Error early warning method and device for transaction service, storage medium and electronic equipment

By classifying and identifying business transaction images and generating error warning results, the problem that transaction data error warning solutions in the existing technology cannot achieve intelligent judgment, and the accuracy and efficiency of error detection in trading business are improved.

CN119006923APending Publication Date: 2024-11-22中国邮政储蓄银行股份有限公司
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411169294.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

Existing transaction data error warning plans rely on artificially customized rules and cannot achieve intelligent judgment and early warning, especially when there are many types of data, they cannot adapt to all scenarios.

Method used

By obtaining business transaction images, entering a classification model to identify transaction types, and performing image occlusion processing on the image, error recognition is performed in combination with the target error recognition model, and error warning results are generated to characterize errors in the business process.

Benefits of technology

It realizes intelligent error warning for trading business processes, improves detection efficiency and classification accuracy, reduces the error rate of traditional data matching warnings, and adapts to multiple trading scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006923A_ABST
    Figure CN119006923A_ABST
Patent Text Reader

Abstract

The invention provides a transaction business error early warning method and device, a storage medium and electronic equipment. Comprising the following steps: acquiring a business transaction image, inputting the business transaction image into a classification model, identifying a business transaction type of the business transaction image, determining a position of to-be-identified information in the business transaction image according to the business transaction type, and performing image covering processing on an area except the position of the to-be-identified information in the business transaction image, according to the business transaction type, business error identification is carried out on the business image to be identified, an error identification result is generated, business error early warning is carried out according to the error identification result, and the error identification result is used for representing that business errors occur in the business handling process. The problem that in the prior art, intelligent judgment and early warning cannot be carried out through existing conditions due to the fact that early warning is carried out according to artificially customized rules in a transaction data error early warning scheme is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of error warning for transaction services, and in particular, to a method, device, storage medium and electronic device for error warning for transaction services. Background Art

[0002] With the rapid development of society and the ever-changing economy, banks, as the country's financial service center, bear great responsibilities. The various potential risks in bank transactions, especially teller transactions, are also increasing. If tellers do not exercise reasonable control when conducting counter transactions, many operational problems will arise, including transaction compliance, information integrity, etc., which will more or less cause losses and impacts. At the same time, the workload will be increased accordingly during the bank's back-end audit and review stage. Therefore, better early warning of operational errors at bank counters has become the top priority of counter business.

[0003] The existing transaction data error warning scheme only uses big data analysis to form an identification code through features, and performs analysis based on the matching degree of the identification code. This is a rule-based judgment. When there are many types of data, manually customized rules cannot meet all scenarios, and intelligent judgment and warning cannot be performed based on existing conditions. Summary of the invention

[0004] The main purpose of this application is to provide a transaction business error warning method, device, storage medium and electronic device, so as to at least solve the problem that the prior art transaction data error warning scheme is based on artificially customized rules for warning, resulting in the inability to make intelligent judgments and warnings based on existing conditions.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for error warning of transaction business is provided, comprising: obtaining a business transaction image, inputting the business transaction image into a classification model, and identifying the business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of the multiple sets of training data includes a historical business transaction image and the business transaction type corresponding to the historical business transaction image; according to the business transaction type, determining the position of the information to be identified in the business transaction image, performing image masking processing on the area outside the position of the information to be identified in the business transaction image, and obtaining a business image to be identified, wherein the information to be identified is information that needs to be detected for business errors in the business transaction image of the business transaction type; according to the business transaction type, performing business error identification on the business image to be identified, generating an error identification result, using the error identification result to perform business error warning, and the error identification result is used to characterize the occurrence of business errors in the business handling process.

[0006] Optionally, according to the business transaction type, business error identification is performed on the business image to be identified, including: according to the business transaction type, inputting the business image to be identified into a target error identification model corresponding to the business transaction type, so as to use the target error identification model to perform error identification processing on the information to be identified, and obtain the error identification result, wherein the target error identification model is obtained by machine learning through multiple groups of error identification training data, and each group of the multiple groups of error identification training data includes historical business images to be identified and the error identification results corresponding to the historical business images to be identified.

[0007] Optionally, the target error recognition model is used to perform recognition processing on the information to be recognized to obtain an error recognition result, including: using the target error recognition model to perform error prediction recognition on each piece of information to be recognized to obtain an error prediction score corresponding to each piece of recognition information; outputting the target recognition information whose error prediction score is greater than a score threshold, and sorting the target recognition information in descending order according to the error prediction score corresponding to the target recognition information to obtain an error prediction sequence; selecting the first N target recognition information in the error prediction sequence, and generating the error recognition result according to the error prediction scores corresponding to the first N target recognition information, wherein the error recognition result includes the error prediction ratio corresponding to the first N target recognition information, and the error prediction ratio is the probability that errors occur in the first N target recognition information.

[0008] Optionally, before acquiring the business transaction image, the method further includes: during the transaction processing, if the generation of the business transaction image is not detected, generating a prompt message, wherein the prompt message is used to remind the lack of the business transaction image for the business processing.

[0009] Optionally, before inputting the business transaction image into the classification model, the method further includes: performing a preprocessing operation on the business transaction image, wherein the preprocessing operation includes adjusting the image clarity.

[0010] Optionally, after obtaining the error prediction score corresponding to each identification information, the method further includes: determining whether the error prediction score corresponding to each identification information is greater than the score threshold; if it is determined that the error prediction scores corresponding to each identification information are all less than or equal to the score threshold, the error identification result is not generated.

