Electronic document detection management system based on construction qualification of high-voltage switch cabinet

Abstract

The invention relates to the technical field of intelligent qualification identification, and particularly discloses an electronic document detection management system based on high-voltage switch cabinet construction qualification, and the system comprises a feature table construction module which is used for identifying electronic qualification documents and constructing a feature table; the qualification rating module is used for receiving qualification rating input by the staff based on the feature table to obtain qualification rating items; the sample pool construction module is used for reading a preset number of features in the feature table, constructing a feature-qualification rating sample and establishing a sample pool; the document identification module is used for training a neural network model based on the sample pool; according to the method, the features are extracted from the big data sample, several features are selected according to the actual situation, the neural network model is trained in combination with the rating of the electronic document by a worker, and the training process of the neural network model is a shallow learning process and is extremely high in efficiency due to the fact that the number of the features is small and the features are extracted in advance.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent qualification identification, and specifically to an electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets. Background Art

[0002] Electric power construction qualification refers to a series of qualifications and conditions that an electric power engineering construction enterprise needs to possess when conducting electric power engineering construction. The number of documents to be reviewed for electric power construction qualification is extremely large, and the workload of pure manual review is very heavy. Therefore, many auxiliary electronic document detection schemes have emerged in the prior art.

[0003] However, on the premise of meeting the major standards, different regions and different units may have different regulations and standards for the construction party, which makes the migration of the electronic document detection scheme extremely low, and its input cost is very high. Therefore, in most cases, the demander still adopts the manual detection method; how to reduce the input cost of the electronic document detection scheme and improve its migratability, so that the intelligent detection scheme can truly help the reviewers is the technical problem to be solved by the technical solution of the present invention. Summary of the Invention

[0004] The purpose of the present invention is to provide an electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets to solve the problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] An electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets, the system includes:

[0007] A feature table construction module, which is used to determine the document arrangement order of the electronic qualification documents, identify the electronic qualification documents based on the document arrangement order, and construct a feature table; each row of the feature table corresponds to an uploader, and each column corresponds to a type of identified feature;

[0008] A qualification rating module, which is used to receive the qualification rating input by the staff based on the feature table to obtain the qualification rating items;

[0009] A sample pool construction module, which is used to read a preset number of features in the feature table, construct samples of feature-qualification rating, and establish a sample pool;

[0010] A document recognition module, which is used to train a neural network model based on the sample pool, and when receiving a new electronic qualification document, identify the new electronic qualification document based on the neural network model to obtain the qualification rating;

[0011] An identification accuracy determination module, which is used to cluster the input of the neural network model every time a qualification rating is output, calculate the variance of each category of input, and judge the identification accuracy of each qualification rating according to the variance;

[0012] A sample update module, which is used to update the samples of each qualification rating in the sample pool according to the identification accuracy.

[0013] As a further solution of the present invention: the feature table construction module includes:

[0014] A first-order determination unit, which is used to obtain the type names of all document types of the electronic qualification documents, and determine the first order according to the type names;

[0015] A second-order determination unit, which is used to query the content volume of each document, and determine the second order according to the content volume; the content volume is determined by the pixel values of each pixel;

[0016] A third-order determination unit, which is used to receive the importance level of each electronic qualification document determined by the management party, and determine the third order according to the importance level;

[0017] An arrangement order determination unit, which is used to determine the document arrangement order of the electronic qualification documents by combining the first order, the second order and the third order;

[0018] A construction execution unit, which is used to identify the electronic qualification documents according to the document arrangement order, extract features, and construct a feature table.

[0019] As a further solution of the present invention: the content of identifying the electronic qualification documents according to the document arrangement order, extracting features, and constructing a feature table includes:

[0020] Read the documents in sequence according to the document arrangement order, traverse each pixel point in the documents, and determine the text box and the image box according to the color value of the pixel point;

[0021] Extract the text in the text box, extract keywords in the text as text features; the text features at least include identity information and license information; the identity information includes contractor information, licensor information and staff information;

[0022] Extract the image features of the image box; the image features at least include an image histogram and an image contour;

[0023] Statistically calculate the text features and the image features according to the feature extraction order, and construct a feature table.

