A method and system for detecting financial and tax data

Abstract

The present invention relates to the technical field of financial systems, and in particular to a method and system for detecting financial and tax data, comprising the following steps: S1: acquiring business management data based on the system; S2: performing document recognition management of each business based on hardware equipment according to the business management data; S3: improving the accuracy of document recognition data using a character separation positioning module; S4: performing business management based on the document recognition results; S5: archiving and storing documents and sorting them, and automatically issuing warnings based on the sorting results. The present invention integrates the document flow, approval flow, image flow, capital flow, and voucher flow into one through a document design platform, image management, and process management, thereby realizing business-finance integration and intelligent tax burden warning; through invoice recognition, intelligent form filling is realized to reduce the error rate of form filling, and through intelligent accounting, the work efficiency of financial personnel is improved and the business information of the enterprise can be reflected in a more real-time manner, thus safeguarding the business development of the enterprise.

Description

Technical Field

[0001] The present invention relates to the technical field of financial systems, and in particular to a financial and taxation data detection method and system. Background Art

[0002] Financial accounting is an important part of enterprise economic accounting. It reflects and supervises the enterprise and its

[0003] Financial accounting is a method of calculating the labor consumption, material consumption and capital occupation and their economic effects in the production and operation process of each accounting unit within the industry. Financial accounting is based on the audited original vouchers, uses accounting subjects, fills in accounting vouchers, registers accounting books, uses currency as the calculation scale, and continuously, systematically and comprehensively records and calculates the economic activities of each accounting unit or accounting item.

[0004] The contents of financial accounting mainly include: calculation of income, expenses and costs; calculation and processing of financial results; other matters that require accounting procedures and accounting calculations, etc.; business personnel provide original documents such as invoices to the finance department for offline verification and accounting; finance personnel need to check the invoices after receiving them, and then perform manual statistics and manually enter the documents for accounting, which requires a large workload and low efficiency; due to the delay of offline business processes, finance cannot conduct timely accounting in the future, resulting in data delays, affecting company decisions and causing certain tax risks. Summary of the invention

[0005] The purpose of the present invention is to solve the defects mentioned in the above background technology by proposing a financial and tax data detection method and system.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A method for detecting financial and tax data is provided, comprising the following steps:

[0008] S1: Obtain business management data based on the system;

[0009] S2: Perform document identification and management for each business based on business management data and hardware equipment;

[0010] S3: Use the character separation positioning module to improve the accuracy of document recognition data;

[0011] S4: Business management based on document recognition results;

[0012] S5: Archive and store documents and organize them, and issue automatic warnings based on the organization results.

[0013] As a preferred technical solution of the present invention: the business management data of S1 includes expense reimbursement data, image management data, customer management data, procurement management data, warehouse management data, sales management data, asset management data, and personnel management data.

[0014] As a preferred technical solution of the present invention: S1 also configures a document template based on the acquired business management data.

[0015] As a preferred technical solution of the present invention: in S2, the image data of the document is collected based on the hardware device, and the document type is identified, the compliance of the document is checked, the process node of the current document is determined and the corresponding document identification process is selected for identification and approval.

[0016] As a preferred technical solution of the present invention: in S2, the identification of the document includes the following steps:

[0017] S2.1: Identify employee QR code;

[0018] S2.2: Identify and query employee documents;

[0019] S2.3: Print employee receipts;

[0020] S2.4: Print employee information.

[0021] As a preferred technical solution of the present invention: in S2.2, the document image is corrected based on a correction algorithm:

[0022]

[0023] Among them, W ij is the gray value of the ith row and jth column of the corrected document image, I ij is the gray value at the i-th row and j-th column of the document image, α is the correction coefficient, m is the number of rows of the document image, and n is the number of columns of the document image;

[0024] The document is also recognized based on the OCR recognition algorithm:

[0025] Select the document recognition benchmark, perform benchmark recognition based on the recognition frame, and obtain the four vertex coordinates of the benchmark (x1, y1), (x2, y2), (x3, y3) and (x4, y4).