[0011] Optionally, the classification model is a VGG16 model, and the target error recognition model is a ResNet+FPN network model.

[0012] According to another aspect of the present application, there is provided an error warning device for transaction business, comprising: a classification unit, for acquiring a business transaction image, inputting the business transaction image into a classification model, and identifying a business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling business, and the classification model is obtained by machine learning through multiple sets of training data sets, and each set of training data in the multiple sets of training data sets includes a historical business transaction image and the business transaction type corresponding to the historical business transaction image; a processing unit, for determining the position of information to be identified in the business transaction image according to the business transaction type, performing image masking processing on the area outside the position of the information to be identified in the business transaction image, and obtaining a business image to be identified, wherein the information to be identified is information that needs to be detected for business errors in the business transaction image of the business transaction type; a first generating unit, for performing business error identification on the business image to be identified according to the business transaction type, generating an error identification result, using the error identification result to perform business error warning, and the error identification result is used to characterize the occurrence of business errors in the business handling process.

[0013] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is run, the device where the computer-readable storage medium is located is controlled to execute any one of the transaction business error warning methods.

[0014] According to another aspect of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include an error warning method for executing any one of the transaction services described.

[0015] Applying the technical solution of the present application, a business transaction image is obtained, the business transaction image is input into a classification model, and the business transaction type of the business transaction image is identified, wherein the business transaction image is an image of a business transaction interface for handling business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of training data in the multiple sets of training data includes historical business transaction images and business transaction types corresponding to the historical business transaction images; according to the business transaction type, the position of the information to be identified in the business transaction image is determined, and the area outside the position of the information to be identified in the business transaction image is subjected to image masking processing to obtain the business image to be identified, wherein the information to be identified is information that requires business error detection for the business transaction image of the business transaction type; according to the business transaction type, business error identification is performed on the business image to be identified, and an error identification result is generated, and a business error warning is issued using the error identification result, and the error identification result is used to characterize the occurrence of business errors in the business handling process. By acquiring business transaction images, a classification model is used to classify the business transaction images according to their business types. The classified business transaction images are input into the corresponding recognition model to generate business images to be identified. Error identification is performed on the business images to be identified for use in business error warnings. This solves the problem in the existing technology that transaction data error warning schemes issue warnings based on artificially customized rules, resulting in the inability to make intelligent judgments and warnings based on existing conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings constituting part of the present application are used to provide a further understanding of the present application. The exemplary embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 A hardware structure block diagram of a mobile terminal for executing an error warning method for a transaction service provided in an embodiment of the present application is shown;

[0018] Figure 2 A schematic diagram of a process flow of an error warning method for a transaction service provided according to an embodiment of the present application is shown;

[0019] Figure 3 A schematic diagram of a process flow of an error warning method for a specific transaction business provided in accordance with an embodiment of the present application is shown;

[0020] Figure 4 A schematic diagram of the network structure of the VGG16 model provided according to an embodiment of the present application is shown;

[0021] Figure 5 A schematic diagram of the network structure of a ResNet+FPN model provided according to an embodiment of the present application is shown;

[0022] Figure 6A structural block diagram of an error warning device for transaction services provided according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0023] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0024] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] As introduced in the background technology, the prior art transaction data error warning scheme uses artificially customized rules to issue warnings, resulting in the inability to make intelligent judgments and warnings through existing conditions. In order to solve the problem that the prior art transaction data error warning scheme uses artificially customized rules to issue warnings, resulting in the inability to make intelligent judgments and warnings through existing conditions, the embodiments of the present application provide a transaction business error warning method, device, storage medium and electronic device.

[0027] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0028] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a mobile terminal of a transaction service error warning method according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is for illustration only and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0029] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the error warning method of the transaction business in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0030] In this embodiment, an error warning method for transaction services running on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0031] Figure 2 1 is a flow chart of an error warning method for a transaction service according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0032] Step S201, obtaining a business transaction image, inputting the business transaction image into a classification model, and identifying a business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling a business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of training data in the multiple sets of training data includes historical business transaction images and business transaction types corresponding to the historical business transaction images;

[0033] Specifically, the business transaction images are classified (the image formats of different business types are inconsistent), and the business transaction types of the business transaction images are identified so that they can be subsequently fed into different detection and recognition models (in order to improve the detection and recognition rates, there are specific network inputs for specific images).

[0034] Step S202, according to the business transaction type, determining the position of the information to be identified in the business transaction image, performing image masking processing on the area outside the position of the information to be identified in the business transaction image, and obtaining the business image to be identified, wherein the information to be identified is information that needs to be subjected to business error detection in the business transaction image of the business transaction type;

[0035] Specifically, after identifying the business transaction type of the business transaction image, the business transaction image is placed in the information extraction model corresponding to the business transaction type, and the important information position in the image is detected, such as transaction type, transaction code, account number and other flow-related information. In order to avoid interference with subsequent recognition, the area outside the obtained detection frame is filled with the background color of the image, such as the bank of account and other unnecessary information. The specific filling method is to directly estimate the background and then fill it through a simple image processing method, and then crop the largest outline of the business image to be identified, and at the same time normalize it to 1024 pixels in height and 512 pixels in width.

[0036] Step S203, according to the business transaction type, business error recognition is performed on the business image to be recognized, and an error recognition result is generated. The error recognition result is used to issue a business error warning. The error recognition result is used to characterize the business error that occurs during the business handling process.

[0037] Specifically, according to the business transaction type of the business transaction image, the business image to be identified is input into the corresponding target error recognition model, the business errors in the business image to be identified are identified, and an error recognition result is generated. The teller is prompted through a pop-up window to indicate the probability of an error in the transaction and give the error category (error recognition result), so that the teller can correct the transaction as soon as possible.