[0024] As a further solution of the present invention: the content of statistically calculating the text features and the image features according to the feature extraction order and constructing a feature table includes:

[0025] The staff pre - inputs a rated feature library for various types of features; the rated features in the rated feature library are the standardized data for each type of feature;

[0026] Verify each feature based on the rated feature library to determine the maximum similarity;

[0027] Statistically analyze all the maximum similarities to obtain a feature group;

[0028] Statistically analyze the feature groups corresponding to each electronic qualification document to obtain a feature table.

[0029] As a further solution of the present invention: the qualification rating module includes:

[0030] A rating receiving unit for receiving the qualification rating of each electronic qualification document by the staff;

[0031] A rating inserting unit for statistically analyzing the qualification ratings and inserting them into the feature table as qualification rating items;

[0032] A feature group comparison unit for regularly extracting electronic qualification documents with the same qualification rating, reading their feature groups, comparing the feature groups, generating a confirmation request according to the comparison result, and sending it to the staff.

[0033] As a further solution of the present invention: the sample pool construction module includes:

[0034] A data extraction unit for extracting other data in the feature table except the qualification rating items to obtain a sample matrix;

[0035] A data processing unit for calculating the covariance matrix of the sample matrix, calculating the eigenvalues and eigenvectors of the covariance matrix;

[0036] A feature selection unit for selecting eigenvectors based on the eigenvalues as features, constructing samples of feature - qualification rating, and establishing a sample pool.

[0037] As a further solution of the present invention: the content of selecting eigenvectors based on the eigenvalues includes:

[0038] Sort the eigenvectors according to the eigenvalues;

[0039] Sequentially select eigenvectors, and calculate the principal component contribution rate and cumulative contribution rate according to the eigenvalues;

[0040] When the output condition is reached, output the selection result; the output condition includes that the number of selected eigenvectors reaches a preset number and the cumulative contribution rate reaches 80%;

[0041] Among them, the calculation process of the principal component contribution rate of the k - th feature is: The cumulative contribution rate of the first m features is: Wherein, is the principal component contribution rate of the k-th feature, λ k is the eigenvalue of the k-th eigenvector after sorting, p is the total number of features, and i is the serial number of the eigenvector; λ i is the eigenvalue of the i-th eigenvector; is the cumulative contribution rate of the first m features.

[0042] As a further solution of the present invention: The document recognition module includes:

[0043] A sample pool splitting unit for splitting the sample pool into a training set and a test set to train the neural network model; the element ratio of the training set to the test set is 8:2;

[0044] A process query unit for querying the recognition process corresponding to the input of the neural network model and extracting the features of the electronic qualification document when a new electronic qualification document is received;

[0045] A model application unit for inputting the extracted features into the neural network model to obtain a qualification rating.

[0046] As a further solution of the present invention: The recognition accuracy determination module includes:

[0047] An input classification unit for classifying the inputs corresponding to the same qualification rating into one category;

[0048] A similarity calculation unit for calculating the similarity between any two inputs in the same category of inputs;

[0049] A variance calculation unit for calculating the average similarity and calculating the similarity variance according to the average similarity;

[0050] An accuracy judgment unit for judging the recognition accuracy of each qualification rating according to the similarity variance; the recognition accuracy is inversely proportional to the similarity variance.

[0051] As a further solution of the present invention: The sample update module includes:

[0052] A feature group query unit for querying all feature groups corresponding to the qualification rating in the feature table;

[0053] A quantity update unit for updating the preset quantity of features in the sample construction process according to the recognition accuracy of the qualification rating.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention extracts features from big data samples, selects several features according to the actual situation, and combines the ratings of electronic documents by staff to train a neural network model. Since the number of features is small and pre-extracted, the training process of the neural network model is a shallow learning process with extremely high efficiency; in this process, each requester only needs to rate a part of the electronic documents to obtain an intelligent recognition model that can assist in work. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings in the following descriptions are only some embodiments of the present invention.

