Bill identification result checking method and device, computer device and storage medium

Abstract

The application relates to a bill identification result verification method and device, equipment and a storage medium, and relates to the technical field of image information digitization. The method comprises the following steps: acquiring an image identification result; the image identification result comprises information obtained by performing image identification on a bill by using an image identification model; performing identification accuracy verification on the image identification result based on a preset accuracy verification rule to obtain an accuracy verification result; the accuracy verification result is used to indicate whether the image identification result passes the verification of the accuracy verification rule or whether the image identification result fails to pass the verification of the accuracy verification rule; and determining a verification confidence of the image identification result based on the accuracy verification result. Through the above method, the accuracy of the image identification result can be automatically verified and the accuracy of the image identification result can be given as a reference, the process of manually determining the accuracy of the image identification result is reduced, and the verification efficiency and accuracy of the accuracy of the image identification result are improved.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of image information digitization, and in particular to a bill recognition result verification method and device, a computer device, and a storage medium. BACKGROUND

[0002] With the development of artificial intelligence technology, OCR (Optical Character Recognition) technology is applied in various fields. In a scene where bill checking is required, the uploaded bill image can be recognized by the OCR technology to obtain relevant information in the bill image.

[0003] In related technologies, the recognition accuracy of an OCR model is usually improved by training the OCR model a large number of times, and the recognition confidence corresponding to the recognition result of the OCR model is determined as the confidence of the recognition result.

[0004] However, due to different training methods or training sample quantities of the OCR model, the recognition ability of the OCR model is uneven, thereby affecting the credibility of the recognition confidence provided by the OCR model, and further affecting the credibility of the recognition result output by the OCR model. Further manual credibility judgment of the recognition result of the OCR model is required, and due to the complexity of the bill information, the accuracy and efficiency of the manual recognition result judgment are low. SUMMARY

[0005] Embodiments of the present application provide a bill recognition result verification method and device, a computer device, and a storage medium, which can automatically verify the accuracy of an image recognition result and provide a reference for the accuracy of the image recognition result, reducing the process of manually determining the accuracy of the image recognition result, and improving the verification efficiency and accuracy of the accuracy of the image recognition result. The technical solution is as follows.

[0006] On the one hand, a bill recognition result verification method is provided, and the method comprises:

[0007] obtaining an image recognition result; the image recognition result comprises information obtained by image recognition of a bill by an image recognition model;

[0008] performing accuracy verification on the image recognition result based on a preset accuracy verification rule to obtain an accuracy verification result; the accuracy verification result is used to indicate that the image recognition result passes the verification of the accuracy verification rule or that the image recognition result does not pass the verification of the accuracy verification rule;

[0009] determining a verification confidence of the image recognition result based on the accuracy verification result.

[0010] In another aspect, a bill identification result verification device is provided, the device comprising:

[0011] a result obtaining module configured to obtain an image identification result, the image identification result comprising information obtained by performing image identification on a bill by using an image identification model;

[0012] a verification module configured to perform identification accuracy verification on the image identification result based on a preset accuracy verification rule to obtain an accuracy verification result, the accuracy verification result being configured to indicate whether the image identification result passes the accuracy verification rule or fails the accuracy verification rule;

[0013] a confidence degree determining module configured to determine a verification confidence degree of the image identification result based on the accuracy verification result.

[0014] In a possible implementation, the image identification result comprises at least two bill fields, the bill fields containing numerical information and text information;

[0015] the accuracy verification rule comprises a numerical verification rule and a text verification rule, the numerical verification rule being configured to indicate a correlation relationship between numerical information of each bill field in a bill, and the text verification rule being configured to indicate a corresponding relationship between standardized text information and text information of a bill field.

[0016] In a possible implementation, the verification module comprises:

[0017] a numerical verification submodule configured to perform accuracy verification on numerical information in each bill field based on the numerical verification rule in the accuracy verification rule;

[0018] a text verification submodule configured to perform accuracy verification on text information in each bill field based on the text verification rule in the accuracy verification rule.

[0019] In a possible implementation, the numerical verification submodule comprises:

[0020] a field determining unit configured to determine a verification field and a calculation field group in each bill field based on the numerical verification rule, the calculation field group comprising at least two bill fields;

[0021] a numerical calculation unit configured to perform calculation on each numerical information in the calculation field group based on the correlation relationship between bill fields in a bill indicated by the numerical verification rule to obtain a calculation result;

[0022] The rule matching unit is configured to determine that the numerical information in each of the bill fields passes the check of the numerical check rule if the calculation result matches the numerical information of the verification field, and determine that the numerical information in each of the bill fields does not pass the check of the numerical check rule if the calculation result does not match the numerical information of the verification field.

[0023] In a possible implementation, the numerical check rule has a numerical rule score, and the text check rule has a text rule score.

[0024] The confidence determination module is configured to,

[0025] obtain an identification confidence corresponding to each of the bill fields output by the image recognition model;

[0026] adjust the identification confidence of each of the bill fields based on a numerical rule score corresponding to the numerical check rule of each numerical information and a text rule score corresponding to the text check rule of each text information, to obtain a check confidence of each of the bill fields.

