Financial data analysis method and system based on image recognition
Through the financial data analysis method based on image recognition, the problem of inaccurate extraction of text information in financial notes is solved. Through image recognition and binary processing, combined with identification thresholds and influencing factors, efficient and accurate identification and matching of financial notes is achieved. , improves data processing efficiency and accuracy.
Patent Information
- Application Number
- CN202410971267.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-19
- Publication Date
- 2024-11-08
AI Technical Summary
The prior art is difficult to effectively extract and identify table text information in financial notes, resulting in low data processing efficiency and poor accuracy. Conventional OCR technology is inaccurate in extracting text information in images and is prone to row and column misalignment.
Using financial data analysis method based on image recognition, the data clarity of cells is determined by collecting financial bill image data, paper edge detection and image recognition, binarization processing and template comparison, and the absolute value of grayscale difference is calculated to determine the data clarity of cells, and The data transmission path is determined based on the identification threshold and the influencing factor, and the user database matching is performed to improve the recognition accuracy.
It improves the recognition accuracy of financial bill image data, reduces the workload of system entry operations, improves staff efficiency and data accuracy, and effectively processes high-quality and low-quality bill information.
Smart Images

Figure CN118918592A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of image recognition, and in particular relates to a financial data analysis method and system based on image recognition. Background Art
[0002] In today's rapidly developing digital business environment, enterprises are facing increasing challenges in financial data processing and management. Traditional manual processing of financial documents, such as invoices, receipts and bills, is not only time-consuming and laborious, but also prone to data input errors, affecting the accuracy and real-time nature of financial data. Therefore, more and more companies are beginning to pay attention to and invest in the application of intelligent recognition of financial document information.
[0003] Intelligent recognition technology has attracted widespread attention from enterprises due to its high efficiency, accuracy and automation. By using OCR and machine learning algorithms, the intelligent recognition system can quickly and accurately identify and extract key information from financial documents, such as amount, date, supplier information, etc. This technology can not only significantly reduce the time and cost of manual input, but also effectively reduce the risk of human error and improve the efficiency and quality of data processing.
[0004] At present, the extraction of financial bill image data has always been a difficult problem for OCR technology. Conventional OCR technology extracts text information from images, but cannot effectively extract and recognize the original format of the text information in the form. It is also necessary to manually fill in the text information to the corresponding position, which greatly reduces work efficiency. In addition, the existing conventional OCR recognition algorithm often results in misalignment of rows and columns when extracting financial statement data.
[0005] In view of the shortcomings of the existing technology, it is urgent to design a financial data analysis method and system based on image recognition to solve the above technical problems. Summary of the invention
[0006] In order to solve the above-mentioned deficiencies in the prior art, the purpose of the present invention is to overcome the existing deficiencies and provide a financial data analysis method based on image recognition, which is characterized by comprising the following steps:
[0007] Collecting enterprise financial bill image data information, performing paper edge detection on the image data information, obtaining a regular first processed image containing the enterprise financial bill, and transmitting the first processed image to a source database;
[0008] Obtain any first processed image stored in the source database, perform image recognition on the first processed image, obtain enterprise financial bill data used to describe any user, and perform recognition threshold detection on the enterprise financial bill data:
[0009] If the recognition degree of the enterprise financial bill data is not lower than the bill recognition threshold, transmitting the enterprise financial bill data and the corresponding first processed image to the first recognition database;
[0010] If the recognition degree of the enterprise financial bill data is lower than the bill recognition threshold, transmitting the corresponding first processed image to the second recognition database;
[0011] Obtain any enterprise financial bill data from the first identification database, perform user database matching on the enterprise financial bill data, and newly enter the successfully matched enterprise financial bill data into the user database.
