Method and device for detecting electronic files of pumped storage engineering, storage medium and terminal

By acquiring and preprocessing the data of the electronic files of the extraction engineering, extracting key information and classifying them, and using multimodal feature extraction network and directional detection sub-models for detection, the problem of low accuracy of the electronic files of the extraction engineering engineering is solved, and the problem of low detection accuracy of the electronic files of the extraction engineering is realized. More targeted and accurate detection results.

CN118861534BActive Publication Date: 2024-11-22THREE GORGES HI TECH INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411354436.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-27
Publication Date
2024-11-22
Estimated Expiration
2044-09-27

AI Technical Summary

Technical Problem

The existing electronic file detection methods cannot meet the diversity detection needs of electronic documents for extraction and storage projects, resulting in low detection accuracy.

Method used

By obtaining the electronic files of the extraction project to be detected and real-time operating environment data, preprocessing and key information extraction, classifying file categories, and calling the multimodal feature extraction network and directional detection sub-model for detection, generating a detection report.

Benefits of technology

It realizes directional detection of electronic documents of different types of storage engineering, improves the accuracy and safety of the detection, and ensures the effectiveness of the detection process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118861534B_ABST
    Figure CN118861534B_ABST
Patent Text Reader

Abstract

The present invention discloses a detection method and device, storage medium, and terminal for electronic files of pumped storage engineering, and relates to the technical field of file management. The main purpose is to solve the problem of low detection accuracy of electronic files of pumped storage engineering. It mainly includes obtaining the electronic files of pumped storage engineering to be detected and real-time operating environment data, and pre-processing the electronic files of pumped storage engineering to be detected to obtain target electronic files of pumped storage engineering; classifying files according to the key information extraction results to obtain target electronic files of pumped storage engineering of different file categories; performing operating environment detection according to the real-time operating environment data, and calling the corresponding pumped storage engineering file detection model for each category of target electronic files of pumped storage engineering to obtain file detection results when the operating environment detection is qualified; generating a detection report of the electronic files of pumped storage engineering to be detected according to the file detection results and the real-time operating environment data. It is used to detect electronic files of pumped storage engineering.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of file management, and in particular to a detection method and device, a storage medium and a terminal for electronic files of a pumping and storage project. Background Art

[0002] With the rapid development of computer technology, the urgency and importance of digital archives construction have become increasingly prominent. The detection of electronic archives has been promoted in more and more fields.

[0003] At present, electronic archive detection is mainly aimed at batch text-based electronic archives or electronic archives in a single technical field. Pumped storage projects (pumped water storage projects) involve multiple professional fields such as water conservancy, electricity, machinery, and construction. In addition, due to the long construction period of pumped storage projects and the large number of participating units, the project files are not only text-based files, but also include design documents, construction documents, and acceptance reports involving different technical fields. The existing electronic archive detection cannot meet the diverse detection needs of electronic files in pumped storage projects, resulting in low accuracy of file detection. Summary of the invention

[0004] In view of this, the present invention provides a method and device for detecting electronic files of pumped storage projects, a storage medium, and a terminal, the main purpose of which is to solve the problem of low detection accuracy of existing electronic files of pumped storage projects.

[0005] According to one aspect of the present invention, a method for detecting electronic files of pumped storage engineering is provided, comprising:

[0006] Acquire the electronic file of the pumped storage project to be detected and the real-time operating environment data, and pre-process the electronic file of the pumped storage project to be detected to obtain the electronic file of the target pumped storage project;

[0007] Extract key information from the target pumped storage project electronic files, and classify the files according to the key information extraction results to obtain target pumped storage project electronic files of different file categories, wherein the file categories include design documents, construction drawings and acceptance reports;

[0008] Performing an operating environment test according to the real-time operating environment data, and when the operating environment test is qualified, calling a corresponding pumped storage engineering file detection model for each category of target pumped storage engineering electronic files to perform a test to obtain a file detection result, wherein the pumped storage engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumped storage engineering file detection models include different directional detection sub-models;

[0009] A test report of the electronic file of the pumped storage project to be tested is generated based on the file test result and the real-time operating environment data.

[0010] Furthermore, for each category of target pumped storage engineering electronic files, the corresponding pumped storage engineering file detection model is called to perform detection to obtain file detection results, including:

[0011] For each category of the target pumping engineering electronic files, the multimodal feature extraction network of the corresponding pumping engineering file detection model is used to extract features of the text and image in the file to obtain file fusion features, which include image-text correlation features, image-text complementarity features and file global features;

[0012] If the target pumping engineering electronic file is a design file, then based on the file fusion features, the directional detection sub-model is used to perform data signature detection, version detection, content consistency detection, metadata integrity detection, image clarity detection and technical specification detection on the target pumping engineering electronic file, and perform sensitive information encryption and access permission configuration;

[0013] If the target pumped storage engineering electronic file is a construction drawing, then based on the file fusion feature, the directional detection sub-model is used to perform drawing authenticity detection, drawing integrity detection and drawing readability detection on the target pumped storage engineering electronic file, and configure drawing encryption and access log data tables;

[0014] If the target pumping project electronic file is an acceptance report, then based on the file fusion features, the directional detection sub-model is used to perform digital signature detection, image and text consistency detection, content integrity detection, data consistency detection and content availability detection on the target pumping project electronic file, and sensitive information encryption and access permission configuration are performed.

[0015] Furthermore, the multimodal feature extraction network includes an image feature extraction network, a text feature extraction network and an image-text feature fusion network. For each category of the target extraction engineering electronic file, the multimodal feature extraction network corresponding to the extraction engineering file detection model is used to extract features from the text and image in the file to obtain the file fusion feature, including:

[0016] For each category of the target extraction engineering electronic file, using the image feature extraction network to extract image space features in the target extraction engineering electronic file;

[0017] Using the text feature extraction network to extract text temporal features in the target extraction engineering electronic file;

[0018] The feature fusion network is used to fuse the image spatial features and the text temporal features to obtain file fusion features.

[0019] Furthermore, before performing the operating environment detection according to the real-time operating environment data, the method further includes:

[0020] Based on the different categories of pumped storage engineering file detection models, test detection is performed on the positive samples of pumped storage engineering electronic files, and operating environment test data of the different categories of pumped storage engineering file detection models are extracted;

[0021] According to the data distribution of the operating environment test data, the operating environment parameter intervals matching the detection processes of different categories of files are calculated, and the operating environment parameter intervals include a CPU load interval, a memory occupancy interval, a network bandwidth rate interval, and a disk read speed interval;

[0022] The performing of operating environment detection according to the real-time operating environment data includes:

[0023] Determine the file category of the target pumped storage project electronic file, and retrieve the operating environment parameter interval matching the file category, and if the real-time operating environment data satisfies the operating environment parameter interval, determine that the system operating environment test is qualified;

[0024] If the real-time operating environment data does not satisfy the operating environment parameter range, it is determined that the system operating environment test fails.

