Image recognition-based project engineering contract classification storage method and device

Abstract

The application discloses an image recognition-based project engineering contract classification storage method and device, and relates to the technical field of data classification. The method comprises the following steps: obtaining an original image file of a project engineering contract, and performing preprocessing to obtain a standardized image to be recognized; inputting the image to be recognized into an image recognition model to generate structured data containing contract text information; analyzing the structured data, recognizing key information in the project engineering contract by using a natural language processing technology and a preset classification rule library, and generating at least one classification label; and automatically storing the original image file and the corresponding structured data in a preset storage path associated with the classification label in a database according to the classification label, and establishing a retrievable index. The method solves the problems of low efficiency, inconsistent classification standards, retrieval difficulty and information island in the prior art.

Description

Technical Field

[0001] This invention relates to the field of data classification technology, and in particular to a method and apparatus for classifying and storing project engineering contracts based on image recognition. Background Technology

[0002] In the management of construction, municipal, and transportation engineering projects, contracts are the core legal documents and economic basis throughout the entire project lifecycle. A large-scale engineering project often involves hundreds or even thousands of contracts, including but not limited to general contracting contracts, specialized subcontracting contracts, design contracts, supervision contracts, procurement contracts, and labor contracts. These contracts are numerous, diverse in type, and complex in content.

[0003] Currently, the management of engineering contracts mainly suffers from the following pain points: 1) Inefficiency: Traditional methods require manual scanning of paper contracts into electronic versions (usually PDF or image format), followed by document management personnel or project assistants reading and understanding each contract, and then manually creating folders on the computer for categorization and storage based on file names or content. This process is time-consuming and labor-intensive, resulting in extremely high management costs for companies with a large volume of contracts.

[0004] 2) Inconsistent classification standards: Manual classification relies on individual understanding of contract content and company classification standards. Different people may have different understandings, which may lead to the same contract being classified into different categories, or errors in the entry of key contract information (such as amount and contracting parties), causing difficulties in subsequent retrieval and statistics.

[0005] 3) Difficulty in retrieval: Stored contract files are typically indexed only by filenames or simple directory structures. When searching for contracts of a specific project, type, or party, searches are limited to vague filenames or navigating through folders, resulting in extremely low efficiency. The massive amounts of contract data are difficult to utilize effectively and cannot provide data support for decision-making.

[0006] 4) Information silos: Key information in the contract (such as contract amount, construction period, and key terms) is locked in unstructured images or PDF files, making it impossible to link and analyze data with other business systems, thus forming information silos. Summary of the Invention

[0007] This invention provides a method and apparatus for classifying and storing project engineering contracts based on image recognition, which solves the problems of low efficiency, inconsistent classification standards, difficulty in retrieval, and information silos in the existing technology.

[0008] In a first aspect, embodiments of the present invention provide a method for classifying and storing project engineering contracts based on image recognition, the method comprising: Obtain the original image file of the project engineering contract, and preprocess the original image file to obtain a standardized image to be identified; The image to be identified is input into a pre-trained image recognition model to identify and extract the text content in the image to generate structured data containing contract text information; The structured data is analyzed, and key information in the project engineering contract is identified using natural language processing technology and a pre-set classification rule base, and at least one classification label is generated. Based on the classification tags, the original image files and their corresponding structured data are automatically stored in the database under a preset storage path associated with the classification tags, and a searchable index is established.

[0009] The technical solution provided in this application has at least the following beneficial effects: It achieves full automation from contract image scanning to final classification and archiving, eliminating the need for manual intervention and reducing contract processing time from hours to minutes, greatly freeing up manpower. Classification based on image recognition models and preset rules avoids errors and inconsistent standards caused by human factors, ensuring the accuracy and consistency of contract classification. Through structured storage and the establishment of multi-dimensional indexes, users can quickly and accurately locate target contracts by combining any number of conditions, just like querying a database, turning static data into dynamic data. The image recognition model extracts key information from contracts, providing a data foundation for subsequent data linkage and analysis with other business systems, thus solving the problem of information silos.