[0012] As a further optimization of the above solution, performing image recognition on the first processed image specifically includes the following:
[0013] Acquire corporate financial bill images for describing any user based on image recognition;
[0014] Perform frame selection and recognition on the corporate financial bill image, and perform binary processing on the image;
[0015] Obtain the enterprise financial bill template stored in the source database, and compare the enterprise financial bill template with the binary image based on the binary image:
[0016] If the absolute value of the grayscale difference between the binary image and the cell range located by the enterprise financial bill template is not greater than the detection threshold, the cell is marked as unqualified data;
[0017] If the absolute value of the grayscale difference between the binary image and the cell range located by the enterprise financial bill template is greater than the detection threshold, the cell is marked as qualified data;
[0018] Obtain the absolute value of the grayscale difference between the binarized image and all cells of the enterprise financial bill template. If any absolute value of the grayscale difference is not greater than the detection threshold, transmit the first processed image corresponding to the binarized image to the second recognition database; otherwise, transmit it to the first recognition database.
[0019] As a further optimization of the above solution, the identification threshold detection of the enterprise financial bill data also includes the following:
[0020] Set a recognition threshold T = 0.4a + 0.2b + 0.2c + 0.2d;
[0021] Detect the influencing factors a, b, c and d of the enterprise financial bill data, record any existing influencing factors as 1, otherwise as 0, calculate the recognition value S of the first processed image, and the recognition threshold T=0.6, then:
[0022] If S ≥ T, transmitting the enterprise financial bill data and the corresponding first processed image to the first recognition database;
[0023] If S<T, transmitting the corresponding first processed image to the second recognition database;
[0024] In the above formula, a is the type of bill, b is the buyer's name, c is the buyer's identification number, and d is the buyer's account information.
[0025] As a further optimization of the above solution, matching the enterprise financial bill data with the user database specifically includes the following:
[0026] Acquire any enterprise financial bill data from the first identification database, wherein the enterprise financial bill data is used to record and describe the enterprise financial bill information;
[0027] Traverse all the influencing factors of the enterprise financial bill data, and retrieve the influencing factors b and c of any enterprise financial bill data;
[0028] If any enterprise financial bill data is retrieved that contains the impact factor b, traverse the user database to match all user registered data associated with the impact factor b; based on all the traversed user registered data, match and enter the impact factor of any enterprise financial bill data;
[0029] If any enterprise financial bill data is retrieved that contains the impact factor c, traverse the user database to match all user registered data associated with the impact factor c; based on all the traversed user registered data, match and enter the impact factor of any enterprise financial bill data;
[0030] If any enterprise financial bill data retrieved contains influencing factors b and c, traverse the user database to match any registered data of all users that simultaneously contain influencing factors b and c; based on all traversed registered data of users, match and enter the influencing factors of any enterprise financial bill data.
[0031] As a further optimization of the above solution, matching the enterprise financial bill data with the user database also includes the following:
[0032] Setting any of the enterprise financial bill data to include influencing factors, linkage factors and candidate factors;
[0033] After traversing the user database to match all registered user data associated with any influencing factor, all influencing factors except the current influencing factor are newly entered into the to-be-processed row based on the influencing factor;
[0034] Obtaining the corresponding linkage factor of any row where the influencing factor of the enterprise financial bill data is successfully matched based on image recognition, and matching the linkage factor based on image recognition with the user database:
[0035] If the linkage factor matches the user database, the linkage factor corresponding to the user database is copied and transmitted and newly entered into the to-be-processed row of the user database;
[0036] If the matching value between the linkage factor and the user database is less than the matching threshold, the linkage factor based on image recognition is newly entered into the to-be-processed row of the user database;
[0037] Based on image recognition, any candidate factor of the enterprise financial bill data is obtained, and the candidate factor based on image recognition is newly entered into a pending row in the user database.
[0038] As a further optimization of the above solution, any existing influencing factor also includes the corresponding influencing factor attribute.
[0039] As a further optimization of the above solution, the enterprise financial bill data is traversed and retrieved for the influencing factor a:
[0040] If the influencing factor a is not found in the enterprise financial bill data, the first processed image corresponding to the enterprise financial bill data is transmitted to the second recognition database;
[0041] If the enterprise financial bill data is found to have an impact factor a, the impact factor a is matched with the enterprise financial bill template stored in the source database, and the successfully matched enterprise financial bill template is copied and transmitted to the first identification database.
