Official seal automatic identification method and system based on enterprise digital application project

Abstract

The embodiment of the invention provides an official seal automatic identification method and system based on an enterprise digital application project, and belongs to the technical field of image identification. The method comprises the following steps: acquiring a to-be-identified file, and processing the to-be-identified file into corresponding image data; calling the trained official seal recognition model based on the image data, and outputting an official seal recognition result; wherein the official seal identification model is obtained by training in the MobileNet based on a YOLO algorithm; and judging whether the current to-be-identified file is signed to be compliant or not based on the official seal identification result. Compared with a traditional manual auditing mode, the scheme has the advantages that the recognition efficiency is remarkably improved, the risk of manual misjudgment or missed judgment is reduced, high precision and speed can be kept when a large number of files are processed, and the method is particularly suitable for compliance management of large enterprises.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to an automatic seal recognition method based on enterprise digital application projects and an automatic seal recognition system based on enterprise digital application projects. Background Art

[0002] The official seal is an important symbol when an enterprise signs legal documents such as contracts and documents externally, and has legal effect. Especially for large group enterprises, numerous projects of subsidiaries and branches will generate a large number of documents that need to be reviewed in daily operations. These documents must be reviewed during the upload process to ensure their compliance and legal validity. Therefore, how to automatically detect whether the official seal is affixed when the document is uploaded has become a key requirement, which can effectively reduce the burden on reviewers and improve the overall operation efficiency of the group. However, the existing review methods mainly rely on manual inspection, which not only takes a long time, but also is difficult to maintain high efficiency when dealing with a large number of documents, and is prone to misjudgment or missed judgment problems caused by fatigue or distraction, thus affecting the accuracy of the review results.

[0003] Although the current technology for automatically recognizing official seals can improve efficiency to a certain extent, there are still some technical bottlenecks in practical applications. First of all, for different types of documents, their formats and the positions where the official seals are affixed may vary, and coupled with the change of the seal angle, this poses a challenge to algorithm design. Secondly, the background text, images or other visual interference information in the document may affect the accurate recognition of the official seal. Especially when the color of the official seal is uneven, the difficulty of recognition will be further increased. In addition, the existing recognition algorithms have high requirements for the clarity of the image. Once the document scanning quality is poor, the recognition rate of the algorithm will drop significantly. These problems limit the applicability of the existing automation solutions in practical scenarios and are difficult to completely replace manual review.

[0004] Therefore, there is an urgent need for an automatic solution that can accurately and quickly recognize official seals in different document formats and complex backgrounds, overcome the interference factors and recognition difficulties in the existing technology, and meet the high-efficiency requirements of large enterprises for project compliance review. Summary of the Invention

[0005] The purpose of the embodiments of the present invention is to provide an automatic seal recognition method and system based on enterprise digital application projects, so as to at least solve the problem that the current solution cannot accurately and quickly recognize official seals in different document formats and complex backgrounds.

[0006] To achieve the above object, a first aspect of the present invention provides an automatic seal recognition method based on enterprise digital application projects, and the method includes: collecting a file to be recognized and processing the file to be recognized into corresponding image data; calling a trained seal recognition model based on the image data and outputting a seal recognition result; wherein, the seal recognition model is trained based on the YOLO algorithm in MobileNet; judging whether the current file to be recognized is signed in compliance based on the seal recognition result.

[0007] Optionally, the processing the file to be recognized into corresponding image data includes: performing non-image data recognition in the file to be recognized as data to be converted; matching and executing corresponding image conversion methods based on the data type of the data to be converted to obtain image data corresponding to each data to be converted.

[0008] Optionally, the method further includes training the seal recognition model; the training rule of the seal recognition model is: collecting historical image information including seal images; pushing each piece of historical image information to the user side for the user to mark the seal area and recycling the marking result; constructing a training sample based on the historical image information with a seal area marked, and performing model training based on the YOLO algorithm in MobileNet to obtain a seal recognition model.

[0009] Optionally, the pushing each piece of historical image information to the user side for the user to mark the seal area includes: opening each piece of historical image information on the user side based on the PPOCRLabel tool and recycling the marking result of the seal area in each piece of historical image information by the user based on the PPOCRLabel tool.

[0010] Optionally, the constructing a training sample based on the historical image information with a seal area marked includes: converting the historical image information with a seal area marked into the VOC format; splitting the historical image information in the VOC format according to a preset ratio to obtain a training set, a validation set and a test set respectively.