[0056] Figure 1 It is a composition structure diagram of an electronic document detection and management system based on the construction qualification of high-voltage switchgear. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0058] Figure 1 It is a flow block diagram of an electronic document detection and management system based on the construction qualification of high-voltage switchgear. In an embodiment of the present invention, an electronic document detection and management system based on the construction qualification of high-voltage switchgear, the system 10 includes:

[0059] A feature table construction module 11, configured to determine the document arrangement order of electronic qualification documents, identify the electronic qualification documents based on the document arrangement order, and construct a feature table; each row of the feature table corresponds to an uploader, and each column corresponds to a type of identified feature;

[0060] A qualification rating module 12, configured to receive the qualification rating input by the staff based on the feature table to obtain a qualification rating item;

[0061] A sample pool construction module 13, configured to read a preset number of features from the feature table, construct samples of feature-qualification ratings, and establish a sample pool;

[0062] A document recognition module 14, configured to train a neural network model based on the sample pool, and when a new electronic qualification document is received, identify the new electronic qualification document based on the neural network model to obtain a qualification rating;

[0063] The recognition accuracy determination module 15 is used to cluster the input of the neural network model every time a qualification rating is output, calculate the variance of each type of input, and determine the recognition accuracy of each qualification rating according to the variance.

[0064] The sample update module 16 is used to update the samples of each qualification rating in the sample pool according to the recognition accuracy.

[0065] In an example of the technical solution of the present invention, an electronic qualification document recognition solution based on a neural network model is provided. Each electronic qualification document uploaded by a user is a document set composed of multiple PDF files and arranged in a certain order. Each document in the document set is recognized in turn, and multiple features can be extracted. The features are the core information in the document, such as the main name or official seal, etc. After the features are extracted, a set composed of the features can be constructed. Each electronic qualification document corresponding to a user will ultimately be converted into a feature set. Indexed by the user's identity label, these feature sets are statistically analyzed to obtain a feature table; the feature table reflects the feature sets of the electronic qualification documents uploaded by multiple users participating in the document detection. By receiving the qualification rating of each electronic qualification document by the staff, a qualification rating item can be introduced into the feature table, so as to obtain the feature - qualification rating of each user.

[0066] Since there are a huge number of documents and a very large number of features in the electronic qualification documents, if all features are used to train the neural network model, the workload will be very large. Therefore, the present invention first selects a part of the features from the features, and then trains the relationship between this part of the features and the qualification rating, thereby generating a neural network model with not very high accuracy. With the trained neural network model, the newly received electronic qualification documents are recognized to obtain the qualification rating; on this basis, the electronic qualification documents with the same qualification rating are statistically analyzed and compared for differences. If the accuracy is high, the electronic qualification documents with the same qualification rating should be similar. If they are not similar, it means that the accuracy is not high, and the number of selected features needs to be increased to update the neural network model. This is a negative feedback architecture, which can obtain a neural network model with sufficient performance at a relatively low cost.

[0067] In an example of the technical solution of the present invention, the working process of the feature table construction module 11 is specifically defined. The feature table construction module includes:

[0068] The first - order determination unit is used to obtain the type names of all document types of the electronic qualification document and determine the first order according to the type names.

[0069] A second-order determination unit for querying the content volume of each type of document and determining the second order according to the content volume; the content volume is determined by the pixel values of each pixel;

[0070] A third-order determination unit for receiving the importance level of each electronic qualification document determined by the management party and determining the third order according to the importance level;

[0071] An arrangement order determination unit for determining the document arrangement order of the electronic qualification documents by combining the first order, the second order, and the third order;

[0072] A construction execution unit for identifying the electronic qualification documents according to the document arrangement order, extracting features, and constructing a feature table.

