Contract Verification System and Method Based on 5G and Optical Character Recognition

Abstract

This application relates to a contract verification system and method based on 5G and optical character recognition, belonging to the field of contract bill of materials verification technology. The system includes an image acquisition device, a computer device, and a price database server. The image acquisition device and the computer device are connected via a 5G private network, and the computer device is communicatively connected to the price database server. During the comparison and recognition process between the paper contract and the contract document, the image acquisition device acquires images of the paper contract, and after data acquisition by the computer device, the table image is recognized to determine the text content in the paper image. This method, based on 5G network and high-precision table text recognition, can quickly and accurately extract text content from paper content images, thereby improving the comparison accuracy of paper and electronic contracts under fully automatic recognition.

Description

Technical Field

[0001] This application relates to the field of contract bill of materials verification technology, and in particular to a contract verification system and method based on 5G and optical character recognition (OCR). Background Technology

[0002] As society develops, the number of contracts within enterprises also increases accordingly. Typically, when a contract involves material procurement, staff need to import the price and material list into the enterprise system to achieve digital office operations.

[0003] In related technologies, price and bill of materials contracts are typically documented in paper form, with backups of one or more versions stored as PDFs. During system entry, the PDF and paper versions are usually manually compared for consistency. Once the content of the PDF file is confirmed to be completely identical to the paper document, the electronic archive is imported into the enterprise system. Alternatively, the paper version can be imported into the system via scanning or text recognition.

[0004] However, manual verification methods in related technologies suffer from low efficiency and low accuracy. Summary of the Invention

[0005] This application relates to a contract verification system and method based on 5G and optical character recognition, which can improve the accuracy of comparing paper contracts and electronic contracts. The technical solution is as follows:

[0006] On the one hand, a contract verification system based on 5G and optical character recognition is provided. The system includes an image acquisition device, a computer device, and a price database server. The image acquisition device and the computer device are connected via a 5G private network, and the computer device is connected to the price database server.

[0007] The image acquisition device is used to acquire images of paper contracts and send the images of paper contracts to the computer device via the 5G private network;

[0008] The computer device is configured to receive the paper contract image via a 5G private network, the paper contract image including a table area; perform table recognition on the paper contract image to locate the table area; generate a table area image based on the table area; input the table area image into a table area text recognition model to output a recognized table, the recognized table being the text content in the paper contract image in tabular form; retrieve the contract document, the contract document corresponding to the paper contract image; compare the text content in the contract document with the text content in the recognized table; in response to the successful text content comparison, generate a price list file; and send the price list file to the price database server.

[0009] The price database server is used to receive and store the price list file.

[0010] On the other hand, a contract verification method based on 5G and optical character recognition is provided. This method is applied to a computer device within the contract verification system based on 5G and optical character recognition as described above. The method includes:

[0011] The paper contract image is received via a 5G private network, and the paper contract image includes a table area;

[0012] Perform table recognition on the paper contract image and locate the table area;

[0013] Generate a table area image based on the table area;

[0014] The table area image is input into the table area text recognition model, and the recognized table is output as the text content in the paper contract image in table form.

[0015] Retrieve the contract document, which corresponds to the image of the paper contract;

[0016] The text content in the contract document is compared with the text content in the recognition table.

[0017] Upon successful comparison of the text content, a price list file is generated; the price list file is then sent to the price database server.

[0018] The beneficial effects of the technical solution provided in this application include at least the following:

[0019] In the process of comparing and recognizing paper contracts with electronic contracts, image acquisition devices are used to capture images of the paper contracts. After data collection via a 5G private network, the computer locates and recognizes the table areas, and then uses a table area text recognition model to identify the text content within the table. Once the content of the paper contract is compared and recognized with that in the electronic contract document, a price list file is generated and stored in the price database server. This method, based on a fast 5G network and high-precision table text recognition, can quickly and accurately extract text content from paper content images, thereby improving the accuracy of comparing paper and electronic contracts in a fully automated manner. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This illustration shows a schematic diagram of the structure of a contract verification system based on 5G and optical character recognition, provided in an exemplary embodiment of this application.