[0042] The present invention also discloses a financial data analysis system based on image recognition, the system comprising:
[0043] A first image processing module is used to collect corporate financial bill image data information, perform paper edge detection on the image data information, obtain a regular first processed image containing corporate financial bills, and transmit the first processed image to a source database;
[0044] An image recognition module is used to obtain any first processed image stored in the source database, perform image recognition on the first processed image, obtain corporate financial bill data used to describe any user, and perform recognition threshold detection on the corporate financial bill data:
[0045] If the recognition degree of the enterprise financial bill data is not lower than the bill recognition threshold, transmitting the enterprise financial bill data and the corresponding first processed image to the first recognition database;
[0046] If the recognition degree of the enterprise financial bill data is lower than the bill recognition threshold, transmitting the corresponding first processed image to the second recognition database;
[0047] The data matching module is used to obtain any enterprise financial bill data from the first identification database, match the enterprise financial bill data with the user database, and enter the successfully matched enterprise financial bill data into the user database.
[0048] As a further optimization of the above solution, the first image processing module specifically includes the following:
[0049] An image acquisition unit, used for acquiring an enterprise financial bill image for describing any user based on image recognition;
[0050] An image processing unit, used to perform frame selection and recognition on the corporate financial bill image and perform binary processing on the image;
[0051] The image judgment unit is used to obtain the enterprise financial bill template stored in the source database, and compare the enterprise financial bill template with the binary image based on the binary image:
[0052] If the absolute value of the grayscale difference between the binary image and the cell range located by the enterprise financial bill template is not greater than the detection threshold, the cell is marked as unqualified data;
[0053] If the absolute value of the grayscale difference between the binary image and the cell range located by the enterprise financial bill template is greater than the detection threshold, the cell is marked as qualified data;
[0054] An image storage unit is used to obtain the absolute value of the grayscale difference between the binary image and all cells of the enterprise financial bill template. If any absolute value of the grayscale difference is not greater than the detection threshold, the first processed image corresponding to the binary image is transmitted to the second recognition database; otherwise, it is transmitted to the first recognition database.
[0055] As a further optimization of the above solution, any existing influencing factor also includes the corresponding influencing factor attribute.
[0056] The present invention adopts the above technical solution, which has the following beneficial effects compared with the prior art:
[0057] 1. The present invention proposes a financial data analysis method based on image recognition, which specifically performs OCR processing on the image data of corporate financial bills, quickly realizes the isolation processing of electronic bills and scanned bills, and quickly extracts keywords according to the OCR recognition results to match the corresponding text information. Based on a complete corporate user database, the accuracy of corporate financial bill image data recognition is further improved, the workload of system entry operations of bill images is reduced to a minimum, and the efficiency of staff and data accuracy are greatly improved.
[0058] 2. In the process of processing the first processed image, the present invention calculates the absolute value of the grayscale difference between the binary image and the corresponding enterprise financial bill template in all cell ranges, so as to quickly determine whether the corresponding cell has data information or is clear, and transmits different recognition results to the first or second recognition database respectively, so as to efficiently process high-quality bill information and accurately process low-quality bill information, with good effect.
[0059] 3. The present invention also sets an influencing factor for the enterprise financial bill data, that is, the influencing factor is a characteristic value that can quickly judge and confirm the bill information. By identifying and detecting the influencing factor, the quality of the invoice to be processed is further improved, and the images below the recognition threshold are transmitted to the second recognition database for manual processing. At the same time, the identified influencing factors are stored in the first recognition database for subsequent information processing.
