Invoice classification method, apparatus, and medium

Abstract

The application provides an invoice classification method and device and a medium, relates to the technical field of computers, and is used for solving the problem that manual identification of signature information on a paper invoice is prone to errors in the prior art.The method comprises the following steps: obtaining to-be-identified signature information data on an invoice and pre-collected contrast signature information data; the to-be-identified signature information data and the contrast signature information data are matched and calculated by using a recursive algorithm and an autonomous learning function to obtain a first identification result of the to-be-identified signature information data; and first classification information of the invoice is obtained according to the first identification result, and the first classification information is used for first classification of the invoice. Through obtaining an accurate identification result of to-be-identified signature information, the application realizes classification and archiving of the invoice, improves the accuracy of identification of signature information, improves the accuracy of classification of the invoice, and can reduce the workload of financial personnel and improve work efficiency.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to an invoice classification method, apparatus, and computer-readable storage medium. Background Technology

[0002] In financial work, the finance department has certain requirements for invoice filing to facilitate future inquiries about invoices. These requirements include requiring the person claiming reimbursement to clearly write down the information such as the person, department, and purpose of the invoice in black ink, and requiring the finance staff to classify and file the invoices after receiving them.

[0003] Currently, the classification of invoices mainly relies on manual identification of reimbursement signatures on paper invoices. In practice, the main drawbacks are that it is time-consuming and labor-intensive, and errors or confusion may occur in invoice classification or invoice number identification. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the above-mentioned shortcomings of the prior art by providing an invoice classification method, apparatus and computer-readable storage medium to solve the problem that errors are easily made in the manual identification of signature information on paper invoices in the prior art.

[0005] In a first aspect, the present invention provides an invoice classification method, the method comprising:

[0006] Obtain the signature information data to be identified on the invoice and the pre-collected signature information data for comparison;

[0007] The signature information data to be identified and the signature information data to be compared are matched and calculated using a recursive algorithm and a self-learning function to obtain the first identification result of the signature information data to be identified.

[0008] The first classification information of the invoice is obtained based on the first identification result, and the first classification information is used to classify the invoice in the first way.

[0009] Optionally, obtaining the signature information data to be identified on the invoice and the pre-collected signature information data for comparison specifically includes:

[0010] Receive the first scan data obtained by scanning the reimbursement signature written by the person making the reimbursement on the invoice, and use the first scan data as the signature information data to be identified;

[0011] The system receives second scan data obtained by scanning the reference signature samples, and uses the second scan data and the reimbursement signature information corresponding to each reference signature sample as reference signature information data. The reference signature samples are reimbursement signatures written by potential reimbursement personnel collected in advance.

[0012] Optionally, the step of matching the signature information data to be identified and the comparison signature information data using a recursive algorithm and a self-learning function to obtain a first identification result for the signature information data to be identified specifically includes:

[0013] Assign the first scan data to X, assign the second scan data to Y, and set the maximum number of recursions and the initial parameters of the recursive algorithm.

[0014] The correlation coefficient between X and Y in the kth recursion is calculated according to the following formula (1):

[0015]

[0016] Where k is the current recursion count, β k Let X be the correlation coefficient between X and Y in the k-th recursion, and let Covariance(X) be the coefficient of correlation between X and Y. k Y k Let E(X) be the covariance of X and Y during the k-th recursion. k ), E(Y k ) represents the expected value of the reimbursement signature information of X and Y during the k-th recursion;

[0017] The output coefficients of X and Y in the kth recursion are calculated according to the following formula (2):

[0018]

[0019] in, Let X and Y be the output coefficients of X and Y in the k-th recursion, p be the preset number of signature information types, and ψ be the output coefficients of X and Y. k This is the self-learning function during the k-th recursion;

[0020] Update the self-learning function for the next recursion iteration in the k-th recursion according to the following formula (3):

[0021]

[0022] Where, ψ k+1 For the self-learning function in the next recursive iteration, β min G This represents the historical recursive minimum value.

