File management method and system based on image recognition and electronic equipment

Abstract

The invention relates to the technical field of archive information security screening, in particular to an archive management method and system based on image recognition and electronic equipment, and the method comprises the steps: obtaining a to-be-recognized picture set corresponding to a to-be-managed archive based on a preset digital database, performing region division and marking on to-be-recognized pictures in the to-be-recognized picture set to obtain to-be-recognized marked regions; obtaining to-be-recognized images according to the to-be-recognized mark area, and sequentially inputting the to-be-recognized images into a preset training model to obtain a to-be-processed feature set corresponding to the to-be-recognized images; classifying and identifying the to-be-processed features in the to-be-processed feature set in sequence to obtain mark information corresponding to the to-be-processed features; and performing security level marking on the to-be-identified image based on the result type to obtain a security level marked image. According to the method, the type of the to-be-managed file can be quickly distinguished, so that manual identification operation is avoided, and the type identification efficiency of the to-be-managed file is improved.

Description

Technical Field

[0001] The present application relates to the technical field of archive information security screening, and in particular to an archive management method, system and electronic equipment for image recognition. Background Art

[0002] The field of archive management covers the collection, organization, preservation, retrieval and reuse of archives. It ensures the security and accessibility of historical data by providing structured information management strategies. With the rapid development of information technology and the advancement of archive digitization, archive management has expanded from traditional physical storage to digital processing of electronic archives. In relevant archive management systems, special attention is paid to the digital protection, retrieval efficiency and data security of archives.

[0003] In order to improve the efficiency of archive management, relevant technologies use AI technologies such as machine learning to enhance the intelligent processing capabilities of archive management, automatically classify and archive archives, identify and extract key information, and optimize search algorithms to improve retrieval efficiency.

[0004] However, the archives to be managed include confidential archives and non-confidential archives. Confidential archives and non-confidential archives are archived mixedly without clear markings in the directory, and it is impossible to directly distinguish confidential archives and non-confidential archives. In addition, due to the huge total amount of digital resources of archives to be managed, manual identification alone is inefficient. Summary of the invention

[0005] In order to facilitate the automatic identification of files to be managed and quickly distinguish the confidential types of files to be managed, the present application provides a file management method, system and electronic device based on image recognition.

[0006] In the first aspect, the present application provides an image recognition archive management method, which adopts the following technical solution: An image recognition archive management method comprises the following steps: Based on a preset digital database, a set of pictures to be identified corresponding to the archives to be managed is obtained, and the pictures to be identified in the set of pictures to be identified are divided into regions and marked to obtain marked regions to be identified; Acquire an image to be identified according to the marked area to be identified, and sequentially input the image to be identified into a preset training model to obtain a set of features to be processed corresponding to the image to be identified, wherein the set of features to be processed includes at least one feature to be processed; Classifying and identifying the features to be processed in the set of features to be processed in sequence to obtain tag information corresponding to the features to be processed, wherein the tag information includes a result type; The image to be identified is marked with a confidentiality level based on the result type to obtain a confidentiality level marked image, and the archive to be managed is managed and stored based on the confidentiality level marked image.

[0007] By adopting the above technical scheme, the pictures to be identified corresponding to the archives to be managed are divided into regions and marked to obtain the marked regions to be identified, and then the images to be identified corresponding to the marked regions to be identified are identified through a preset training model to obtain a set of features to be processed, and different result types are obtained by classifying and identifying the features to be processed in the set of features to be processed, and the images to be identified are marked based on the result types to obtain confidentiality-level marked images, thereby improving the management of the archives to be managed and facilitating the automatic identification of the archives to be managed. By obtaining the result type of the confidentiality-level marked images, the confidential types of the archives to be managed can be quickly distinguished, thereby avoiding manual identification operations and improving the type identification efficiency of the archives to be managed.