[0026] Calculate the height and width of each character:

[0027]

[0028]

[0029] Where H is the height of each character in the document, L is the width of each character, and N is the number of characters contained in the recognition box;

[0030] Select a single character in the recognition box as a candidate character, extract the features of the candidate character, and identify the candidate character based on the KNN algorithm:

[0031]

[0032] Among them, D is the recognition result of candidate characters, K is the number of candidate characters, X k is the extracted candidate recognition character feature data, Y k It is the character feature data stored in the database.

[0033] As a preferred technical solution of the present invention: the KNN algorithm is sorted based on the value of the character recognition result D, and the character data stored in the database corresponding to the smallest value of the character recognition result D is selected as the final recognition result.

[0034] As a preferred technical solution of the present invention: the specific operation steps of using the character separation and positioning module in S3 to improve the accuracy of document recognition data are as follows:

[0035] S3.1: Use U-Net semantic classification network to locate characters on financial documents;

[0036] S3.2: Then, the characters located by the U-Net semantic classification network are binarized, and the processing results are filtered and rotated for correction;

[0037] S3.3: Segment the characters after image binarization and filter rotation correction to form multiple character regions with clearer classification;

[0038] S3.4: Add a BN layer to the convolutional neural network to obtain an improved convolutional neural network, and use the improved convolutional neural network to recognize the segmented character area to improve the character recognition speed and accuracy;

[0039] S3.5: An improved MobileNet is obtained by omitting the extension layer of the linear financial document in MobileNet, and the improved MobileNet is used to recognize the segmented and located characters to further improve the character recognition accuracy.

[0040] As a preferred technical solution of the present invention: in S4, intelligent form filling is performed based on the document recognition result to achieve corresponding business management. In S5, the document process and image invoice are archived, and vouchers and reports are automatically generated. In addition, intelligent tax burden warning is performed according to the industry attributes and tax law requirements of the enterprise.

[0041] A financial and taxation data detection system is provided, comprising:

[0042] Data acquisition module: used to obtain business management data;

[0043] Document identification module: performs document identification management for each business based on business management data and hardware equipment;

[0044] Character separation and positioning module: locates and segments characters on financial documents to improve character recognition accuracy;

[0045] Business management module: business management based on document recognition results;

[0046] Arrangement and warning module: archive and store documents and organize them, and issue automatic warnings based on the organization results.

[0047] Compared with the prior art, the financial and tax data detection method and system provided by the present invention have the following beneficial effects:

[0048] The present invention integrates the document flow, approval flow, image flow, capital flow and voucher flow into one through the document design platform, image management and process management, thus realizing the integration of business and finance and intelligent early warning of tax burden. Through invoice recognition, intelligent form filling reduces the error rate of form filling, and through intelligent accounting, the work efficiency of financial personnel is improved and the business information of the enterprise can be reflected in a more real-time manner, thus safeguarding the business development of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A method flow chart of a preferred embodiment of the present invention;

[0050] Figure 2 It is a system block diagram of a preferred embodiment of the present invention.

[0051] The meanings of the marks in the figure are as follows: 100, data acquisition module; 200, document recognition module; 300, business management module; 400, sorting and warning module. DETAILED DESCRIPTION

[0052] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of this embodiment can be combined with each other. The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only 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 are within the scope of protection of the present invention.

[0053] Reference Figure 1The preferred embodiment of the present invention provides a method for detecting financial and tax data, comprising the following steps:

[0054] S1: Obtain business management data based on the system;

[0055] S2: Perform document identification and management for each business based on business management data and hardware equipment;

[0056] S3: Business management based on document recognition results;

[0057] S4: Archive and store documents and organize them, and issue automatic warnings based on the organization results.

[0058] The business management data of S1 includes expense reimbursement data, image management data, customer management data, procurement management data, warehouse management data, sales management data, asset management data, and personnel management data.