[0027] The confidence determination module is configured to,

[0028] obtain a number of numerical check formulas included in the numerical check rule, the numerical check formulas being used to indicate an association relationship between the numerical information of each of the bill fields in the bill;

[0029] when the number of the numerical check formulas included in the numerical check rule is greater than a number threshold, adjust the identification confidence of each of the fields according to a first adjustment rule to obtain the check confidence of each of the bill fields, the first adjustment rule being provided with a first numerical rule score of the numerical check rule and a first text rule score of the text check rule;

[0030] when the number of the numerical check formulas included in the numerical check rule is less than or equal to the number threshold, adjust the identification confidence of each of the fields according to a second adjustment rule to obtain the check confidence of each of the bill fields, the second adjustment rule being provided with a second numerical rule score of the numerical check rule and a second text rule score of the text check rule.

[0031] In a possible implementation, the apparatus further includes:

[0032] The labeling module is configured to, when it is determined that the image recognition result includes a target bill field with a check confidence less than a target confidence threshold, label the bill field in the image recognition result in a target form.

[0033] In another aspect, a computer device is provided, which includes a processor and a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to implement the above-mentioned bill identification result verification method.

[0034] In another aspect, a computer readable storage medium is provided, which stores at least one computer program, the computer program being loaded and executed by a processor to implement the above-mentioned bill identification result verification method.

[0035] In another aspect, a computer program product is provided, which includes at least one computer program, the computer program being loaded and executed by a processor to implement the bill identification result verification method provided in various optional implementation manners.

[0036] The technical solutions provided in the present application can include the following beneficial effects:

[0037] The bill identification result verification method provided in the embodiments of the present application, after obtaining the image identification result of the bill obtained by the image recognition model after image recognition, uses the preset accuracy verification rule to verify the accuracy of the image identification result, and determines the verification confidence of the image identification result according to the accuracy verification result indicating whether the image identification result passes or fails the accuracy verification rule verification; through the above method, the accuracy of the image identification result can be automatically verified and the accuracy of the image identification result can be given as a reference after obtaining the image identification result, reducing the process of manually determining the accuracy of the image identification result, and improving the verification efficiency and accuracy of the accuracy of the image identification result.

[0038] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0039] The accompanying drawings incorporated in the specification and forming a part thereof illustrate embodiments consistent with the present application and together with the description are used to explain the principles of the application.

[0040] Figure 1 An application schematic diagram of applying the OCR technology to the medical claim settlement scene in the prior art is shown;

[0041] Figure 2 An application schematic diagram of the bill identification method provided in an exemplary embodiment of the present application in the medical claim settlement scene is shown;

[0042] Figure 3 A flowchart of the bill identification result verification method provided in an exemplary embodiment of the present application is shown;

[0043] Figure 4 A flow chart of a bill recognition result verification method provided by an example embodiment of the present application is shown;

[0044] Figure 5 A schematic diagram of a rule configuration interface provided by an example embodiment of the present application is shown;

[0045] Figure 6 A block diagram of a bill recognition result verification device provided by an example embodiment of the present application is shown;

[0046] Figure 7 is a structural block diagram of a computer device according to an example embodiment;

[0047] Figure 8 is a structural block diagram of a computer device according to an example embodiment. DETAILED DESCRIPTION

[0048] The example embodiments will be described in detail herein with reference to the drawings. When the following description refers to accompanying drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following example embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0049] First, the terms involved in the present application are explained:

[0050] OCR (Optical Character Recognition): refers to the process in which an electronic device (such as a scanner or digital camera) checks characters printed on paper and then translates the shapes into computer text using character recognition methods, i.e., the process of scanning text materials and then analyzing and processing image files to obtain text and layout information.

[0051] In some business scenarios, image recognition can be performed using OCR technology to improve business processing efficiency. The business scenarios can include medical insurance claim scenarios and account checking scenarios. Taking the medical insurance claim scenario as an example, Figure 1 An application schematic diagram of applying OCR technology to the medical claim scenario in the prior art is shown. As shown in Figure 1As shown, the process of the OCR recognition service applied to the medical claim system is implemented as follows: the claim business system initiates an OCR recognition request, calls an OCR recognition model to perform image recognition on the received medical bill, obtains OCR recognition data, returns the OCR recognition data to the claim business system, and the corresponding claim business system receives the OCR recognition data, and subsequently, a human being judges the accuracy of the OCR recognition data. In the above process, since the recognition accuracy of different OCR recognition models does not reach a unified accuracy level, the human being needs to judge and confirm the recognition result returned by the OCR recognition model to determine the accuracy of the OCR recognition result, thereby improving the accuracy of business processing. However, the efficiency of the human being in judging the recognition result is low, and the probability of human judgment error is high, and the accuracy is poor.

[0052] Based on this, the embodiment of the present application provides a bill recognition result verification method for automatically verifying the accuracy of the image recognition result obtained after image recognition of the bill, so as to reduce the verification operation required by the human being and improve the verification efficiency and accuracy. Figure 2 An application schematic diagram of the bill recognition method provided by an example embodiment of the present application in a medical claim scene is shown in FIG. 1. Figure 2 As shown in FIG. 1, in the medical claim scene, a bill recognition method provided by an example embodiment of the present application is applied. Figure 1 Based on the prior art shown in FIG. 1, after the claim business system receives the OCR recognition data, a step 210 of performing accuracy verification on the OCR recognition data by an accuracy verification rule is added to obtain the credibility of the image recognition result and feed back to the relevant personnel, so that the relevant personnel can manually review based on the result of the rule verification.