[0010] In one optional implementation, the original image file of the project engineering contract is obtained, and the original image file is preprocessed to obtain a standardized image to be identified, including: Obtain the original image files of the project's engineering contract; The original image file is subjected to image correction to obtain the corrected image file; The corrected image file is then enhanced to obtain an enhanced image file; Noise removal is performed on the enhanced image file to obtain a standardized image to be identified.

[0011] In one alternative implementation, the image recognition model is constructed based on an OCR algorithm optimized by the Mask R-CNN architecture, and the image recognition model includes a basic backbone network constructed based on the ResNeXt-101 algorithm, an attention module constructed based on the CBAM algorithm and placed after at least one residual block in the basic backbone network, a neck network constructed based on the PANet algorithm, a region proposal network, a region feature alignment module, a branch network, and a text recognition module constructed based on the CRNN algorithm, which are connected in sequence. In one alternative implementation, the training method for the image recognition model includes: Collect several engineering contract images of different formats and qualities, and annotate the positions and contents of the text lines in the engineering contract images to construct training and validation datasets. An initial image recognition model is constructed using an OCR algorithm optimized with the Mask R-CNN architecture. The initial image recognition model is trained end-to-end using the training dataset to optimize the model parameters and obtain an optimized image recognition model. The optimized image recognition model is validated using the validation dataset. If the accuracy is greater than the accuracy threshold, the final image recognition model is output; otherwise, training continues.

[0012] In one optional implementation, the image to be recognized is input into a pre-trained image recognition model to recognize and extract the text content from the image, generating structured data containing contract text information, including: The image to be identified is input into a pre-trained image recognition model, and a multi-scale feature map set of the image to be identified is extracted using a basic backbone network. The attention module is used to perform attention weighting on the multi-scale feature map set to obtain an attention-enhanced multi-scale feature map set of the image to be recognized. Using a neck network, multi-scale fusion is performed on a multi-scale feature map set enhanced by attention mechanism to obtain a fused feature map of the image to be recognized. A region proposal network is used to generate several candidate regions for the fused feature map. The region feature alignment module is used to refine the features of each candidate region to obtain the corresponding candidate region features. Using a branch network, multi-task prediction is performed on the features of each candidate region to obtain the corresponding class probability, refined bounding box coordinates, and segmentation mask. Based on the fused feature map of the image to be recognized, the category probabilities of several candidate regions, the refined bounding box coordinates, and the segmentation mask, the text recognition module is used to perform text recognition and obtain structured data containing contract text information.

[0013] In one optional implementation, the structured data is analyzed using natural language processing techniques and a pre-defined classification rule base to identify key information in the project contract and generate at least one classification label, including: The structured data is analyzed for layout to identify different areas such as titles, paragraphs, tables, and signature areas; Using a pre-trained named entity recognition model, extract several named entities of preset entity types corresponding to key information in the project engineering contract from different regions of the structured data; The system matches several named entities corresponding to the identified key information with a preset classification rule library, and generates at least one classification label based on the matching results.

[0014] In one alternative implementation, the named entity recognition model is built based on the BiLSTM-CRF algorithm.

[0015] In one optional implementation, if the matching result is a fuzzy match or a conflict, the manual review interface is activated, and the structured data and corresponding key information are pushed to the review terminal for manual assistance in determining the classification label.

[0016] In one optional implementation, based on the classification tags, the original image files and their corresponding structured data are automatically stored in a pre-defined storage path in the database associated with the classification tags, and a searchable index is established, including: Based on the category tags, query the preset "tag-storage path" mapping table to determine the unique storage directory of the project engineering contract in the database; The original image file, the extracted structured data, the identified key information, and the generated classification labels are stored as a complete record in the storage directory. Based on the key information and classification tags in the project engineering contract, a full-text search index and a relational database index are established for the records.