[0025] Furthermore, the test report of the electronic file of the pumped storage project to be tested is generated based on the file test result and the real-time operating environment data, including:

[0026] Acquire detection process operating environment data, and determine operating environment difference data between the detection process operating environment data and the real-time operating environment data;

[0027] If the operating environment difference data satisfies the operating environment difference interval, a normal operating environment status report is generated, and a test report of the electronic file of the pumped storage project to be tested is generated based on the normal operating environment status report and the test result;

[0028] If the operating environment difference data does not satisfy the operating environment difference interval, abnormal information matching the operating environment difference data is identified from the system abnormal data mapping relationship set, and a test report of the electronic file of the pumped storage project to be tested is generated based on the abnormal information and the test result.

[0029] Furthermore, extracting key information from the target pumping and storage engineering electronic file includes:

[0030] Extracting project-related text from the target pumping engineering electronic file using a text information extraction network model to obtain key text information, wherein the text information extraction network model is trained based on corpus samples in a pumping engineering-related corpus;

[0031] The image information in the target pumping engineering electronic file is extracted using an image information extraction network model to obtain key image information, wherein the image information extraction network model includes a multi-layer convolutional network for feature extraction at different scales;

[0032] A key information extraction result is generated according to the text key information and the image key information.

[0033] Furthermore, the preprocessing of the electronic file of the pumped storage project to be detected to obtain the target electronic file of the pumped storage project includes:

[0034] Using a text processing model, filtering the text in the electronic file of the pumping engineering to be detected for noise words, and filtering the images in the electronic file of the pumping engineering to be detected for noise images;

[0035] Hash value matching is performed on the electronic files of the pumping engineering to be detected after filtering the noise words and noise images, and the files with the same hash value are deduplicated;

[0036] After the file deduplication process, the remaining electronic files of the pumping engineering to be tested are respectively subjected to text, image and metadata anomaly detection;

[0037] The electronic files of pumped-storage projects to be inspected with normal results for all items are determined as target pumped-storage projects electronic files, and an early warning is issued for the electronic files of pumped-storage projects to be inspected with abnormal results for any inspection item.

[0038] According to another aspect of the present invention, a detection device for pumped storage engineering electronic files is provided, comprising:

[0039] An acquisition module is used to acquire the electronic file of the pumped storage project to be detected and the real-time operating environment data, and pre-process the electronic file of the pumped storage project to be detected to obtain the electronic file of the target pumped storage project;

[0040] A classification module is used to extract key information from the target pumped storage project electronic files, and classify the files according to the key information extraction results to obtain target pumped storage project electronic files of different file categories, wherein the file categories include design documents, construction drawings and acceptance reports;

[0041] A detection module is used to perform an operating environment detection according to the real-time operating environment data, and when the operating environment detection is qualified, for each category of target pumped storage engineering electronic files, call the corresponding pumped storage engineering file detection model to perform detection to obtain a file detection result, wherein the pumped storage engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumped storage engineering file detection models include different directional detection sub-models;

[0042] A generation module is used to generate a test report of the electronic file of the pumped storage project to be tested based on the file detection result and real-time operating environment data.

[0043] Furthermore, the detection module includes:

[0044] The first extraction unit is used to extract features of text and image in each category of the target extraction engineering electronic file by using a multimodal feature extraction network of a corresponding extraction engineering file detection model to obtain a file fusion feature, wherein the file fusion feature includes an image-text correlation feature, an image-text complementarity feature, and a file global feature;

[0045] A first detection unit is used for, if the target pumping engineering electronic file is a design file, using the directional detection sub-model to perform data signature detection, version detection, content consistency detection, metadata integrity detection, image clarity detection and technical specification detection on the target pumping engineering electronic file according to the file fusion feature, and to perform sensitive information encryption and access permission configuration;

[0046] A second detection unit is used for, if the target pumped storage engineering electronic file is a construction drawing, using the directional detection sub-model to perform drawing authenticity detection, drawing integrity detection and drawing readability detection on the target pumped storage engineering electronic file according to the file fusion feature, and configuring drawing encryption and access log data table;

[0047] The third detection unit is used to perform digital signature detection, image and text consistency detection, content integrity detection, data consistency detection and content availability detection on the target pumping project electronic file based on the file fusion features, if the target pumping project electronic file is an acceptance report, and to perform sensitive information encryption and access permission configuration using the directional detection sub-model.

[0048] Furthermore, in a specific application scenario, the extraction unit is specifically used to extract the image spatial features of the target extraction engineering electronic file in each category by using the image feature extraction network; extract the text temporal features of the target extraction engineering electronic file by using the text feature extraction network; and fuse the image spatial features and the text temporal features by using the feature fusion network to obtain file fusion features.

[0049] Furthermore, the device also includes:

[0050] A testing module, used to perform test detection on positive samples of pumped storage engineering electronic files based on different categories of the pumped storage engineering file detection models, and extract operating environment test data of different categories of pumped storage engineering file detection models;

[0051] A calculation module, for calculating, based on the data distribution of the operating environment test data, operating environment parameter intervals matching different types of file detection processes, wherein the operating environment parameter intervals include a CPU load interval, a memory occupancy interval, a network bandwidth rate interval, and a disk read speed interval;

[0052] The detection module further includes:

[0053] A first determination unit is used to determine the file category of the target pumped storage project electronic file, and retrieve the operating environment parameter interval matching the file category, and if the real-time operating environment data satisfies the operating environment parameter interval, determine that the system operating environment detection is qualified;

[0054] The second determining unit is configured to determine that the system operating environment detection fails if the real-time operating environment data does not satisfy the operating environment parameter range.

[0055] Furthermore, the generation module includes:

[0056] An acquisition unit, used to acquire detection process operating environment data, and determine operating environment difference data between the detection process operating environment data and the real-time operating environment data;

[0057] A first generating unit is configured to generate a normal operating environment status report if the operating environment difference data satisfies the operating environment difference interval, and generate a test report of the electronic file of the pumped storage project to be tested based on the normal operating environment status report and the test result;

[0058] The second generating unit is used to identify abnormal information matching the operating environment difference data from the system abnormal data mapping relationship set if the operating environment difference data does not satisfy the operating environment difference interval, and generate a test report of the electronic file of the pumped storage project to be tested based on the abnormal information and the test result.

[0059] Furthermore, the classification module includes:

[0060] A second extraction unit is used to extract project-related text in the target pumping engineering electronic file using a text information extraction network model to obtain text key information, wherein the text information extraction network model is trained based on corpus samples in a pumping engineering related corpus;

[0061] A third extraction unit is used to extract image information from the target pumping engineering electronic file using an image information extraction network model to obtain key image information, wherein the image information extraction network model includes a multi-layer convolutional network for feature extraction at different scales;

[0062] The third generating unit is used to generate a key information extraction result according to the text key information and the image key information.

[0063] Furthermore, the acquisition module further includes:

[0064] A filtering unit, used for filtering the text in the electronic file of the pumping engineering to be detected by using a text processing model, and filtering the image in the electronic file of the pumping engineering to be detected by using a noise word filtering model;

[0065] A deduplication unit is used to match hash values ​​of the electronic files of the pumping engineering to be detected after filtering the noise words and noise images, and to perform deduplication processing on the files with the same hash values;

[0066] The fourth detection unit is used to perform text, image and metadata anomaly detection on the remaining electronic files of the pumping engineering to be detected after the file deduplication processing;

[0067] The third determination unit is used to determine the electronic file of the pumped storage project to be detected with normal results in all items as the target pumped storage project electronic file, and to issue an early warning for the electronic file of the pumped storage project to be detected with abnormal results in any detection item.