[0023] After reaching the preset maximum number of recursions, obtain the E(X) with the smallest output coefficient across all recursion counts. k The first recognition result is the signature information data to be identified.

[0024] Optionally, the reimbursement signature includes: invoice purpose, name of the person requesting reimbursement, and reimbursement number;

[0025] The first identification result includes: the invoice purpose information, the name information of the person claiming reimbursement, and the reimbursement number information corresponding to the signature information data to be identified;

[0026] The first category information includes: information for classifying invoices according to their purpose, and information for sorting invoices according to the name of the person seeking reimbursement and the reimbursement number.

[0027] Optionally, the method further includes:

[0028] The system obtains the invoice information data itself from the invoice, identifies the invoice information data to obtain a second identification result, and obtains the second classification information of the invoice based on the second identification result. The second classification information is used to classify the invoice in a second way.

[0029] Optionally, the step of obtaining the invoice information data itself on the invoice, identifying the invoice information data to obtain a second identification result, and obtaining the second classification information of the invoice based on the second identification result specifically includes:

[0030] Receive third scan data of the text on the invoice itself and fourth scan data of the QR code on the invoice;

[0031] The third scan data is identified to obtain invoice type information, and the fourth scan data is verified through the online tax system to obtain invoice authenticity information;

[0032] The second category information of the invoice is obtained based on the invoice type information and the invoice authenticity information.

[0033] Optionally, after obtaining the first classification information and / or the second classification information of the invoice, the method further includes:

[0034] Output the scanned image of the invoice, along with the first and / or second category information, as an editable document;

[0035] Receive inspection information input by the user after inspecting the editable document;

[0036] Invoices are classified based on the first and / or second classification information and inspection information.

[0037] In a second aspect, the present invention provides an invoice sorting device, including a computer device, the computer device comprising:

[0038] The acquisition module is used to acquire the signature information data to be identified on the invoice and the pre-collected signature information data for comparison;

[0039] The identification module, connected to the acquisition module, is used to perform matching calculations on the signature information data to be identified and the signature information data to be compared using a recursive algorithm and a self-learning function, so as to obtain the first identification result of the signature information data to be identified.

[0040] The classification information module, connected to the recognition module, is used to obtain first classification information of the invoice based on the first recognition result. The first classification information is used to classify the invoice.

[0041] Optionally, the apparatus further includes an invoice classifier, the invoice classifier comprising:

[0042] The classification execution module, connected to the classification information module, is used to classify invoices according to the first classification information.

[0043] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the invoice classification method as described above.

[0044] This invention provides an invoice classification method, apparatus, and computer-readable storage medium. By introducing computer technology to identify signature information on invoices, and by comparing the signature information to be identified with pre-collected reference signature information through machine learning, an accurate identification result of the signature information to be identified can be obtained. This identification result can then be used to help financial personnel classify and archive invoices, improving the accuracy of signature information identification and thus the accuracy of invoice classification. It also reduces the workload of financial personnel and improves work efficiency. Attached Figure Description

[0045] Figure 1 This is a flowchart of an invoice classification method according to an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram of the structure of an invoice sorting device according to an embodiment of the present invention;

[0047] Figure 3 This is a flowchart of another invoice classification method in an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of the structure of a computer device for an invoice sorting apparatus according to an embodiment of the present invention. Detailed Implementation

[0049] To enable those skilled in the art to better understand the technical solution of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0050] It is understood that the specific embodiments and accompanying drawings described herein are merely for explaining the invention and are not intended to limit the invention.

[0051] It is understood that, without conflict, the various embodiments and features in the embodiments of the present invention can be combined with each other.

[0052] It is understood that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, while the parts unrelated to the present invention are not shown in the drawings.

[0053] It is understood that each unit or module involved in the embodiments of the present invention may correspond to only one entity structure, or may be composed of multiple entity structures, or multiple units or modules may be integrated into one entity structure.

[0054] It is understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of this invention may occur in a different order than that marked in the accompanying drawings.