[0060] 4. The present invention accurately groups financial bill information according to data characteristics, matches according to corresponding characteristics, and the influencing factors are directly entered items; further for linkage factor analysis, when the parameter information corresponding to the linkage factor is identified by OCR technology and matches the information value stored in the user database, the information value stored in the user database is directly copied, and this selection is saved and entered as full accuracy; otherwise, the linkage factor based on image recognition is newly entered into the pending row of the user database, and there may be recognition errors. The technology of the present invention will effectively reduce the staff's auditing efforts on financial bill information, improve their work efficiency and accuracy, and have good positive effects. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:
[0062] Figure 1 It is a schematic diagram of a process of the present invention;
[0063] Figure 2 Another schematic diagram of the process of the present invention;
[0064] Figure 3 Another schematic diagram of the process of the present invention. DETAILED DESCRIPTION
[0065] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0066] like Figure 1-3 As shown, the embodiment of the present invention discloses a financial data analysis method based on image recognition, which specifically includes the following:
[0067] Collecting enterprise financial bill image data information, performing paper edge detection on the image data information, obtaining a regular first processed image containing the enterprise financial bill, and transmitting the first processed image to a source database;
[0068] Obtain any first processed image stored in the source database, perform image recognition on the first processed image, obtain enterprise financial bill data used to describe any user, and perform recognition threshold detection on the enterprise financial bill data:
[0069] If the recognition degree of the enterprise financial bill data is not lower than the bill recognition threshold, the enterprise financial bill data and the corresponding first processed image are transmitted to the first recognition database;
[0070] If the recognition degree of the enterprise financial bill data is lower than the bill recognition threshold, the corresponding first processed image is transmitted to the second recognition database;
[0071] Obtain any enterprise financial bill data from the first identification database, match the enterprise financial bill data with the user database, and enter the successfully matched enterprise financial bill data into the user database.
[0072] Specifically, OCR refers to the technology of converting the text in paper documents into black and white dot matrix image files by optical means for printed characters, and converting the text in the image files into text format through recognition software, and sending them to word processing software for further editing and processing. At present, the extraction of financial bill image data has always been a difficult point in OCR technology. Conventional OCR technology extracts text information from images, but cannot effectively extract and recognize the original format of text information in tables. It is necessary to manually fill in the text information in the corresponding position, which greatly reduces work efficiency. In addition, the existing conventional OCR recognition algorithm for extracting financial statement data often results in row and column misalignment.
[0073] The present invention proposes a financial data analysis method based on image recognition, which specifically performs OCR processing on the image data of corporate financial bills, quickly realizes the isolation processing of electronic bills and scanned bills, and quickly extracts keywords according to the OCR recognition results to match the corresponding text information. Based on a complete corporate user database, the accuracy of corporate financial bill image data recognition is further improved, the workload of system entry operations of bill images is reduced to a minimum, and the efficiency of staff and data accuracy are greatly improved.
[0074] Specifically, performing image recognition on the first processed image specifically includes the following steps:
[0075] Based on image recognition, an image of a corporate financial bill is obtained to describe any user; more specifically, the corporate financial bill image of the present invention can be materials for company accounting, bank reconciliation, account settlement, account book query and printout, payment currency, customer reconciliation, invoicing, reimbursement, etc.;
[0076] Perform frame selection and recognition on the corporate financial bill images and perform binary processing on the images;
[0077] Get the enterprise financial bill template stored in the source database, and compare the enterprise financial bill template with the binary image based on the binary image:
[0078] If the absolute value of the grayscale difference between the binary image and the enterprise financial bill template within the cell range is not greater than the detection threshold, the cell is marked as unqualified data;
[0079] If the absolute value of the grayscale difference between the binary image and the enterprise financial bill template within the cell range is greater than the detection threshold, the cell is marked as qualified data;
[0080] Obtain the absolute value of the grayscale difference between the binary image and all cells of the enterprise financial bill template. If any absolute value of the grayscale difference is not greater than the detection threshold, transmit the first processed image corresponding to the binary image to the second recognition database; otherwise, transmit it to the first recognition database.
[0081] During the processing of the first processed image, the present invention calculates the absolute value of the grayscale difference between the binary image and the corresponding enterprise financial bill template in all cell ranges, thereby quickly determining whether the corresponding cell has data information or is clear, and transmits different recognition results to the first or second recognition database respectively, so as to efficiently process high-quality bill information and accurately process low-quality bill information, with good results.
[0082] It should be noted here that the present invention uses OCR recognition technology to convert the image of the bill information into text, which is specifically manifested as binary image processing. By identifying the cell borders of the bill information, the corresponding recognition table is quickly generated, and the cells involved in the present invention are not in the same format.
