A method and system for intelligent recognition of assertion elements in bank receipts

By using the LeNet-5 convolutional neural network model to identify elements in bank statement PDF files, the problem of insufficient recognition accuracy in existing technologies is solved, and the entire process of bank statement testing is automated.

CN115841673BActive Publication Date: 2026-11-17CHINA CITIC BANK CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202211549366.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-12-05
Publication Date
2026-11-17
Estimated Expiration
2042-12-05

AI Technical Summary

Technical Problem

The existing technology does not have enough accuracy in recognizing elements in PDF files of bank receipts, which causes assertion verification to fail in automated testing and prevents it from executing smoothly.

Method used

The LeNet-5 convolutional neural network model is used to identify elements in bank receipt PDF files. The element regions are divided, converted into pixel data, and initial and precise identification is performed. The results are verified using verification rules and printed on the receipt file.

Benefits of technology

It has achieved fully automated testing of bank receipts, improving the accuracy of element recognition and the success rate of automated testing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115841673B_ABST
    Figure CN115841673B_ABST
Patent Text Reader

Abstract

This invention relates to a method and system for intelligent recognition of assertion elements on bank receipts, comprising: acquiring a bank receipt PDF file; dividing the bank receipt PDF file into several element regions; converting the element regions into corresponding pixel data according to their positions; inputting the pixel data as input data into a LeNet-5 convolutional neural network model for initial recognition, obtaining initial recognition results for each element region; printing the initial recognition results to the corresponding element regions; acquiring element regions whose initial recognition results are assertion verification failures and pending assertion verification results; adjusting the parameters of the LeNet-5 convolutional neural network model for accurate recognition, and printing the accurate recognition results to the corresponding element regions. This invention optimizes the element recognition method for bank receipt PDF files and supports accurate verification of assertion elements, thereby completing the entire process of automated testing for bank receipts.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of image processing and text recognition technology, and in particular to an intelligent recognition method and system for assertion elements on bank receipts. Background Technology

[0002] Bank receipts are valid proof of transactions conducted by individuals or organizations at banks. They serve as the primary basis for businesses to prepare accounting vouchers and are widely used in banking systems. There are various types of bank receipts, such as deposit and withdrawal receipts, transfer receipts, fee receipts, interest receipts, tax payment receipts, and foreign exchange settlement receipts. During software testing of banking systems, many transactions involve testing receipt files. We need to extract all elements from the receipt files and compare them with corresponding fields in the system database to verify the correctness of all elements in the receipt files.

[0003] Currently, the comparison process for receipt testing requires manual completion, which is often time-consuming and inefficient. We need to introduce automated testing to automate the entire testing process. During automation, the automatic identification and extraction of elements from the receipt documents is crucial. Only with sufficiently accurate extracted data can the automated testing proceed smoothly. However, common element recognition methods for PDF files often suffer from insufficient accuracy. Insufficient accuracy can lead to errors in element name extraction or element data extraction. Either type of error will cause assertions to fail, thus halting the automated testing process.

[0004] Convolutional Neural Networks (CNNs) are a type of feedforward neural network that incorporates convolutional computations and has a deep structure. They are one of the representative algorithms in deep learning. CNNs have representation learning capabilities and can perform both supervised and unsupervised learning. The LeNet-5 model was the first CNN to be successfully applied to digit recognition problems, achieving an accuracy of approximately 99.2% in handling both digit recognition and character recognition issues. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes an intelligent method and system for identifying assertion elements in bank receipts. Addressing the pain points of automated testing of bank receipts, this invention optimizes the element recognition method for PDF receipt files. It utilizes the LeNet-5 model in convolutional neural networks to intelligently identify assertion elements on the receipts, extracts the identified data, compares it with field values ​​in the database, and directly prints the verification results onto the receipt file. Furthermore, it supports precise verification of assertion elements, thereby completing the entire automated testing process for bank receipts.

[0006] To achieve the above objectives, the technical solution adopted by the present invention includes:

[0007] A method for intelligent recognition of assertion elements in bank receipts, wherein the assertion elements include business type, transaction date, payer name, payer account number, payer's bank name, and payer's bank code, including:

[0008] Obtain the bank statement PDF file;

[0009] The bank statement PDF file is divided into several element regions;

[0010] Based on the position of the element region, the element region is converted into corresponding pixel data;

[0011] The pixel data is input into the LeNet-5 convolutional neural network model for initial recognition, and the initial recognition results of each element region are obtained according to the first verification rule.