[0022] Figure 2 The illustration shows a flowchart of a contract verification system method based on 5G and optical character recognition provided in an exemplary embodiment of this application.

[0023] Figure 3 This illustration shows a flowchart of another contract verification method based on 5G and optical character recognition provided in an exemplary embodiment of this application.

[0024] Figure 4 A schematic diagram of the structure of a text recognition sub-model provided in an exemplary embodiment of this application is shown. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0026] Figure 1 This illustration shows a schematic diagram of a contract verification system based on 5G and optical character recognition, provided in an exemplary embodiment of this application. Please refer to [the diagram]. Figure 1The schematic diagram of this contract verification system based on 5G and optical character recognition includes an image acquisition device 110, a computer device 120, and a price database server 130. The image acquisition device and the computer device are connected via a 5G private network, and the computer device is connected to the price database server.

[0027] In this application embodiment, the image acquisition device can be implemented as a camera, scanner, or other device capable of acquiring the content of paper contracts. Optionally, when the image acquisition device is implemented as a camera, the acquired file is a JPG image file; when the image acquisition device is implemented as a scanner, the acquired file is a portable document file in PDF format. In this application, the example of an image acquisition device acquiring a portable document file in PDF format will be used for illustration.

[0028] Optionally, the computer equipment and image acquisition equipment are connected via a 5G private network to further improve data transmission speed compared to conventional methods. In this embodiment, the computer equipment is equipped with Robotic Process Automation (RPA) system software. Through the RPA system, the computer can realize a streamlined contract identification and verification process. In this embodiment, the computer equipment also has data storage capabilities to adapt to the reading of contract documents. In this case, the computer equipment needs to be configured with a 5G communication module.

[0029] In this embodiment, the price database server is communicatively connected to a computer device and is implemented as a dedicated server for storing price list files. Optionally, the price database server has a front-end platform interface that allows users to access price list files and is connected to the enterprise's management platform. This embodiment does not limit the actual functions and application methods of the price database server.

[0030] Based on the above description, the image acquisition device is used to acquire images of paper contracts and transmit them to a computer via a 5G private network. The computer receives the paper contract images via the 5G private network; the paper contract images include table areas; it performs table recognition on the paper contract images to locate the table areas; it generates table area images based on the table areas; it inputs the table area images into a table area text recognition model and outputs a recognized table, which is the text content in the paper contract image in tabular form; it retrieves the contract document, which corresponds to the paper contract image; it compares the text content in the contract document with the text content in the recognized table; in response to a successful text content comparison, it generates a price list file; and it sends the price list file to a price database server. The price database server receives and stores the price list files.

[0031] In summary, the system provided in this application, during the comparison and recognition process between paper contracts and contract documents, acquires paper images using an image acquisition device, and after data acquisition via a 5G private network using a computer device, locates the table area and recognizes the table image. It then uses a table area text recognition model to identify the text content within the table, determining the text content in the paper image. After comparing and recognizing the content of the paper contract with the content in the contract document, a price list file is generated and stored in the price database server. This method, based on a fast 5G network and high-precision table text recognition, can quickly and accurately extract text content from paper content images, thereby improving the accuracy of comparing paper and electronic contracts under fully automatic recognition.

[0032] Figure 2 This application illustrates a flowchart of a contract verification system method based on 5G and optical character recognition, provided in an exemplary embodiment, illustrating how this method can be applied to, for example... Figure 1 In the computer equipment within the contract verification system based on 5G and optical character recognition, the method includes:

[0033] Step 201: Receive the paper contract image via the 5G private network. The paper contract image includes a table area.

[0034] In this embodiment of the application, the computer device is equipped with a 5G communication module for receiving images of paper contracts.

[0035] This application is specifically designed for identifying bill of materials (BOM) contracts. Therefore, corresponding to the actual presentation of the BOM contract, the paper contract image should contain a table area. If the paper contract image does not contain a table area, it can be filtered out through a table location process.

[0036] Step 202: Perform table recognition on the paper contract image and locate the table area.

[0037] This process involves locating and extracting the areas in a paper contract image that require text recognition. This application embodiment does not limit the method for determining the table area. Optionally, the computer device determines the table area based on feature extraction; or, the computer device determines the table area by determining edges.