[0053] Figure 3 A flowchart of a bill recognition result verification method provided by an example embodiment of the present application is shown in FIG. 2. The method can be executed by a computer device, which can be implemented as a server or a terminal. Figure 3 As shown in FIG. 2, the bill recognition result verification method can include the following steps:

[0054] Step 310: Obtain an image recognition result. The image recognition result includes information obtained by performing image recognition on a bill by an image recognition model.

[0055] The image recognition model can be implemented as an OCR recognition model. The bill can be a bill in a corresponding business scenario. For example, in a medical claim scenario, the bill can be a medical bill. In an account checking scenario, the bill can be a purchase bill, and so on. Different business scenarios can have corresponding bill types, which are not limited by the present application.

[0056] At step 320, the accuracy verification rule is preset, and the accuracy of the image recognition result is verified based on the accuracy verification rule to obtain an accuracy verification result. The accuracy verification result is used to indicate whether the image recognition result passes the accuracy verification rule or whether the image recognition result fails to pass the accuracy verification rule.

[0057] The accuracy verification rule can be preconfigured based on the bill type and the bill content, and is used to verify whether the image recognition result conforms to the bill content under the bill type, such as an operation rule, a text record rule, and the like. Different business scenarios can have different accuracy verification rule configurations, and the same business scenario can also correspond to different verification requirements and different accuracy verification rule configurations. The present application does not limit this.

[0058] At step 330, the verification confidence of the image recognition result is determined based on the accuracy verification result.

[0059] The verification confidence is used to indicate the credibility of the image recognition result, so that relevant personnel can directly determine whether manual verification of the image recognition result is needed based on the verification confidence, which can reduce the process of manually verifying the content of the image recognition result, thereby improving the verification efficiency.

[0060] In summary, the bill recognition result verification method provided by the embodiments of the present application, after obtaining the image recognition result of the bill obtained by the image recognition model, the accuracy of the image recognition result is verified by using the preset accuracy verification rule, and the verification confidence of the image recognition result is determined according to the accuracy verification result indicating whether the image recognition result passes or fails the accuracy verification rule. Through the above method, after obtaining the image recognition result, the accuracy of the image recognition result can be automatically verified and the accuracy of the image recognition result can be given as a reference, which reduces the process of manually determining the accuracy of the image recognition result, and improves the verification efficiency and accuracy of the accuracy of the image recognition result.

[0061] In the embodiment of the present application, the image recognition result of the bill can include at least two bill fields, each of which contains numerical information and text information. Correspondingly, when setting the accuracy checking rule, the accuracy checking rule can include a numerical checking rule and a text checking rule. The numerical checking rule indicates the correlation between the bill fields, and the text checking rule indicates the correspondence between the standardized text information and the text information of the bill field. The correlation between the numerical information of the bill fields can include an operation relationship, a comparison relationship, etc. The standardized text information can be general text information corresponding to a business scenario. For example, in a medical claim scenario, the standardized text information can be extracted from a medical experience document or a claim experience document, and is used to standardize the text content in the corresponding business scenario. In this case, Figure 4 A flowchart of a bill recognition result checking method provided by an exemplary embodiment of the present application is shown. The method can be executed by a computer device, which can be implemented as a server or a terminal. As shown in Figure 4 The bill recognition result checking method can include the following steps:

[0062] Step 410: Obtain an image recognition result. The image recognition result includes information obtained by image recognition of a bill through an image recognition model.

[0063] Step 420: Perform accuracy checking on the numerical information in each bill field based on the numerical checking rule in the accuracy checking rule.

[0064] Optionally, the process of performing accuracy checking on the numerical information in the image recognition result can be implemented as follows:

[0065] Determine the verification field and the calculation field group in each bill field based on the numerical checking rule. The calculation field group includes at least two bill fields.

[0066] Calculate each numerical information in the field group based on the correlation between the bill fields indicated by the numerical checking rule, and obtain a calculation result.

[0067] If the calculation result matches the numerical information of the verification field, it is determined that the numerical information in each bill field passes the numerical checking rule. If the calculation result does not match the numerical information of the verification field, it is determined that the numerical information in each bill field does not pass the numerical checking rule.

[0068] In the embodiment of the present application, the relevant personnel can configure the accuracy checking rule based on actual needs, and the computer device receives the rule configuration operation, and generates the accuracy checking rule based on the rule configuration operation. Optionally, the rule configuration interface can be provided in the computer device, and the rule configuration operation performed by the relevant personnel in the rule configuration interface is received. Taking the configuration of the numerical value checking rule as an example, the verification field and the calculation field group can be indicated in the numerical value checking rule, and taking the medical claim scenario as an example, Figure 5 The schematic diagram of the rule configuration interface provided by an example embodiment of the present application is shown as follows, Figure 5 In the rule configuration interface, the association relationship of each ticket field in the calculation field group and the verification field (i.e., the selection result value as shown in Figure 5 ) can be configured. The association relationship of each ticket field in the calculation field group can be represented by a numerical value checking formula, such as Figure 5 The numerical value checking formula is “total amount of money = personal account payment + personal cash payment + medical insurance pooling fund payment + additional fund payment”, and the verification field is “total expense amount”. The numerical value of the total amount of money obtained according to the numerical value checking formula is compared with the numerical value of the total expense amount to determine the accuracy of the numerical value information of each ticket field.