[0012] Print the initial recognition results to the corresponding element area;

[0013] The initial identification results include assertion verification passed, assertion verification failed, and assertion verification result pending.

[0014] Obtain the element region whose initial identification result is assertion verification failure or assertion verification result pending;

[0015] Adjust the parameters of the LeNet-5 convolutional neural network model to accurately identify the element regions whose initial identification results are assertion verification failure and assertion verification results pending, and obtain accurate identification results according to the second verification rule;

[0016] Print the accurate recognition results to the corresponding element area.

[0017] Furthermore, the LeNet-5 convolutional neural network model has a total of 7 layers, of which the first 4 layers consist of alternating convolutional and pooling layers, and the last 3 layers are fully connected layers.

[0018] Furthermore, the first verification rule includes: if all characters in the element area pass, the initial recognition result is that the assertion verification is passed;

[0019] If two or more characters in the element area fail the verification, the initial recognition result is an assertion verification failure.

[0020] If only one character in the element area fails verification, the initial recognition result is "assertion verification result pending".

[0021] Furthermore, the accurate identification results include assertion verification passing and assertion verification failing.

[0022] Furthermore, the second verification rule includes: if all characters in the element area pass, the accurate recognition result is that the assertion verification passes;

[0023] If a character verification fails within the element region, the accurate recognition result is an assertion verification failure.

[0024] This invention also relates to an intelligent recognition system for assertion elements on bank receipts, comprising:

[0025] The acquisition module is used to acquire bank statement PDF files;

[0026] The segmentation module is used to segment the bank receipt PDF file to generate several element regions;

[0027] The conversion module is used to convert the element region into corresponding pixel data according to the position of the element region;

[0028] The first processing module is used to input the pixel data as input data into the LeNet-5 convolutional neural network model for initial recognition, and obtain the initial recognition result of each element region according to the first verification rule.

[0029] The first output module is used to print the initial recognition result to the corresponding element area; the initial recognition result includes assertion verification passed, assertion verification failed, and assertion verification result pending.

[0030] The second processing module is used to adjust the parameters of the LeNet-5 convolutional neural network model, accurately identify the element regions whose initial identification results are assertion verification failure and assertion verification results pending, and obtain accurate identification results according to the second verification rule.

[0031] The second output module is used to print the accurate recognition result to the corresponding element area.

[0032] The present invention also relates to a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0033] The present invention also relates to an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to execute the computer program stored in the memory and perform the method described above.

[0034] The beneficial effects of this invention are as follows:

[0035] This invention optimizes the element recognition method for bank receipt PDF files. It uses the LeNet-5 model in convolutional neural networks to intelligently identify assertion elements in the receipt, extracts the identified data, compares it with field values ​​in the database, and prints the verification results directly on the receipt file. It also supports precise verification of assertion elements, thereby completing the entire process of automated testing for bank receipts. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the intelligent recognition method for assertion elements on bank receipts according to the present invention.

[0037] Figure 2 This is a schematic diagram of the structure of an intelligent recognition system for assertion elements on bank receipts according to the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] The first aspect of this invention relates to a process flow as follows: Figure 1 The method shown here is an intelligent recognition method for assertion elements in bank receipts. The assertion elements include the transaction type, transaction date, payer name, payer account number, payer's bank name, and payer's bank code, and include:

[0040] Obtain the bank statement PDF file;

[0041] The bank statement PDF file is divided into several element regions;

[0042] Specifically, in the application embodiment, the bank receipt PDF file is preprocessed to convert it into pixel data that can be used as input image data for a convolutional neural network. First, the bank receipt PDF file is divided into element regions according to different receipt type templates. For example, the payer's name element region in a transfer receipt is:

[0043] name XXXXX Co., Ltd.

[0044] The currency and amount element area is as follows:

[0045] Currency and Amount Eighty-nine thousand two hundred and twenty-two yuan and sixty-eight fen (RMB 89,222.8)

[0046] Based on the position of the element region, the element region is converted into corresponding pixel data;

[0047] The pixel data is input into the LeNet-5 convolutional neural network model for initial recognition, and the initial recognition results of each element region are obtained according to the first verification rule.