[0083] Specifically, the identification threshold detection of enterprise financial bill data also includes the following:
[0084] Set a recognition threshold T = 0.4a + 0.2b + 0.2c + 0.2d;
[0085] Detect the influencing factors a, b, c and d of the enterprise financial bill data, record any existing influencing factors as 1, otherwise as 0, calculate the recognition value S of the first processed image, and the recognition threshold T = 0.6, then:
[0086] If S ≥ T, transmit the enterprise financial bill data and the corresponding first processed image to the first recognition database;
[0087] If S<T, transmitting the corresponding first processed image to the second recognition database;
[0088] In the above formula, a is the type of bill, b is the buyer's name, c is the buyer's identification number, and d is the buyer's account information.
[0089] More specifically, the present invention also sets an influencing factor for the enterprise financial bill data, that is, the influencing factor is a characteristic value that can quickly judge and confirm the bill information. By identifying and detecting the influencing factor, the quality of the invoice to be processed is further improved, and the image below the recognition threshold T is transmitted to the second recognition database for manual processing. At the same time, the identified influencing factors a, b, c and d are stored in the first recognition database for subsequent information processing.
[0090] Specifically, the user database matching of enterprise financial bill data includes the following:
[0091] Acquire any enterprise financial bill data from the first identification database, where the enterprise financial bill data is used to record and describe enterprise financial bill information;
[0092] Traverse all the influencing factors of the enterprise financial bill data, and retrieve the influencing factors b and c of any enterprise financial bill data;
[0093] If any enterprise financial bill data is retrieved that contains the impact factor b, traverse the user database to match all user registered data associated with the impact factor b; based on all the traversed user registered data, match and enter the impact factor of any enterprise financial bill data;
[0094] If any enterprise financial bill data is retrieved that contains the impact factor c, traverse the user database to match all user registered data associated with the impact factor c; based on all the traversed user registered data, match and enter the impact factor of any enterprise financial bill data;
[0095] If any enterprise financial bill data retrieved contains influencing factors b and c, traverse the user database to match any registered data of all users that contain influencing factors b and c; based on all traversed registered data of users, match and enter the influencing factors of any enterprise financial bill data.
[0096] More specifically, the present invention provides a specific embodiment, by obtaining any enterprise financial bill data of the first identification database, and traversing all influencing factors, mainly traversing influencing factors b and c. Due to the data characteristics of influencing factors b and c, that is, influencing factor c is the unique tax identification number of any enterprise user, influencing factor b is any enterprise name, and each influencing factor b corresponds to a unique influencing factor c; when traversing influencing factors b and c, the influencing factors traversed to the user database are marked, so that all corresponding bill information entered into the user database can be obtained. More specifically, if there is an enterprise whose influencing factor b is X Company, and influencing factor c is 91XXXX, and X Company was formerly known as XY Company, then according to influencing factor b, all the row information of the company named X Company can be matched; according to influencing factor c, all the row information of the company's names can be matched. Based on the design of the present invention, data that can briefly describe the bill information can be quickly obtained according to the characteristic value of the enterprise bill information, so as to quickly match the financial bills to be processed. It should be noted here that if any enterprise financial bill data retrieved does not contain influencing factors b and c, the corresponding financial bill will be transmitted to the second identification database; if any enterprise financial bill data retrieved contains influencing factors b or c, the user database is traversed and no influencing factors b or c are matched, all influencing factors of the enterprise financial bill data will be newly entered into the user database to be processed rows.
[0097] Specifically, matching the enterprise financial bill data to the user database also includes the following:
[0098] Set any enterprise financial bill data including influencing factors, linkage factors and candidate factors;
[0099] After traversing the user database to match all registered user data associated with any influencing factor, all influencing factors except the current influencing factor are newly entered into the to-be-processed row based on the influencing factor;
[0100] Based on image recognition, obtain the corresponding linkage factor of any row where the influencing factor of the enterprise financial bill data is successfully matched, and match the linkage factor based on image recognition with the user database:
[0101] If the linkage factor matches the user database value and is not less than the matching threshold, the linkage factor corresponding to the user database is copied and transferred and newly entered into the user database to be processed row;
[0102] If the linkage factor matches the user database value less than the matching threshold, the linkage factor based on image recognition will be newly entered into the user database to be processed row;
[0103] Based on image recognition, any candidate factors of enterprise financial bill data are obtained, and the candidate factors based on image recognition are newly entered into the pending row of the user database.