[0068] According to another aspect of the present invention, there is provided a storage medium storing at least one executable instruction, wherein the executable instruction enables a processor to execute operations corresponding to the above-mentioned method for detecting electronic files of pumping and storage engineering.

[0069] According to another aspect of the present invention, there is provided a terminal, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0070] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned method for detecting electronic files of pumping and storage engineering.

[0071] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0072] The present invention provides a detection method and device, storage medium and terminal for electronic files of pumped storage engineering. The embodiment of the present invention obtains the electronic files of pumped storage engineering to be detected and real-time operation environment data, and pre-processes the electronic files of pumped storage engineering to be detected to obtain target electronic files of pumped storage engineering; extracts key information of the target electronic files of pumped storage engineering, and classifies files according to the key information extraction results to obtain target electronic files of pumped storage engineering of different file categories, wherein the file categories include design files, construction drawings and acceptance reports; performs operation environment detection according to the real-time operation environment data, and when the operation environment detection is qualified, for each category of target electronic files of pumped storage engineering, calls a corresponding pumped storage engineering file detection model for detection to obtain file detection results, wherein the pumped storage engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumped storage engineering file detection models include different directional detection sub-models; generates a detection report of the electronic files of pumped storage engineering to be detected according to the file detection results and the real-time operation environment data, realizes directional detection of files of different categories, makes the detection of the target electronic files of pumped storage engineering more targeted and accurate, and at the same time, the operation environment detection ensures the safety and effectiveness of the detection process, thereby greatly improving the accuracy of the detection of electronic files of pumped storage engineering.

[0073] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0075] Figure 1A flow chart of a method for detecting electronic files of a pumped storage project provided by an embodiment of the present invention is shown;

[0076] Figure 2 A flow chart of another method for detecting electronic files of pumped storage engineering provided by an embodiment of the present invention is shown;

[0077] Figure 3 A block diagram showing a device for detecting electronic files of pumped storage engineering provided by an embodiment of the present invention is shown;

[0078] Figure 4 A schematic diagram of the structure of a terminal provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0079] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0080] Aiming at the technical problem of low accuracy of existing pumped storage project detection. The embodiment of the present invention provides a method for detecting electronic files of pumped storage projects, such as Figure 1 As shown, the method includes:

[0081] 101. Obtain electronic files of the pumped storage project to be tested and real-time operating environment data, and pre-process the electronic files of the pumped storage project to be tested to obtain electronic files of the target pumped storage project.

[0082] In an embodiment of the present invention, the electronic file of the pumped storage project to be detected is an electronic file of the construction management file of the pumped storage project that needs to be detected by the electronic file, which can be collected from various project management systems, databases or archives related to the pumped storage project, including but not limited to design documents, construction drawings, and acceptance reports. The collection process of the electronic file of the pumped storage project to be detected can specifically perform text recognition on the electronic files in various project management systems, databases or archives through a neural network model to identify the project information such as the project name, design unit and date corresponding to the electronic file, and use the electronic file whose project information matches the target pumped storage project as the electronic file of the pumped storage project to be detected. After obtaining the electronic file of the pumped storage project to be detected, the electronic file is pre-processed such as data cleaning, text deduplication, file deduplication, and file format standardization to ensure the consistency of the electronic file data.

[0083] 102. Extract key information from the target pumping and storage engineering electronic files, and classify the files based on the key information extraction results to obtain target pumping and storage engineering electronic files of different file categories.

[0084] In the embodiment of the present invention, different and targeted detection models are used to detect electronic files of pumping and storage engineering of different categories. Therefore, it is necessary to classify the target electronic files of pumping and storage engineering before file detection, so that the electronic files of pumping and storage engineering of different categories can be detected by using the detection model matching with it. Among them, the file categories include design files, construction drawings and acceptance reports. The file classification process specifically includes: extracting key information of the target electronic files of pumping and storage engineering, and the key information includes file name, file metadata (modifier, date and version), text content, image information, etc., and then automatically classifying and processing the electronic files according to the extracted key information. Specifically, the file content can be identified based on the neural network model, and then classified based on the preset classification strategy. For example, if the file name includes "design" and includes image information, it can be identified as a design file; if the file name includes "construction" and includes image information, it can be identified as a construction drawing, etc. It is also possible to extract key information and classify the electronic files based on the pre-trained classification model to directly determine the category of the electronic files. Among them, the basic model of the classification model can be a Bert model, a support vector machine model, etc., and the training samples include design document samples, construction drawing samples and acceptance report samples.

[0085] It should be noted that the naming principles of engineering project files and the relatively fixed content formats in the files can be easily classified based on the key information extraction method, which can reduce the data processing volume and server resource consumption in the file classification process.

[0086] 103. Perform an operating environment test based on the real-time operating environment data, and if the operating environment test is qualified, call the corresponding pumped storage project file detection model for each category of target pumped storage project electronic files to perform detection and obtain file detection results.

[0087] In the embodiment of the present invention, a pumping engineering file detection model suitable for different categories of target pumping engineering electronic files is constructed in advance according to the detection requirements of different categories of target pumping engineering electronic files. Among them, the pumping engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumping engineering file detection models include different directional detection sub-models. That is to say, for the detection model of each category of target pumping engineering electronic files, the network structure and feature extraction process of the multimodal feature extraction network are the same (there are differences in the parameters of the network or the model training samples), but the directional detection sub-model is constructed for different detection requirements. Due to the different emphasis of different categories of pumping engineering electronic files in the detection process, for example, the design file has higher requirements for the text context semantics and the clarity of the image; the acceptance report pays more attention to the authenticity of the image and text, and the construction drawing has higher requirements for the matching of text and image. Therefore, for different categories of target pumping engineering electronic files, different directional detection sub-models are constructed, so that the detection process can maximize the focus of the detection, which is conducive to more accurately meeting the detection requirements. In addition, since the target pumping engineering electronic files mostly include multiple modalities such as plain text, text in images, and images, the multimodal feature extraction network can comprehensively and accurately extract the file content, further improving the accuracy of directional detection, thereby improving the accuracy of file detection.

[0088] It should be noted that before testing the file, the operating environment also needs to be tested. The operating environment and the system operating environment of the execution file detection process, for example, the system's CPU load, memory usage, network bandwidth, and disk input and output (IO) and other operating data. By testing the operating environment before executing the file detection, it is possible to promptly discover that the system operating environment does not meet the detection conditions (CPU load is too high) or that there are security risks in the system operating environment (abnormal memory access), thereby effectively ensuring the effective and safe execution of electronic file detection for pumped storage projects.

[0089] 104. Generate a test report of the electronic file of the pumped storage project to be tested based on the file test result and the real-time operating environment data.