[0048] Specifically, in the application embodiment, the LeNet-5 model has a total of 7 layers, in the following order: convolutional layer, pooling layer, convolutional layer, pooling layer, fully connected layer, fully connected layer, and fully connected layer. The convolutional layer is the core of the convolutional neural network, treating each neuron as a filter and using a sliding window to compute local data. The pooling layer is used to compress data and parameters to reduce overfitting. The fully connected layer connects all neurons with weights, making the model computation more accurate; it is usually placed at the end of the convolutional neural network.

[0049] The structural details of each layer are as follows:

[0050] The first layer (convolutional layer): The input to this layer is the original image pixels. The LeNet-5 model accepts an input layer size of 32×32×1, splitting the original pixel data into multiple 32×32 pixel data as input. The first convolutional layer has a filter size of 5×5, a depth of 6, and a stride of 1.

[0051] The second layer (pooling layer): The input of this layer is the output of the first layer, which is a 28×28×6 node matrix. The filter size is 2×2, and the stride of both the length and width is 2.

[0052] The third layer (convolutional layer): The input matrix of this layer is 14×14×6, the filter size is 5×5, the depth is 16, and the stride is 1;

[0053] Fourth layer (pooling layer): The input matrix size of this layer is 10×10×16, the filter size is 2×2, and the stride is 2;

[0054] Fifth layer (fully connected layer): The input matrix of this layer is 5×5×16, and the number of output nodes is 120;

[0055] Layer 6 (Fully Connected Layer): This layer has 120 input nodes and 84 output nodes;

[0056] The seventh layer (fully connected layer): This layer has 84 input nodes and 10 output nodes.

[0057] The original pixel data corresponding to each region element in the return file is split into multiple 32×32 pixel data as input, and the region element data can be output after passing through 7 levels.

[0058] This invention collects historical data of different types of bank receipts as a training set. After determining the number of layers according to the above model structure, it uses the backpropagation algorithm (BP algorithm) to iteratively train the convolutional neural network model. After the training converges, the weights of the convolutional neural network model can be obtained, which are the coefficients corresponding to the connections between nodes. After the model training is completed, the element data of the bank receipt file region to be identified is collected as input to obtain the output result.

[0059] Print the initial recognition results to the corresponding element area;

[0060] Specifically, in this embodiment of the application, the initial identification results of assertion elements are obtained through the LeNet-5 model described above. After extracting the data after the initial identification and comparing it with the field values ​​in the database, the initial identification results of each element region are obtained according to the first verification rule. The initial identification results are then printed directly on the return receipt file.

[0061] Specifically, the initial recognition results include assertion verification passed, assertion verification failed, and assertion verification result pending. The first verification rule is that if all characters in the element area pass, the initial recognition result is assertion verification passed.

[0062] If two or more characters in the element area fail the verification, the initial recognition result is an assertion verification failure.

[0063] If only one character in the element area fails verification, the initial recognition result is "assertion verification result pending".

[0064] Obtain the element region whose initial identification result is assertion verification failure or assertion verification result pending;

[0065] Adjust the parameters of the LeNet-5 convolutional neural network model to accurately identify the element regions whose initial identification results are assertion verification failure and assertion verification results pending, and obtain accurate identification results according to the second verification rule;

[0066] Print the accurate recognition results to the corresponding element area.

[0067] Specifically, in this embodiment, after selecting re-verification, precise identification will be initiated for the element regions that need to be re-verified. Precise identification also utilizes the aforementioned LeNet-5 model, but the model parameters need to be adjusted to achieve the purpose of precise verification. The specific parameter adjustment scheme is as follows:

[0068] In the first layer (convolutional layer), the input layer size of the LeNet-5 model is adjusted to 32×32×1, that is, the original pixel data is split into multiple 16×16 pixel data as input;

[0069] In the second layer (pooling layer), the input layer is adjusted to a 14×14×6 node matrix;

[0070] In the third layer (convolutional layer), the input layer is adjusted to a 7×7×6 node matrix;

[0071] No adjustments were made to the remaining layers.

[0072] The accurate recognition result is obtained according to the second verification rule, and the accurate recognition result is printed to the corresponding element area. Specifically, the second verification rule includes that if all characters in the element area pass, the accurate recognition result is an assertion verification passed.