[0055] It is understood that the flowcharts and block diagrams of this invention illustrate the possible architecture, functions, and operations of systems, apparatuses, devices, and methods according to various embodiments of this invention. Each block in the flowchart or block diagram may represent a unit, module, program segment, or code, containing executable instructions for implementing the specified function. Furthermore, each block or combination of blocks in the block diagram and flowchart can be implemented using a hardware-based system to achieve the specified function, or using a combination of hardware and computer instructions.

[0056] It is understood that the units and modules involved in the embodiments of the present invention can be implemented by software or by hardware. For example, the units and modules can be located in a processor.

[0057] Example 1:

[0058] like Figure 1 As shown, Embodiment 1 of the present invention provides an invoice classification method, the method comprising:

[0059] S11. Obtain the signature information data to be identified on the invoice and the pre-collected signature information data for comparison;

[0060] S12. The signature information data to be identified and the signature information data to be compared are matched and calculated using a recursive algorithm and a self-learning function to obtain the first identification result of the signature information data to be identified.

[0061] S13. Obtain the first classification information of the invoice based on the first identification result. The first classification information is used to classify the invoice in the first way.

[0062] Specifically, the method described in this embodiment is applied to, for example, Figure 2The invoice sorting device shown, particularly the computer device 1 used therein, compares the signature information to be identified with pre-collected reference signature information through machine learning to obtain an accurate recognition result for the signature information to be identified. Then, based on the first classification information obtained from this recognition result, it helps financial personnel to classify and archive invoices. One example is that the computer device 1 controls an invoice sorter 2 to automatically classify paper invoices. For this purpose, an invoice sorter 2 needs to be designed. One possible design method is that the invoice sorter 2 includes a scanning camera, a photosensitive reader, an invoice stacking brush, invoice transfer rollers, and other related components. Since the mechanical structure of the invoice sorter 2 is not the subject of this invention, it will not be described in detail. The classification information obtained by the computer device 1 can also be used to classify electronic images of invoices, such as scanning images of electronic invoices or paper invoices. The method described in this embodiment can improve the accuracy of signature information recognition, thereby improving the accuracy of invoice classification, reducing the workload of financial personnel, and improving work efficiency.

[0063] Optionally, obtaining the signature information data to be identified on the invoice and the pre-collected signature information data for comparison specifically includes:

[0064] Receive the first scan data obtained by scanning the reimbursement signature written by the person making the reimbursement on the invoice, and use the first scan data as the signature information data to be identified;

[0065] The system receives second scan data obtained by scanning the reference signature samples, and uses the second scan data and the reimbursement signature information corresponding to each reference signature sample as reference signature information data. The reference signature samples are reimbursement signatures written by potential reimbursement personnel collected in advance.

[0066] Specifically, the method described in this embodiment can obtain signature information by analyzing the pigment in the ink of a ballpoint pen. Previously, the person seeking reimbursement was required to mark the invoice with a black pen, for example, "Purpose + Person's Name + Number," such as "Travel + Zhang San + No. 1." The invoice is placed in the invoice classifier 2, and the scanning camera of the invoice classifier 2 scans the invoice to obtain a scanned image. The photosensitive reader acquires data from the pigment in the black pen, such as high-quality acetylene black, to obtain the first scan data. At this time, the scan data of other inks on the original invoice is not acquired. The signature of the potential reimbursement person can be obtained in the same way.

[0067] Optionally, the step of matching the signature information data to be identified and the comparison signature information data using a recursive algorithm and a self-learning function to obtain a first identification result for the signature information data to be identified specifically includes:

[0068] Assign the first scan data to X, assign the second scan data to Y, and set the maximum number of recursions and the initial parameters of the recursive algorithm.