[0069] When setting the numerical value checking rule, multiple numerical value checking formulas can be set, so that when identifying the accuracy of the numerical value information in each ticket field contained in the image recognition result based on the numerical value checking rule, the numerical value information in each ticket field can be checked by multiple different numerical value checking formulas at the same time. In this case, the computer device can determine multiple verification fields and multiple calculation field groups from each ticket field based on different numerical value checking formulas, that is, corresponding to different numerical value checking formulas, the same ticket field can be a verification field and a field in the calculation field group, so as to perform different numerical value checking on the numerical value information in each ticket field. For example, the numerical value checking rule contains the numerical value checking formula with the verification field “total expense amount” as shown in Figure 5 , and also contains the numerical value checking formula with the verification field “reimbursement ratio value”, which is “reimbursement ratio = pooling fund payment amount / (total amount of money - personal account payment - personal cash payment - other payment) * 100%”. It should be noted that the above numerical value checking formula is only illustrative, and different numerical value checking formulas can be set based on different ticket content and different checking needs, which are not limited in the present application.

[0070] Optionally, the rule configuration interface can further be configured to configure a bill type corresponding to the accuracy checking rule, the bill type being used to indicate a bill type to which the accuracy checking rule is applicable; illustratively, the bill type can include outpatient and inpatient, and different accuracy checking rules can be configured for different bill types during the accuracy checking rule configuration, so that the accuracy checking rule is more adaptable to the corresponding application scenario; based on this, before the image recognition result is checked for accuracy based on the preset accuracy checking rule, the method further includes:

[0071] obtaining a bill type of the bill;

[0072] obtaining a preset accuracy checking rule corresponding to the bill type.

[0073] Optionally, the rule configuration interface can be configured to configure the validity time and the invalidity time of the accuracy checking rule, so as to limit the effective time range of the accuracy checking rule; based on this, before the image recognition result is checked based on the accuracy checking rule, it is further needed to determine whether the accuracy checking rule is invalid, that is, to determine whether the current time is within the effective time range indicated by the validity time and the invalidity time corresponding to the accuracy checking rule, if the current time is within the effective time range, it indicates that the accuracy checking rule is valid, and the accuracy checking rule can be used for checking, if the current time is not within the effective time range, it indicates that the accuracy checking rule has been invalidated, and the accuracy checking rule cannot be used for checking; optionally, after it is determined that a certain accuracy checking rule has been invalidated, the computer device can delete the invalidated accuracy checking rule.

[0074] In step 430, the text information in each bill field is checked for accuracy based on the text checking rule in the accuracy checking rule.

[0075] Due to different naming ways for the same consumption item, the bill text messages of the same consumption item on different bills are different; for example, in the medical claim scenario, due to the naming differences of medical expenses in different places, the naming text of the medical bill amount item may be different, for example, some places call the registration fee as the diagnosis fee, and the two are actually the same consumption item, therefore, when the computer device checks the image recognition result, the accuracy of the text information in each bill field can be checked based on the standardized text information and the corresponding bill text information, wherein, the text information in the bill field matches the standardized text information in the text checking rule, or the bill text information corresponding to the standardized text information in the text checking rule, it is confirmed that the text information in the bill field passes the text checking rule; when the text information in the bill field does not match the standardized text information in the text checking rule and the bill text information corresponding to the standardized text information, it is confirmed that the text information in the bill field does not pass the text checking rule. Illustratively, if the text checking rule indicates that "diagnosis fee" and "registration fee" have a corresponding relationship, the text information contained in the bill field is "diagnosis fee", which does not match "diagnosis fee" and "registration fee", indicating that the text information in the bill field is incorrectly identified.

[0076] In the embodiments of the present application, the text checking rule can be divided into business experience checking rule and claim experience checking rule, wherein the standardized text in the business experience checking rule is the general text in the corresponding business scenario, and the standardized text in the claim experience checking rule is the general text in the corresponding claim scenario.

[0077] Optionally, when it is confirmed that the text information in the bill field is accurately identified, and the text information in the bill field matches the bill text information corresponding to the standardized text information, in order to facilitate subsequent statistical processing, the computer device can replace the text information in the bill field with the standardized text information corresponding to the bill text information.

[0078] Step 440, obtaining the identification confidence corresponding to each bill field output by the image recognition model.

[0079] The identification confidence is the probability value corresponding to each bill field in the image recognition result when the image recognition model identifies the bill and outputs the image recognition result, for example, when the image recognition model outputs "personal cash payment of 90 yuan", the probability value corresponding to "personal cash payment of 90 yuan" is 95%, indicating that the content on the bill has a 95% probability of containing "personal cash payment of 90 yuan", the above probability value can be regarded as the identification confidence of the bill field.

[0080] Step 450, based on the numerical rule score corresponding to the numerical check rule of each numerical information and the text rule score corresponding to the text check rule of each text information, the recognition confidence of each bill field is adjusted to obtain the check confidence of each bill field.