[0017] Secondly, embodiments of the present invention provide a project engineering contract classification and storage device based on image recognition, used to implement a project engineering contract classification and storage method, the device comprising: An image preprocessing unit is used to acquire the original image files of the project engineering contract and preprocess the original image files to obtain a standardized image to be recognized. The image recognition unit is used to input the image to be recognized into a pre-trained image recognition model, recognize and extract the text content in the image to be recognized, and generate structured data containing contract text information; The data classification unit is used to analyze the structured data, use natural language processing technology and a preset classification rule base to identify key information in the project engineering contract, and generate at least one classification label. The data storage unit is used to automatically store the original image file and its corresponding structured data into a preset storage path in the database associated with the classification label, and to establish a searchable index.

[0018] A third aspect of this invention provides an electronic device, which includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, such that the at least one processor can perform the method proposed in the first aspect of the present invention.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in the first aspect of the present invention. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the steps of a project engineering contract classification and storage method based on image recognition, provided in an embodiment of the present invention. Figure 3 This is a functional unit diagram of a project engineering contract classification and storage device based on image recognition provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0022] The present invention will be further described below with reference to the accompanying drawings.

[0023] Reference Figure 1 , Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention.

[0024] like Figure 1As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0025] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0026] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating device, a data storage module, a network communication module, a user interface module, and electronic programs.

[0027] exist Figure 1 In the electronic device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device. The electronic device calls the image recognition-based project engineering contract classification storage device stored in the memory 1005 through the processor 1001 and executes the image recognition-based project engineering contract classification storage method provided in the embodiment of the present invention.

[0028] Reference Figure 2 The present invention provides a method for classifying and storing project engineering contracts based on image recognition, the method comprising: S201: Obtain the original image file of the project engineering contract, and preprocess the original image file to obtain a standardized image to be identified; S202: Input the image to be recognized into a pre-trained image recognition model to recognize and extract the text content in the image to be recognized, and generate structured data containing contract text information; S203: Analyze the structured data, use natural language processing technology and a preset classification rule base to identify key information in the project engineering contract, and generate at least one classification label; S204: Based on the classification labels, the original image files and their corresponding structured data are automatically stored in the database under a preset storage path associated with the classification labels, and a searchable index is established.

[0029] The technical solution provided in this application has at least the following beneficial effects: It achieves full automation from contract image scanning to final classification and archiving, eliminating the need for manual intervention and reducing contract processing time from hours to minutes, greatly freeing up manpower. Classification based on image recognition models and preset rules avoids errors and inconsistent standards caused by human factors, ensuring the accuracy and consistency of contract classification. Through structured storage and the establishment of multi-dimensional indexes, users can quickly and accurately locate target contracts by combining any number of conditions, just like querying a database, turning static data into dynamic data. The image recognition model extracts key information from contracts, providing a data foundation for subsequent data linkage and analysis with other business systems, thus solving the problem of information silos.

[0030] In one optional implementation, the original image file of the project engineering contract is obtained, and the original image file is preprocessed to obtain a standardized image to be identified, including: S2011: Obtain the original image files of the project engineering contract; S2012: Perform image correction on the original image file to obtain a corrected image file; It is worth noting that image correction includes performing perspective or affine transformations on the image to correct tilt, curvature, and perspective distortion. S2013: Perform image enhancement on the corrected image file to obtain an enhanced image file; It is worth noting that the image enhancement uses adaptive threshold binarization, contrast stretching, or gamma correction algorithms to enhance the contrast between text and background in the image. S2014: Remove noise from the enhanced image file to obtain a standardized image to be recognized; It is worth noting that noise removal uses average filtering or Gaussian filtering algorithms to remove interference information such as scanning noise and stains from the image.

[0031] In this embodiment, the original image file of the paper contract is acquired by means of a document scanner, a high-speed scanner, or a mobile device. Due to limitations in shooting or scanning conditions, the original image may have problems such as tilt, blurriness, uneven lighting, and cluttered background. Therefore, the original image is first preprocessed, including: image correction using image processing algorithms to correct perspective distortion; image enhancement to improve the contrast between text and background; and noise removal to filter out irrelevant interference. The goal of preprocessing is to output a clear, upright, and standardized image suitable for machine recognition.