[0069] The correlation coefficient between X and Y in the kth recursion is calculated according to the following formula (1):

[0070]

[0071] Where k is the current recursion count, β k Let X be the correlation coefficient between X and Y in the k-th recursion, and let Covariance(X) be the coefficient of correlation between X and Y. k Y k Let E(X) be the covariance of X and Y during the k-th recursion. k ), E(Y k ) represents the expected value of the reimbursement signature information of X and Y during the k-th recursion;

[0072] The output coefficients of X and Y in the kth recursion are calculated according to the following formula (2):

[0073]

[0074] in, Let X and Y be the output coefficients of X and Y in the k-th recursion, p be the preset number of signature information types, and ψ be the output coefficients of X and Y. k This is the self-learning function during the k-th recursion;

[0075] Update the self-learning function for the next recursion iteration in the k-th recursion according to the following formula (3):

[0076]

[0077] Where, ψ k+1 For the self-learning function in the next recursive iteration, β min G This represents the historical recursive minimum value.

[0078] After reaching the preset maximum number of recursions, obtain the E(X) with the smallest output coefficient across all recursion counts. k The first recognition result is the signature information data to be identified.

[0079] Specifically, the method described in this embodiment realizes rapid recognition and iteration through multiple trials of machine learning to obtain a recognition result of signature information. Since it is impossible to exhaust all possible reimbursement signature contents in the control signature samples, it is necessary to match the signature information data to be recognized with the control signature information data through machine learning. Specifically, the principle of infinite proximity may be applied to calculate the correlation coefficient, then calculate the deviation, and realize machine learning and iteration through circulation to obtain an infinitely close solution, and quickly and accurately output the first recognition result when the deviation result is the minimum. The specific iterative calculation formula has been described above. For the iterative calculation process, the maximum number of iterations is set to 50, and the initial number of iterations is 0. When k=0, the first scan data is assigned to X, and the second scan data is assigned to Y, the reimbursement signature information corresponding to Y is known, and the reimbursement signature information of X can be randomly selected by a machine within a preset range. When k≠0, each data is a value updated after machine learning iteration. The above formulas (1)-(3) are circularly calculated from k=0 to 50, and after the maximum number of recursions is reached, the solution corresponding to the historical minimum recursion value is output.

[0080] Optionally, the reimbursement signature includes: invoice purpose, name of the reimbursement personnel, and reimbursement number;

[0081] The first recognition result includes: invoice purpose information, reimbursement personnel name information and reimbursement number information corresponding to the signature information data to be recognized;

[0082] The first classification information includes: information for classifying invoices according to invoice purpose information, and information for sorting invoices according to reimbursement personnel name information and reimbursement number information.

[0083] Specifically, after the method described in this embodiment recognizes the reimbursement signature information written on the invoice, it performs a first stacking process on the invoice through the invoice classifier 2 according to the marked invoice purpose, and further sorts the invoices based on the stacking to obtain a first classification result, for example, for travel: Zhang San No. 1, Zhang San No. 2, Zhang San No. 3, etc. The numbers of the reimbursement number information are arranged from low to high, for example: Arabic numerals 1, 2, 3, 4, 5..., Chinese numerals "First", "Second", "Third"..., Chinese characters "Yi", "Er", "San"..., or "Yi", "Er", "San"——"⑴", "⑵", "⑶"——"1", "2", "3"——"⑴", "⑵", "⑶"——"①", "②", "③", etc.

[0084] Optionally, the method further comprises:

[0085] Acquiring original invoice information data on an invoice, recognizing the original invoice information data to obtain a second recognition result, and obtaining second classification information of the invoice according to the second recognition result, where the second classification information is used to perform a second classification on the invoice.

[0086] Optionally, the step of obtaining the invoice information data itself on the invoice, identifying the invoice information data to obtain a second identification result, and obtaining the second classification information of the invoice based on the second identification result specifically includes:

[0087] Receive third scan data of the text on the invoice itself and fourth scan data of the QR code on the invoice;

[0088] The third scan data is identified to obtain invoice type information, and the fourth scan data is verified through the online tax system to obtain invoice authenticity information;

[0089] The second category information of the invoice is obtained based on the invoice type information and the invoice authenticity information.