[0008] In one embodiment, the result type includes at least one item of data, and the step of sequentially classifying and identifying the features to be processed in the set of features to be processed to obtain tag information corresponding to the features to be processed includes the following steps: Determine whether there is marking information for the feature to be processed; If there is no marking information for the feature to be processed, key feature comparison is performed on the feature to be processed in sequence to obtain a corresponding category data set, wherein the category data set includes at least one category data, and the category data corresponds to the feature to be processed one by one; Determining whether the category data in the category data set are the same; If the category data in the category data set are the same, taking one of the category data as the result type corresponding to the image to be identified; If the category data in the category data set are different, a tag category is obtained according to the category data and a preset tag category, and the tag category is used as the result type of the image to be identified.

[0009] By adopting the above technical scheme, the category data set is obtained by comparing the key features of the features to be processed in turn, and the category data is judged to obtain the result type corresponding to the current image to be identified, and the category data corresponding to each feature to be processed is gradually and accurately marked, and the confidential area of ​​the archives to be managed is accurately found, thereby improving the management security of the archives to be managed.

[0010] In one embodiment, the step of sequentially comparing the key features of the features to be processed to obtain a corresponding category data set includes the following steps: Determine whether the feature to be processed is a processable feature; If the feature to be processed is a processable feature, performing key feature comparison based on the feature to be processed to obtain category data corresponding to the feature to be processed; If the feature to be processed is not of a processable type, the feature to be processed is segmented to obtain corresponding branch features, and key feature comparison is performed on the branch features to obtain corresponding branch result types, and the branch result types are marked according to a preset confidentiality level to obtain category data corresponding to the feature to be processed.

[0011] By adopting the above technical scheme, it is determined whether the feature to be processed is a processable feature. If so, a key feature comparison is performed based on the feature to be processed to obtain the category data corresponding to the feature to be processed. If not, it is necessary to first perform feature segmentation on the feature to be processed to obtain branch features, and perform key feature comparison on the branch features respectively to obtain the corresponding branch result type. The branch result type is marked according to the preset confidentiality level to obtain the category data corresponding to the feature to be processed, and then the category data is obtained in different ways according to the different features to be processed. This can improve the accuracy of obtaining the category data corresponding to the feature to be processed, accurately find the confidential areas in the archives to be managed, and thus improve the management security of the archives to be managed.

[0012] In one embodiment, the determining whether the feature to be processed is a processable feature comprises the following steps: Acquire a feature identifier according to the feature to be processed, and determine whether the feature identifier is unique, wherein the feature identifier represents the type of the feature to be processed; If the feature identifier is unique, determining that the feature to be processed is a processable feature; If the feature identifier is not unique, it is determined that the feature to be processed is not a processable feature.

[0013] By adopting the above technical scheme, for complex images, there are features to be processed with multiple feature identifiers, and it is necessary to obtain branch features, and perform key feature comparison on the branch features respectively to obtain corresponding branch result types, and mark the branch result types according to preset confidentiality levels to obtain category data corresponding to the features to be processed, so as to accurately screen out features to be processed with mixed types, obtain corresponding category data by different methods, and improve the accuracy of obtaining category data of features to be processed.

[0014] In one embodiment, the step of dividing and marking the to-be-recognized pictures in the to-be-recognized picture set to obtain the to-be-recognized marked areas includes the following steps: Scanning the to-be-recognized pictures in sequence according to preset intervals based on preset scanning routes to obtain a plurality of sets of scanning area sets, wherein the scanning area sets include scanning areas and data types corresponding to the scanning areas; The scanning areas having the same data type and being continuous are taken as pending areas, and the pending areas are marked to obtain the marked areas to be identified.

[0015] By adopting the above technical solution, the set of pictures to be identified is scanned in sequence to obtain the corresponding set of scan areas, and then the scan areas with the same data type and continuous are used as pending areas, and the pending areas are marked to obtain the marked areas to be identified, so as to facilitate the subsequent analysis of the marked areas to be identified, improve the efficiency of obtaining the images to be identified, facilitate the subsequent processing of the images to be identified, and be able to quickly distinguish the confidential types of the files to be managed, thereby improving the management efficiency of the files to be managed.

[0016] In one embodiment, the step of acquiring the image to be identified based on the marked area to be identified comprises the following steps: Matching a corresponding preprocessing model in a preprocessing database according to the mark area to be identified, wherein the preprocessing database stores a plurality of groups of preprocessing models, and the preprocessing model is used to perform clear processing on the mark area to be identified; A set of images to be recognized is generated based on the preprocessing model, wherein the set of images to be recognized includes at least one image to be recognized.