[0090] In an embodiment of the present invention, the file detection result includes an authenticity detection result, an integrity detection result, an availability detection result, and a security detection result, that is, for each category of the target pumping engineering electronic file, the above four dimensions are detected respectively. Among them, the specific contents of each detection are different for files of different categories. For example, for authenticity detection, the acceptance report mainly detects the authenticity of the electronic signature, the consistency between the acceptance image and the acceptance result description text; and the construction drawings mainly detect the authenticity of the watermark and whether the image has tampering features. After completing various tests on the target pumping engineering electronic file, the test results and the real-time operating environment data are used together to generate a test report of the pumping engineering electronic file to be detected. With the execution of the data processing process of the electronic file detection, the real-time operating environment data of the system is also constantly changing. Writing the real-time operating environment data into the test report can facilitate the monitoring of the detection process and ensure that the detection process of the electronic file is not interfered by any abnormal factors, thereby further improving the accuracy and credibility of the file detection.

[0091] In an embodiment of the present invention, for further explanation and limitation, as Figure 2 As shown, the steps call the corresponding pumped storage engineering file detection model for detection of each category of target pumped storage engineering electronic files to obtain file detection results, including:

[0092] 201. For each category of the target pumping engineering electronic files, use the multimodal feature extraction network of the corresponding pumping engineering file detection model to extract features from the text and image in the file to obtain file fusion features.

[0093] 202. If the target pumping and storage engineering electronic file is a design file, then based on the file fusion features, the directional detection sub-model is used to perform data signature detection, version detection, content consistency detection, metadata integrity detection, image clarity detection and technical specification detection on the target pumping and storage engineering electronic file, and perform sensitive information encryption and access permission configuration.

[0094] 203. If the target pumped storage project electronic file is a construction drawing, then based on the file fusion features, the directional detection sub-model is used to perform drawing authenticity detection, drawing integrity detection and drawing readability detection on the target pumped storage project electronic file, and the drawing encryption and access log data table are configured.

[0095] 204. If the target pumping and storage project electronic file is an acceptance report, then based on the file fusion characteristics, the directional detection sub-model is used to perform digital signature detection, image and text consistency detection, content integrity detection, data consistency detection and content availability detection on the target pumping and storage project electronic file, and perform sensitive information encryption and access permission configuration.

[0096] In an embodiment of the present invention, the file fusion features include image-text correlation features, image-text complementarity features, and file global features. The text and image in the target extraction engineering electronic file are extracted by a multimodal feature extraction network, and the correlation information between the image features and the text features, i.e., the image-text correlation features, such as the text content of the image description, is captured by feature fusion; the overall semantic information in the file can also be captured by the fusion between the text features and the image features, including the complementary information between the text and image features, i.e., the image-text complementarity features; finally, all the image features and text features extracted from the file are used as the file global features.

[0097] After completing the extraction of image features and text features, the directional detection sub-model detects the target extraction engineering electronic files based on the image-text correlation features, image-text complementarity features, file global features, and pre-configured feature matching samples for feature matching (such as digital signature feature samples, official seal watermark feature samples, etc.), wherein the feature matching samples can be stored in the partition space of the model or retrieved from the blockchain, which is not specifically limited in the embodiment of the present invention. Among them, the directional detection sub-model is constructed according to the detection characteristics of different categories of target extraction engineering electronic files, including a variety of algorithms for detecting authenticity, integrity, availability and security. Moreover, the detection contents of the authenticity, integrity, availability and security of different categories of files are also different.

[0098] Authenticity detection includes consistency detection of images and texts, verification of digital signatures or watermarks, and version detection. The consistency detection of images and texts is based on the feature similarity of the image-text correlation features to determine the consistency of the image and text descriptions in the file. For example, whether the on-site photos in the acceptance report match the construction progress described in the text; whether the equipment installation diagram in the construction drawing is consistent with the equipment installation steps described in the text, etc. The verification of digital signatures or watermarks is based on the deep-level image features and the contextual features of the text in the image-text complementarity features to verify whether the file content remains consistent during the digital signature signing or watermarking process. If the digital signature or watermark image is recorded in the blockchain, the image space feature vector of the digital signature or watermark image is compared to detect whether the file has been tampered with. Version detection is to obtain the access log of the file and detect whether the current version is the latest version in the historical version.

[0099] Integrity testing includes file content integrity testing and metadata integrity testing. Among them, file content integrity testing can calculate the hash value of the file through a hash algorithm, and compare the calculated hash value with the preset hash value. If the difference in the hash value is less than the preset difference threshold, the file is determined to be complete. For example, the text description in the construction drawing is 5 drawings, but the actual number of drawings is 4, so there are omissions and missing drawings; whether the acceptance report contains chapters such as the acceptance process, test results, conclusions and suggestions. If there is missing content, the integrity test is abnormal. Metadata integrity testing is to test the file's title, date, author, version number and other information. If any file information is missing, the test result is abnormal.

[0100] Usability testing includes text semantic testing, image quality testing, and standardization testing. Among them, text semantic testing is to use natural language models to test text features, for example, to calibrate whether the text semantic expression is complete, whether the context semantic logic is smooth, whether the punctuation is used correctly, etc. Image quality testing is to use convolutional neural networks to detect the texture and edge features in the image features to determine whether the texture of each test target is clear and the edges are complete. For example, it can detect whether the drawings are blurred, whether there are missing design structures in the drawings, and whether necessary annotations in the drawings exist.

[0101] Security testing includes sensitive word encryption configuration and access permission control. Sensitive word encryption configuration is to determine whether the file contains sensitive words based on text features, and encrypt and hide sensitive words for protection. In addition, access permissions can be set for the encryption and hiding of sensitive words so that specific people can view the sensitive word part. Access permission control is to determine the control level of the file based on the global features of the file, and configure access permissions based on the control level. For example, if the watermark image feature of the file represents confidentiality or the text feature represents confidentiality, the control level of the file is determined according to the preset confidentiality and the mapping relationship between different confidentiality levels and control levels, so as to configure access permissions.

[0102] It should be noted that due to the different content composition characteristics of different target pumping engineering electronic files, the detection focus on the file content is also different. For example, the design file includes detailed engineering design instructions, technical specifications, drawings and calculation files, with a similar proportion of text and images, and the text is mostly technical descriptions and the images are mostly design drawings. Its detection needs to focus on the standardization of the text, the clarity and integrity of the image, and the consistency between the text and the image. The construction drawings contain detailed construction drawings, indicating the size, materials and construction methods; they are mainly images, with less text, and are mostly annotations and explanatory texts. Its detection process is more inclined to the detection of images, and the integrity, standardization and clarity of the images are higher. The acceptance report includes the acceptance process, test results, conclusions and suggestions, etc. It has more text and can contain images such as charts, photos and signatures. Its detection focuses more on the authenticity identification of the image and the matching of the text and the image to detect the tampering of the file. By using different directional detection sub-models for detection based on the content characteristics of engineering electronic files of different categories of targets, the detection algorithm of the model can be highly matched with the detection requirements, the training process of the model can be more targeted, and the learning ability can be stronger, thereby greatly improving the efficiency and accuracy of detection.