[0104] More specifically, the present invention also provides an embodiment, according to the financial bill information, the existing conventional bill information is divided into influencing factors, linkage factors and candidate factors, wherein the influencing factors are characteristic values that can quickly judge and confirm the bill information, the linkage factors are possible change items, and the candidate factors are inevitable change items. That is, the present invention accurately groups the financial bill information according to the data characteristics, matches according to the corresponding characteristics, and the influencing factors are directly entered items; further for the linkage factor analysis, when the parameter information corresponding to the linkage factor is identified by OCR technology and matches the information value stored in the user database, the information value stored in the user database is directly copied, and this selection is saved and entered as complete accuracy; otherwise, the linkage factor based on image recognition is newly entered into the pending row of the user database, and there may be recognition errors. The technology of the present invention will effectively reduce the staff's review of financial bill information, improve their work efficiency and accuracy, and have good positive effects. It should be noted here that although the present invention gives a description of the influencing factors a, b, c and d, the practical application of the present invention should include but not be limited to the types of influencing factors a, b, c and d, and be adjusted according to different types of financial bill information. The same applies to linkage factors and selected factors, which will not be described in detail here.
[0105] Specifically, any existing influencing factor also includes the corresponding influencing factor attribute. More specifically, the influencing factor attribute of the present invention is specifically used for image recognition processing, that is, after identifying the influencing factor attribute, the identification and locking of the cell is completed, so as to further determine the specific situation of the influencing factor of the enterprise financial bill information.
[0106] Specifically, the impact factor a on the traversal retrieval of corporate financial bill data is:
[0107] If the influencing factor a is not found in the enterprise financial bill data, the first processed image corresponding to the enterprise financial bill data is transmitted to the second recognition database;
[0108] If the enterprise financial bill data is found to have an impact factor a, the impact factor a is matched with the enterprise financial bill template stored in the source database, and the successfully matched enterprise financial bill template is copied and transmitted to the first identification database.
[0109] More specifically, although the present invention provides an embodiment of a financial data analysis method based on image recognition, which can quickly identify and further analyze corporate financial bill data based on the influencing factor a, the present invention also supports the pre-setting of different corporate financial bill templates, so as to quickly match according to different influencing factors a, thereby avoiding major differences caused by different layouts and affecting the efficiency of financial bill entry.
[0110] The present invention also discloses an embodiment, a financial data analysis system based on image recognition, comprising the following:
[0111] The first image processing module is used to collect corporate financial bill image data information, perform paper edge detection on the image data information, obtain a regular first processed image containing corporate financial bills, and transmit the first processed image to a source database;
[0112] An image recognition module is used to obtain any first processed image stored in the source database, perform image recognition on the first processed image, obtain enterprise financial bill data used to describe any user, and perform recognition threshold detection on the enterprise financial bill data:
[0113] If the recognition degree of the enterprise financial bill data is not lower than the bill recognition threshold, the enterprise financial bill data and the corresponding first processed image are transmitted to the first recognition database;
[0114] If the recognition degree of the enterprise financial bill data is lower than the bill recognition threshold, the corresponding first processed image is transmitted to the second recognition database;
[0115] The data matching module is used to obtain any enterprise financial bill data from the first identification database, match the enterprise financial bill data with the user database, and enter the successfully matched enterprise financial bill data into the user database.
[0116] Specifically, the first image processing module specifically includes the following:
[0117] An image acquisition unit, used for acquiring an enterprise financial bill image for describing any user based on image recognition;
[0118] An image processing unit is used to perform frame selection and recognition on the corporate financial bill image and perform binary processing on the image;
[0119] The image judgment unit is used to obtain the enterprise financial bill template stored in the source database, and compare the enterprise financial bill template with the binary image based on the binary image:
[0120] If the absolute value of the grayscale difference between the binary image and the enterprise financial bill template within the cell range is not greater than the detection threshold, the cell is marked as unqualified data;
[0121] If the absolute value of the grayscale difference between the binary image and the enterprise financial bill template within the cell range is greater than the detection threshold, the cell is marked as qualified data;
[0122] The image storage unit is used to obtain the absolute value of the grayscale difference between the binary image and all cells of the enterprise financial bill template. If any absolute value of the grayscale difference is not greater than the detection threshold, the binary image corresponding to the first processed image is transmitted to the second recognition database; otherwise, it is transmitted to the first recognition database.