[0038] For example, the error recognition results may be 50% probability of forgetting to submit the ID card, 30% probability of missing signature, and 20% probability of no attachment.

[0039] Through customized image classification and detection, the detection efficiency can be improved, while removing extra information can improve classification accuracy, and the results of traditional data matching warnings are more accurate.

[0040] Through this embodiment, a business transaction image is obtained, and the business transaction image is input into a classification model to identify the business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of training data in the multiple sets of training data includes historical business transaction images and business transaction types corresponding to the historical business transaction images; according to the business transaction type, the position of the information to be identified in the business transaction image is determined, and the area outside the position of the information to be identified in the business transaction image is subjected to image masking processing to obtain the business image to be identified, wherein the information to be identified is information that requires business error detection for the business transaction image of the business transaction type; according to the business transaction type, business error identification is performed on the business image to be identified to generate an error identification result, and the error identification result is used to issue a business error warning, and the error identification result is used to characterize the occurrence of business errors in the business handling process. By acquiring business transaction images, a classification model is used to classify the business transaction images according to their business types. The classified business transaction images are input into the corresponding recognition model to generate business images to be identified. Error identification is performed on the business images to be identified for use in business error warnings. This solves the problem in the existing technology that transaction data error warning schemes issue warnings based on artificially customized rules, resulting in the inability to make intelligent judgments and warnings based on existing conditions.

[0041] During the specific implementation process, business error identification is performed on the business image to be identified according to the business transaction type, including: according to the business transaction type, the business image to be identified is input into the target error identification model corresponding to the business transaction type, and the target error identification model is used to perform error identification processing on the information to be identified to obtain an error identification result, wherein the target error identification model is obtained by machine learning through multiple groups of error identification training data, and each group of error identification training data in the multiple groups of error identification training data includes historical business images to be identified and error identification results corresponding to the historical business images to be identified.

[0042] Specifically, a target error recognition model is used to perform recognition processing on information to be recognized to obtain an error recognition result, including: using the target error recognition model to perform error prediction recognition on each piece of information to be recognized to obtain an error prediction score corresponding to each piece of recognition information; outputting target recognition information whose error prediction score is greater than a score threshold, and sorting the target recognition information in descending order according to the error prediction scores corresponding to the target recognition information to obtain an error prediction sequence; selecting the first N target recognition information in the error prediction sequence, and generating an error recognition result according to the error prediction scores corresponding to the first N target recognition information, the error recognition result including the error prediction proportion corresponding to the first N target recognition information, and the error prediction proportion is the probability that errors occur in the first N target recognition information.

[0043] Among them, N can be set to 3, 4, or 5, and can be set according to specific identification requirements; the score threshold can be set according to different business types. For example, the score threshold can be set to 60 points. When the identification information exceeds 60 points, it is considered that the identification information may be erroneous.

[0044] The method extracts identification information whose error prediction scores exceed the score threshold, sorts the extracted identification information in descending order according to the error prediction scores, obtains an error prediction sequence, selects the first N in the error prediction sequence as identification information with a greater probability of error, and calculates the probability of error in the first N identification information according to the error prediction scores of the first N identification information. For example, when N is 3, after calculation, it is determined that there is a 50% probability of forgetting to hand in the ID card, a 30% probability of missing a signature, and a 20% probability of no attachment. The three identification information and the corresponding probabilities are reminded to the teller through a page pop-up window, and the three identification information are verified to determine whether an error occurs.

[0045] More specifically, before acquiring the business transaction image, the method further includes: during the transaction processing, if the generation of the business transaction image is not detected, generating a prompt message, wherein the prompt message is used to remind the lack of the business transaction image for the business processing.

[0046] When the method fails to detect the business transaction image, it will prompt the teller that the transaction flow is missing an image and ask whether to retake it. At the same time, the reminder can be ignored and closed without any other operation. The reminder can reduce the operation of transmitting less images.

[0047] Furthermore, before inputting the business transaction image into the classification model, the method further includes: performing a preprocessing operation on the business transaction image, wherein the preprocessing operation includes adjusting the image clarity.

[0048] This method automatically acquires transaction images when the counter equipment detects image input, and then regularizes the clarity of the images, for example, all images are uniformly adjusted to a 512*1024 format.

[0049] Furthermore, after obtaining the error prediction score corresponding to each identification information, the method also includes: determining whether the error prediction score corresponding to each identification information is greater than a score threshold; if it is determined that the error prediction score corresponding to each identification information is less than or equal to the score threshold, no error recognition result is generated.

[0050] When the method determines that the error prediction score corresponding to each piece of identification information is less than or equal to the score threshold, it is considered that there is no transaction error in the business transaction image, no error identification result is generated, and other business transaction images continue to be identified.

[0051] Specifically, the classification model is the VGG16 model, and the target error recognition model is the ResNet+FPN network model.

