obtain an image of a predetermined size; The text OCR module receives the image of the predetermined size from the data communication module, and recognizes text information on the image of the predetermined size; The keyword extraction module receives the text information from the text OCR module, extracts keywords from the text information and transmits the keywords to the data communication module; the keyword extraction module determines whether the keywords are complete; If so, the keyword extraction module transmits the keyword to the data communication module; If not, the keyword extraction module generates second rotation information and transmits the second rotation information to the text OCR module; the text OCR module is further used to rotate the image of the predetermined size according to the second rotation information to obtain a rotated image and recognize text information on the rotated image; as well as The data communication module receives the keyword from the keyword extraction module and transmits the keyword to the graphical user interface.

7. The method for identifying transportation license information according to claim 6, characterized in that: The transport certificate information recognition system also includes a seal fingerprint detection module; The seal fingerprint detection module receives the image of the predetermined size from the data communication module and detects seal information, fingerprint information or first rotation information of the image of the predetermined size.

8. The method for identifying transportation license information according to claim 6, characterized in that: The text OCR module receives the first rotation information from the seal fingerprint detection module, and the text OCR module is further used to rotate the image of the predetermined size according to the first rotation information to obtain a rotated image and recognize text information on the rotated image; The keyword extraction module extracts the keyword in the form of regular expressions according to the text information, and replaces and corrects the erroneous text in the keyword.

9. The method for identifying transportation license information according to claim 6, characterized in that: The text OCR module uses the det_r50_db++_icdar15 pre-trained model of PaddleOCR and performs secondary training, and introduces the differentiable binary DB++ detection algorithm to detect text information; Alternatively, the text OCR module uses PaddleOCR's ch_PP-OCRv3_rec pre-trained model and performs secondary training, uses the SVTR algorithm for feature enhancement, and uses the knowledge distillation method to compress the model to improve the recognition speed.

Images