[0038] Step 203: Generate a table area image based on the table area.

[0039] Optionally, after recognizing the table region, the computer device obtains a table region image by cropping, and the table region image corresponds to a portion of the paper contract image.

[0040] Step 204: Input the table area image into the table area text recognition model and output the recognized table, which is the text content in the paper contract image in table format.

[0041] In this embodiment, the text recognition model for the table region is a neural network model. Optionally, this neural network model has an active learning function, meaning that after the computer device receives the basic parameters of the model and establishes a basic model, the computer device can adjust the model structure using labeled samples, ultimately enabling the model to accurately output results. Optionally, after inputting the table region image into the recognition model, the computer device will obtain a recognition table, which outputs the text content corresponding to the table region in the paper contract image in tabular form.

[0042] Step 205: Retrieve the contract document, which corresponds to the image of the paper contract.

[0043] Once the text content corresponding to the table area is determined, the computer device will retrieve the contract document. Optionally, the computer device may retrieve the contract document based on the contract number, or the contract name, or a combination of the contract name and number, or other contract identifiers corresponding to the contract. In this embodiment, the retrieved contract document is an editable document in docx, xlsx, or other formats.

[0044] Step 206: Compare the text content in the contract document with the text content in the recognition table.

[0045] Optionally, the comparison process includes comparing the number of cells in the table, comparing the positional distribution of cells in the table, and comparing the text content within each cell. In this embodiment, the text comparison is considered successful only when the text content in the contract document is completely consistent with the text content in the recognition table.

[0046] Step 207: In response to the successful text content comparison, a price list file is generated and sent to the price database server.

[0047] Once the text comparison is successful, the computer device generates a price list file and sends it to the server for storage and subsequent retrieval.

[0048] In summary, the method provided in this application, during the comparison and recognition of paper contracts and contract documents, involves acquiring paper images using an image acquisition device, collecting data via a 5G private network on a computer, locating table areas and recognizing table images, and then using a table area text recognition model to identify the text content in the table. After comparing and recognizing the content of the paper contract with the content in the contract document, a price list file is generated and stored in the price database server. This method, based on a fast 5G network and high-precision table text recognition, can quickly and accurately extract text content from paper content images, thereby improving the accuracy of comparing paper and electronic contracts under fully automatic recognition.

[0049] Figure 3 This application illustrates a flowchart of another contract verification method based on 5G and optical character recognition, provided in an exemplary embodiment, illustrating how this method can be applied to applications such as... Figure 1 The following is an example of a contract verification system based on 5G and optical character recognition. The method includes:

[0050] Step 301: Send scan data to the image acquisition device.

[0051] In this embodiment, the image acquisition device is implemented as a scanner. The scan data is used to indicate the scanning size of the image acquisition device; that is, the image acquisition device needs to determine the scanning position and the file size of the generated scanned document based on the scan data.

[0052] After receiving the scanned data, the image acquisition device will acquire the image of the paper contract based on the scanned data.

[0053] Step 302: Receive the paper contract image via the 5G private network. The paper contract image includes a table area.

[0054] This process corresponds to the process shown in step 202, and will not be described in detail here.

[0055] Step 303: Perform horizontal and vertical line recognition and focus recognition on the paper contract image to obtain the horizontal and vertical line recognition results and focus recognition results.

[0056] Step 304: Based on the horizontal and vertical line recognition results and the focus recognition results, perform table existence verification.

[0057] In this embodiment of the application, the determination of the valid content within the paper contract image, i.e. the table area, will be performed by sequentially identifying horizontal and vertical lines and focusing the data to verify whether the table exists.

[0058] Step 305: In response to the validation result indicating that the table exists, locate the table area.

[0059] Optionally, when it is determined that the table does not exist, an alarm is triggered through the RPA system software to indicate that there is no table area in the image.

[0060] Step 306: Generate a cell coordinate system based on the table area.

[0061] Step 307: Locate the cell coordinates based on the cell coordinate system.

[0062] Step 308: Optimize cell merging based on cell coordinates.