[0081] In the embodiment of the present application, the numerical check rule has a numerical rule score, and the text check rule has a text rule score. Among them, the business experience rule and the claim experience rule in the text check rule can correspond to different text rule scores; the rule score corresponding to each rule can be set by relevant personnel based on actual needs, and the present application does not limit this.

[0082] In the embodiment of the present application, the computer device can be provided with different adjustment rules based on the number of numerical check formulas contained in the numerical check rule. In an optional manner,

[0083] The number of numerical check formulas contained in the numerical check rule is obtained, and the numerical check formula is used to indicate the association relationship between the numerical information of each bill field in the bill;

[0084] When the number of numerical check formulas contained in the numerical check rule is greater than the number threshold, the recognition confidence of each field is adjusted according to the first adjustment rule to obtain the check confidence of each bill field; the first adjustment rule is provided with the first numerical rule score of the numerical check rule and the first text rule score of the text check rule;

[0085] When the number of numerical check formulas contained in the numerical check rule is less than or equal to the number threshold, the recognition confidence of each field is adjusted according to the second adjustment rule to obtain the check confidence of each bill field; the second adjustment rule is provided with the second numerical rule score of the numerical check rule and the second text rule score of the text check rule.

[0086] The first adjustment rule and the second adjustment rule are different, which is reflected in the different settings of the first numerical rule score and the first text rule score contained in the first adjustment rule and the second numerical score and the second text rule score contained in the second adjustment rule. The present application does not limit the specific settings of each rule score in the first adjustment rule and the second adjustment rule.

[0087] Further, in each adjustment rule (including the first adjustment rule and the second adjustment rule), different adjustment sub-rules can be set based on the recognition confidence of the bill field. For example, when the recognition confidence of a first bill field is less than or equal to a recognition confidence threshold, the recognition confidence of the first bill field is adjusted according to a first sub-adjustment rule; when the recognition confidence of the first bill field is greater than the recognition confidence threshold, the recognition confidence of the first bill field is adjusted according to a second sub-adjustment rule. The first bill field is any one of the bill fields. In different adjustment sub-rules, the rule scores corresponding to each adjustment sub-rule are different. In different adjustment rules, the rule scores corresponding to each adjustment sub-rule can also be different. The rule scores corresponding to each adjustment sub-rule include rule scores of different accuracy verification rules in each adjustment sub-rule, that is, the rule scores of the same accuracy verification rule in different adjustment sub-rules can be different, and the rule scores of each accuracy verification rule can also be different.

[0088] Taking adjustment of the recognition confidence of information in the bill as an example, Table 1 shows a first adjustment rule according to an example embodiment of the present application.

[0089] Table 1

[0090]

[0091] As shown in Table 1, if the number threshold is 2, the first adjustment rule is the adjustment rule corresponding to the number of the numerical value verification formula in the numerical value verification rule being greater than 2; when the recognition confidence threshold is 0, the first sub-adjustment rule corresponds to the recognition confidence being 0, and the second sub-adjustment rule corresponds to the recognition confidence not being 0, and under each sub-rule, different accuracy verification rules correspond to different rule scores; based on the different sources of the text verification rules, the text verification rules are divided into two categories, namely business experience rules and claim experience rules, and the rule scores corresponding to different types of text verification rules can also be different. The adjustment manner of the first sub-adjustment rule to the recognition confidence is different from the adjustment manner of the second sub-adjustment rule to the recognition confidence. For example, the adjustment manner shown in Table 1, the first sub-adjustment rule obtains a verification confidence by adding rule scores to the recognition confidence; the second sub-adjustment rule obtains a verification confidence by weighting the recognition confidence based on the rule score.

[0092] It should be noted that the rule score of each check rule shown in Table 1 is the rule score that can be obtained when the information in the bill field matches the corresponding check rule, and the score corresponding to the check rule is 0 when the information in the bill field does not match the corresponding check rule. The rule score corresponding to each check rule in Table 1 is only illustrative, and the present application does not limit this.

[0093] Table 2 shows a second adjustment rule shown in an example embodiment of the present application.

[0094] Table 2

[0095]

[0096] As shown in Table 2, when the number threshold is 2, the second adjustment rule is the adjustment rule corresponding to when the number of numerical value check formulas in the numerical value check rule is less than or equal to 2. Among them, when the recognition confidence threshold is 0, the first sub-adjustment rule corresponds to when the recognition confidence is 0, and the second sub-adjustment rule corresponds to when the recognition confidence is not 0. As can be known from the comparison of Table 1 and Table 2, the first sub-adjustment rule and the second sub-adjustment rule in the first adjustment rule are different from the first sub-adjustment rule and the second sub-adjustment rule in the second adjustment rule.

[0097] In a possible implementation, the method further includes:

[0098] When it is determined that the image recognition result contains a target bill field with a check confidence less than a target confidence threshold, the bill field in the image recognition result is labeled in a target form.

[0099] When the check confidence is less than the target confidence threshold, it indicates that the recognition result of the target bill field may be inaccurate, and the target bill field is displayed in a target form to remind relevant personnel. Among them, the target form can mean that the target bill field in the image recognition result is highlighted, displayed with a border, displayed in a target color, displayed in a target font size, displayed in a target font, etc. The above display forms can be used in combination, and the implementation of each display form can be set based on actual needs, and the present application does not limit this.