[0073] If a character verification fails within the element region, the accurate recognition result is an assertion verification failure.

[0074] Another aspect of the present invention relates to an intelligent recognition and processing system for assertion elements in bank receipts, the structure of which is as follows: Figure 2 As shown, it includes:

[0075] The acquisition module is used to acquire bank statement PDF files;

[0076] The segmentation module is used to segment the bank receipt PDF file to generate several element regions;

[0077] The conversion module is used to convert the element region into corresponding pixel data according to the position of the element region;

[0078] The first processing module is used to input the pixel data as input data into the LeNet-5 convolutional neural network model for initial recognition, and obtain the initial recognition result of each element region according to the first verification rule.

[0079] The first output module is used to print the initial recognition result to the corresponding element area; the initial recognition result includes assertion verification passed, assertion verification failed, and assertion verification result pending.

[0080] The second processing module is used to adjust the parameters of the LeNet-5 convolutional neural network model, accurately identify the element regions whose initial identification results are assertion verification failure and assertion verification results pending, and obtain accurate identification results according to the second verification rule.

[0081] The second output module is used to print the accurate recognition result to the corresponding element area.

[0082] By using this system, the aforementioned computational processing methods can be executed and the corresponding technical effects can be achieved.

[0083] Embodiments of the present invention also provide a computer-readable storage medium capable of implementing all steps of the intelligent identification method for assertion elements of bank receipts in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the intelligent identification method for assertion elements of bank receipts in the above embodiments.

[0084] Embodiments of the present invention also provide an electronic device for performing the above-described method. As an implementation device for the method, the electronic device includes at least a processor and a memory. In particular, the memory stores data and related computer programs required for performing the method. The processor calls the data and programs in the memory to execute all the steps of the method and obtain the corresponding technical effect.

[0085] Preferably, the electronic device may include a bus architecture, which may include any number of interconnected buses and bridges. The bus will include various circuits linked together by one or more processors and memories. The bus may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface between the bus and the receiver and transmitter. The receiver and transmitter may be the same element, i.e., a transceiver, providing a unit for communicating with various other systems over a transmission medium. The processor is responsible for managing the bus and general processing, while the memory may be used to store data used by the processor during operation.

[0086] Additionally, the electronic device may further include components such as a communication module, an input unit, an audio processor, a display, and a power supply. The processor (or controller, operating control) used may include a microprocessor or other processor device and / or logic device, which receives input and controls the operation of various components of the electronic device; the memory may be one or more of a buffer, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices, which can store the aforementioned data information, and may also store programs for executing the information, and the processor can execute the program stored in the memory to achieve information storage or processing, etc.; the input unit is used to provide input to the processor, for example, it can be a button or touch input device; the power supply is used to provide power to the electronic device; the display is used to display images and text, for example, it can be an LCD display. The communication module is a transmitter / receiver that transmits and receives signals via an antenna. The communication module (transmitter / receiver) is coupled to the processor to provide input signals and receive output signals, which can be the same as in conventional mobile communication terminals. Based on different communication technologies, multiple communication modules can be incorporated into the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module (transmitter / receiver) is also coupled to a speaker and microphone via an audio processor to provide audio output through the speaker and receive audio input from the microphone, thereby enabling typical telecommunications functions. The audio processor can include any suitable buffer, decoder, amplifier, etc. Furthermore, the audio processor is coupled to a central processing unit, enabling on-device recording via the microphone and on-device playback of stored sound via the speaker.