[0073] The above content considers the recognition order of electronic qualification documents. The simplest way is that the staff sends a template and the user fills it in. At this time, the default order in the template is the arrangement order of the documents in the electronic qualification documents; this method is very conventional, but there is one drawback, that is, if there is a problem with the electronic qualification document itself, then during the feature recognition process, the recognition will only jump out when the problematic document is recognized. When the problematic document is at the back, the amount of ineffective recognition is very large. In the above solution, on the basis of the basic order (the first order), the order is adjusted according to the content volume and the importance level, so that the execution subject of this method first recognizes important documents, can detect problematic documents faster, and reduces the amount of ineffective recognition to a certain extent.

[0074] As a preferred embodiment of the technical solution of the present invention, the content of identifying the electronic qualification documents according to the document arrangement order, extracting features, and constructing a feature table includes:

[0075] Read the documents in sequence according to the document arrangement order, traverse each pixel point in the document, and determine the text box and the image box according to the color value of the pixel point;

[0076] Extract the text in the text box, extract keywords in the text as text features; the text features at least include identity information and license information; the identity information includes contractor information, licensor information, and staff information;

[0077] Extract the image features of the image box; the image features at least include an image histogram and an image contour;

[0078] Statistically count the text features and the image features according to the feature extraction order, and construct a feature table.

[0079] After determining the document arrangement order, identify the electronic qualification documents according to the document arrangement order, and use basic image recognition algorithms to locate text boxes and image boxes. The text boxes contain identity information, and the image boxes contain official seals. Extract the text and keywords in the text in the text box as text features. The text features include identity information and license information. The identity information includes contractor information, licensor information, and staff information, including name, rank, working years, and specialty, etc.; the image features are relatively simple, including contour features and pixel color value distribution features of the entire image, and can be extracted using existing extraction schemes.

[0080] Further, the content of constructing the feature table by statistically counting text features and image features in the order of feature extraction includes:

[0081] Pre-input the rated feature library for various features by the staff; the rated features in the rated feature library are the standardized data of each feature;

[0082] Verify each feature based on the rated feature library to determine the maximum similarity;

[0083] Statistically count all the maximum similarities to obtain a feature group;

[0084] Statistically count the feature groups corresponding to each electronic qualification document to obtain a feature table.

[0085] Regarding the statistical process of text features and image features, since the number of features and the data structure of the features are very complex, the statistical process is very difficult. Therefore, the above content introduces a standardization process, that is, a standard data is preset for each feature, and the extracted features are compared with the standard data. At this time, a percentage value indicating whether they are similar can represent the feature, which makes the feature table become a table composed of pure numbers.

[0086] Specifically, regarding the standard data, in actual applications, the standard data corresponding to the same feature is very likely to be not unique. This application allows the staff to provide more standard data. When comparing, multiple percentage values indicating whether they are similar will be calculated, and the maximum value (the maximum similarity) can be selected.

[0087] As a preferred embodiment of the technical solution of the present invention, the qualification rating module 12 includes:

[0088] A rating receiving unit for receiving the qualification ratings of each electronic qualification document by the staff;

[0089] A rating inserting unit for statistically counting the qualification ratings and inserting them into the feature table as qualification rating items;

[0090] A feature group comparison unit, which is used to regularly extract electronic qualification documents with the same qualification rating, read their feature groups, compare the feature groups, generate a confirmation request according to the comparison result, and send it to the staff.

[0091] The working process of the qualification rating module 12 is very simple. It only needs to receive the qualification ratings of each electronic qualification document from the staff. After receiving the qualification rating, it inserts it into the feature table. In this process, the present invention will also regularly extract electronic qualification documents with the same qualification rating, and then compare whether they are similar. If the difference is very large, it is very likely that the staff input is incorrect. At this time, the staff is reminded to confirm whether there is an error.

[0092] As a preferred embodiment of the technical solution of the present invention, the sample pool construction module 13 includes:

[0093] A data extraction unit, which is used to extract other data in the feature table except the qualification rating item to obtain a sample matrix;

[0094] A data processing unit, which is used to calculate the covariance matrix of the sample matrix, and calculate the eigenvalues and eigenvectors of the covariance matrix;

[0095] A feature selection unit, which is used to select eigenvectors based on the eigenvalues as features, construct samples of feature - qualification rating, and establish a sample pool.