[0011] Optionally, performing model training based on the YOLO algorithm in MobileNet to obtain a seal recognition model includes: performing model training based on the training set and the YOLO algorithm in MobileNet to obtain an initial seal recognition model; performing fine-tuning of the initial seal recognition model based on the validation set to obtain a seal recognition model; after obtaining the seal recognition model, the method further includes: performing performance testing of the seal recognition model based on the test set.

[0012] Optionally, the performance test of the official seal recognition model based on the test set includes: performing the IoU test and mAP test of the recognition model based on the test set respectively, and judging whether the performance of the official seal recognition model meets the expectation based on the test results; if the performance of the official seal recognition model meets the expectation, performing subsequent official seal recognition based on the official seal recognition model; if the performance of the official seal recognition model does not meet the expectation, re-performing model training.

[0013] Optionally, the judgment of whether the current file to be recognized is signed in compliance based on the official seal recognition result includes: judging the number of official seals and the types of official seals in the current file to be recognized based on the official seal recognition result respectively; if the number of official seals meets the target number of audited official seals and the types of official seals meet the target types of official seals, it is determined that the current file to be recognized is signed in compliance; if the number of official seals does not meet the target number of audited official seals, or the types of official seals do not meet the target types of official seals, it is determined that the current file to be recognized is not signed in compliance.

[0014] The second aspect of the present invention provides an automatic official seal recognition system based on enterprise digital application projects, and the system includes: a collection unit for collecting the file to be recognized and processing the file to be recognized into corresponding image data; an identification unit for calling the trained official seal recognition model based on the image data and outputting an official seal recognition result; wherein, the official seal recognition model is trained based on the YOLO algorithm in MobileNet; an audit unit for judging whether the current file to be recognized is signed in compliance based on the official seal recognition result.

[0015] On the other hand, the present invention provides a computer-readable storage medium, and instructions are stored on the computer-readable storage medium, and when the instructions run on a computer, the computer is made to execute the above-mentioned automatic official seal recognition method based on enterprise digital application projects.

[0016] Through the above technical solutions, the solution of the present invention realizes the efficient detection of official seals in documents by processing the file to be recognized into image data and using the pre-trained official seal recognition model. The official seal recognition model is trained on MobileNet based on the YOLO algorithm and has efficient image feature extraction and processing capabilities, and can quickly and accurately identify the position and status of official seals in documents. By outputting the official seal recognition result, the system can further judge whether the document is stamped with an official seal, so as to realize the automatic detection of the compliance of document signing. Compared with the traditional manual review method, this solution significantly improves the recognition efficiency, reduces the risk of manual misjudgment or missed judgment, and can maintain high accuracy and speed when processing a large number of documents, and is especially suitable for the compliance management of large enterprises.

[0017] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent specific embodiment part. Description of the Drawings

[0018] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and form a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present invention, but do not constitute a limitation to the embodiments of the present invention. In the accompanying drawings:

[0019] Figure 1 is a flowchart of the steps of a method for automatically recognizing official seals based on enterprise digital application projects provided by an embodiment of the present invention;

[0020] Figure 2 is a system structure diagram of a system for automatically recognizing official seals based on enterprise digital application projects provided by an embodiment of the present invention. Specific Embodiments

[0021] The following will describe in detail the specific embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not used to limit the present invention.

[0022] Figure 1 is a method flowchart of a method for automatically recognizing official seals based on enterprise digital application projects provided by an embodiment of the present invention. As Figure 1 shown, an embodiment of the present invention provides a method for automatically recognizing official seals based on enterprise digital application projects, and the method includes:

[0023] Step S10: Collect the file to be recognized and process the file to be recognized into corresponding image data.

[0024] Specifically, non-image data recognition is performed in the file to be recognized as data to be converted; based on the data type of the data to be converted, the corresponding image conversion method is matched and executed to obtain the image data corresponding to each data to be converted.

[0025] In the embodiments of the present invention, the content in the file usually includes two types of data: image data and non-image data, such as text, tables, etc. In order to achieve comprehensive recognition of the file content, it is necessary to be able to distinguish and process these two types of data. Among them, non-image data such as text information cannot be directly used for official seal recognition, so this type of data needs to be specifically processed and converted into image data for subsequent unified processing and recognition.

[0026] The specific implementation method is as follows: First, non-image data in the file to be recognized is identified, and elements such as text, tables, and symbols are extracted as the data to be converted. Then, based on the data types of these data to be converted, corresponding image conversion methods are matched and executed. For example, for text data, optical character recognition (OCR) technology is used to convert it into an image, or an image representation of the text is generated according to the layout rules. Table data can be presented in a graphical form in a similar way. After this process is completed, all the content in the file will be converted into image data, forming a unified image input for subsequent official seal recognition processing.