[0123] It should be noted that the technical means for implementing another embodiment of the present invention, a financial data analysis system based on image recognition, is the same as that of a financial data analysis method based on image recognition, and thus will not be described in detail herein.
[0124] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other.
Claims
1. A financial data analysis method based on image recognition, characterized in that , including the following steps: Collecting enterprise financial bill image data information, performing paper edge detection on the image data information, obtaining a regular first processed image containing the enterprise financial bill, and transmitting the first processed image to a source database; Obtain any first processed image stored in the source database, perform image recognition on the first processed image, obtain enterprise financial bill data used to describe any user, and perform recognition threshold detection on the enterprise financial bill data: If the recognition degree of the enterprise financial bill data is not lower than the bill recognition threshold, transmitting the enterprise financial bill data and the corresponding first processed image to the first recognition database; If the recognition degree of the enterprise financial bill data is lower than the bill recognition threshold, transmitting the corresponding first processed image to the second recognition database; Obtain any enterprise financial bill data from the first identification database, perform user database matching on the enterprise financial bill data, and newly enter the successfully matched enterprise financial bill data into the user database.
2. The financial data analysis method based on image recognition according to claim 1 is characterized in that , performing image recognition on the first processed image specifically includes the following steps: Acquire corporate financial bill images for describing any user based on image recognition; Perform frame selection and recognition on the corporate financial bill image, and perform binary processing on the image; Obtain the enterprise financial bill template stored in the source database, and compare the enterprise financial bill template with the binary image based on the binary image: If the absolute value of the grayscale difference between the binary image and the cell range located by the enterprise financial bill template is not greater than the detection threshold, the cell is marked as unqualified data; If the absolute value of the grayscale difference between the binary image and the cell range located by the enterprise financial bill template is greater than the detection threshold, the cell is marked as qualified data; Obtaining the absolute value of the grayscale difference between the binary image and all cells of the enterprise financial bill template, and if any absolute value of the grayscale difference is not greater than the detection threshold, transmitting the first processed image corresponding to the binary image to the second recognition database; Otherwise, it is transmitted to the first identification database.
3. A financial data analysis method based on image recognition according to claim 2, characterized in that , the identification threshold detection of the enterprise financial bill data also includes the following: Set a recognition threshold T = 0.4a + 0.2b + 0.2c + 0.2d; Detect the influencing factors a, b, c and d of the enterprise financial bill data, record any existing influencing factors as 1, otherwise as 0, calculate the recognition value S of the first processed image, and the recognition threshold T=0.6, then: If S ≥ T, transmitting the enterprise financial bill data and the corresponding first processed image to the first recognition database; If S<T, transmitting the corresponding first processed image to the second recognition database; In the above formula, a is the type of bill, b is the buyer's name, c is the buyer's identification number, and d is the buyer's account information.
4. The method for analyzing financial data based on image recognition according to claim 3, characterized in that , matching the enterprise financial bill data to the user database specifically includes the following: Acquire any enterprise financial bill data from the first identification database, wherein the enterprise financial bill data is used to record and describe the enterprise financial bill information; Traverse all the influencing factors of the enterprise financial bill data, and retrieve the influencing factors b and c of any enterprise financial bill data; If any enterprise financial bill data is retrieved that contains the impact factor b, traverse the user database to match all user registered data associated with the impact factor b; Based on all the registered data of users traversed, the influencing factors of any enterprise financial bill data are matched and entered; If any enterprise financial bill data is retrieved that contains the impact factor c, traverse the user database to match all user registered data associated with the impact factor c; Based on all the registered data of users traversed, the influencing factors of any enterprise financial bill data are matched and entered; If any enterprise financial bill data is retrieved that contains the impact factors b and c, traverse the user database to match any user registered data that contains both the impact factors b and c; Based on the traversed registered data of all users, the influencing factors of any enterprise financial bill data are matched and entered.