[0090] Specifically, the method described in this embodiment, in addition to identifying reimbursement signature information, also identifies invoice types and invoice authenticity, and further improves the invoice classification effect by using invoice type information and invoice authenticity information, counts the quantity of each type of invoice, and punches holes in counterfeit invoices.

[0091] Optionally, after obtaining the first classification information and / or the second classification information of the invoice, the method further includes:

[0092] Output the scanned image of the invoice, along with the first and / or second category information, as an editable document;

[0093] Receive inspection information input by the user after inspecting the editable document;

[0094] Invoices are classified based on the first and / or second classification information and inspection information.

[0095] Specifically, the method described in this embodiment also provides manual inspection. After the machine automatically obtains the classification information, it outputs an editable document to the finance personnel, who then verify the classification information. If a classification error is found, the incorrectly classified invoices are marked with fluorescent markers.

[0096] Example 1: A more specific example is as follows Figure 3 As shown, it includes:

[0097] (1) Place the invoice into the invoice classifier 2 and obtain the first classification information; Specifically, start the invoice classifier 2, place the invoice marked with "purpose, personnel name, number" into the invoice classifier 2, obtain the black pen high-quality acetylene black marking information data by scanning the thermal sensor, and the computer device 1 compares the information collected and set in the early stage with the text recorded by the black pen to obtain the first classification information of the invoice. The comparison method is to use iterative calculation. The specific process is as before and will not be repeated.

[0098] (2) The invoices are classified for the first time according to the invoice purpose information in the first classification information; specifically, the invoices are classified according to the invoice purpose obtained by automatic identification, such as travel, transportation, heating, trade union, party dues, etc., and the invoice classifier 2 is controlled to perform invoice stacking processing according to the invoice purpose information.

[0099] (3) Sort the invoices according to the name information and reimbursement number information of the reimbursement personnel in the first category information; Specifically, based on the piling in step (2), the carbon black mark that has been identified is divided, the combination of "name + number" is further identified, the name of the reimbursement personnel on the invoice is automatically identified in different pilings, and the invoices with the same "name" are arranged according to the number order after the name, for example, business trip: Zhang San + 1, Zhang San + 2, Zhang San + 3, Li Si + 1, Li Si + 2, etc.

[0100] (4) Classify the invoices a second time according to the invoice type information; specifically, for the scanned results of the invoices, only the text of the invoice itself is recognized, and the number of invoices with the three types of invoices, namely “VAT general invoice”, “VAT special invoice” and “special invoice”, is counted. In the already sorted categories, without disrupting the order, the invoices are classified according to the invoice types given by the tax bureau, such as 3 VAT general invoices, 5 VAT special invoices, etc.

[0101] (5) Verify the authenticity of invoices through the online tax system; Specifically, all invoices, whether "general VAT invoices", "special VAT invoices" or "special invoices", have a QR code on them. By scanning the QR code and connecting to the online tax system, the authenticity of the invoices can be quickly verified. Therefore, after completing the above identification and classification process, the QR code information on each invoice is transmitted to the tax invoice system by scanning the QR code on the invoice to distinguish the authenticity of the invoice. If it is a genuine invoice, proceed to the next step. If there are some problems with the invoice (such as no invoice found or incorrect information), the invoice is punched.

[0102] (6) Receive the user's verification information on the invoice classification information; specifically, after classification, the classified invoice content is output as an editable document through the scanning system. The user selects the output device, such as a mobile phone, computer, etc., to output the invoice. The user screens the invoice classification, number, authenticity, etc. If any errors are found, fluorescent markings are directly applied to the invoice to distinguish the invoices and facilitate manual verification later. If there are no problems, the final invoice classification result is obtained.

[0103] Example 2:

[0104] like Figure 2 and 4As shown, Embodiment 2 of the present invention provides an invoice sorting device, including a computer device 1, wherein the computer device 1 includes:

[0105] The acquisition module 11 is used to acquire the signature information data to be identified on the invoice and the pre-collected signature information data for comparison;

[0106] The identification module 12, connected to the acquisition module 11, is used to perform matching calculations on the signature information data to be identified and the signature information data to be compared using a recursive algorithm and a self-learning function, so as to obtain the first identification result of the signature information data to be identified.