[0027] Based on the solution of the present invention, by uniformly converting non-image data in the file into image data, consistent processing and recognition of all the content of the file can be performed, avoiding the risk of only processing partial image data and ignoring other important information. This process ensures the accuracy and comprehensiveness of official seal recognition in the case of diverse file types and complex formats. In addition, customized conversion methods are adopted for different types of data, further improving the efficiency of data conversion and the quality of image generation, ensuring the input quality of the official seal recognition model, and thus improving the reliability and accuracy of the entire recognition process.

[0028] Step S20: Invoke the trained official seal recognition model based on the image data and output the official seal recognition result.

[0029] Specifically, the method further includes training the official seal recognition model; the training rule of the official seal recognition model is: collecting historical image information containing official seal images; pushing each piece of historical image information to the user side for the user to mark the official seal area, and recovering the marking results; constructing training samples based on the historical images with official seal area markings, and performing model training in MobileNet based on the YOLO algorithm to obtain the official seal recognition model.

[0030] In the embodiment of the present invention, the method further includes training the official seal recognition model; the training rule of the official seal recognition model is: collecting historical image information containing official seal images; pushing each piece of historical image information to the user side for the user to mark the official seal area, and recovering the marking results; constructing training samples based on the historical images with official seal area markings, and performing model training in MobileNet based on the YOLO algorithm to obtain the official seal recognition model.

[0031] In the embodiment of the present invention, the training process of the model is based on the collection of historical image information and user annotation feedback. First, a large amount of historical image information containing official seals is collected as the initial training data source. These historical images may be from real contracts, document scans, etc. To further improve the accuracy of the model and the positioning accuracy of the official seal area, these images are pushed to the user side for the user to annotate the specific area of the official seal. Through the user's annotation feedback, these annotation results can be recovered, and training samples are constructed based on these historical images with the official seal area annotated.

[0032] Furthermore, after the training samples are constructed, the model training will be carried out based on MobileNetV3 in combination with the YOLO algorithm. As a lightweight convolutional neural network, MobileNetV3 has the advantages of simple structure and low computational cost, and is suitable for model applications in mobile devices or resource-constrained environments. The YOLO (You Only Look Once) algorithm is famous for its efficiency in object detection tasks and can perform object localization and classification simultaneously in an image. Combining the advantages of these two technologies, the official seal recognition model not only maintains high accuracy but also has excellent computational performance, especially suitable for the automatic recognition of official seals when processing a large number of documents.

[0033] Based on the solution of the present invention, through the user's annotation of the official seal area, the model can obtain accurate annotation data, ensuring the high quality of the training samples, thereby improving the recognition accuracy of the model in practical applications. Based on the lightweight design of MobileNetV3 and the efficient characteristics of the YOLO algorithm, the model achieves a good balance between performance and accuracy. This enables the model to not only run in a server environment but also operate efficiently in resource-constrained environments such as mobile devices, meeting the official seal recognition needs of enterprises in various scenarios.

[0034] Preferably, the pushing of each historical image information to the user side for the user to annotate the official seal area includes: opening each historical image information on the user side based on the PPOCRLabel tool and recovering the user's annotation results for the official seal area in each historical image information based on the PPOCRLabel tool.

[0035] In the embodiments of the present invention, the collected historical image information is pushed to the user side, and the user can use the PPOCRLabel tool to label the official seal areas in these images. PPOCRLabel is an open-source OCR (Optical Character Recognition) annotation tool, which is widely used in the annotation process of text and object detection tasks. Through this tool, the user can conveniently and accurately mark the specific position and scope of the official seal in the image. The PPOCRLabel tool can intuitively display the image information, allowing the user to manually select the area where the official seal is located and label these areas with annotation tags. The interface of this tool is simple and the operation is easy. It supports image input and output in multiple formats, so it is applicable to various types of scanned images of documents. After the user completes the annotation, the tool will automatically save these annotation results to form an annotation dataset containing the official seal areas. By recycling these annotation data, high-quality training samples are formed, which are further used for the training of the official seal recognition model.