[0017] By adopting the above technical solution, based on the processing of the preprocessing model, clear images to be identified are obtained, thereby forming a corresponding set of images to be identified, which is convenient for subsequent operations, improves the efficiency of obtaining classified marked images, and can improve the management security of the archives to be managed.

[0018] In one embodiment, after obtaining the feature identifier according to the feature to be processed, the following steps are also included: Determine whether the feature identifier corresponding to the feature to be processed is a text identifier; If the feature identifier is a text identifier, a plurality of feature words to be identified are obtained based on the feature to be processed based on feature identification technology; Screening the feature words to be identified in a preset database, and determining whether corresponding comparison feature words can be obtained in the preset database; If the corresponding comparison feature word can be obtained in the preset database, the feature word to be identified is screened in the feature set to be processed to obtain a screening position, wherein the screening position represents the position of the comparison feature word in the image to be identified; A corresponding comparison result is obtained according to the comparison feature word, and the comparison result is used to mark the feature to be processed based on the screening position to obtain marking information.

[0019] By adopting the above technical solution, by pre-marking based on the feature words to be identified, the subsequent identification of each feature to be processed is reduced, the efficiency of obtaining confidentiality-level marked images is improved, and the work efficiency of the security and confidentiality of the files to be managed is improved.

[0020] In the second aspect, the present application provides an image recognition archive management system, which adopts the following technical solutions: An image recognition archive management system, comprising: An image acquisition module, wherein the image acquisition module obtains a set of to-be-recognized pictures corresponding to the to-be-managed archives based on a preset digital database, divides and marks the to-be-recognized pictures in the to-be-recognized picture set into regions, so as to obtain a to-be-recognized marked region; A feature extraction module, wherein the feature extraction module obtains an image to be identified according to the marked area to be identified, and sequentially inputs the image to be identified into a preset training model to obtain a set of features to be processed corresponding to the image to be identified, wherein the set of features to be processed includes at least one feature to be processed; A classification and identification module, wherein the classification and identification module sequentially classifies and identifies the features to be processed in the feature set to be processed to obtain tag information corresponding to the image to be identified, wherein the tag information includes a result type; A marking generation module is used to mark the image to be identified based on the result type to obtain a classified image, and to manage and store the archive to be managed based on the classified image.

[0021] By adopting the above technical solution, the image acquisition module collects the pictures to be identified from the digitized paper archives according to the set rules, divides and marks the pictures to be identified, generates the marked areas to be identified, and ensures that the acquired pictures meet the subsequent processing requirements. The feature extraction module is responsible for extracting the set of features to be processed corresponding to each type of images to be identified, so that the classification and identification module can classify and identify the images to be identified based on the features to be processed, and the marking generation module marks the images to be identified based on the marking information, thereby facilitating the subsequent scanning of the digitized paper archives, and can quickly distinguish the confidential types of the archives to be processed, thereby improving the work efficiency of the security and confidentiality of the archives to be managed.

[0022] In one embodiment, the confidentiality level mark includes a hyperlink mark, the mark information also includes image features and location information, and also includes a result output module, and the result output module is used to display the result type, image features and location information in a list form on a display screen for staff to view.

[0023] In a third aspect, the present application provides an electronic device, which adopts the following technical solution: An electronic device, comprising a processor and a memory coupled to each other, wherein the memory stores a computer program that can be run on the processor; When the computer program is executed by the processor, the image recognition archive management method as described in the first aspect is implemented.