[0052] This method uses the VGG16 model as the classification model and adjusts the parameters of the VGG16 model, including:

[0053] 1) The input image size is 224x224x3, and it is convolved twice with 64 3x3 convolution kernels with 3 channels, a step size of 1, padding=same padding, and then activated by ReLU. The output size is 224x224x64;

[0054] 2) After max pooling, the filter is 2x2, the step size is 2, the image size is halved, and the size after pooling becomes 112x112x64;

[0055] 3) After 128 3x3 convolution kernels, two convolutions, and ReLU activation, the size becomes 112x112x128;

[0056] 4) max pooling, the size becomes 56x56x128;

[0057] 5) After 256 3x3 convolution kernels, three convolutions, and ReLU activation, the size becomes 56x56x256;

[0058] 6) max pooling, the size becomes 28x28x256;

[0059] 7) After 512 3x3 convolution kernels, three convolutions, and ReLU activation, the size becomes 28x28x512;

[0060] 8) max pooling, the size becomes 14x14x512;

[0061] 9) After 512 3x3 convolution kernels, three convolutions, and ReLU, the size becomes 14x14x512;

[0062] 10) max pooling, the size becomes 7x7x512;

[0063] 11) Then Flatten() flattens the data into a vector, turning it into a one-dimensional 51277=25088.

[0064] 12) Then pass through two layers of 1x1x4096 and one layer of 1x1x1000 fully connected layer (three layers in total), and then activate by ReLU;

[0065] 13) Finally, 1000 prediction results are output through softmax (here is the number of possible classifications of the image type).

[0066] The ResNet+FPN network model is used as the target error recognition model. ResNet is a deep residual network structure that can better solve the gradient vanishing and gradient exploding problems that occur during deep neural network training, making the network easier to train and converge; FPN can effectively capture target information of different scales by building a pyramid structure on feature maps at different levels, improving the performance of target detection and recognition; combining ResNet and FPN can fully utilize the deep feature extraction capability of ResNet and the multi-scale feature fusion capability of FPN to improve the recognition accuracy and robustness of the model.

[0067] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the error warning method for the transaction business of the present application will be described in detail below in combination with specific embodiments.

[0068] Among them, ResNet is a feature extraction and classification model that uses a deep network to extract high-dimensional features from the incoming image and then classifies the features.

[0069] Some existing solutions are based on big data loan risk control methods, which conduct risk assessment by obtaining the basic information data of the loan object in the bank and performing big data analysis and matching. A big data-based loan risk control method is described, which obtains the basic information data of the loan object in the bank, and the basic information data includes the loan data of the loan object corresponding to the loan account and the supplementary data of other accounts of the loan object except the loan account; a sample identification code is generated according to the instance sample data formed by each loan record in the loan data; a multi-dimensional behavior action analysis is performed based on the sample identification code and the supplementary data to generate an action identification code corresponding to the instance sample data; the suspected result of the instance sample data is identified based on the action identification code, and the suspected result is stored in the corresponding database. The present invention can automatically reduce the large-scale purposeless verification through the big data multi-dimensional analysis technology, and improve the efficiency and accuracy of risk supervision. The invention only forms an identification code through big data analysis through features, and analyzes according to the matching degree of the identification code, which belongs to the rule-based judgment. When there are many types of data, the artificially customized rules cannot meet all scenarios, and intelligent judgment and early warning cannot be performed through existing conditions.

[0070] Some solutions in the prior art provide a bank counter business monitoring system, including a model design monitoring module, a teller operation monitoring module, a teller credential monitoring module, a duty monitoring module and a knowledge base monitoring module. The multiple modules work together to achieve supervision of the entire process of the bank counter. The present invention can realize the risk warning management of the entire process and each link of the bank counter operation, facilitate the management of the bank's counter operation risk pre-control, and reduce the risk of the bank counter operation. Although the risk warning management of the entire process and each link of the bank counter operation can be realized, the various modules designed by the invention only perform simple matching through a database or a question library, and the rule coverage is insufficient. In addition, each piece of information must be manually reviewed, and all result judgments and identifications require manual operation by the auditor, and it is time-consuming and laborious to make intelligent judgments through models.

[0071] The common warning method used in existing technologies is to simply analyze big data and obtain results through simple data matching.

[0072] This embodiment relates to a specific transaction error warning method, such as Figure 3 As shown, the following steps are included:

[0073] Step 1: When the counter equipment detects image input, it automatically obtains the transaction image, and then regularizes the image. All images are uniformly adjusted to 512*1024 format. If no image is detected, the teller will be prompted that the transaction is missing an image and whether to retake it. At the same time, the reminder can be ignored and closed without any other operation. This reminder can reduce the operation of transmitting less images.

[0074] Step 2: Send the images obtained in the first step to a simple classification model to classify the transmitted images (the image formats of different business types are inconsistent) and distinguish which format the transaction images belong to, so that they can be subsequently sent to different detection and recognition models (in order to improve the detection and recognition rates, there are specific network inputs for specific images, and the specific detection and recognition models are consistent);

[0075] Step 3: Send the different classification results in the second step into different information extraction models to detect the important information locations in the image, such as transaction type, transaction code, account number and other flow-related information.

[0076] Step 4: Data preprocessing. In order to avoid interference with subsequent recognition, all areas outside the detection box obtained in the third step are filled with unnecessary information such as the bank of account opening, etc. according to the background color of the image. The specific filling method is to estimate the background and then fill it through a simple image processing method, and then crop the largest outline and normalize it to 1024 pixels in height and 512 pixels in width.

[0077] Step 5: Input the processed transaction image into the recognition model (target error recognition model) for classification. At the same time, based on the classification results, output the results with a score of more than 60 to the front end of the counter. Make predictions based on the probability among the output results. For example, there are three prediction errors with a score of more than 60 points. Based on the scores of these three prediction errors, normalize them and determine the percentage of each prediction error. For example, there is a 50% probability of forgetting to hand in the ID card, a 30% probability of missing a signature, and a 20% probability of no attachments. The teller can choose to verify all operations according to the given prompts.