[0096] The core of the sample pool construction process is to select several features from the features of multiple users, and then construct samples of feature - qualification rating.

[0097] Specifically, the content of selecting eigenvectors based on the eigenvalues includes:

[0098] Sort the eigenvectors according to the eigenvalues;

[0099] Select eigenvectors in sequence, and calculate the principal component contribution rate and cumulative contribution rate according to the eigenvalues;

[0100] When the output condition is reached, output the selection result; the output conditions include that the number of selected eigenvectors reaches the preset number and the cumulative contribution rate reaches 80%.

[0101] For a sample containing p users and n features, calculating its covariance matrix is a basic matrix operation process, which is not difficult for those skilled in the art. Then, p eigenvalues can be obtained from this covariance matrix and sorted from large to small. Each eigenvalue corresponds to an eigenvector (column vector); then, select eigenvectors in sequence and calculate the principal component contribution rate and cumulative contribution rate.

[0102] Among them, the calculation process of the principal component contribution rate of the k - th feature is: The cumulative contribution rate of the first m features is as follows: In the formula, is the principal component contribution rate of the k-th feature, and λ k is the eigenvalue of the k-th eigenvector after sorting, p is the total number of features, and i is the serial number of the eigenvector; λ i is the eigenvalue of the i-th eigenvector; is the cumulative contribution rate of the first m features

[0103] When the cumulative contribution rate is large enough or the number of eigenvectors is large enough, the selection process is terminated.

[0104] As a preferred embodiment of the technical solution of the present invention, the document recognition module 14 includes:

[0105] A sample pool splitting unit, configured to split the sample pool into a training set and a test set for training a neural network model; the element ratio of the training set to the test set is 8:2;

[0106] A process query unit, configured to query the recognition process corresponding to the input of the neural network model and extract the features of the electronic qualification document when a new electronic qualification document is received;

[0107] A model application unit, configured to input the extracted features into the neural network model to obtain a qualification rating.

[0108] The above content is the training process of the neural network model. When the sample pool has been constructed, and the number of features is small and the extraction process is clear, the training process is a shallow learning process with extremely low training difficulty.

[0109] It should be noted that the input of the trained neural network model is the features extracted by the same feature extraction method (the method provided by the feature table construction module 11), and the input is the qualification rating.

[0110] As a preferred embodiment of the technical solution of the present invention, the recognition accuracy determination module 14 includes:

[0111] An input classification unit, configured to classify the inputs corresponding to the same qualification rating into one category;

[0112] A similarity calculation unit, configured to calculate the similarity between any two inputs in the same category;

[0113] A variance calculation unit, configured to calculate the average similarity and calculate the similarity variance according to the average similarity;

[0114] An accuracy judgment unit, configured to judge the recognition accuracy of each qualification rating according to the similarity variance; the recognition accuracy is inversely proportional to the similarity variance.

[0115] In an example of the technical solution of the present invention, a specific recognition accuracy determination scheme is provided. First, query the inputs (multiple features) with the same qualification rating to obtain an input set, calculate the similarity between any two inputs (the comparison method adopts the comparison method of features and standard data), so as to convert the input set into the similarity between each pair, calculate the variance of the similarity, and the variance reflects the difference situation of the same type of inputs. In the framework of the technical solution of the present invention, the greater the variance, the lower the recognition accuracy. Although this relationship may not be correct, the purpose of this application is to improve the accuracy of the neural network model, which means that as long as the recognition accuracy is not definitely the highest, then the accuracy of the neural network model is improved.

[0116] Finally, the sample update module 15 includes:

[0117] A feature group query unit, configured to query all feature groups corresponding to the qualification rating in the feature table;

[0118] A quantity update unit, configured to update the preset quantity of features in the sample construction process according to the recognition accuracy of the qualification rating.

[0119] The update process of this application is to adjust the quantity of selected features, that is, the output condition in the content of selecting feature vectors based on the feature values in the sample pool construction module 13.