[0059] The S1 also configures a document template based on the acquired business management data.

[0060] In S2, the image data of the document is collected based on the hardware device, and the document type is identified to check whether the document is compliant, determine the process node of the current document and select the corresponding document identification process for identification and approval.

[0061] In S2, the identification of the document includes the following steps:

[0062] S2.1: Identify employee QR code;

[0063] S2.2: Identify and query employee documents;

[0064] S2.3: Print employee receipts;

[0065] S2.4: Print employee information.

[0066] Correct the document image based on the correction algorithm:

[0067]

[0068] Among them, W ij is the gray value of the ith row and jth column of the corrected document image, I ij is the gray value at the i-th row and j-th column of the document image, α is the correction coefficient, m is the number of rows of the document image, and n is the number of columns of the document image;

[0069] The document is also recognized based on the OCR recognition algorithm:

[0070] Select the document recognition benchmark, perform benchmark recognition based on the recognition frame, and obtain the four vertex coordinates of the benchmark (x1, y1), (x2, y2), (x3, y3) and (x4, y4).

[0071] Calculate the height and width of each character:

[0072]

[0073]

[0074] Where H is the height of each character in the document, L is the width of each character, and N is the number of characters contained in the recognition box;

[0075] Select a single character in the recognition box as a candidate character, extract the features of the candidate character, and identify the candidate character based on the KNN algorithm:

[0076]

[0077] Among them, D is the recognition result of candidate characters, K is the number of candidate characters, X k is the extracted candidate recognition character feature data, Y k It is the character feature data stored in the database.

[0078] The KNN algorithm performs sorting based on the value of the character recognition result D, and selects the character data stored in the database corresponding to the smallest value of the character recognition result D as the final recognition result.

[0079] The specific steps for using the character separation positioning module in S3 to improve the accuracy of document recognition data are as follows:

[0080] S3.1: Use the U-Net semantic classification network to locate the characters on the financial documents. This network not only solves the problem that the traditional algorithm is difficult to extract the character area of ​​the data in this paper, but also makes the character positioning more accurate;

[0081] S3.2: Then, the characters located by the U-Net semantic classification network are binarized, and the processing results are filtered and rotated for correction;

[0082] S3.3: Segment the characters after image binarization and filter rotation correction to form multiple character regions with clearer classification;

[0083] S3.4: A BN layer is added to the convolutional neural network to obtain an improved convolutional neural network, and the improved convolutional neural network is used to recognize the segmented character areas to improve the character recognition speed and accuracy. The improved convolutional neural network has a faster convergence speed of the LSS curve during the training process, achieves a higher training speed, and has good recognition accuracy, which is more conducive to character recognition in financial documents.

[0084] S3.5: An improved MobileNet is obtained by omitting the extension layer of linear financial documents in MobileNet, and the improved MobileNet is used to recognize the segmented and located characters to further improve the character recognition accuracy. Although omitting the extension layer of linear financial documents increases the number of training rounds of MobileNet, the training time of each round is reduced, and the overall training time is also shortened. In addition, the recognition accuracy of the improved MobileNe is also improved, which can further improve the character recognition accuracy on financial documents. The recognition rate is maintained at 98.7%, and the entire automatic recognition process time is maintained at 0.8 seconds.

[0085] In S4, intelligent form filling is performed based on the document recognition result to achieve corresponding business management.

[0086] In the S5, the document process and the image invoice are archived, and vouchers and reports are automatically generated. In addition, intelligent tax burden warning is carried out according to the industry attributes and tax law requirements of the enterprise.

[0087] Reference Figure 2 , providing a financial and tax data detection system, including:

[0088] Data acquisition module 100: used to acquire business management data;

[0089] Document identification module 200: performs document identification management for each business based on hardware devices according to business management data;

[0090] Character separation and positioning module 300: locates and segments characters on financial documents to improve character recognition accuracy;

[0091] Business management module 400: performs business management based on the document recognition result;

[0092] Arrangement and warning module 500: archives and stores documents and arranges them, and issues automatic warnings based on the arrangement results.