[0107] The classification information module 13 is connected to the identification module 12 and is used to obtain the first classification information of the invoice based on the first identification result. The first classification information is used to classify the invoice.

[0108] Optionally, the device further includes an invoice classifier 2, the invoice classifier 2 comprising:

[0109] The classification execution module, connected to the classification information module, is used to classify invoices according to the first classification information.

[0110] Optionally, module 11 specifically includes:

[0111] The first receiving unit is used to receive the first scan data obtained by scanning the reimbursement signature written by the person making the reimbursement on the invoice, and to use the first scan data as the signature information data to be identified.

[0112] The second receiving unit is used to receive the second scan data obtained by scanning the reference signature sample, and to use the second scan data and the reimbursement signature information corresponding to each reference signature sample as reference signature information data. The reference signature sample is the reimbursement signature written by potential reimbursement personnel collected in advance.

[0113] Optionally, the invoice classifier 2 further includes:

[0114] The scanning module is used to scan the pigment base of the reimbursement signature written by the person submitting the reimbursement on the invoice to obtain the first scan data.

[0115] Optionally, the identification module 12 specifically includes:

[0116] An initialization unit is used to assign the first scan data to X, the second scan data to Y, and set the maximum number of recursions and the initial parameters of the recursive algorithm.

[0117] The first calculation unit is used to calculate the correlation coefficient between X and Y in the k-th recursion according to the following formula (1):

[0118]

[0119] Where k is the current recursion count, β k Let X be the correlation coefficient between X and Y in the k-th recursion, and let Covariance(X) be the coefficient of correlation between X and Y. k Y k Let E(X) be the covariance of X and Y during the k-th recursion. k ), E(Y k Let ) be the expected value of the reimbursement signature information of X and Y at the k-th recursion;

[0120] The second calculation unit is used to calculate the output coefficients of X and Y in the kth recursion according to the following formula (2):

[0121]

[0122] in, Let X and Y be the output coefficients of X and Y in the k-th recursion, p be the preset number of signature information types, and ψ be the output coefficients of X and Y. k This is the self-learning function during the k-th recursion;

[0123] The third calculation unit is used to update the self-learning function for the next recursion number at the k-th recursion according to the following formula (3):

[0124]

[0125] Where, ψ k+1 For the self-learning function in the next recursive iteration, β min G This represents the historical recursive minimum value.

[0126] The first result unit is used to obtain the E(X) with the smallest output coefficient across all recursion counts, up to the preset maximum recursion count. k The first recognition result is the signature information data to be identified.

[0127] Optionally, the reimbursement signature includes: invoice purpose, name of the person requesting reimbursement, and reimbursement number;

[0128] The first identification result includes: the invoice purpose information, the name information of the person claiming reimbursement, and the reimbursement number information corresponding to the signature information data to be identified;

[0129] The first category information includes: information on first grouping invoices according to invoice purpose information, and information on sorting invoices according to the name of the person claiming reimbursement and the reimbursement number information.

[0130] Optionally, the classification execution module performs a first classification on the invoices based on the first classification information, specifically including:

[0131] The invoices are first grouped according to their purpose information, and then sorted according to the name of the person claiming reimbursement and the reimbursement number.

[0132] Optionally,

[0133] Module 11 is also used to obtain the invoice information data itself on the invoice;

[0134] The identification module 12 is also used to identify the information data of the invoice itself to obtain a second identification result;

[0135] The classification information module 13 is also used to obtain second classification information of the invoice based on the second identification result, and the second classification information is used to classify the invoice in a second way.

[0136] Optionally, the classification execution module is also used to classify invoices a second time based on the second classification information.