[0120] All functions that can be realized by the electronic document detection and management system based on the high-voltage switch cabinet construction qualification are completed by computer devices. The computer devices include one or more processors and one or more memories. At least one program code is stored in the one or more memories, and the program code is loaded and executed by the one or more processors to implement the electronic document detection and management system based on the high-voltage switch cabinet construction qualification.

[0121] The processor fetches instructions from the memory one by one, analyzes the instructions, and then completes corresponding operations according to the requirements of the instructions, generating a series of control commands to make each part of the computer act automatically, continuously and coordinately, becoming an organic whole, realizing the input of the program, the input of data, and the operation and output of results. All arithmetic operations or logical operations generated in this process are completed by the arithmetic unit; the memory includes a read-only memory (ROM), and the read-only memory is used to store computer programs, and a protection device is provided outside the memory.

[0122] Exemplarily, the computer program can be divided into one or more modules, and one or more modules are stored in the memory and executed by the processor to complete the present invention. One or more modules can be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.

[0123] Those skilled in the art can understand that the description of the above service device is only an example and does not constitute a limitation on the terminal device. It may include more or fewer components than the above description, or combine some components, or different components. For example, it may include input / output devices, network access devices, buses, etc.

[0124] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The above processor is the control center of the above terminal device, and connects various parts of the entire user terminal through various interfaces and lines.

[0125] The above memory can be used to store computer programs and / or modules. The above processor realizes various functions of the above terminal device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as information collection template display function, product information release function, etc.); the data storage area can store data created according to the use of the berth status display system (such as product information collection templates corresponding to different product types, product information to be released by different product providers, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disks, memory, plug-in hard disks, smart media cards (SMCs), secure digital (SD) cards, flash cards, at least one magnetic disk storage device, flash device, or other volatile solid-state storage devices.

[0126] If the modules / units integrated in the terminal device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable medium. Based on this understanding, to implement all or part of the modules / units in the above-described embodiment system of the present invention, it can also be completed by instructing relevant hardware through a computer program. The above computer program can be stored in a computer-readable medium. When the computer program is executed by a processor, the functions of the above various system embodiments can be realized. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0127] It should be noted that in this article, the term "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article, or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article, or device including that element.

[0128] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. An electronic document detection and management system based on the construction qualification of high-voltage switchgear, characterized in that, The system includes: A feature table construction module, which is used to determine the document arrangement order of electronic qualification documents, identify the electronic qualification documents based on the document arrangement order, and construct a feature table; each row of the feature table corresponds to an uploader, and each column corresponds to a type of identified feature; A qualification rating module, which is used to receive the qualification rating input by the staff based on the feature table to obtain qualification rating items; A sample pool construction module, which is used to read a preset number of features in the feature table, construct samples of feature - qualification ratings, and establish a sample pool; A document recognition module, which is used to train a neural network model based on the sample pool, and when receiving a new electronic qualification document, identify the new electronic qualification document based on the neural network model to obtain a qualification rating; An identification accuracy determination module, which is used to cluster the input according to the qualification rating every time the neural network model outputs a qualification rating, calculate the variance of each type of input, and judge the identification accuracy of each qualification rating according to the variance; A sample update module, which is used to update the samples of each qualification rating in the sample pool according to the identification accuracy.

2. The electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets according to claim 1, characterized in that The feature table construction module includes: A first - order determination unit, which is used to obtain the type names of all document types of electronic qualification documents and determine the first order according to the type names; A second - order determination unit, which is used to query the content volume of each document and determine the second order according to the content volume; the content volume is determined by the pixel values of each pixel; A third - order determination unit, which is used to receive the importance level of each electronic qualification document determined by the management party and determine the third order according to the importance level; An arrangement order determination unit, which is used to combine the first order, the second order, and the third order to determine the document arrangement order of electronic qualification documents; A construction execution unit, which is used to identify the electronic qualification documents according to the document arrangement order, extract features, and construct a feature table.