[0063] Steps 306 to 308 describe the method for extracting features of the table area when its existence is confirmed. Since the subsequent comparison process is a cell-by-cell text content comparison, in this embodiment, for identifiable cells within the table area, cell division and cell identifier determination will be performed. Optionally, the cell identifier corresponds to the coordinates of the cell's location. In one example, the identifier for the cell in the first row and first column is "11," and the identifier for the cell in the second row and first column is "21." This application does not limit the specific presentation format of the cell identifier and cell coordinates.

[0064] Step 309: Generate an image of the table area after cell merging optimization, corresponding to the table area.

[0065] In this embodiment of the application, the table area image is marked with multiple cells and their corresponding positions. In the subsequent recognition process, this embodiment of the application can directly input the table area image into the recognition model, or it can be input into the recognition model in a distributed manner.

[0066] Step 310: Input the image of the table region into the input terminal of the text detection sub-model of the table region text recognition model.

[0067] This application embodiment illustrates the method of directly inputting a table area image into a table area text recognition model. In other embodiments of this application, the computer device inputs each cell into the table area text recognition model, obtains the cell recognition results, and then concatenates the cell recognition results according to the cell coordinates to generate a recognition table.

[0068] In this embodiment, the table region text recognition name includes interconnected text detection sub-models, text correction sub-models, text recognition sub-models, and table structuring sub-models. Therefore, inputting the table region image into the table region text recognition model is equivalent to inputting the table region image into the input terminal of the text detection sub-model.

[0069] Step 311: The text detection sub-model outputs the results based on the DBnet algorithm.

[0070] In this embodiment of the application, the table content in the text detection process varies in length and has line breaks, which is irregular text detection. Therefore, the DBnet algorithm is used to perform text detection in order to determine the location of the text.

[0071] Step 312: Using the text correction sub-model, a reference rectangle is set, and the position of the text within the text area is corrected based on the reference rectangle to obtain the correct text content.

[0072] In this embodiment, the text correction sub-model is used to adjust characters that are tilted or excessively large in the text. Optionally, in this sub-model, the image after reading the text box image is corrected into a rectangular frame, and the text direction is judged and corrected at the same time; that is, if the text is tilted, it will be straightened.

[0073] Step 313: The CRNN algorithm with CTC decoding is applied to the text recognition sub-model to perform text recognition on the positive text content, and the text recognition content is obtained.

[0074] In this embodiment, the text recognition sub-model is divided into two main categories based on the shape of the text to be recognized: regular text recognition and irregular text recognition. Regular text mainly refers to printed fonts, scanned text, etc., and the text is generally located at a horizontal position. Irregular text is often not at a horizontal position and has problems such as curvature, occlusion, and blurriness. Regular text recognition is mainly based on the CRNN algorithm with CTC decoding. The CRNN algorithm does not require character segmentation of the sample data and can recognize text sequences of arbitrary length. Figure 4 As shown, the text recognition sub-model includes a sequentially connected convolutional layer 410, a recurrent layer 420, and a transcription layer 430.

[0075] Step 314: The text recognition content is structured using a table structuring sub-model, and the recognition table is output.

[0076] In this embodiment of the application, after the text recognition is completed, the text recognition content is structurally adjusted and output based on the cell position content in the table image, and finally the recognition table is obtained.

[0077] Step 315: Retrieve the contract documents.

[0078] Step 316: Compare the text content in the contract document with the text content in the recognition table.

[0079] Step 317: In response to the successful text content comparison, a price list file is generated.

[0080] It should be noted that in some embodiments of this application, the comparison can only be passed if the comparison result is 100% similar, that is, completely identical. Therefore, the computer device can display and arrange the reasons for the comparison failure to adapt to the subsequent manual verification and adjustment process.

[0081] Step 318: Send the price list file to the price database server.

[0082] In summary, the method provided in this application, during the comparison and recognition of paper contracts and contract documents, involves acquiring paper images using an image acquisition device, collecting data via a 5G private network on a computer, locating table areas and recognizing table images, and then using a table area text recognition model to identify the text content in the table. After comparing and recognizing the content of the paper contract with the content in the contract document, a price list file is generated and stored in the price database server. This method, based on a fast 5G network and high-precision table text recognition, can quickly and accurately extract text content from paper content images, thereby improving the accuracy of comparing paper and electronic contracts under fully automatic recognition.