[0078] This embodiment is applicable to a wide range of scenarios:

[0079] 1) During the counter transaction process, by configuring hardware facilities, transaction information can be obtained and the high-definition camera or scanner can be linked to obtain image information. During the transaction, a pop-up window can be used to remind the teller how likely the transaction is to go wrong and give the type of error, so as to facilitate the correction of the transaction in the first time, reduce the error rate, and reduce the incidence of in-process review and post-audit.

[0080] 2) During the transaction review, potential problems that may arise in the transaction are given based on the transaction code and related image information uploaded by the transaction, providing reference opinions to the auditor. The auditor can quickly verify the compliance of the transaction based on the prompts. If there are errors, immediately contact the transaction personnel to make modifications and supplements to ensure the correctness of the transaction, increase the success rate of the review during the process, and improve efficiency.

[0081] 3) In the post-audit phase, auditors can quickly analyze transaction details based on the error warnings prompted by the model, compare transactions for errors through warning prompts, and quickly perform audit operations to improve work efficiency and reduce labor costs.

[0082] The classification model used in this embodiment is designed as VGG16. The specific network structure of the VGG16 model is as follows: Figure 4 As shown. Because the images of different business types are quite different, it is easy to implement multi-classification of this feature. In order to improve the classification accuracy and the speed of the entire network training, the VGG16 model is used. The specific parameter settings are as follows. The final output dimension can be appropriately adjusted according to the actual number of image types:

[0083] 1) The input image size is 224x224x3, and it is convolved twice with 64 3x3 convolution kernels with 3 channels, a step size of 1, padding=same padding, and then activated by ReLU. The output size is 224x224x64;

[0084] 2) After max pooling, the filter is 2x2, the step size is 2, the image size is halved, and the size after pooling becomes 112x112x64;

[0085] 3) After 128 3x3 convolution kernels, two convolutions, and ReLU activation, the size becomes 112x112x128;

[0086] 4) max pooling, the size becomes 56x56x128;

[0087] 5) After 256 3x3 convolution kernels, three convolutions, and ReLU activation, the size becomes 56x56x256;

[0088] 6) max pooling, the size becomes 28x28x256;

[0089] 7) After 512 3x3 convolution kernels, three convolutions, and ReLU activation, the size becomes 28x28x512;

[0090] 8) max pooling, the size becomes 14x14x512;

[0091] 9) After 512 3x3 convolution kernels, three convolutions, and ReLU, the size becomes 14x14x512;

[0092] 10) max pooling, the size becomes 7x7x512;

[0093] 11) Then Flatten() flattens the data into a vector, turning it into a one-dimensional 51277=25088.

[0094] 11) After two layers of 1x1x4096 and one layer of 1x1x1000 fully connected layers (three layers in total), the network is activated by ReLU.

[0095] 12) Finally, 1000 prediction results are output through softmax (here is the number of possible classifications of the image type).

[0096] Among them, the target error recognition model corresponding to each transaction business is set, and the target error recognition model can adopt the ResNet+FPN network model, in which the network structure of the model is as follows: Figure 5 As shown in the figure, 1x1 indicates a convolution operation with a convolution kernel of 1*1, 3x3 indicates a convolution operation with a convolution kernel of 3*3, + indicates that the feature maps are directly combined (concat), and 0.5x indicates upsampling, that is, the length and width of the feature map are halved after each convolution. The feature map finally extracted is directly classified and recognized.

[0097] This embodiment realizes the human-machine coupling mode of teller operation by using image classification and recognition, thereby improving work efficiency and reducing the error rate. By using the background estimation method to remove the customized key information detection model design of the image with additional information, only the important information area of ​​the image voucher is used for detection and recognition, which can reduce interference information and improve the accuracy of classification. The classification detection and recognition of image pictures can reduce the error rate of counter transactions, in-process supervision and post-audit, and improve work efficiency.

[0098] In addition, this embodiment can improve detection efficiency through customized image classification and detection, while removing extra information to improve classification accuracy, and the results are more accurate than traditional data matching warnings.

[0099] Through experimental comparison, the use of customized image information detection model improves the accuracy by 5% compared with the unified detection model. At the same time, by removing the interference of additional information and only using the more complete image classification accuracy of the main information area of ​​the flow, the accuracy is increased by 10%. Therefore, the customized image detection designed in this scheme can better realize the error warning of the counter.

[0100] The embodiment of the present application also provides an error warning device for transaction services. It should be noted that the error warning device for transaction services of the embodiment of the present application can be used to execute the error warning method for transaction services provided by the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions thereof will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0101] The following is an introduction to the error warning device for transaction services provided in an embodiment of the present application.

[0102] Figure 6 Schematic diagram of an error warning device for transaction services according to an embodiment of the present application. Figure 6 As shown, the device includes: a classification unit 61, a processing unit 62, and a first generation unit 63.

[0103] The classification unit 61 is used to obtain a business transaction image, input the business transaction image into a classification model, and identify the business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling a business, and the classification model is obtained by machine learning through multiple sets of training data sets, and each set of training data in the multiple sets of training data sets includes historical business transaction images and business transaction types corresponding to the historical business transaction images;

[0104] Specifically, the business transaction images are classified (the image formats of different business types are inconsistent), and the business transaction types of the business transaction images are identified so that they can be subsequently fed into different detection and recognition models (in order to improve the detection and recognition rates, there are specific network inputs for specific images).