3. The electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets according to claim 2, characterized in that, The content of identifying the electronic qualification documents according to the document arrangement order, extracting features, and constructing a feature table includes: Reading the documents in sequence according to the document arrangement order, traversing each pixel point in the document, and determining the text box and the image box according to the color value of the pixel point; Extracting the text in the text box, extracting keywords in the text as text features; the text features at least include identity information and license information; the identity information includes contractor information, licensor information, and staff information; Extracting the image features of the image box; the image features at least include an image histogram and an image contour; Statistically counting the text features and the image features according to the feature extraction order to construct a feature table.

4. The electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets according to claim 3, characterized in that, The content of statistically counting the text features and the image features according to the feature extraction order to construct a feature table includes: Pre - inputting a rated feature library about various features by the staff; the rated features in the rated feature library are the standardized data of each feature; Verifying each feature based on the rated feature library to determine the maximum similarity; Statistically counting all the maximum similarities to obtain a feature group; Statistically counting the feature groups corresponding to each electronic qualification document to obtain a feature table.

5. The electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets according to claim 1, wherein The qualification rating module includes: A rating receiving unit, which is used to receive the qualification rating of each electronic qualification document by the staff; A rating insertion unit, which is used to count the qualification ratings, and as qualification rating items, insert them into the feature table; A feature group comparison unit, which is used to regularly extract electronic qualification documents with the same qualification ratings, read their feature groups, compare the feature groups, generate a confirmation request according to the comparison result, and send it to the staff.

6. The electronic document detection and management system based on the construction qualification of high-voltage switch cabinets according to claim 4, characterized in that, The sample pool construction module includes: A data extraction unit, which is used to extract other data in the feature table except the qualification rating items to obtain a sample matrix; A data processing unit, which is used to calculate the covariance matrix of the sample matrix, and calculate the eigenvalues and eigenvectors of the covariance matrix; A feature selection unit, which is used to select eigenvectors based on the eigenvalues as features, construct samples of feature - qualification ratings, and establish a sample pool.

7. The electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets according to claim 5, characterized in that, The content of selecting eigenvectors based on the eigenvalues includes: Sort the eigenvectors according to the eigenvalues; Select eigenvectors in turn, and calculate the principal component contribution rate and cumulative contribution rate according to the eigenvalues; When the output conditions are met, output the selection result; the output conditions include that the number of selected eigenvectors reaches the preset number and the cumulative contribution rate reaches 80%; Among them, the calculation process of the principal component contribution rate of the k-th feature is as follows: The cumulative contribution rate of the first m features is: In the formula, is the principal component contribution rate of the k-th feature, λ k is the eigenvalue of the k-th eigenvector after sorting, p is the total number of features, and i is the serial number of the eigenvector; λ i is the eigenvalue of the i-th eigenvector; is the cumulative contribution rate of the first m features.

8. The electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets according to claim 5, characterized in that, The document recognition module includes: A sample pool splitting unit, which is used to split the sample pool into a training set and a test set to train a neural network model; the element ratio of the training set to the test set is 8:2; A process query unit, which is used to query the recognition process corresponding to the input of the neural network model and extract the features of the electronic qualification document when receiving a new electronic qualification document; A model application unit, which is used to input the extracted features into the neural network model to obtain a qualification rating.

9. The electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets according to claim 5, wherein The recognition accuracy determination module includes: An input classification unit, which is used to classify the inputs corresponding to the same qualification rating into one category; A similarity calculation unit, which is used to calculate the similarity between any two inputs in the same category of inputs; A variance calculation unit, which is used to calculate the average similarity and calculate the similarity variance according to the average similarity; An accuracy judgment unit, which is used to judge the recognition accuracy of each qualification rating according to the similarity variance; the recognition accuracy is inversely proportional to the similarity variance.

10. The electronic document detection and management system based on the construction qualification of high-voltage switchgear cabinets according to claim 6, characterized in that, The sample update module includes: A feature group query unit, which is used to query all feature groups corresponding to the qualification rating in the feature table; A quantity update unit, which is used to update the preset quantity of features in the sample construction process according to the recognition accuracy of the qualification rating.

Images