[0083] The method provided in this application uses 5G network transmission, which has a fast transmission speed.

[0084] The method provided in this application combines OCR recognition technology with RPA (Robotic Process Automation) to improve the accuracy and efficiency of the contract bill of materials.

[0085] The method provided in this application embodiment saves the computational load of computer equipment by performing pre-recognition of table areas and cropping of images.

[0086] The method provided in this application modularizes the model, enabling the text content within the table area to be gradually identified. It also personalizes different model structures according to the actual situation of the bill of materials, further improving the accuracy of recognition.

[0087] The above are merely optional embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A 5G and optical character recognition based contract verification system, characterized in that, The contract verification system based on 5G and optical character recognition comprises an image acquisition device, a computer device and a price database server, the image acquisition device and the computer device are connected through a 5G private network, and the computer device is connected with the price database server; The image acquisition device is used for acquiring a paper contract image and sending the paper contract image to the computer device through the 5G private network. The computer device is used for receiving the paper contract image through the 5G private network, the paper contract image comprises a table region, performing table recognition on the paper contract image, positioning the table region, generating a table region picture based on the table region, inputting the table region picture into a table region text recognition model, and outputting to obtain an identified table, the identified table is text content in the paper contract image in the form of a table, calling a contract document corresponding to the paper contract image, comparing the text content in the contract document with the text content in the identified table, and generating a price list file in response to the text content comparison passing; The price list file is sent to the price database server; The price database server is used for receiving and storing the price list file. The table region comprises at least one cell, and the computer device is further used for generating a cell coordinate system based on the table region, positioning a cell coordinate based on the cell coordinate system, and performing cell merging optimization based on the cell coordinate. A table region picture corresponding to the table region and subjected to cell merging optimization is generated.

2. The system of claim 1, wherein, The image acquisition device is realized as a scanner. The computer device is further used for sending scanning data to the image acquisition device, the scanning data being used for instructing a scanning size of the image acquisition device, and the image acquisition device is further used for receiving the scanning data and acquiring the paper contract image based on the scanning data.

3. The system of claim 1, wherein, The computer device is further used for performing horizontal and vertical line recognition and focus recognition on the paper contract image to obtain horizontal and vertical line recognition results and focus recognition results, performing table existence verification based on the horizontal and vertical line recognition results and the focus recognition results, and positioning the table region in response to a verification result of the table existence verification indicating that a table exists.

4. The system of claim 1, wherein The computer device is further used for inputting cells into the table region text recognition model one by one to output cell recognition results, and generating the identified table based on the cell recognition results. The table region text recognition model comprises text detection sub-models, text correction sub-models, text recognition sub-models and table structure sub-models connected with each other.

5. The system of claim 1, wherein, ​ The computer device is further configured to input the table region picture into an input end of a text detection sub-model of a table region text recognition model; output a text detection sub-model output result based on a DBnet algorithm through the text detection sub-model, the text detection sub-model output result being used to indicate the position of the text in the table region; set a reference rectangular frame through the text correction sub-model, and correct the position of the text in the text region based on the reference rectangular frame to obtain correctly positioned text content; recognize the correctly positioned text content through the text recognition sub-model by applying a CRNN algorithm of a CTC decoding mode to obtain text recognition content; structure the text recognition content through the table structure sub-model to obtain the recognized table.

6. The system of claim 5, wherein, The text recognition sub-model comprises sequentially connected convolution layers, recurrent layers and transcription layers.

7. A 5G and optical character recognition based contract verification method, characterized by, The method is applied to a computer device in a 5G-based optical character recognition contract verification system as claimed in any one of claims 1 to 6, and the method comprises: receiving a paper contract image through a 5G private network, the paper contract image comprising a table region; performing table recognition on the paper contract image to locate the table region; generating a table region picture based on the table region; inputting the table region picture into a table region text recognition model to output a recognized table, the recognized table being text content in the paper contract image in the form of a table; calling a contract document, the contract document corresponding to the paper contract image; comparing the text content in the contract document with the text content in the recognized table; generating a price list file in response to the text content comparison being passed; and sending the price list file to a price library server.

Images