[0137] Optionally,

[0138] The acquisition module 11 further includes: a third / fourth receiving unit, used to receive third scan data of the text on the invoice itself and fourth scan data of the QR code on the invoice;

[0139] The identification module 12 further includes: a second / third identification result unit, used to identify the third scan data to obtain invoice type information, and to verify the fourth scan data through the online tax system to obtain invoice authenticity information;

[0140] The second category of information specifically includes: information on second-level grouping and quantity statistics of invoices based on invoice type information, and information on first-level marking of invoices based on invoice authenticity information.

[0141] Optionally, the classification execution module is further configured to perform a second classification of the invoices based on the second classification information, specifically including:

[0142] The invoices are then sorted and counted in quantity based on their type, and marked as genuine or counterfeit based on their authenticity.

[0143] Optionally, the device further includes:

[0144] The output module is used to output the scanned image of the invoice, along with the first category information and / or the second category information, into an editable document;

[0145] The fifth receiving unit is used to receive inspection information input by the user after inspecting the editable document;

[0146] The control module is used to control the classification of invoices based on the first classification information and / or the second classification information and the inspection information.

[0147] Example 2 provides an invoice sorting device, which includes at least a computer device 1 corresponding to the method of Example 1, and may also include invoice sorting machinery, i.e., an invoice sorter 2. The device can achieve accurate sorting of invoices.

[0148] Additionally, Embodiment 2 provides a computer device that can also be described as including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the invoice classification method as described in Embodiment 1.

[0149] The memory is connected to the processor. The memory can be flash memory, read-only memory or other types of memory. The processor can be a central processing unit or a microcontroller.

[0150] Example 3:

[0151] Embodiment 3 of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the invoice classification method as described in Embodiment 1.

[0152] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules, or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory or other memory technologies, CD-ROM (Compact Disc Read-Only Memory), DVD or other optical disc storage, cartridges, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer.

[0153] Example 3 provides a computer-readable storage medium in which a stored computing program, when executed, is used to accurately classify invoices.

[0154] Embodiments 1-3 of the present invention provide an invoice classification method, apparatus, and computer-readable storage medium. By introducing computer technology to identify signature information on invoices, and by comparing the signature information to be identified with pre-collected reference signature information through machine learning, an accurate identification result of the signature information to be identified can be obtained. This identification result can then be used to help financial personnel classify and archive invoices, thereby improving the accuracy of signature information identification and thus improving the accuracy of invoice classification. It can also reduce the workload of financial personnel and improve work efficiency.

[0155] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for classifying invoices, characterized in that, The method includes: Obtain the signature information data to be identified on the invoice and the pre-collected signature information data for comparison, specifically including: The system receives the first scanned data obtained by scanning the reimbursement signature written by the person submitting the expense report on the invoice, and uses this first scanned data as the signature information data to be identified. Receive the second scan data obtained by scanning the reference signature sample, and use the second scan data and the reimbursement signature information corresponding to each reference signature sample as the reference signature information data. The reference signature sample is the reimbursement signature written by potential reimbursement personnel collected in advance. The signature information data to be identified and the reference signature information data are matched using a recursive algorithm and a self-learning function to obtain the first identification result of the signature information data to be identified, specifically including: Assign the first scan data to X, assign the second scan data to Y, and set the maximum number of recursions and the initial parameters of the recursive algorithm. The correlation coefficient between X and Y in the kth recursion is calculated according to the following formula (1): ， Where k is the current recursion count, Let X be the correlation coefficient between X and Y in the k-th recursion. Let X and Y be the covariances of X and Y in the k-th recursion. , Let X and Y be the expected values ​​of their reimbursement signature information during the k-th recursion. Calculate the output coefficients of X and Y in the kth recursion according to the following formula (2): ， in, Let X and Y be the output coefficients of X and Y in the k-th recursion, and p be the preset number of signature information types. This is the self-learning function during the k-th recursion; Update the self-learning function for the next recursion iteration in the k-th recursion according to the following formula (3): ， in, For the self-learning function in the next recursive iteration, This represents the historical recursive minimum value. After reaching the preset maximum number of recursions, obtain the output coefficient with the smallest output coefficient across all recursion counts. The first recognition result of the signature information data to be identified; The first classification information of the invoice is obtained based on the first identification result, and the first classification information is used to classify the invoice in the first way.