[0105] The processing unit 62 is used to determine the position of the information to be identified in the business transaction image according to the business transaction type, and perform image masking processing on the area outside the position of the information to be identified in the business transaction image to obtain the business image to be identified, wherein the information to be identified is information that needs to be subjected to business error detection in the business transaction image of the business transaction type;

[0106] Specifically, after identifying the business transaction type of the business transaction image, the business transaction image is placed in the information extraction model corresponding to the business transaction type, and the important information position in the image is detected, such as transaction type, transaction code, account number and other flow-related information. In order to avoid interference with subsequent recognition, the area outside the obtained detection frame is filled with the background color of the image, such as the bank of account and other unnecessary information. The specific filling method is to directly estimate the background and then fill it through a simple image processing method, and then crop the largest outline of the business image to be identified, and at the same time normalize it to 1024 pixels in height and 512 pixels in width.

[0107] The first generating unit 63 is used to perform business error recognition on the business image to be recognized according to the business transaction type, generate an error recognition result, use the error recognition result to issue a business error warning, and the error recognition result is used to characterize the business error occurring in the business handling process.

[0108] Specifically, according to the business transaction type of the business transaction image, the business image to be identified is input into the corresponding target error recognition model, the business errors in the business image to be identified are identified, and an error recognition result is generated. The teller is prompted through a pop-up window to indicate the probability of an error in the transaction and give the error category (error recognition result), so that the teller can correct the transaction as soon as possible.

[0109] In this embodiment, the classification unit is used to obtain a business transaction image, input the business transaction image into the classification model, and identify the business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling business, and the classification model is obtained by machine learning through multiple sets of training data sets, and each set of training data in the multiple sets of training data sets includes historical business transaction images and business transaction types corresponding to the historical business transaction images; the processing unit is used to determine the position of the information to be identified in the business transaction image according to the business transaction type, and perform image masking processing on the area outside the position of the information to be identified in the business transaction image to obtain the business image to be identified, wherein the information to be identified is information that requires business error detection for the business transaction image of the business transaction type; the first generating unit is used to perform business error identification on the business image to be identified according to the business transaction type, generate an error identification result, use the error identification result to issue a business error warning, and the error identification result is used to characterize the occurrence of business errors in the business handling process. By acquiring business transaction images, a classification model is used to classify the business transaction images according to their business types. The classified business transaction images are input into the corresponding recognition model to generate business images to be identified. Error identification is performed on the business images to be identified for use in business error warnings. This solves the problem in the existing technology that transaction data error warning schemes issue warnings based on artificially customized rules, resulting in the inability to make intelligent judgments and warnings based on existing conditions.

[0110] As an optional solution, the processing unit includes a processing module, which is used to input the business image to be identified into a target error recognition model corresponding to the business transaction type according to the business transaction type, so as to perform error recognition processing on the information to be identified using the target error recognition model to obtain an error recognition result, wherein the target error recognition model is obtained by machine learning through multiple groups of error recognition training data, and each group of error recognition training data in the multiple groups of error recognition training data includes historical business images to be identified and error recognition results corresponding to the historical business images to be identified.

[0111] An optional solution, the processing module includes an identification submodule, an output submodule and a selection submodule; the identification submodule is used to perform error prediction and identification on each information to be identified by adopting a target error identification model to obtain an error prediction score corresponding to each identification information; the output submodule is used to output the target identification information whose error prediction score is greater than the score threshold, and the target identification information is sorted in descending order according to the error prediction score corresponding to the target identification information to obtain an error prediction sequence; the selection submodule is used to select the first N target identification information in the error prediction sequence, and generate an error identification result according to the error prediction scores corresponding to the first N target identification information, and the error identification result includes the error prediction proportion corresponding to the first N target identification information, and the error prediction proportion is the probability of error in the first N target identification information.

[0112] Specifically, identification information whose error prediction scores exceed a score threshold is extracted, and the extracted identification information is sorted in descending order according to the error prediction scores to obtain an error prediction sequence. The first N pieces of the error prediction sequence are selected as identification information with a greater probability of error, and the probability of error in the first N pieces of identification information is calculated based on their error prediction scores.

[0113] In an optional solution, the device also includes a second generation unit, which is used to generate a prompt message before obtaining the business transaction image and during the transaction process, if the business transaction image generation is not detected, wherein the prompt message is used to remind the business transaction image that is missing for business processing.

[0114] Specifically, when the business transaction image cannot be detected, the teller will be prompted that the transaction is missing an image and whether to take a retake. At the same time, the reminder can be ignored and closed without any other operation. This reminder can reduce the operation of transmitting less images.

[0115] In an optional solution, the device further includes a preprocessing unit for performing a preprocessing operation on the business transaction image before inputting the business transaction image into the classification model, wherein the preprocessing operation includes adjusting the image clarity.

[0116] Specifically, when the counter equipment detects image input, it automatically acquires the transaction image and then regularizes the image clarity, for example, all images are uniformly adjusted to a 512*1024 format.

[0117] In an optional scheme, the processing unit also includes a first determination module and a second determination module. The first determination module is used to determine whether the error prediction score corresponding to each identification information is greater than a score threshold after obtaining the error prediction score corresponding to each identification information; the second determination module is used to not generate an error recognition result when it is determined that the error prediction score corresponding to each identification information is less than or equal to the score threshold.

[0118] Specifically, when it is determined that the error prediction score corresponding to each identification information is less than or equal to the score threshold, it is considered that there is no transaction error in the business transaction image, and no error identification result is generated, and other business transaction images continue to be identified.

[0119] An optional solution is that the classification model is the VGG16 model and the target error recognition model is the ResNet+FPN network model.