[0093] In this embodiment, the system application layer is divided into two parts: business and finance. The business part mainly includes expense reimbursement, image management, customer management, procurement management, warehouse management, sales management, asset management, and personnel management modules; the financial part mainly includes accounting vouchers, voucher preparation, balance sheet, income statement, consolidated statement, account balance sheet, beginning balance, cash flow statement, month-end closing, and tax burden warning.

[0094] The system enables seamless integration of business and finance through full business digital management processes, document management, and image management, thereby realizing intelligent financial accounting and tax burden early warning.

[0095] It is also equipped with a document design platform, an image management platform and a process management platform to realize the five-in-one flow of document, approval flow, image flow, capital flow and voucher flow, laying the foundation for achieving business and financial integration and intelligent tax burden early warning.

[0096] In order to better integrate the document design platform, image management platform and process management platform in the system.

[0097] Perform image acquisition and recognition of documents based on hardware equipment, collect image data of documents based on hardware equipment, identify document types, check whether documents are compliant, determine the process node of the current document and select the corresponding document recognition process for identification and approval.

[0098] The identification and approval of documents include identifying employee QR codes; identifying and querying employee documents; printing employee documents and printing employee information.

[0099] During the document recognition and approval process, the document image is corrected based on the correction algorithm:

[0100]

[0101] Among them, W ij is the gray value of the ith row and jth column of the corrected document image, I ij is the gray value at the i-th row and j-th column of the document image, α is the correction coefficient, m is the number of rows of the document image, and n is the number of columns of the document image;

[0102] Correcting the document image based on the correction algorithm can enhance the character area, suppress the flying character area in the document image, enhance the contrast between the character area and the non-character area, and improve the recognition accuracy of the characters.

[0103] When identifying and querying employee documents, the document is also identified based on the OCR recognition algorithm:

[0104] Select the document recognition benchmark, perform benchmark recognition based on the recognition frame, and obtain the four vertex coordinates of the benchmark (x1, y1), (x2, y2), (x3, y3) and (x4, y4).

[0105] Calculate the height and width of each character:

[0106]

[0107]

[0108] Where H is the height of each character in the document, L is the width of each character, and N is the number of characters contained in the recognition box;

[0109] Select a single character in the recognition box as a candidate character, extract the features of the candidate character, and identify the candidate character based on the KNN algorithm:

[0110]

[0111] Among them, D is the recognition result of candidate recognition characters, the number of candidate recognition characters is 15, X k is the extracted candidate recognition character feature data, Y k It is the character feature data stored in the database.

[0112] The KNN algorithm sorts the values ​​of the character recognition results D and selects the character data stored in the database corresponding to the smallest value of the character recognition result D as the final recognition result.

[0113] Based on OCR technology, the outline of the characters is recognized, and then based on the KN algorithm, the specific recognition of the characters is performed. It can be applied to the problem of multi-classification of characters, and the recognition process is simple and the recognition speed is fast. Through invoice recognition, intelligent form filling is realized to reduce the error rate of form filling.

[0114] Through the document design platform, different business document types such as expense reimbursement, purchase payment, sales order, purchase order, warehouse entry and exit order can be designed; and for different business document types, different business approval processes can be designed through the process management platform, which can be flexibly designed according to the approval amount authority, project leader authority, etc. to meet the needs of users; the image management platform can ensure the validity and legality of the invoices uploaded by users, and the system can realize intelligent form filling without manual filling and data aggregation; at the same time, the user's last invoice and other materials are classified and managed; for each business document type, rules are set in accordance with accounting standards and accounting law requirements to meet the needs of intelligent accounting, and according to the attributes of different industries and tax law requirements, the system has designed corresponding tax burden warnings. Intelligent accounting improves the work efficiency of financial personnel and can respond to business information in a more real-time manner, which protects the business development of enterprises.