2. The method according to claim 1, characterized in that, The reimbursement signature includes: invoice purpose, name of the person requesting reimbursement, and reimbursement number; The first identification result includes: the invoice purpose information, the name information of the person claiming reimbursement, and the reimbursement number information corresponding to the signature information data to be identified; The first category information includes: information for classifying invoices according to their purpose, and information for sorting invoices according to the name of the person seeking reimbursement and the reimbursement number.

3. The method according to any one of claims 1-2, characterized in that, The method further includes: The system obtains the invoice information data itself from the invoice, identifies the invoice information data to obtain a second identification result, and obtains the second classification information of the invoice based on the second identification result. The second classification information is used to classify the invoice in a second way.

4. The method according to claim 3, characterized in that, The process of obtaining the invoice information data itself, identifying the invoice information data to obtain a second identification result, and obtaining the second classification information of the invoice based on the second identification result specifically includes: Receive third scan data of the text on the invoice itself and fourth scan data of the QR code on the invoice; The third scan data is identified to obtain invoice type information, and the fourth scan data is verified through the online tax system to obtain invoice authenticity information; The second category information of the invoice is obtained based on the invoice type information and the invoice authenticity information.

5. The method according to claim 3, characterized in that, After obtaining the first and / or second classification information of the invoice, the method further includes: Output the scanned image of the invoice, along with the first and / or second category information, as an editable document; Receive inspection information input by the user after inspecting the editable document; Invoices are classified based on the first and / or second classification information and inspection information.

6. An invoice sorting device, characterized in that, Includes computer equipment, said computer equipment comprising: The acquisition module is used to acquire the signature information data to be identified on the invoice and the pre-collected signature information data for comparison, specifically including: The first receiving unit is used to receive the first scanned data obtained by scanning the reimbursement signature written by the person submitting the reimbursement on the invoice, and to use the first scanned data as the signature information data to be identified. The second receiving unit is used to receive the second scan data obtained by scanning the reference signature sample, and to use the second scan data and the reimbursement signature information corresponding to each reference signature sample as reference signature information data. The reference signature sample is the reimbursement signature written by potential reimbursement personnel collected in advance. The identification module, connected to the acquisition module, is used to perform matching calculations between the signature information data to be identified and the reference signature information data using a recursive algorithm and a self-learning function to obtain a first identification result for the signature information data to be identified. Specifically, it includes: An initialization unit is used to assign the first scan data to X, the second scan data to Y, and set the maximum number of recursions and the initial parameters of the recursive algorithm. The first calculation unit is used to calculate the correlation coefficient between X and Y in the k-th recursion according to the following formula (1): ， Where k is the current recursion count, Let X be the correlation coefficient between X and Y in the k-th recursion. Let X and Y be the covariances of X and Y in the k-th recursion. , Let X and Y be the expected values ​​of their reimbursement signature information during the k-th recursion. The second calculation unit is used to calculate the output coefficients of X and Y in the kth recursion according to the following formula (2): ， in, Let X and Y be the output coefficients of X and Y in the k-th recursion, and p be the preset number of signature information types. This is the self-learning function during the k-th recursion; The third computational unit is used to update the self-learning function for the next recursion number at the k-th recursion according to the following formula (3): ， in, For the self-learning function in the next recursive iteration, This represents the historical recursive minimum value. The first result unit is used to obtain the output coefficient with the smallest output coefficient across all recursion counts, up to the preset maximum recursion count. The first recognition result of the signature information data to be identified; The classification information module, connected to the recognition module, is used to obtain first classification information of the invoice based on the first recognition result. The first classification information is used to classify the invoice.

7. The apparatus according to claim 6, characterized in that, The apparatus further includes an invoice classifier, the invoice classifier comprising: The classification execution module, connected to the classification information module, is used to classify invoices according to the first classification information.

8. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the invoice classification method as described in any one of claims 1-5.

Images