[0024] In summary, the present application includes at least one of the following beneficial technical effects: 1. Divide and mark the image to be identified corresponding to the archive to be managed to obtain the marked area to be identified, and then identify the image to be identified corresponding to the marked area to be identified through the preset training model to obtain a set of features to be processed, and classify and identify the features to be processed in the set of features to be processed to obtain different result types, and mark the image to be identified based on the result type to obtain a confidentiality-level marked image, which is convenient for automatic identification of the archive to be managed. By obtaining the result type of the confidentiality-level marked image, the confidential type of the archive to be managed can be quickly distinguished, thereby avoiding manual identification operations and improving the efficiency of type identification of the archive to be managed; 2. Scan the set of pictures to be identified in sequence to obtain the corresponding set of scanning areas, then use the scanning areas with the same data type and continuous as the pending areas, and mark the pending areas to obtain the marked areas to be identified, so as to facilitate the subsequent analysis of the marked areas to be identified and improve the overall efficiency of obtaining the images to be identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a block diagram of the archive management method of image recognition provided by the embodiment of the present application; Figure 2 It is a block diagram of a method for obtaining a marked area to be identified provided in an embodiment of the present application; Figure 3 is a schematic diagram of a preset scanning route provided in an embodiment of the present application; Figure 4 is another scanning schematic diagram of a preset scanning route provided in an embodiment of the present application; Figure 5 It is a block diagram of a method for obtaining result types provided in an embodiment of the present application; Figure 6It is a schematic diagram of the structure of the archive management system provided in this embodiment; Figure 7 It is a structural block diagram of the electronic device provided in this embodiment.

[0026] Explanation of the accompanying drawings: 10, image acquisition module; 20, feature extraction module; 30, classification and identification module; 40, label generation module; 50, result output module; 61, processor; 62, memory; 63, computer program. DETAILED DESCRIPTION

[0027] To more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. However, it should be understood by those of ordinary skill in the art that the present application can be implemented without these details. In some cases, in order to avoid unnecessary descriptions that make various aspects of the present application obscure, well-known methods, processes, systems, components and / or circuits that have been described at a higher level will not be described in detail. For those of ordinary skill in the art, it is obvious that various changes can be made to the embodiments disclosed in the present application, and without departing from the principles and scope of the present application, the general principles defined in the present application can be applied to other embodiments and application scenarios. Therefore, the present application is not limited to the embodiments shown, but conforms to the broadest scope consistent with the scope claimed for protection of the present application.

[0028] The embodiment of the present application discloses an archive management method for image recognition, which is applied to the archive management system. Specifically, the archive management system marks the confidentiality level of the archives to be managed, and then outputs a list containing information such as the confidential file number and storage location by querying the paper digitized database and parsing the file path, so as to facilitate the archive management personnel to perform subsequent separation and separate storage operations.

[0029] like Figure 1 As shown, the image recognition archive management method includes the following steps: S100, based on a preset digital database, a set of to-be-recognized pictures corresponding to the to-be-managed archives is obtained, and the to-be-recognized pictures in the set of to-be-recognized pictures are divided into regions and marked to obtain marked regions to be recognized.

[0030] Among them, the preset digital database stores the digital data corresponding to the paper archives. The specific formation method of the preset digital database can be to scan the paper archives for PDF storage or to store them by manual input. For paper archives with official seals, they can be directly obtained by PDF scanning. The archives to be managed represent archives that need to be classified and marked. The set of pictures to be identified represents the set of pictures to be identified corresponding to the archives to be managed in the preset digital database. The marked area to be identified here represents the data formed by dividing and marking each picture to be identified. There is more than one marked area to be identified for the same picture to be identified. Each marked area to be identified obtained by the picture to be identified does not overlap, and the marked area to be identified is tiled to cover the picture to be identified.

[0031] Combination Figure 2 It should be noted that the steps of dividing and marking the to-be-recognized pictures in the to-be-recognized picture set to obtain the to-be-recognized marked areas include the following steps: S110, scanning the to-be-recognized picture in sequence according to a preset interval based on a preset scanning route to obtain a plurality of sets of scanning area sets.

[0032] S120, taking the scanning areas with the same data type and continuous as the pending areas, and marking the pending areas to obtain the marked areas to be identified.

[0033] The preset interval represents the size of the scanning area set in advance. The setting of the preset interval is specifically set according to the size of the image to be identified. In order to facilitate subsequent identification, the size of the preset interval must be smaller than the size of the image to be identified.