[0032] In one alternative implementation, the image recognition model is constructed based on an Optical Character Recognition (OCR) algorithm optimized with a Mask Region-based Convolutional Neural Network (Mask R-CNN) architecture. The image recognition model includes a base backbone network constructed based on the Residual Next-101 (ResNeXt-101) algorithm, an attention module constructed based on the CBAM algorithm and placed after at least one residual block in the base backbone network, a neck network constructed based on the Path Aggregation Network (PANet) algorithm, a region proposal network, a region feature alignment module, a branch network, and a text recognition module constructed based on the Convolutional Recurrent Neural Network (CRNN) algorithm, all connected in sequence. In one alternative implementation, the training method for the image recognition model includes: A-1: Collect several engineering contract images of different formats and qualities, and annotate the positions and contents of the text lines in the engineering contract images to construct training and validation datasets. A-2: Construct an initial image recognition model using an OCR algorithm optimized with the Mask R-CNN architecture; A-3: Use the training dataset to train the initial image recognition model end-to-end, optimize the model parameters, and obtain an optimized image recognition model; A-4: Use the aforementioned validation dataset to validate the optimized image recognition model. If the accuracy is greater than the accuracy threshold, output the final image recognition model; otherwise, continue training.

[0033] In one optional implementation, the image to be recognized is input into a pre-trained image recognition model to recognize and extract the text content from the image, generating structured data containing contract text information, including: S2021: Input the image to be identified into a pre-trained image recognition model, and use a basic backbone network to extract a multi-scale feature map set of the image to be identified; S2022: Using an attention module, attention weights are applied to the multi-scale feature map set to obtain an attention-enhanced multi-scale feature map set of the image to be recognized; S2023: Using a neck network, multi-scale fusion is performed on a multi-scale feature map set enhanced by attention mechanism to obtain a fused feature map of the image to be recognized; S2024: Use a region proposal network to generate several candidate regions for the fused feature map; S2025: Use the region feature alignment module to refine the features of each candidate region to obtain the corresponding candidate region features; S2026: Using a branch network, perform multi-task prediction on the features of each candidate region to obtain the corresponding class probability, refined bounding box coordinates and segmentation mask; S2027: Based on the fused feature map of the image to be recognized, the category probabilities of several candidate regions, the refined bounding box coordinates, and the segmentation mask, the text recognition module is used to perform text recognition and obtain structured data containing contract text information.

[0034] In one optional implementation, the structured data is analyzed using natural language processing techniques and a pre-defined classification rule base to identify key information in the project contract and generate at least one classification label, including: S2031: Perform layout structure analysis on the structured data to identify different areas of the title, paragraphs, tables, and signature area; S2032: Using a pre-trained named entity recognition model, extract several named entities of preset entity types corresponding to key information in the project engineering contract from different regions of the structured data; It is worth noting that the entity type should include at least the project name, contract type, contracting party, contractor, contract amount, signing date, project location, and contract duration. S2033: Match several named entities corresponding to the identified key information with a preset classification rule library, and generate at least one classification label based on the matching results; It is worth noting that the classification tags include primary tags (such as construction contracts, design contracts, and procurement contracts) and secondary tags (such as building construction projects, municipal projects, and transportation projects). The rules in the classification rule base can be: "IF Contract type entity contains 'construction' THEN Primary tag = 'Construction contract'"; "IF Project name entity contains 'metro' or 'rail transit' THEN Secondary tag = 'Transportation project'". The identified entities are matched with the rule base to automatically generate one or more classification tags for the current contract. These tags constitute the basis for subsequent automatic storage.

[0035] In one alternative implementation, the named entity recognition model is built on a Bidirectional Long Short-Term Memory (BiLSTM)-Conditional Random Field (CRF) algorithm.

[0036] In one optional implementation, if the matching result is a fuzzy match or a conflict, the manual review interface is activated, and the structured data and corresponding key information are pushed to the review terminal for manual assistance in determining the classification label. It is worth noting that the results of manual review are used as new sample data to iteratively optimize the named entity recognition model and classification rule base.