[0120] Specifically, the ResNet+FPN network model is used as the target error recognition model. ResNet is a deep residual network structure that can better solve the gradient vanishing and gradient exploding problems that occur during deep neural network training, making the network easier to train and converge; FPN can effectively capture target information of different scales by building a pyramid structure on feature maps at different levels, improving the performance of target detection and recognition; combining ResNet and FPN can fully utilize the deep feature extraction capability of ResNet and the multi-scale feature fusion capability of FPN to improve the recognition accuracy and robustness of the model.

[0121] The error warning device for transaction business includes a processor and a memory, and the above-mentioned classification unit, processing unit, first generation unit, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions. The above-mentioned modules are all located in the same processor; or, the above-mentioned modules are located in different processors in any combination.

[0122] The processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the kernel parameters can be adjusted to solve the problem that the existing technology of early warning schemes for transaction data errors is based on artificially customized rules, resulting in the inability to make intelligent judgments and early warnings based on existing conditions.

[0123] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0124] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is run, the device where the computer-readable storage medium is located is controlled to execute the error warning method of the transaction business.

[0125] Specifically, the error warning methods for transaction business include:

[0126] Step S201, obtaining a business transaction image, inputting the business transaction image into a classification model, and identifying a business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling a business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of training data in the multiple sets of training data includes historical business transaction images and business transaction types corresponding to the historical business transaction images;

[0127] Step S202, according to the business transaction type, determining the position of the information to be identified in the business transaction image, performing image masking processing on the area outside the position of the information to be identified in the business transaction image, and obtaining the business image to be identified, wherein the information to be identified is information that needs to be subjected to business error detection in the business transaction image of the business transaction type;

[0128] Step S203, according to the business transaction type, business error recognition is performed on the business image to be recognized, and an error recognition result is generated. The error recognition result is used to issue a business error warning. The error recognition result is used to characterize the business error that occurs during the business handling process.

[0129] An embodiment of the present invention provides a processor, which is used to run a program, wherein the error warning method of the transaction business is executed when the program is running.

[0130] Specifically, the error warning methods for transaction business include:

[0131] Step S201, obtaining a business transaction image, inputting the business transaction image into a classification model, and identifying a business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling a business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of training data in the multiple sets of training data includes historical business transaction images and business transaction types corresponding to the historical business transaction images;

[0132] Step S202, according to the business transaction type, determining the position of the information to be identified in the business transaction image, performing image masking processing on the area outside the position of the information to be identified in the business transaction image, and obtaining the business image to be identified, wherein the information to be identified is information that needs to be subjected to business error detection in the business transaction image of the business transaction type;

[0133] Step S203, according to the business transaction type, business error recognition is performed on the business image to be recognized, and an error recognition result is generated. The error recognition result is used to issue a business error warning. The error recognition result is used to characterize the business error that occurs during the business handling process.

[0134] An embodiment of the present invention provides a device, the device including a processor, a memory, and a program stored in the memory and executable on the processor, and when the processor executes the program, at least the following steps are implemented:

[0135] Step S201, obtaining a business transaction image, inputting the business transaction image into a classification model, and identifying a business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling a business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of training data in the multiple sets of training data includes historical business transaction images and business transaction types corresponding to the historical business transaction images;

[0136] Step S202, according to the business transaction type, determining the position of the information to be identified in the business transaction image, performing image masking processing on the area outside the position of the information to be identified in the business transaction image, and obtaining the business image to be identified, wherein the information to be identified is information that needs to be subjected to business error detection in the business transaction image of the business transaction type;

[0137] Step S203, according to the business transaction type, business error recognition is performed on the business image to be recognized, and an error recognition result is generated. The error recognition result is used to issue a business error warning. The error recognition result is used to characterize the business error that occurs during the business handling process.

[0138] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0139] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing at least the following method steps:

[0140] Step S201, obtaining a business transaction image, inputting the business transaction image into a classification model, and identifying a business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling a business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of training data in the multiple sets of training data includes historical business transaction images and business transaction types corresponding to the historical business transaction images;

[0141] Step S202, according to the business transaction type, determining the position of the information to be identified in the business transaction image, performing image masking processing on the area outside the position of the information to be identified in the business transaction image, and obtaining the business image to be identified, wherein the information to be identified is information that needs to be subjected to business error detection in the business transaction image of the business transaction type;

[0142] Step S203, according to the business transaction type, business error recognition is performed on the business image to be recognized, and an error recognition result is generated. The error recognition result is used to issue a business error warning. The error recognition result is used to characterize the business error that occurs during the business handling process.

[0143] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0144] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0145] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0146] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0147] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0148] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0149] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0150] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0151] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0152] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0153] 1) A transaction error warning method of the present application includes: obtaining a business transaction image, inputting the business transaction image into a classification model, and identifying a business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of training data in the multiple sets of training data includes historical business transaction images and business transaction types corresponding to the historical business transaction images; according to the business transaction type, determining the position of information to be identified in the business transaction image, performing image masking processing on the area outside the position of the information to be identified in the business transaction image, and obtaining a business image to be identified, wherein the information to be identified is information that needs to be detected for business errors in the business transaction image of the business transaction type; according to the business transaction type, performing business error identification on the business image to be identified, generating an error identification result, using the error identification result to perform business error warning, and the error identification result is used to characterize the occurrence of business errors in the business handling process. By acquiring business transaction images, a classification model is used to classify the business transaction images according to their business types. The classified business transaction images are input into the corresponding recognition model to generate business images to be identified. Error identification is performed on the business images to be identified for use in business error warnings. This solves the problem in the existing technology that transaction data error warning schemes issue warnings based on artificially customized rules, resulting in the inability to make intelligent judgments and warnings based on existing conditions.