[0036] Based on the solution of the present invention, by the user using the PPOCRLabel tool to annotate the official seal area, the accuracy and efficiency of the annotation are significantly improved. Compared with the automatic annotation method, manual annotation can more accurately calibrate the official seal area in the case of complex backgrounds or interference, avoiding mislabeling or missing labeling, thereby improving the quality of the model training data. At the same time, the use of the PPOCRLabel tool simplifies the user operation process, enabling the annotation work of a large number of images to be completed quickly. With these accurate annotation data, the construction of the training samples is more accurate, which helps to improve the detection accuracy of the official seal recognition model in the actual scenario.

[0037] Preferably, constructing the training sample based on the historical image information with the official seal area annotation includes: converting the historical image information with the official seal area annotation into the VOC format; splitting the historical image information in the VOC format according to a preset ratio to obtain a training set, a validation set, and a test set respectively.

[0038] In the embodiments of the present invention, first, the historical image information with the official seal area annotation is converted into the VOC (Pascal Visual Object Classes) format. The VOC format is a widely used object detection data format, which includes image files and corresponding annotation files (such as XML format), where the annotation files record detailed information such as the category and position (represented in the form of a bounding box) of the target objects in the image. By converting the annotated official seal area into the VOC format, the system can standardize the processing of these annotation data to make it applicable to mainstream object detection algorithms, such as YOLO.

[0039] Furthermore, after converting to the VOC format, the system will split these data according to a preset ratio to generate a training set, a validation set, and a test set. The training set is used for the initial training of the model, the validation set is used to adjust the model parameters during the training process, and the test set is used to finally evaluate the performance of the model. Through this data splitting method, the model can be fully learned during the training process, and at the same time, through the evaluation of the validation set and the test set, it can be ensured that the model has good generalization ability and will not overfit to the training data. For example, it is allocated according to the ratio of 70% training set, 15% validation set, and 15% test set.

[0040] Based on the solution of the present invention, by converting the labeled data into the VOC format, data processing can be carried out with a unified standard, improving data compatibility and processing efficiency, and ensuring that the labeled information can be smoothly used in the subsequent model training process. A reasonable method for splitting the training set, validation set, and test set ensures the scientificity and reliability of model training. The training set ensures that the model has sufficient learning data, while the validation set can help the model adjust and optimize parameters during the training process, avoiding the model's over-reliance on the training data. Finally, the introduction of the test set provides a reliable basis for the final performance evaluation of the model, enabling the model to be fully tested and verified before deployment and having good performance in actual applications.

[0041] Preferably, model training is performed based on the YOLO algorithm in MobileNet to obtain a seal recognition model, including: in MobileNet, model training is performed based on the training set and the YOLO algorithm to obtain an initial seal recognition model; the initial seal recognition model is fine-tuned based on the validation set to obtain a seal recognition model; after obtaining the seal recognition model, the method further includes; performing a performance test on the seal recognition model based on the test set.

[0042] Furthermore, the performance test on the seal recognition model based on the test set includes: respectively performing an IoU test and an mAP test on the recognition model based on the test set, and judging whether the performance of the seal recognition model meets the expectations based on the test results; if the performance of the seal recognition model meets the expectations, subsequent seal recognition is performed based on the seal recognition model; if the performance of the seal recognition model does not meet the expectations, model training is re-executed.

[0043] In the embodiment of the present invention, during the model training stage, the system performs initial training on the MobileNet model based on the training set and the YOLO algorithm. The goal of the initial training is to enable the model to learn various features of the seal, including shape, color, position, etc., through a large amount of training data, so as to have a preliminary seal recognition ability.

[0044] Furthermore, after the initial training is completed, the model is fine-tuned based on the validation set. The role of the validation set is to help evaluate the performance of the model and optimize the model parameters on this basis to avoid overfitting or underfitting. At this stage, the model will further learn and adjust according to the data in the validation set to improve its adaptability to unknown data. Through this fine-tuning process, the model can better handle the seal recognition task in complex scenarios, such as seals at different angles, positions, and backgrounds.

[0045] Furthermore, after obtaining the fine-tuned seal recognition model, the system will also perform a performance test on the model based on the test set. This step is a key link before model deployment to ensure that the model can maintain high recognition ability in actual applications. The specific performance tests include two key indicators: IoU (Intersection over Union) test and mAP (mean Average Precision) test. The IoU test is used to evaluate the overlap between the predicted seal position by the model and the actual seal position. The higher the IoU value, the better the accuracy of the model's positioning; while the mAP test measures the overall precision of the model at different thresholds. The higher the mAP value, the more accurately the model can maintain high recognition accuracy in various scenarios.