[0100] Taking the number of numerical value check formulas in the numerical value check rule being less than or equal to 2 as an example, i.e., adjusting the recognition confidence of each bill field in the image recognition result according to the second adjustment rule shown in Table 2, if the target confidence threshold is 0.6, then if the recognition confidence of a bill field is 0, even if it matches all check rules, the check confidence obtained by adjusting the recognition confidence based on the second adjustment rule is 0.5, which is lower than the target confidence threshold, and the bill field is labeled in a target form to prompt relevant personnel to review.

[0101] In summary, the bill recognition result verification method provided by the embodiments of the present application, after obtaining the image recognition result of the bill obtained by the image recognition model, uses the preset accuracy verification rule to verify the accuracy of the image recognition result, and determines the verification confidence of the image recognition result according to the accuracy verification result indicating whether the image recognition result passes or fails the accuracy verification rule verification; through the above method, the accuracy of the image recognition result can be automatically verified and the reference of the accuracy of the image recognition result can be given after obtaining the image recognition result, reducing the process of manually determining the accuracy of the image recognition result, and improving the verification efficiency and accuracy of the accuracy of the image recognition result.

[0102] When verifying the image recognition result, the numerical value information is verified by the numerical value verification rule, the text information is verified by the text verification rule, and the recognition confidence of each bill field in the image recognition result is adjusted based on the score corresponding to the numerical value verification rule and the score corresponding to the text verification rule, thereby improving the verification accuracy of the image recognition result.

[0103] Figure 6 A block diagram of a bill recognition result verification device provided by an exemplary embodiment of the present application is shown, which can execute to implement all or part of the steps of the embodiments shown in Figure 3 or Figure 4 The bill recognition result verification device can include: Figure 6

[0104] The result obtaining module 610 is configured to obtain an image recognition result; the image recognition result includes information obtained by image recognition of a bill by an image recognition model;

[0105] The verification module 620 is configured to verify the accuracy of the image recognition result based on a preset accuracy verification rule to obtain an accuracy verification result; the accuracy verification result is used to indicate whether the image recognition result passes the verification of the accuracy verification rule or the image recognition result fails the verification of the accuracy verification rule;

[0106] The confidence determination module 630 is configured to determine the verification confidence of the image recognition result based on the accuracy verification result.

[0107] In a possible implementation, the image recognition result includes at least two bill fields, and the bill field contains numerical value information and text information;

[0108] ​The accuracy checking rule includes a numerical value checking rule and a text checking rule; the numerical value checking rule indicates a correlation relationship between numerical value information of each ticket field in the ticket, and the text checking rule is used for indicating a corresponding relationship between standardized text information and text information of the ticket field.

[0109] In a possible implementation, the checking module 620 includes:

[0110] A numerical value checking submodule is configured to perform accuracy checking on numerical value information in each ticket field based on the numerical value checking rule in the accuracy checking rule.

[0111] A text checking submodule is configured to perform accuracy checking on text information in each ticket field based on the text checking rule in the accuracy checking rule.

[0112] In a possible implementation, the numerical value checking submodule includes:

[0113] A field determining unit is configured to determine, based on the numerical value checking rule, a verification field and a calculation field group in each ticket field; the calculation field group includes at least two ticket fields.

[0114] A numerical value calculating unit is configured to calculate each numerical value information in the calculation field group based on a correlation relationship between each ticket field in the ticket indicated by the numerical value checking rule, to obtain a calculation result.

[0115] A rule matching unit is configured to determine, if the calculation result matches numerical value information of the verification field, that the numerical value information in each ticket field passes the checking of the numerical value checking rule; and determine, if the calculation result does not match the numerical value information of the verification field, that the numerical value information in each ticket field does not pass the checking of the numerical value checking rule.

[0116] In a possible implementation, the numerical value checking rule has a numerical value rule score, and the text checking rule has a text rule score.

[0117] The confidence degree determining module 630 is configured to,

[0118] obtain an identification confidence degree corresponding to each ticket field output by the image recognition model;

[0119] adjust the identification confidence degree of each ticket field based on a numerical value rule score corresponding to the numerical value checking rule of each numerical value information and a text rule score corresponding to the text checking rule of each text information, to obtain a checking confidence degree of each ticket field.

[0120] The confidence determination module 630 is configured to,

[0121] obtain a number of numerical verification formulas included in the numerical verification rule, the numerical verification formulas being used to indicate an association relationship between numerical information of each bill field in the bill;

[0122] when the number of the numerical verification formulas included in the numerical verification rule is greater than a number threshold, adjusting the recognition confidence of each field according to a first adjustment rule to obtain a verification confidence of each bill field; the first adjustment rule is provided with a first numerical rule score of the numerical verification rule and a first text rule score of the text verification rule;

[0123] when the number of the numerical verification formulas included in the numerical verification rule is less than or equal to the number threshold, adjusting the recognition confidence of each field according to a second adjustment rule to obtain a verification confidence of each bill field; the second adjustment rule is provided with a second numerical rule score of the numerical verification rule and a second text rule score of the text verification rule.

[0124] In a possible implementation, the apparatus further includes:

[0125] The labeling module is configured to, when it is determined that the image recognition result includes a target bill field with a verification confidence less than a target confidence threshold, label the bill field in the image recognition result in a target form.