[0154] 2) A transaction error warning device of the present application includes: a classification unit, which is used to obtain a business transaction image, input the business transaction image into a classification model, and identify the business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling business, and the classification model is obtained by machine learning through multiple sets of training data sets, and each set of training data in the multiple sets of training data sets includes historical business transaction images and business transaction types corresponding to the historical business transaction images; a processing unit, which is used to determine the position of information to be identified in the business transaction image according to the business transaction type, and perform image masking processing on the area outside the position of the information to be identified in the business transaction image to obtain a business image to be identified, wherein the information to be identified is information that needs to be detected for business errors in the business transaction image of the business transaction type; a first generating unit, which is used to perform business error identification on the business image to be identified according to the business transaction type, generate an error identification result, use the error identification result to perform business error warning, and the error identification result is used to characterize the occurrence of business errors in the business handling process. By acquiring business transaction images, a classification model is used to classify the business transaction images according to their business types. The classified business transaction images are input into the corresponding recognition model to generate business images to be identified. Error identification is performed on the business images to be identified for use in business error warnings. This solves the problem in the existing technology that transaction data error warning schemes issue warnings based on artificially customized rules, resulting in the inability to make intelligent judgments and warnings based on existing conditions.

[0155] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A transaction error warning method, characterized in that: include: Acquire a business transaction image, input the business transaction image into a classification model, and identify a business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling a business, and the classification model is obtained by machine learning through multiple sets of training data, and each set of the multiple sets of training data includes a historical business transaction image and the business transaction type corresponding to the historical business transaction image; According to the business transaction type, the position of the information to be identified in the business transaction image is determined, and the area outside the position of the information to be identified in the business transaction image is subjected to image masking processing to obtain the business image to be identified, wherein the information to be identified is information that needs to be subjected to business error detection in the business transaction image of the business transaction type; According to the business transaction type, business error identification is performed on the business image to be identified to generate an error identification result, and the error identification result is used to issue a business error warning. The error identification result is used to characterize the business error that occurs during the business processing process.

2. The method according to claim 1, characterized in that: According to the business transaction type, business error identification is performed on the business image to be identified, including: According to the business transaction type, the business image to be identified is input into a target error recognition model corresponding to the business transaction type, so as to use the target error recognition model to perform error recognition processing on the information to be identified, and obtain the error recognition result, wherein the target error recognition model is obtained by machine learning through multiple groups of error recognition training data, and each group of the multiple groups of error recognition training data includes historical business images to be identified and the error recognition results corresponding to the historical business images to be identified.

3. The method according to claim 2, characterized in that The target error recognition model is used to perform recognition processing on the information to be recognized to obtain an error recognition result, including: Using the target error recognition model to perform error prediction recognition on each of the to-be-recognized information, and obtaining an error prediction score corresponding to each piece of recognized information; Outputting the target recognition information whose error prediction score is greater than the score threshold, and sorting the target recognition information in descending order according to the error prediction score corresponding to the target recognition information to obtain an error prediction sequence; The first N target identification information in the error prediction sequence are selected, and the error identification result is generated according to the error prediction scores corresponding to the first N target identification information. The error identification result includes the error prediction proportions corresponding to the first N target identification information, and the error prediction proportions are the probabilities that errors occur in the first N target identification information.

4. The method according to claim 1, characterized in that Before acquiring the business transaction image, the method further includes: During the transaction process, if the generation of the business transaction image is not detected, a prompt message is generated, wherein the prompt message is used to remind the user that the business transaction image for the business process is missing.

5. The method according to claim 1, characterized in that Before inputting the business transaction image into the classification model, the method further includes: Perform a preprocessing operation on the business transaction image, wherein the preprocessing operation includes image clarity adjustment.

6. The method according to claim 3, characterized in that After obtaining the error prediction score corresponding to each piece of identification information, the method further includes: Determining whether the error prediction score corresponding to each piece of identification information is greater than the score threshold; When it is determined that the error prediction scores corresponding to each piece of identification information are all less than or equal to the score threshold, the error identification result is not generated.

7. The method according to claim 2, characterized in that: The classification model is a VGG16 model, and the target error recognition model is a ResNet+FPN network model.

8. An error warning device for transaction services, characterized in that: include: a classification unit, configured to obtain a business transaction image, input the business transaction image into a classification model, and identify a business transaction type of the business transaction image, wherein the business transaction image is an image of a business transaction interface for handling a business, and the classification model is obtained by machine learning through multiple sets of training data sets, and each set of training data in the multiple sets of training data sets includes a historical business transaction image and the business transaction type corresponding to the historical business transaction image; a processing unit, configured to determine the position of the information to be identified in the business transaction image according to the business transaction type, and perform image masking processing on the area outside the position of the information to be identified in the business transaction image to obtain the business image to be identified, wherein the information to be identified is information that needs to be subjected to business error detection in the business transaction image of the business transaction type; The first generating unit is used to perform business error identification on the business image to be identified according to the business transaction type, generate an error identification result, use the error identification result to issue a business error warning, and the error identification result is used to characterize the business error occurring in the business handling process.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the error warning method for transaction business described in any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include an error warning method for executing the transaction business described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Abnormal service identification method and device

    CN113095392A

  • Early warning method, device and equipment based on voucher image

    CN113298182A

  • Transaction risk determination method, system and device and computer readable storage medium

    CN115630842A