[0046] Furthermore, based on these test results, the system will determine whether the performance of the seal recognition model meets the expectations. If the model performance meets the set standards, such as the IoU value and mAP value reaching the expected thresholds (e.g., IoU value greater than 0.5, mAP greater than 95%), the model can enter the actual application stage for subsequent seal recognition tasks. If the model performance does not meet the standard, the system will re-execute the model training and further adjust the model parameters according to the test feedback until the model meets the expected performance requirements. This iterative optimization process ensures the reliability and stability of the model before deployment.

[0047] Based on the solution of the present invention, through the systematic model training, fine-tuning and testing processes, the efficiency and accuracy of the official seal recognition model are ensured. First, based on the lightweight design of MobileNet and combined with the object detection ability of the YOLO algorithm, the model can quickly perform official seal recognition in resource-constrained environments, especially suitable for running on mobile devices or embedded systems. In addition, the introduction of the validation set and the fine-tuning process make the performance of the model more robust in actual applications, and it can handle complex image scenarios and diverse official seal styles. In the testing stage, through the evaluation of multiple performance indicators such as IoU and mAP, the overall performance of the model in official seal positioning and recognition accuracy is ensured. If the model fails to meet the expected standards in the test, the system will further improve the model performance through retraining and tuning. This iterative optimization mechanism not only improves the reliability of the model but also continuously improves the performance of the model to ensure its applicability in different application scenarios.

[0048] Preferably, the trained official seal recognition model is integrated into the file upload system. After integration, the system can automatically trigger the official seal recognition model to detect the official seals in the file when the user uploads a file that needs to be subject to compliance detection. The system will intuitively display the number of official seals contained in the document. In addition, the system also provides user-friendly operation options. For example, when the official seal is not correctly recognized, the user can mark a note to request manual review, ensuring the flexibility of the system and the usability in complex scenarios.

[0049] Step S30: Determine whether the currently to-be-recognized file is signed in compliance based on the official seal recognition result.

[0050] Specifically, based on the official seal recognition result, respectively determine the number and type of official seals in the currently to-be-recognized file; if the number of official seals meets the target number of official seals for review and the type of official seals meets the target type of official seals, it is determined that the currently to-be-recognized file is signed in compliance; if the number of official seals does not meet the target number of official seals for review or the type of official seals does not meet the target type of official seals, it is determined that the currently to-be-recognized file is not signed in compliance.

[0051] In the embodiment of the present invention, after the official seal recognition model completes the detection of the official seals in the file, the system will further analyze the number and type of official seals in the file according to the recognition result. This process can ensure that the official seals stamped on the file conform to the expected review standards to determine whether the file meets the signing compliance requirements.

[0052] Further, check the number of official seals in the current file to be recognized and determine whether it meets the number of official seals required by the target review. For example, some legal documents may require multiple official seals to be affixed for validity, such as when both the company and its subsidiary seal the document. The system compares the number of official seals in the recognition result with the expected standard. If the number of official seals is insufficient or exceeds the requirement, the system will directly determine that the document signing is non-compliant. It will also verify the type of official seal recognized to ensure that the affixed official seal belongs to the target official seal type. Each company usually has different types of official seals, such as contract seals, financial seals, etc., and different types of official seals have different legal effects. By comparing the types of official seals, the system can further ensure the legality of document signing. For example, certain documents must use contract seals instead of financial seals or other official seals. If the type of official seal recognized does not meet the expectation, the system will also determine that the document signing is non-compliant.

[0053] Based on the solution of the present invention, through the automated verification of the number and type of official seals, the compliance of document signing can be quickly and accurately determined. Compared with the traditional manual review method, this method significantly improves the review speed, especially enabling efficient automation when processing a large number of documents. It can reduce compliance issues caused by human misjudgment or overlooking details of official seals. In addition, this automated process can ensure the consistency of review standards, maintaining a stable judgment standard regardless of the number of documents processed and avoiding inconsistencies in manual operations.

[0054] Figure 2 It is the system structure diagram of the official seal automatic recognition system based on the enterprise digital application project provided by an embodiment of the present invention. As Figure 2 shown, an embodiment of the present invention provides an official seal automatic recognition system based on the enterprise digital application project. The system includes: a recognition unit for calling the trained official seal recognition model based on the image data and outputting the official seal recognition result; wherein, the official seal recognition model is trained based on the YOLO algorithm in MobileNet; and a review unit for judging whether the current file to be recognized is signed in compliance based on the official seal recognition result.