[0126] To sum up, the bill recognition result verification apparatus provided by the embodiments of the present application, after obtaining the image recognition result of the bill obtained by the image recognition model, uses the preset accuracy verification rule to verify the accuracy of the image recognition result, and determines the verification confidence of the image recognition result according to the accuracy verification result indicating whether the image recognition result passes or fails the accuracy verification rule. Through the above method, after obtaining the image recognition result, the accuracy of the image recognition result can be automatically verified and the accuracy of the image recognition result can be given as a reference, reducing the process of manually determining the accuracy of the image recognition result, and improving the verification efficiency and accuracy of the accuracy of the image recognition result.

[0127] Figure 7A structural block diagram of a computer device 700 is shown according to an example embodiment of the present application. The computer device can be implemented as a server in the above-mentioned solutions of the present application. The computer device 700 includes a central processing unit (CPU) 701, a system memory 704 including a random access memory (RAM) 702 and a read-only memory (ROM) 703, and a system bus 705 connecting the system memory 704 and the central processing unit 701. The computer device 700 also includes a mass storage device 706 for storing an operating system 709, application programs 710, and other program modules 711.

[0128] Without loss of generality, the computer readable medium can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes RAM, ROM, erasable programmable read-only memory (EPROM), electrically-erasable programmable read-only memory (EEPROM) flash or other solid state memory technology, CD-ROM, digital versatile disks (DVD), or other optical storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices. Of course, those skilled in the art will recognize that the computer storage media described above can not be limited to the foregoing examples. The system memory 704 and the mass storage device 706 described above can be collectively referred to as memory.

[0129] According to various embodiments of the present disclosure, the computer device 700 can also operate in connection with a remote computer through a network such as the Internet. That is, the computer device 700 can be connected to a network 708 through a network interface unit 707 connected to the system bus 705, or can be connected to other types of networks or remote computer systems (not shown) using the network interface unit 707.

[0130] The memory further comprises at least one instruction, at least one program, a code set or an instruction set stored in the memory, and the central processor 701 implements all or part of the steps of the bill identification result verification method shown in each of the embodiments by executing the at least one instruction, at least one program, code set or instruction set.

[0131] Figure 8 A structural block diagram of a computer device 800 shown in an example embodiment of the present application is shown. The computer device 800 can be implemented as the backend service node described above, such as a smart phone, a tablet computer, a notebook computer, a desktop computer, etc. The computer device 800 can also be referred to as a user device, a portable terminal, a laptop terminal, a desktop terminal, and other names.

[0132] Generally, the computer device 800 comprises a processor 801 and a memory 802.

[0133] The processor 801 can comprise one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 801 can be implemented in at least one of a hardware form of a DSP (Digital Signal Processing), a FPGA (Field-Programmable Gate Array), a PLA (Programmable Logic Array). The processor 801 can also comprise a main processor and a coprocessor, the main processor being a processor for processing data in an awake state, also referred to as a CPU (Central Processing Unit), and the coprocessor being a low-power processor for processing data in a standby state. In some embodiments, the processor 801 can be integrated with a GPU (Graphics Processing Unit) for rendering and drawing content required to be displayed by a display screen. In some embodiments, the processor 801 can further comprise an AI (Artificial Intelligence) processor for processing computing operations related to machine learning.

[0134] The memory 802 may include one or more computer-readable storage media, which may be non-transitory. The memory 802 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 802 is used to store at least one instruction, which is executed by the processor 801 to implement all or part of the steps in the ticket recognition result verification method shown in the method embodiments of this application.

[0135] In some embodiments, the computer device 800 may also optionally include a peripheral device interface 803 and at least one peripheral device. The processor 801, memory 802, and peripheral device interface 803 can be connected via a bus or signal line. Each peripheral device can be connected to the peripheral device interface 803 via a bus, signal line, or circuit board. Specifically, the peripheral device includes at least one of the following: a radio frequency circuit 804, a display screen 805, a camera assembly 806, an audio circuit 807, and a power supply 808.

[0136] In some embodiments, the computer device 800 further includes one or more sensors 809. The one or more sensors 809 include, but are not limited to, an accelerometer 810, a gyroscope 811, a pressure sensor 812, an optical sensor 813, and a proximity sensor 814.

[0137] Those skilled in the art will understand that Figure 8 The structure shown does not constitute a limitation on the computer device 800, and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0138] In one exemplary embodiment, a computer-readable storage medium is also provided, which stores at least one computer program that is loaded and executed by a processor to implement all or part of the steps in the above-described document recognition result verification method. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, or optical data storage device, etc.

[0139] In one exemplary embodiment, a computer program product is also provided, comprising at least one computer program that is loaded and executed by a processor. Figure 3 or Figure 4All or part of the steps of the bill identification result verification method shown in any embodiment.

[0140] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the application being indicated by the following claims.

[0141] It is to be understood that the application is not limited to the precise construction herein described and as shown in the attached drawings, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application. The scope of the application is to be indicated by the appended claims, rather than the description and examples.