[0103] In one embodiment of the present invention, for further explanation and limitation, the step is to extract features of the contents in the target pumping engineering electronic files of each category using a multimodal feature extraction network of a corresponding pumping engineering file detection model to obtain file fusion features, including:

[0104] For each category of the target extraction engineering electronic file, using the image feature extraction network to extract image space features in the target extraction engineering electronic file;

[0105] Using the text feature extraction network to extract text temporal features in the target extraction engineering electronic file;

[0106] The feature fusion network is used to fuse the image spatial features and the text temporal features to obtain file fusion features.

[0107] In an embodiment of the present invention, the basic model of the image feature extraction network may be a deep feedforward neural network (CNN: Convolutional Neural Network), which is used to extract spatial features of images in a file to obtain image spatial features (image features). The basic model of the text feature extraction network may be a recurrent neural network (RNN: Recurrent Neural Network), which is used to extract time series features of text in a file to obtain text time series features (text features). The network of the feature fusion network may include a multi-layer multimodal feature splicing subnetwork, a multimodal feature weighting subnetwork and a tensor fusion subnetwork, and may also use an adversarial network or a Transformer model to learn a self-attention mechanism to achieve fusion of multimodal feature data. The training process of the above model may be forward propagated according to the input data, calculate the loss function, and use the back propagation algorithm to update the model parameters so that the prediction results of the model gradually approach the true label, and the model parameters are adjusted by the stochastic gradient descent method to reduce the loss function and improve the model learning performance. By fusing the multimodal features in the file, we can better capture the consistency and correlation between image features and text features, and provide a more accurate and direct detection basis for the subsequent authenticity, integrity and readability detection of the file, thereby effectively improving the accuracy and efficiency of the authenticity, integrity and readability detection of the file.

[0108] In one embodiment of the present invention, for further explanation and limitation, before the step of performing the operating environment detection according to the real-time operating environment data, the method further includes:

[0109] Based on the different categories of pumped storage engineering file detection models, test detection is performed on the positive samples of pumped storage engineering electronic files, and operating environment test data of the different categories of pumped storage engineering file detection models are extracted;

[0110] According to the data distribution of the operating environment test data, the operating environment parameter intervals matching the detection processes of different categories of files are calculated;

[0111] The performing of operating environment detection according to the real-time operating environment data includes:

[0112] Determine the file category of the target pumped storage project electronic file, and retrieve the operating environment parameter interval matching the file category, and if the real-time operating environment data satisfies the operating environment parameter interval, determine that the system operating environment test is qualified;

[0113] If the real-time operating environment data does not satisfy the operating environment parameter range, it is determined that the system operating environment test fails.

[0114] In the embodiment of the present invention, the operating environment parameter intervals corresponding to target pumped storage engineering electronic files of different categories are different, and the operating environment parameter intervals are determined based on the test detection results of the positive samples of the pumped storage engineering electronic files, that is, the corresponding pumped storage engineering file detection models are used for multiple detections for the positive samples of the pumped storage engineering electronic files of different categories, and the operating environment data in each detection process is tested, so as to determine the use data of the CPU load, memory occupancy, network bandwidth rate and disk reading speed in the model test process according to the distribution interval of the operating environment data obtained by the test, and then determine the parameter interval that the operating environment needs to meet before the file detection starts according to the difference between the limit parameters of the system operation and the use data, that is, the operating environment parameter interval. Among them, the operating environment parameter interval includes the CPU load interval, the memory occupancy interval, the network bandwidth rate interval and the disk reading speed interval. The positive sample of the pumped storage engineering electronic file is a sample of the pumped storage engineering electronic file with a normal detection result. After determining the operating environment parameter intervals of target pumped storage engineering electronic files of different categories, in response to the instruction to detect the file, the corresponding operating environment parameter interval is retrieved according to the file category of the target pumped storage engineering electronic file, so as to detect the current real-time operating environment data according to this interval range.

[0115] It should be noted that the pumped storage engineering file detection models used in the detection of different types of target pumped storage engineering electronic files have different requirements for the operating environment. For example, the data processing volume of the convolutional neural network and the recurrent neural network of the pumped storage engineering file detection model of the design file is greater than that of the acceptance report, and the requirements for the CPU and memory are higher, and thus the requirements for the operating environment are also higher. By testing different types of pumped storage engineering file detection models and determining the consumption of system operating resources during the detection process, the requirements for the operating environment of the detection process of different types of target pumped storage engineering electronic files can be inferred, which greatly improves the matching accuracy of the operating environment parameter range and the detection process, as well as the accuracy of the operating environment detection, thereby ensuring that the detection of target pumped storage engineering electronic files is effectively executed.

[0116] In one embodiment of the present invention, for further explanation and limitation, the step of generating a test report of the electronic file of the pumped storage project to be tested based on the file test result and the real-time operating environment data includes:

[0117] Acquire detection process operating environment data, and determine operating environment difference data between the detection process operating environment data and the real-time operating environment data;

[0118] If the operating environment difference data satisfies the operating environment difference interval, a normal operating environment status report is generated, and a test report of the electronic file of the pumped storage project to be tested is generated based on the normal operating environment status report and the test result;

[0119] If the operating environment difference data does not satisfy the operating environment difference interval, abnormal information matching the operating environment difference data is identified from the system abnormal data mapping relationship set, and a test report of the electronic file of the pumped storage project to be tested is generated based on the abnormal information and the test result.

[0120] In the embodiment of the present invention, the real-time operating environment data currently obtained is the operating environment data after the target pumping and storage engineering electronic file detection is completed. In order to ensure the authenticity and reliability of the file detection process, the detection process operating environment data in the file detection process is collected, and the difference between the detection process operating environment data and the real-time operating environment data before the detection is started is calculated to obtain the operating environment difference data. Among them, the operating environment difference data includes the CPU load difference value, the memory occupancy difference value, the network bandwidth rate difference value and the disk read speed difference value. In the above calculation process, the detection process operating environment data can take the average value or peak value of the data to participate in the calculation, and the embodiment of the present invention does not make specific restrictions. If the operating environment difference data meets the operating environment difference interval, it indicates that the operating environment of the detection process is normal, otherwise, it indicates that the operating environment of the detection process is abnormal. Among them, the operating environment difference interval can be determined based on the operating environment test data in the historical detection process. If the operating environment difference data does not meet the operating environment difference interval, the operating environment difference data is matched with the mapping relationship in the abnormal data mapping relationship set to identify the specific abnormal type or cause of the system. Among them, the abnormal data mapping relationship set includes pre-constructed mapping relationships between different abnormal information and different operating environment difference data.

[0121] By comparing the system operating environment before file detection with the system operating environment during file detection, it is possible to detect abnormal conditions in the detection process, such as interference of other plug-ins with the detection process, other malicious intrusions, etc., thereby ensuring the authenticity and reliability of the target pumped storage project electronic file detection process.

[0122] In one embodiment of the present invention, for further explanation and limitation, the step of extracting key information from the target pumped storage project electronic file includes:

[0123] Using a text information extraction network model to extract project-related text from the target pumping and storage project electronic file to obtain key text information;

[0124] Using an image information extraction network model to extract image information from the target pumping and storage engineering electronic file to obtain key image information;

[0125] A key information extraction result is generated according to the text key information and the image key information.