[0037] In one optional implementation, based on the classification tags, the original image files and their corresponding structured data are automatically stored in a pre-defined storage path in the database associated with the classification tags, and a searchable index is established, including: S2041: Based on the category tags, query the preset "tag-storage path" mapping table to determine the unique storage directory of the project engineering contract in the database; In this embodiment, the tag combination (first-level tag = 'construction contract', second-level tag = 'building construction project') may correspond to the path " / contract library / construction contract / building construction project / " in the database. The final storage location is determined by querying this table based on the generated tags. S2042: Store the original image file, extracted structured data, identified key information, and generated classification labels as a complete record in the storage directory; S2043: Based on the key information and category tags in the project engineering contract, such as project name, contract number, contracting party, contract amount, signing date, and all category tags as key fields, establish a full-text search index and a relational database index for the record; In this embodiment, to achieve fast retrieval, a powerful index is built for the record using key information entities (such as project name and contract amount) and category tags as key fields. This can be a B-Tree index in a relational database or a full-text search index in a search engine such as Elasticsearch. In the future, users can perform precise queries in milliseconds using any one or more keywords (such as "find all 'construction contracts' with 'AAA' exceeding '10 million'").

[0038] This invention also provides a project engineering contract classification and storage device based on image recognition, referring to... Figure 3 The diagram shows a functional unit diagram of a project engineering contract classification and storage device 300 based on image recognition according to the present invention. The device may include the following units: Image preprocessing unit 301 is used to acquire the original image file of the project engineering contract and preprocess the original image file to obtain a standardized image to be recognized. Image recognition unit 302 is used to input the image to be recognized into a pre-trained image recognition model, recognize and extract the text content in the image to be recognized, and generate structured data containing contract text information; The data classification unit 303 is used to analyze the structured data, use natural language processing technology and a preset classification rule base to identify key information in the project engineering contract, and generate at least one classification label. The data storage unit 304 is used to automatically store the original image file and its corresponding structured data into a preset storage path in the database associated with the classification label, and to establish a searchable index.

[0039] Based on the same inventive concept, another embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. Memory, used to store computer programs; When a processor executes a program stored in memory, it implements the image recognition-based project engineering contract classification and storage method of the present invention.

[0040] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. The communication interface is used for communication between the aforementioned terminal and other devices. The memory can include Random Access Memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory can also be at least one storage device located remotely from the aforementioned processor.

[0041] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0042] Furthermore, to achieve the above objectives, embodiments of the present invention also propose a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the image recognition-based project engineering contract classification and storage method of the present invention.

[0043] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable vehicles (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0044] The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (apparatus), and computer program products according to embodiments of the invention. It should 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 terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0045] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0046] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0047] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. "" and / or "" indicate that either one or both can be selected. Furthermore, the terms "includes," "contains," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the statement "includes a..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the element.

[0048] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered 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 classifying and storing project engineering contracts based on image recognition, characterized in that, The method includes: Obtain the original image file of the project engineering contract, and preprocess the original image file to obtain a standardized image to be identified; The image to be identified is input into a pre-trained image recognition model to identify and extract the text content in the image to generate structured data containing contract text information; The structured data is analyzed, and key information in the project engineering contract is identified using natural language processing technology and a pre-set classification rule base, and at least one classification label is generated. Based on the classification tags, the original image files and their corresponding structured data are automatically stored in the database under a preset storage path associated with the classification tags, and a searchable index is established.

2. The project engineering contract classification and storage method based on image recognition according to claim 1, characterized in that, Obtain the original image file of the project engineering contract, and preprocess the original image file to obtain a standardized image to be recognized, including: Obtain the original image files of the project's engineering contract; The original image file is subjected to image correction to obtain the corrected image file; The corrected image file is then enhanced to obtain an enhanced image file; Noise removal is performed on the enhanced image file to obtain a standardized image to be identified.

3. The project engineering contract classification and storage method based on image recognition according to claim 2, characterized in that, The image recognition model is constructed based on an OCR algorithm optimized by the Mask R-CNN architecture. The image recognition model includes a basic backbone network constructed based on the ResNeXt-101 algorithm, an attention module constructed based on the CBAM algorithm and set after at least one residual block in the basic backbone network, a neck network constructed based on the PANet algorithm, a region proposal network, a region feature alignment module, a branch network, and a text recognition module constructed based on the CRNN algorithm, which are connected in sequence.