Claims

1. A method of verifying a result of a ticket identification, characterized by, The method comprises: obtaining an image recognition result; the image recognition result comprises information obtained by image recognition of a bill by an image recognition model; performing accuracy verification on the image recognition result based on a preset accuracy verification rule to obtain an accuracy verification result; the accuracy verification result is used to indicate that the image recognition result passes the verification of the accuracy verification rule or that the image recognition result fails to pass the verification of the accuracy verification rule; determining a verification confidence of the image recognition result based on the accuracy verification result; the image recognition result comprises at least two bill fields, and the bill field contains numerical information and text information; the accuracy verification rule comprises a numerical verification rule and a text verification rule; the numerical verification rule indicates the correlation between the numerical information of each bill field in the bill, and the text verification rule is used to indicate the corresponding relationship between the standardized text information and the text information of the bill field; the numerical verification rule has a numerical rule score, and the text verification rule has a text rule score; determining the verification confidence of the image recognition result based on the accuracy verification result comprises: obtaining an identification confidence corresponding to each bill field output by the image recognition model; adjusting the identification confidence of each bill field based on the numerical rule score corresponding to the numerical verification rule of each numerical information and the text rule score corresponding to the text verification rule of each text information to obtain the verification confidence of each bill field, comprising: obtaining the number of numerical verification formulas contained in the numerical verification rule, the numerical verification formula being used to indicate the correlation between the numerical information of each bill field in the bill; when the number of numerical verification formulas contained in the numerical verification rule is greater than a number threshold, adjusting the identification confidence of each field according to a first adjustment rule to obtain the verification confidence of each bill field; the first adjustment rule specifies a first numerical rule score of the numerical verification rule and a first text rule score of the text verification rule; when the number of numerical verification formulas contained in the numerical verification rule is less than or equal to the number threshold, adjusting the identification confidence of each field according to a second adjustment rule to obtain the verification confidence of each bill field; the second adjustment rule specifies a second numerical rule score of the numerical verification rule and a second text rule score of the text verification rule.

2. The method of claim 1, wherein, performing accuracy verification on the image recognition result based on the accuracy verification rule comprises: performing accuracy verification on the numerical information in each bill field based on the numerical verification rule in the accuracy verification rule; performing accuracy verification on the text information in each bill field based on the text verification rule in the accuracy verification rule.

3. The method of claim 2, wherein, performing accuracy verification on the numerical information in the image recognition result based on the numerical verification rule in the accuracy verification rule comprises: determine a verification field in each of the bill fields and a calculation field group based on the numerical verification rule; the calculation field group contains at least two of the bill fields; calculate each numerical information in the calculation field group based on the association relationship between each bill field in the bill indicated by the numerical verification rule, to obtain a calculation result; if the calculation result matches the numerical information of the verification field, it is determined that the numerical information in each of the bill fields passes the verification of the numerical verification rule; if the calculation result does not match the numerical information of the verification field, it is determined that the numerical information in each of the bill fields does not pass the verification of the numerical verification rule.

4. The method of claim 1, wherein, The method further comprises: when it is determined that the image recognition result contains a target bill field with a verification confidence less than a target confidence threshold, labeling the bill field in the image recognition result in a target form.

5. A ticket identification result verification device characterized by comprising: The device comprises: a result obtaining module configured to obtain an image recognition result; the image recognition result comprises information obtained by image recognition of a bill by an image recognition model; a verification module configured to perform identification accuracy verification on the image recognition result based on a preset accuracy verification rule, to obtain an accuracy verification result; the accuracy verification result is used to indicate that the image recognition result passes the verification of the accuracy verification rule or that the image recognition result does not pass the verification of the accuracy verification rule; a confidence determining module configured to determine a verification confidence of the image recognition result based on the accuracy verification result; the image recognition result comprises at least two bill fields, and the bill field contains numerical information and text information; the accuracy verification rule comprises a numerical verification rule and a text verification rule; the numerical verification rule indicates the association relationship between the numerical information of each bill field in the bill, and the text verification rule is used to indicate the corresponding relationship between the standardized text information and the text information of the bill field; the numerical verification rule has a numerical rule score, and the text verification rule has a text rule score; the confidence determining module is configured to obtain an identification confidence corresponding to each of the bill fields output by the image recognition model; based on the numerical rule score corresponding to the numerical verification rule of each numerical information and the text rule score corresponding to the text verification rule of each text information, the identification confidence of each of the bill fields is adjusted to obtain the verification confidence of each of the bill fields; the confidence determining module is further configured to obtain the number of numerical verification formulas contained in the numerical verification rule; the numerical verification formula is used to indicate the association relationship between the numerical information of each bill field in the bill; when the number of numerical verification formulas contained in the numerical verification rule is greater than a number threshold, the identification confidence of each field is adjusted according to a first adjustment rule to obtain the verification confidence of each of the bill fields; the first adjustment rule specifies a first numerical rule score of the numerical verification rule and a first text rule score of the text verification rule. When the number of the numerical verification formulas included in the numerical verification rules is less than or equal to the number threshold, the identification confidence of each field is adjusted according to a second adjustment rule to obtain the verification confidence of each bill field; the second adjustment rule specifies a second numerical rule score of the numerical verification rule and a second text rule score of the text verification rule.

6. A computer device, comprising: The computer device includes a processor and a memory, and the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor to implement the bill identification result verification method as claimed in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer readable storage medium stores at least one computer program, and the computer program is loaded and executed by the processor to implement the bill identification result verification method as claimed in any one of claims 1 to 4.

Images