[0126] In an embodiment of the present invention, the key information extraction result includes image type, text keywords, file name, etc. In order to accurately classify the target pumping engineering electronic file, the project-related text is identified through a text information extraction network model to obtain text key information. Among them, the text information extraction network model is obtained by fine-tuning and training based on the corpus samples in the pumping engineering related corpus, which can accurately capture the text key information related to the stage of the pumping engineering project in the file, and better identify the professional terms and industry specifications in the file. The text key information includes the project name, text keywords and project stage, etc. Then, the image information extraction network model extracts the image information in the target pumping engineering electronic file. Among them, the image information extraction network model includes a multi-layer convolutional network for feature extraction at different scales. The multi-layer convolutional network extracts features of different scales, which can improve the model's recognition ability for complex image structures, thereby improving the accuracy of key information extraction.

[0127] In one embodiment of the present invention, for further explanation and limitation, the step of preprocessing the electronic file of the pumped storage project to be detected to obtain the target pumped storage project electronic file includes:

[0128] Using a text processing model, filtering the text in the electronic file of the pumping engineering to be detected for noise words, and filtering the images in the electronic file of the pumping engineering to be detected for noise images;

[0129] Hash value matching is performed on the electronic files of the pumping engineering to be detected after filtering the noise words and noise images, and the files with the same hash value are deduplicated;

[0130] After the file deduplication process, the remaining electronic files of the pumping engineering to be tested are respectively subjected to text, image and metadata anomaly detection;

[0131] The electronic files of pumped-storage projects to be inspected with normal results for all items are determined as target pumped-storage projects electronic files, and an early warning is issued for the electronic files of pumped-storage projects to be inspected with abnormal results for any inspection item.

[0132] In an embodiment of the present invention, in order to ensure the efficiency and accuracy of subsequent file detection, the electronic file of the pumping engineering to be detected is preprocessed in advance. Based on the text processing model, HTML tags, special characters, stop words, etc. in the file text are removed, and stem extraction or word form normalization is performed to reduce the noise interference of the text. Among them, the text processing model can be a natural language model (NLP: Natural Language Processing). Through computer vision technology, non-essential information such as noise and watermarks in the image is detected and removed to improve the quality and readability of the image. Among them, the computer vision technology can be a convolutional neural network model, image restoration technology, etc. The text processing model is also used to calculate the similarity between files, and merge or delete texts with text similarity higher than a threshold. Hash calculation is performed on the text content or file, and the hash values ​​of the two files are compared. The same hash values ​​are regarded as the same data, and the file deduplication operation can be performed. After the above method, it also includes standardization processing such as unit standardization and file naming normalization of the date and time format in the file. Then, the text, image and metadata in the file are respectively detected for abnormality to filter out files with abnormal detection results. Text anomaly detection includes combining an anomaly detection algorithm (Isolation Forest) based on the improved BERT model to identify abnormal patterns and formats that do not conform to the standard in the text, and using Siamese Network to perform similarity analysis on the text to detect duplicate content and inconsistencies. Image anomaly detection includes combining the convolutional neural network model with image forensics technology to identify traces of tampering in the image. Among them, the training samples of the convolutional neural network model are forged images generated based on the generative adversarial network. Metadata anomaly detection includes extracting the metadata of the file (such as creation time, modification time, file size, etc.), using the long short-term memory network model (LSTM: Long Short-Term Memory) to model the metadata in time series, and detecting whether there is abnormal modification behavior and inconsistent metadata. If so, the result of metadata detection anomaly is obtained.

[0133] The present invention provides a detection method for electronic files of pumped storage engineering. The embodiment of the present invention obtains the electronic files of pumped storage engineering to be detected and real-time operation environment data, and pre-processes the electronic files of pumped storage engineering to be detected to obtain target electronic files of pumped storage engineering; extracts key information of the target electronic files of pumped storage engineering, and classifies files according to the key information extraction results to obtain target electronic files of pumped storage engineering of different file categories, wherein the file categories include design files, construction drawings and acceptance reports; performs operation environment detection according to the real-time operation environment data, and when the operation environment detection is qualified, calls a corresponding pumped storage engineering file detection model for detection for each category of target electronic files of pumped storage engineering to obtain file detection results, wherein the pumped storage engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumped storage engineering file detection models include different directional detection sub-models; generates a detection report of the electronic files of pumped storage engineering to be detected according to the file detection results and the real-time operation environment data, realizes directional detection of files of different categories, makes the detection of the target electronic files of pumped storage engineering more targeted and more accurate, and at the same time, the operation environment detection ensures the safety and effectiveness of the detection process, thereby greatly improving the accuracy of the detection of the electronic files of pumped storage engineering.

[0134] Furthermore, as a response to the above Figure 1 The implementation of the method shown in the figure, the embodiment of the present invention provides a detection device for pumping engineering electronic files, such as Figure 3 As shown, the device comprises:

[0135] The acquisition module 31 is used to acquire the electronic file of the pumped storage project to be detected and the real-time operating environment data, and pre-process the electronic file of the pumped storage project to be detected to obtain the electronic file of the target pumped storage project;

[0136] A classification module 32 is used to extract key information from the target pumped storage engineering electronic files, and classify the files according to the key information extraction results to obtain target pumped storage engineering electronic files of different file categories, wherein the file categories include design documents, construction drawings and acceptance reports;

[0137] The detection module 33 is used to perform an operating environment detection according to the real-time operating environment data, and when the operating environment detection is qualified, for each category of target pumped storage engineering electronic files, call the corresponding pumped storage engineering file detection model to perform detection to obtain a file detection result, wherein the pumped storage engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumped storage engineering file detection models include different directional detection sub-models;

[0138] The generating module 34 is used to generate a detection report of the electronic file of the pumped storage project to be detected according to the file detection result and the real-time operating environment data.

[0139] Furthermore, the detection module 33 includes:

[0140] The first extraction unit is used to extract features of text and image in each category of the target extraction engineering electronic file by using a multimodal feature extraction network of a corresponding extraction engineering file detection model to obtain a file fusion feature, wherein the file fusion feature includes an image-text correlation feature, an image-text complementarity feature, and a file global feature;

[0141] A first detection unit is used for, if the target pumping engineering electronic file is a design file, using the directional detection sub-model to perform data signature detection, version detection, content consistency detection, metadata integrity detection, image clarity detection and technical specification detection on the target pumping engineering electronic file according to the file fusion feature, and to perform sensitive information encryption and access permission configuration;

[0142] A second detection unit is used for, if the target pumped storage engineering electronic file is a construction drawing, using the directional detection sub-model to perform drawing authenticity detection, drawing integrity detection and drawing readability detection on the target pumped storage engineering electronic file according to the file fusion feature, and configuring drawing encryption and access log data table;

[0143] The third detection unit is used to perform digital signature detection, image and text consistency detection, content integrity detection, data consistency detection and content availability detection on the target pumping project electronic file based on the file fusion features, if the target pumping project electronic file is an acceptance report, and to perform sensitive information encryption and access permission configuration using the directional detection sub-model.