[0087] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0088] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0089] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0090] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the functions specified in one or more boxes. Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0091] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for intelligent recognition of assertion elements of a bank return, the assertion elements including a business type, a transaction date, a payer name, a payer account number, a payer opening bank name, and a payer opening bank number, characterized in that, include: Obtain the bank statement PDF file; The bank statement PDF file is divided into several element regions; Based on the position of the element region, the element region is converted into corresponding pixel data; The original pixel data corresponding to each element region in the return receipt file is split into multiple 32×32 pixel data as input; the pixel data is input into the LeNet-5 convolutional neural network model for initial recognition; the initial recognition result of the assertion element is obtained through the LeNet-5 model; the data after the initial recognition is extracted and compared with the field values ​​in the database; and the initial recognition result of each element region is obtained according to the first verification rule. Print the initial recognition results to the corresponding element area; The initial identification results include assertion verification passed, assertion verification failed, and assertion verification result pending. The first verification rule is: if all characters in the element area pass, the initial recognition result is that the assertion verification is passed; If two or more characters in the element area fail the verification, the initial recognition result is an assertion verification failure. If only one character in the element area fails the verification, the initial recognition result is "assertion verification result pending". Obtain the element region whose initial identification result is assertion verification failure or assertion verification result pending; Adjust the parameters of the LeNet-5 convolutional neural network model to accurately identify the element regions whose initial identification results are assertion verification failure and assertion verification results pending, and obtain accurate identification results according to the second verification rule; Print the accurate recognition results to the corresponding element area; The specific parameter adjustment plan is as follows: In the first layer (convolutional layer), the input layer size of the LeNet-5 model is adjusted to 32×32×1, that is, the original pixel data is split into multiple 16×16 pixel data as input; In the second layer (pooling layer), the input layer is adjusted to a 14×14×6 node matrix; In the third layer (convolutional layer), the input layer is adjusted to a 7×7×6 node matrix; The second verification rule includes: if all characters in the element area pass, the accurate recognition result is that the assertion verification passes; If a character verification fails within the element region, the accurate recognition result is an assertion verification failure.

2. The method as described in claim 1, characterized in that, The LeNet-5 convolutional neural network model has 7 layers, with the first 4 layers consisting of alternating convolutional and pooling layers, and the last 3 layers being fully connected layers.

3. The method as described in claim 2, characterized in that, The first verification rule includes: if all characters in the element area pass, the initial recognition result is that the assertion verification is passed; If two or more characters in the element area fail the verification, the initial recognition result is an assertion verification failure. If only one character in the element area fails verification, the initial recognition result is "assertion verification result pending".

4. The method as described in claim 3, characterized in that, The accurate identification results include assertion verification passing and assertion verification failing.

5. A smart recognition system for assertion elements in bank receipts, characterized in that, include: The acquisition module is used to acquire bank statement PDF files; The segmentation module is used to segment the bank receipt PDF file to generate several element regions; The conversion module is used to convert the element region into corresponding pixel data according to the position of the element region; The first processing module is used to split the original pixel data corresponding to each element area in the return receipt file into multiple 32×32 pixel data as input; The pixel data is input into the LeNet-5 convolutional neural network model for initial recognition. The LeNet-5 model obtains the initial recognition result of the assertion element. The data after the initial recognition is extracted and compared with the field values ​​in the database. The initial recognition result of each element region is obtained according to the first verification rule. The first output module is used to print the initial recognition result to the corresponding element area; the initial recognition result includes assertion verification passed, assertion verification failed, and assertion verification result pending; the first verification rule is: if all characters in the element area pass, the initial recognition result is assertion verification passed; If two or more characters in the element area fail the verification, the initial recognition result is an assertion verification failure. If only one character in the element area fails the verification, the initial recognition result is "assertion verification result pending". The second processing module is used to adjust the parameters of the LeNet-5 convolutional neural network model, accurately identify the element regions whose initial identification results are assertion verification failure and assertion verification results pending, and obtain accurate identification results according to the second verification rule. The second output module is used to print the accurate recognition result to the corresponding element area; The specific parameter adjustment plan is as follows: In the first layer (convolutional layer), the input layer size of the LeNet-5 model is adjusted to 32×32×1, that is, the original pixel data is split into multiple 16×16 pixel data as input; In the second layer (pooling layer), the input layer is adjusted to a 14×14×6 node matrix; In the third layer (convolutional layer), the input layer is adjusted to a 7×7×6 node matrix; The second verification rule includes: if all characters in the element area pass, the accurate recognition result is that the assertion verification passes; If a character verification fails within the element region, the accurate recognition result is an assertion verification failure.

6. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method of any one of claims 1 to 4.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor is configured to execute a computer program stored in a memory, performing the method according to any one of claims 1 to 4.

Citation Information

Patent Citations

  • Hand-written experiment report recognition method for performance evaluation

    CN107463866A

  • Training method and device of target object recognition model and electronic equipment

    CN111325117A

  • Banking business receipt identification method and identification device

    CN114419654A