[0034] The preset scanning route is the scanning route for the picture to be identified. The preset scanning route is specifically determined according to the size of the preset interval. When the borders of the preset intervals are smaller than the borders of the picture to be identified, the scanning route may include scanning along the short borders of the picture to be identified or scanning along the long borders of the picture to be identified. When the borders of the preset intervals are equal to one border of the picture to be identified, scanning is performed to the opposite side based on the equal borders of the picture to be identified.

[0035] For example, Figure 3 The picture to be identified B, the preset interval A1, and the preset interval A2 are shown. When the preset interval of this embodiment adopts the preset interval A1, the preset scanning route scans the picture to be identified B according to the mark F1 in the figure until the entire picture to be identified B is scanned. When the preset interval of this embodiment adopts the preset interval A2, the preset scanning route scans the picture to be identified B according to the mark F2 in the figure until the entire picture to be identified B is scanned.

[0036] For example, when the preset interval of this embodiment is shown as the preset interval A3, the preset scanning route is as follows: Figure 4 As shown in the mark F3 in FIG. 1 , the image B to be identified is scanned.

[0037] It should be noted here that the preset scanning route is not limited to the scanning method provided in the present application, and other scanning methods may also be used. For example, when the preset interval adopts the preset interval A3, it can be scanned sequentially along the short edge of the image B to be identified, as long as the scanning of the image to be identified can be completed quickly.

[0038] The scan area set includes the scan area and the data type corresponding to the scan area. In order to facilitate the acquisition of the image to be identified, the scan areas with the same data type and continuous are regarded as the same pending area, and a feature area marking box is added to the pending area. Each marking box is given a unique ID to facilitate subsequent tracking and processing, and the position coordinates of the marking box, the picture number to which it belongs and other information are recorded in a special metadata file to ensure the traceability of the data.

[0039] It should be noted here that after obtaining the set of pictures to be identified based on the archive to be managed, the pictures to be identified need to be numbered according to the order of the paper documents of the archive to be managed.

[0040] In addition, for the collection of the picture set to be identified, an efficient image acquisition script is specifically used to batch obtain the picture set to be identified from the paper archive digitization results database in a parallel processing manner. During the acquisition process, a target detection algorithm based on deep learning is used, such as the improved CRAFT (Character Region Awareness for Text Detection) algorithm for text detection, and each picture to be identified in the picture set to be identified is automatically analyzed to obtain the pending area, which includes but is not limited to the text area, seal area, etc., and the feature area mark box is accurately added to the pending area to obtain the marked area to be identified.

[0041] It should be noted here that the specific marking method can use a hyperlink marking point to mark the position coordinates and the picture number corresponding to the undetermined area. Clicking the hyperlink marking point can display the corresponding feature area marking box.

[0042] S200, obtaining an image to be identified according to the marked area to be identified, and inputting the image to be identified into a preset training model in sequence to obtain a set of features to be processed corresponding to the image to be identified.

[0043] The image to be identified represents an image that needs to be feature extracted, and the image to be identified is an image from which the identification mark area is extracted, and the image to be identified can be directly used for subsequent processing. The preset training model is a model for extracting features from the image to be identified, and the feature set to be processed includes at least one feature to be processed, which represents the features obtained by identifying the image to be identified, and the feature to be processed can be a text feature and a seal image feature.

[0044] It should be noted here that the preset training model is an extracted trained deep neural network model, which combines the advantages of convolutional neural network (CNN) and recurrent neural network (RNN), and combines the connectionist temporal classification (CTC) algorithm for transcription to generate a comprehensive and accurate set of features to be processed.

[0045] For example, the features of printed text are extracted by alternating multiple CNN convolutional layers and pooling layers to extract features such as font, font size, and layout format of the text. For example, convolution kernels of different scales are used to extract features of font strokes, and the feature dimension is reduced through pooling operations while maintaining key feature information. The layout format of the text is determined by analyzing the statistical features of character spacing and line spacing.

[0046] For example, the seal image features are extracted by using the convolution operation of CNN, combined with color space conversion and texture analysis algorithms, to extract the shape, color, texture and other features of the seal. For example, the outline shape of the seal is identified by using an edge detection algorithm, the color features of the seal are extracted by color histogram analysis, and the texture features of the seal are extracted by using methods such as gray level co-occurrence matrix.