[0144] Furthermore, in a specific application scenario, the extraction unit is specifically used to extract the image spatial features of the target extraction engineering electronic file in each category by using the image feature extraction network; extract the text temporal features of the target extraction engineering electronic file by using the text feature extraction network; and fuse the image spatial features and the text temporal features by using the feature fusion network to obtain file fusion features.

[0145] Furthermore, the device also includes:

[0146] A testing module, used to perform test detection on positive samples of pumped storage engineering electronic files based on different categories of the pumped storage engineering file detection models, and extract operating environment test data of different categories of pumped storage engineering file detection models;

[0147] A calculation module, for calculating, based on the data distribution of the operating environment test data, operating environment parameter intervals matching different types of file detection processes, wherein the operating environment parameter intervals include a CPU load interval, a memory occupancy interval, a network bandwidth rate interval, and a disk read speed interval;

[0148] The detection module 33 further includes:

[0149] A first determination unit is used to determine the file category of the target pumped storage project electronic file, and retrieve the operating environment parameter interval matching the file category, and if the real-time operating environment data satisfies the operating environment parameter interval, determine that the system operating environment detection is qualified;

[0150] The second determining unit is configured to determine that the system operating environment detection fails if the real-time operating environment data does not satisfy the operating environment parameter range.

[0151] Furthermore, the generating module 34 includes:

[0152] An acquisition unit, used to acquire detection process operating environment data, and determine operating environment difference data between the detection process operating environment data and the real-time operating environment data;

[0153] A first generating unit is configured to generate a normal operating environment status report if the operating environment difference data satisfies the operating environment difference interval, and generate a test report of the electronic file of the pumped storage project to be tested based on the normal operating environment status report and the test result;

[0154] The second generating unit is used to identify abnormal information matching the operating environment difference data from the system abnormal data mapping relationship set if the operating environment difference data does not satisfy the operating environment difference interval, and generate a test report of the electronic file of the pumped storage project to be tested based on the abnormal information and the test result.

[0155] Further, the classification module 32 includes:

[0156] A second extraction unit is used to extract project-related text in the target pumping engineering electronic file using a text information extraction network model to obtain text key information, wherein the text information extraction network model is trained based on corpus samples in a pumping engineering related corpus;

[0157] A third extraction unit is used to extract image information from the target pumping engineering electronic file using an image information extraction network model to obtain key image information, wherein the image information extraction network model includes a multi-layer convolutional network for feature extraction at different scales;

[0158] The third generating unit is used to generate a key information extraction result according to the text key information and the image key information.

[0159] Furthermore, the acquisition module 31 further includes:

[0160] A filtering unit, used for filtering the text in the electronic file of the pumping engineering to be detected by using a text processing model, and filtering the image in the electronic file of the pumping engineering to be detected by using a noise word filtering model;

[0161] A deduplication unit is used to match hash values ​​of the electronic files of the pumping engineering to be detected after filtering the noise words and noise images, and to perform deduplication processing on the files with the same hash values;

[0162] The fourth detection unit is used to perform text, image and metadata anomaly detection on the remaining electronic files of the pumping engineering to be detected after the file deduplication processing;

[0163] The third determination unit is used to determine the electronic file of the pumped storage project to be detected with normal results in all items as the target pumped storage project electronic file, and to issue an early warning for the electronic file of the pumped storage project to be detected with abnormal results in any detection item.

[0164] The present invention provides a detection device for electronic files of pumped storage engineering. The embodiment of the present invention obtains the electronic files of pumped storage engineering to be detected and real-time operation environment data, and pre-processes the electronic files of pumped storage engineering to be detected to obtain target electronic files of pumped storage engineering; extracts key information of the target electronic files of pumped storage engineering, and classifies files according to the key information extraction results to obtain target electronic files of pumped storage engineering of different file categories, wherein the file categories include design files, construction drawings and acceptance reports; performs operation environment detection according to the real-time operation environment data, and when the operation environment detection is qualified, for each category of target electronic files of pumped storage engineering, calls a corresponding pumped storage engineering file detection model for detection to obtain file detection results, wherein the pumped storage engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumped storage engineering file detection models include different directional detection sub-models; generates a detection report of the electronic files of pumped storage engineering to be detected according to the file detection results and the real-time operation environment data, realizes directional detection of files of different categories, makes the detection of the target electronic files of pumped storage engineering more targeted and more accurate, and at the same time, the operation environment detection ensures the safety and effectiveness of the detection process, thereby greatly improving the accuracy of the detection of the electronic files of pumped storage engineering.

[0165] According to one embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the method for detecting electronic files of pumping and storage engineering in any of the above method embodiments.

[0166] Figure 4 A schematic diagram of the structure of a terminal provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the terminal.

[0167] like Figure 4 As shown, the terminal may include: a processor (processor) 402 , a communication interface (Communications Interface) 404 , a memory (memory) 406 , and a communication bus 408 .

[0168] The processor 402 , the communication interface 404 , and the memory 406 communicate with each other via a communication bus 408 .

[0169] The communication interface 404 is used for network communication with other devices such as a client or other servers.

[0170] The processor 402 is used to execute the program 410, and specifically can execute the relevant steps in the above-mentioned embodiment of the method for detecting electronic files of pumped storage engineering.

[0171] Specifically, the program 410 may include program codes, which include computer operation instructions.

[0172] The processor 402 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention. The one or more processors included in the terminal may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0173] The memory 406 is used to store the program 410. The memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0174] The program 410 may be specifically configured to enable the processor 402 to perform the following operations:

[0175] Acquire the electronic file of the pumped storage project to be detected and the real-time operating environment data, and pre-process the electronic file of the pumped storage project to be detected to obtain the electronic file of the target pumped storage project;

[0176] Extract key information from the target pumped storage project electronic files, and classify the files according to the key information extraction results to obtain target pumped storage project electronic files of different file categories, wherein the file categories include design documents, construction drawings and acceptance reports;

[0177] Performing an operating environment test according to the real-time operating environment data, and when the operating environment test is qualified, calling a corresponding pumped storage engineering file detection model for each category of target pumped storage engineering electronic files to perform a test to obtain a file detection result, wherein the pumped storage engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumped storage engineering file detection models include different directional detection sub-models;

[0178] A test report of the electronic file of the pumped storage project to be tested is generated based on the file test result and the real-time operating environment data.