4. The project engineering contract classification and storage method based on image recognition according to claim 3, characterized in that, The training method for the image recognition model includes: Collect several engineering contract images of different formats and qualities, and annotate the positions and contents of the text lines in the engineering contract images to construct training and validation datasets. An initial image recognition model is constructed using an OCR algorithm optimized with the Mask R-CNN architecture. The initial image recognition model is trained end-to-end using the training dataset to optimize the model parameters and obtain an optimized image recognition model. The optimized image recognition model is validated using the validation dataset. If the accuracy is greater than the accuracy threshold, the final image recognition model is output; otherwise, training continues.

5. The project engineering contract classification and storage method based on image recognition according to claim 4, characterized in that, The image to be recognized is input into a pre-trained image recognition model to identify and extract the text content from the image, generating structured data containing contract text information, including: The image to be identified is input into a pre-trained image recognition model, and a multi-scale feature map set of the image to be identified is extracted using a basic backbone network. The attention module is used to perform attention weighting on the multi-scale feature map set to obtain an attention-enhanced multi-scale feature map set of the image to be recognized. Using a neck network, multi-scale fusion is performed on a multi-scale feature map set enhanced by attention mechanism to obtain a fused feature map of the image to be recognized. A region proposal network is used to generate several candidate regions for the fused feature map. The region feature alignment module is used to refine the features of each candidate region to obtain the corresponding candidate region features. Using a branch network, multi-task prediction is performed on the features of each candidate region to obtain the corresponding class probability, refined bounding box coordinates, and segmentation mask. Based on the fused feature map of the image to be recognized, the category probabilities of several candidate regions, the refined bounding box coordinates, and the segmentation mask, the text recognition module is used to perform text recognition and obtain structured data containing contract text information.

6. The project engineering contract classification and storage method based on image recognition according to claim 5, characterized in that, The structured data is analyzed using natural language processing technology and a pre-defined classification rule base to identify key information in the project contract and generate at least one classification label, including: The structured data is analyzed for layout to identify different areas such as titles, paragraphs, tables, and signature areas; Using a pre-trained named entity recognition model, extract several named entities of preset entity types corresponding to key information in the project engineering contract from different regions of the structured data; The system matches several named entities corresponding to the identified key information with a preset classification rule library, and generates at least one classification label based on the matching results.

7. The project engineering contract classification and storage method based on image recognition according to claim 6, characterized in that, The named entity recognition model is built based on the BiLSTM-CRF algorithm.

8. The project engineering contract classification and storage method based on image recognition according to claim 7, characterized in that, If the matching result is a fuzzy match or a conflict, the manual review interface will be activated, and the structured data and corresponding key information will be pushed to the review terminal for manual assistance in determining the classification label.

9. The project engineering contract classification and storage method based on image recognition according to claim 8, characterized in that, Based on the classification tags, the original image files and their corresponding structured data are automatically stored in a pre-defined storage path in the database associated with the classification tags, and a searchable index is established, including: Based on the category tags, query the preset "tag-storage path" mapping table to determine the unique storage directory of the project engineering contract in the database; The original image file, the extracted structured data, the identified key information, and the generated classification labels are stored as a complete record in the storage directory. Based on the key information and classification tags in the project engineering contract, a full-text search index and a relational database index are established for the records.

10. A project engineering contract classification and storage device based on image recognition, used to implement the project engineering contract classification and storage method as described in any one of claims 1-9, characterized in that, The device includes: An image preprocessing unit is used to acquire the original image files of the project engineering contract and preprocess the original image files to obtain a standardized image to be recognized. The image recognition unit is used to input the image to be recognized into a pre-trained image recognition model, recognize and extract the text content in the image to be recognized, and generate structured data containing contract text information; The data classification unit is used to analyze the structured data, use natural language processing technology and a preset classification rule base to identify key information in the project engineering contract, and generate at least one classification label. The data storage unit is used to automatically store the original image file and its corresponding structured data into a preset storage path in the database associated with the classification label, and to establish a searchable index.

Images