[0055] An embodiment of the present invention also provides a computer-readable storage medium. Instructions are stored on this computer-readable storage medium, and when running on a computer, they cause the computer to execute the above-mentioned official seal automatic recognition method based on the enterprise digital application project.

[0056] Those skilled in the art can understand that all or part of the steps in the methods of the above-described embodiments can be completed by instructing relevant hardware through a program, and this program is stored in a storage medium, including several instructions for causing a single-chip microcomputer, a chip, or a processor to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0057] The optional embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. Additionally, it should be noted that, in the above specific embodiments, the various specific technical features described can be combined in any suitable manner without conflict. To avoid unnecessary repetition, the embodiments of the present invention will not separately describe various possible combination methods.

[0058] Furthermore, any combination can be made among the various different embodiments of the present invention as long as it does not violate the idea of the embodiments of the present invention, and it should equally be regarded as the content disclosed in the embodiments of the present invention.

Claims

1. A method for automatic identification of official seals based on enterprise digital application projects, characterized in that, The method includes: Collect the file to be recognized and process the file to be recognized into corresponding image data; Call the trained official seal recognition model based on the image data and output the official seal recognition result; where The official seal recognition model is obtained by training based on the YOLO algorithm in MobileNet; Judge whether the current file to be recognized is signed in compliance based on the official seal recognition result.

2. The method according to claim 1, wherein The processing of the file to be recognized into corresponding image data includes: Perform non-image data recognition in the file to be recognized as the data to be converted; Based on the data type of the data to be converted, match and execute the corresponding image conversion method to obtain the image data corresponding to each data to be converted.

3. The method according to claim 1, wherein The method further includes training the official seal recognition model; The training rules of the official seal recognition model are: Collect historical image information containing official seal images; Push each piece of historical image information to the user side for the user to mark the official seal area and recycle the marking results; Construct training samples based on the historical images with official seal area markings, and perform model training based on the YOLO algorithm in MobileNet to obtain the official seal recognition model.

4. The method according to claim 3, characterized in that, The pushing of each piece of historical image information to the user side for the user to mark the official seal area includes: Open each piece of historical image information on the user side based on the PPOCRLabel tool and recycle the user's marking results of the official seal area in each piece of historical image information based on the PPOCRLabel tool.

5. The method according to claim 3, characterized in that The constructing of training samples based on the historical image information with official seal area markings includes: Convert the historical image information with official seal area markings into the VOC format; Split the historical image information in the VOC format according to a preset ratio to obtain a training set, a validation set, and a test set respectively.

6. The method according to claim 5, characterized in that, Performing model training based on the YOLO algorithm in MobileNet to obtain the official seal recognition model includes: In MobileNet, perform model training based on the training set and the YOLO algorithm to obtain the initial official seal recognition model; Fine-tune the initial official seal recognition model based on the validation set to obtain the official seal recognition model; After obtaining the official seal recognition model, the method further includes; Perform performance testing on the official seal recognition model based on the test set.

7. The method according to claim 6, wherein The performance testing of the official seal recognition model based on the test set includes: Perform the IoU test and mAP test of the recognition model based on the test set respectively, and judge whether the performance of the official seal recognition model meets the expectations based on the test results; If the performance of the official seal recognition model meets the expectations, perform subsequent official seal recognition based on the official seal recognition model; If the performance of the official seal recognition model does not meet the expectations, re-perform model training.

8. The method according to claim 1, characterized in that, The judging whether the current file to be recognized is signed in compliance based on the official seal recognition result includes: Judge the number of official seals and the type of official seals in the current file to be recognized respectively based on the official seal recognition result; If the number of official seals meets the target number of audited official seals and the type of official seals meets the target type of official seals, it is determined that the current file to be recognized is signed in compliance; If the number of official seals does not meet the target number of audited official seals, or the type of official seals does not meet the target type of official seals, it is determined that the current file to be recognized is signed non-compliance.

9. An automatic seal recognition system based on enterprise digital application projects, characterized in that, The system includes: The acquisition unit is used to acquire the file to be recognized and process the file to be recognized into corresponding image data; The recognition unit is used to call the trained official seal recognition model based on the image data and output the official seal recognition result; among them, The official seal recognition model is obtained by training based on the YOLO algorithm in MobileNet; The review unit is used to judge whether the current file to be recognized is signed in compliance based on the official seal recognition result.

10. A computer-readable storage medium, characterized in that, Instructions are stored on the computer-readable storage medium, and when running on a computer, the computer is made to execute the official seal automatic recognition method based on the enterprise digital application project described in any one of claims 1-8.

Images