[0179] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0180] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting electronic files of pumped storage engineering, characterized in that: include: Acquire the electronic file of the pumped storage project to be detected and the real-time operating environment data, and pre-process the electronic file of the pumped storage project to be detected to obtain the electronic file of the target pumped storage project; Extract key information from the target pumped storage project electronic files, and classify the files according to the key information extraction results to obtain target pumped storage project electronic files of different file categories, wherein the file categories include design documents, construction drawings and acceptance reports; Performing an operating environment test according to the real-time operating environment data, and when the operating environment test is qualified, calling a corresponding pumped storage engineering file detection model for each category of target pumped storage engineering electronic files to perform a test to obtain a file detection result, wherein the pumped storage engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumped storage engineering file detection models include different directional detection sub-models; Generate a test report of the electronic file of the pumped storage project to be tested based on the file test result and the real-time operating environment data; The method of calling the corresponding pumped storage engineering file detection model for each category of target pumped storage engineering electronic files for detection to obtain the file detection results includes: For each category of the target pumping engineering electronic files, the multimodal feature extraction network of the corresponding pumping engineering file detection model is used to extract features of the text and image in the file to obtain file fusion features, which include image-text correlation features, image-text complementarity features and file global features; If the target pumping engineering electronic file is a design file, then based on the file fusion features, the directional detection sub-model is used to perform data signature detection, version detection, content consistency detection, metadata integrity detection, image clarity detection and technical specification detection on the target pumping engineering electronic file, and perform sensitive information encryption and access permission configuration; If the target pumped storage engineering electronic file is a construction drawing, then based on the file fusion feature, the directional detection sub-model is used to perform drawing authenticity detection, drawing integrity detection and drawing readability detection on the target pumped storage engineering electronic file, and configure drawing encryption and access log data tables; If the target pumping project electronic file is an acceptance report, then based on the file fusion features, the directional detection sub-model is used to perform digital signature detection, image and text consistency detection, content integrity detection, data consistency detection and content availability detection on the target pumping project electronic file, and sensitive information encryption and access permission configuration are performed.

2. The method according to claim 1, characterized in that The multimodal feature extraction network includes an image feature extraction network, a text feature extraction network and an image-text feature fusion network. For each category of the target extraction engineering electronic file, the multimodal feature extraction network corresponding to the extraction engineering file detection model is used to extract features from the text and image in the file to obtain the file fusion feature, including: For each category of the target extraction engineering electronic file, using the image feature extraction network to extract image space features in the target extraction engineering electronic file; Using the text feature extraction network to extract text temporal features in the target extraction engineering electronic file; The feature fusion network is used to fuse the image spatial features and the text temporal features to obtain file fusion features.

3. The method according to claim 2, characterized in that Before performing the operating environment detection according to the real-time operating environment data, the method further includes: Based on the different categories of pumped storage engineering file detection models, test detection is performed on the positive samples of pumped storage engineering electronic files, and operating environment test data of the different categories of pumped storage engineering file detection models are extracted; According to the data distribution of the operating environment test data, the operating environment parameter intervals matching the detection processes of different categories of files are calculated, and the operating environment parameter intervals include a CPU load interval, a memory occupancy interval, a network bandwidth rate interval, and a disk read speed interval; The performing of operating environment detection according to the real-time operating environment data includes: Determine the file category of the target pumped storage project electronic file, and retrieve the operating environment parameter interval matching the file category, and if the real-time operating environment data satisfies the operating environment parameter interval, determine that the system operating environment test is qualified; If the real-time operating environment data does not satisfy the operating environment parameter range, it is determined that the system operating environment test fails.

4. The method according to claim 1, characterized in that: The method of generating a test report of the electronic file of the pumped storage project to be tested based on the file test result and the real-time operating environment data includes: Acquire detection process operating environment data, and determine operating environment difference data between the detection process operating environment data and the real-time operating environment data; If the operating environment difference data satisfies the operating environment difference interval, a normal operating environment status report is generated, and a test report of the electronic file of the pumped storage project to be tested is generated based on the normal operating environment status report and the test result; If the operating environment difference data does not satisfy the operating environment difference interval, abnormal information matching the operating environment difference data is identified from the system abnormal data mapping relationship set, and a test report of the electronic file of the pumped storage project to be tested is generated based on the abnormal information and the test result.

5. The method according to claim 1, characterized in that The step of extracting key information from the target pumping and storage engineering electronic file includes: Extracting project-related text from the target pumping engineering electronic file using a text information extraction network model to obtain key text information, wherein the text information extraction network model is trained based on corpus samples in a pumping engineering-related corpus; The image information in the target pumping engineering electronic file is extracted using an image information extraction network model to obtain key image information, wherein the image information extraction network model includes a multi-layer convolutional network for feature extraction at different scales; A key information extraction result is generated according to the text key information and the image key information.

6. The method according to claim 1, characterized in that The preprocessing of the electronic file of the pumped storage project to be detected to obtain the target electronic file of the pumped storage project includes: Using a text processing model, filtering the text in the electronic file of the pumping engineering to be detected for noise words, and filtering the images in the electronic file of the pumping engineering to be detected for noise images; Hash value matching is performed on the electronic files of the pumping engineering to be detected after filtering the noise words and noise images, and the files with the same hash value are deduplicated; After the file deduplication process, the remaining electronic files of the pumping engineering to be tested are respectively subjected to text, image and metadata anomaly detection; The electronic files of pumped-storage projects to be inspected with normal results for all items are determined as target pumped-storage projects electronic files, and an early warning is issued for the electronic files of pumped-storage projects to be inspected with abnormal results for any inspection item.

7. A detection device for electronic files of pumped storage engineering, characterized in that: include: An acquisition module is used to acquire the electronic file of the pumped storage project to be detected and the real-time operating environment data, and pre-process the electronic file of the pumped storage project to be detected to obtain the electronic file of the target pumped storage project; A classification module is used to extract key information from the target pumped storage project electronic files, and classify the files according to the key information extraction results to obtain target pumped storage project electronic files of different file categories, wherein the file categories include design documents, construction drawings and acceptance reports; The detection module is used to perform an operating environment detection according to the real-time operating environment data, and when the operating environment detection is qualified, for each category of target pumped storage engineering electronic files, call the corresponding pumped storage engineering file detection model for detection to obtain a file detection result, wherein the pumped storage engineering file detection model includes a multimodal feature extraction network and a directional detection sub-model, and different pumped storage engineering file detection models include different directional detection sub-models; wherein, for each category of target pumped storage engineering electronic files, call the corresponding pumped storage engineering file detection model for detection to obtain a file detection result, specifically including: For each category of the target pumping engineering electronic files, the multimodal feature extraction network of the corresponding pumping engineering file detection model is used to extract features of the text and image in the file to obtain file fusion features, which include image-text correlation features, image-text complementarity features and file global features; If the target pumping engineering electronic file is a design file, then based on the file fusion features, the directional detection sub-model is used to perform data signature detection, version detection, content consistency detection, metadata integrity detection, image clarity detection and technical specification detection on the target pumping engineering electronic file, and perform sensitive information encryption and access permission configuration; If the target pumped storage engineering electronic file is a construction drawing, then based on the file fusion feature, the directional detection sub-model is used to perform drawing authenticity detection, drawing integrity detection and drawing readability detection on the target pumped storage engineering electronic file, and configure drawing encryption and access log data tables; If the target pumping and storage engineering electronic file is an acceptance report, then based on the file fusion features, the directional detection sub-model is used to perform digital signature detection, image and text consistency detection, content integrity detection, data consistency detection and content availability detection on the target pumping and storage engineering electronic file, and perform sensitive information encryption and access permission configuration; A generation module is used to generate a test report of the electronic file of the pumped storage project to be tested based on the file detection result and real-time operating environment data.

8. A storage medium storing at least one executable instruction, wherein the executable instruction enables a processor to execute an operation corresponding to the method for detecting electronic files of pumped storage engineering according to any one of claims 1 to 6.

9. A terminal, comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the method for detecting electronic files of pumped storage engineering as described in any one of claims 1-6.