[0115] This system has a unified technical base. Through the effective combination of PaaS platform and SaaS platform, the system adopts SaaS platform for processing in the case of the same usage scenarios for most users. Users can download and use it directly. For some users with special management needs, adjustments and permission allocation can be made through PaaS platform to solve the public and private use problems. It can effectively meet the public and private domain applications in enterprise industry solutions.

[0116] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0117] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for detecting financial and tax data, characterized in that: The steps include: S1: Obtain business management data based on the system; S2: Perform document identification and management for each business based on business management data and hardware equipment; S3: Use the character separation positioning module to improve the accuracy of document recognition data; S4: Business management based on document recognition results; S5: archive and store documents and organize them, and automatically warn based on the organization results; the business management data in S1 includes expense reimbursement data, image management data, customer management data, procurement management data, warehouse management data, sales management data, asset management data, and personnel management data; in S2, the image data of the document is collected based on the hardware device, and the document type is identified, whether the document is compliant, the process node of the current document is determined, and the corresponding document recognition process is selected for recognition and approval; In S2.2, the document image is corrected based on the correction algorithm: Among them, W ij is the gray value of the ith row and jth column of the corrected document image, I ij is the gray value at the i-th row and j-th column of the document image, α is the correction coefficient, m is the number of rows of the document image, and n is the number of columns of the document image; The document is also recognized based on the OCR recognition algorithm: Select the document recognition benchmark, perform benchmark recognition based on the recognition frame, and obtain the four vertex coordinates of the benchmark (x1, y1), (x2, y2), (x3, y3) and (x4, y4). Calculate the height and width of each character: Where H is the height of each character in the document, L is the width of each character, and N is the number of characters contained in the recognition box; Select a single character in the recognition box as a candidate character, extract the features of the candidate character, and identify the candidate character based on the KNN algorithm: Among them, D is the recognition result of candidate characters, K is the number of candidate characters, X k is the extracted candidate recognition character feature data, Y k Character feature data stored in the database; The specific operation steps of using the character separation positioning module in S3 to improve the accuracy of document recognition data are as follows: S3.1: Use U-Net semantic classification network to locate characters on financial documents; S3.2: Then, the characters located by the U-Net semantic classification network are binarized, and the processing results are filtered and rotated for correction; S3.3: Segment the characters after image binarization and filter rotation correction to form multiple character regions with clearer classification; S3.4: Add a BN layer to the convolutional neural network to obtain an improved convolutional neural network, and use the improved convolutional neural network to recognize the segmented character area to improve the character recognition speed and accuracy; S3.5: The extended layer of the linear financial document is omitted in MobileNet to obtain an improved MobileNet, and the improved MobileNet is used to recognize the segmented and located characters to further improve the character recognition accuracy; in S2, the recognition of the document includes the following steps: S2.1: Identify employee QR code; S2.2: Identify and query employee documents; S2.3: Print employee receipts; S2.4: Print employee information; The KNN algorithm is sorted based on the value of the character recognition result D, and the character data stored in the database corresponding to the smallest value of the character recognition result D is selected as the final recognition result; in S4, intelligent form filling is performed based on the document recognition result to achieve corresponding business management, and in S5, the document process and image invoice are archived, and vouchers and reports are automatically generated, and intelligent tax burden warning is performed according to the industry attributes and tax law requirements of the enterprise.

2. The method for detecting financial and tax data according to claim 1, characterized in that: The S1 also configures a document template based on the acquired business management data.

3. A fiscal and taxation data detection system, based on the fiscal and taxation data detection method according to claim 1 or 2, characterized in that: include: Data acquisition module (100): used to acquire business management data; Document identification module (200): performs document identification management for each business based on the business management data and hardware equipment; Character separation and positioning module (300): locates and segments characters on financial documents to improve character recognition accuracy; Business management module (400): performs business management based on the document recognition result; Arrangement and warning module (500): performs archiving and storage operations and arrangement operations on documents, and automatically issues warnings based on the arrangement results.

Images