5. The method for analyzing financial data based on image recognition according to claim 4, characterized in that , matching the enterprise financial bill data to the user database also includes the following: Setting any of the enterprise financial bill data to include influencing factors, linkage factors and candidate factors; After traversing the user database to match all registered user data associated with any influencing factor, all influencing factors except the current influencing factor are newly entered into the to-be-processed row based on the influencing factor; Obtaining the corresponding linkage factor of any row where the influencing factor of the enterprise financial bill data is successfully matched based on image recognition, and matching the linkage factor based on image recognition with the user database: If the linkage factor matches the user database, the linkage factor corresponding to the user database is copied and transmitted and newly entered into the to-be-processed row of the user database; If the matching value between the linkage factor and the user database is less than the matching threshold, the linkage factor based on image recognition is newly entered into the to-be-processed row of the user database; Based on image recognition, any candidate factor of the enterprise financial bill data is obtained, and the candidate factor based on image recognition is newly entered into a pending row in the user database.
6. The method for analyzing financial data based on image recognition according to claim 5, characterized in that ,Any existing influencing factor also includes the corresponding influencing factor attribute.
7. The method for analyzing financial data based on image recognition according to claim 6, characterized in that , traverse and retrieve the impact factor a of the enterprise financial bill data: If the influencing factor a is not found in the enterprise financial bill data, the first processed image corresponding to the enterprise financial bill data is transmitted to the second recognition database; If the enterprise financial bill data is found to have an impact factor a, the impact factor a is matched with the enterprise financial bill template stored in the source database, and the successfully matched enterprise financial bill template is copied and transmitted to the first identification database.
8. A financial data analysis system based on image recognition, characterized in that , the system comprises the following: A first image processing module is used to collect corporate financial bill image data information, perform paper edge detection on the image data information, obtain a regular first processed image containing corporate financial bills, and transmit the first processed image to a source database; An image recognition module is used to obtain any first processed image stored in the source database, perform image recognition on the first processed image, obtain corporate financial bill data used to describe any user, and perform recognition threshold detection on the corporate financial bill data: If the recognition degree of the enterprise financial bill data is not lower than the bill recognition threshold, transmitting the enterprise financial bill data and the corresponding first processed image to the first recognition database; If the recognition degree of the enterprise financial bill data is lower than the bill recognition threshold, transmitting the corresponding first processed image to the second recognition database; The data matching module is used to obtain any enterprise financial bill data from the first identification database, match the enterprise financial bill data with the user database, and enter the successfully matched enterprise financial bill data into the user database.
9. A financial data analysis system based on image recognition according to claim 8, characterized in that ,The first image processing module specifically includes the following: An image acquisition unit, used for acquiring an enterprise financial bill image for describing any user based on image recognition; An image processing unit, used to perform frame selection and recognition on the corporate financial bill image and perform binary processing on the image; The image judgment unit is used to obtain the enterprise financial bill template stored in the source database, and compare the enterprise financial bill template with the binary image based on the binary image: If the absolute value of the grayscale difference between the binary image and the cell range located by the enterprise financial bill template is not greater than the detection threshold, the cell is marked as unqualified data; If the absolute value of the grayscale difference between the binary image and the cell range located by the enterprise financial bill template is greater than the detection threshold, the cell is marked as qualified data; An image storage unit, used for obtaining the absolute value of the grayscale difference between the binary image and all cells of the enterprise financial bill template, and if any absolute value of the grayscale difference is not greater than the detection threshold, transmitting the first processed image corresponding to the binary image to the second recognition database; Otherwise, it is transmitted to the first identification database.
10. A financial data analysis system based on image recognition according to claim 9, characterized in that ,Any existing influencing factor also includes the corresponding influencing factor attribute.
Citation Information
Cited By
Enterprise financial data analysis management system and method based on cloud computing
CN119991319A