[0047] For example, to extract handwritten text features, CNN is used to extract local features such as stroke length, angle, and curvature, and then the stroke sequence is modeled through a bidirectional long short-term memory network (Bi-LSTM) to learn writing style features and fully capture the time series information and contextual dependencies in the handwriting process.

[0048] Acquiring an image to be identified based on the marked area to be identified includes the following steps: S210, matching a corresponding preprocessing model in a preprocessing database according to the marked area to be identified.

[0049] S220: Generate a set of images to be recognized based on the preprocessing model.

[0050] Among them, the preprocessing database stores several groups of preprocessing models, which are used to clearly process the mark area to be identified. The set of images to be identified includes at least one image to be identified. The image to be identified mentioned here is an image that can be used for subsequent identification. Some pictures may be unclear when scanned. Therefore, different preprocessing models are automatically assigned according to the characteristics of different target image sets to be classified by using image recognition and classification technology. By analyzing the image features corresponding to the mark area to be identified, such as color, texture, etc., the type of image is determined, so as to select the appropriate preprocessing model.

[0051] It should be noted here that these image preprocessing models use a combination of techniques such as image denoising, background removal and occlusion removal. For example, for printed text images, denoising techniques such as Gaussian filtering are used to remove noise from the image. For images with classified seals, image segmentation techniques are used to separate the seal from the background, and then background removal is performed. For handwritten text images, morphological processing and other methods are used to remove occlusions. For seal images, a special seal contour enhancement algorithm is used to highlight the seal contour through edge detection and contour extraction techniques, while removing background interference.

[0052] S300, classifying and identifying the features to be processed in the feature set to be processed in turn to obtain label information corresponding to the features to be processed.

[0053] The tag information includes the result type, the features to be identified, and the corresponding position information, etc. The features to be processed in the feature set to be processed are classified and identified. Specifically, based on the bidirectional long short-term memory network (Bi-LSTM) model, combined with the attention mechanism and transfer learning technology, the feature set to be processed is efficiently and accurately classified. The model is pre-trained on a large-scale labeled data set.

[0054] It should be noted here that specific situations require specific analysis, so the training data of the model can be specifically adjusted according to the scheme of this embodiment to improve the accuracy and generalization ability of classification. The specific classification process is as follows, and the result types include secret type, confidential type, top secret type, and non-confidential type. The model automatically determines the result type of the feature to be identified by learning the feature to be identified in the feature set to be processed.

[0055] S400, marking the image to be identified based on the result type to obtain a classified image, and managing and storing the archive to be managed based on the classified image.

[0056] Among them, the confidentiality mark image represents an image with marking information. The archive management system can obtain the result type, features to be identified and corresponding location information corresponding to the image to be identified by scanning the marking information of the confidentiality mark image corresponding to the image to be identified, and can quickly find the location of confidential archives, thereby facilitating the archive management personnel to perform subsequent stripping and separate storage operations.

[0057] It should be noted here that in the classification of the image to be identified based on the result type, the image to be identified is marked based on the classification standard corresponding to the feature set to be processed and the model, and the image to be identified is marked using image annotation technology. The specific marking method can be to use a hyperlink mark point to mark the feature to be identified and the result type corresponding to the image to be identified, and click the hyperlink mark point to display the corresponding marking information.

[0058] Reference Figure 5 In one embodiment, the result type includes at least one data, and the features to be processed in the feature set to be processed are classified and identified in turn to obtain the tag information corresponding to the features to be processed, including the following steps: S310, determining whether there is any marking information on the feature to be processed.

[0059] S320: If there is no marking information for the feature to be processed, key feature comparison is performed on the features to be processed in sequence to obtain a corresponding category data set.

[0060] S330, determining whether the category data in the category data set are the same.

[0061] S340: If the category data in the category data set are the same, one of the category data is used as the result type corresponding to the image to be identified.

[0062] S350: If the category data in the category data set are different, obtain the tag category according to the category data set and the preset tag category, and use the tag category as the result type of the image to be recognized.

[0063] The category data set includes at least one category data, and the category data corresponds to the feature to be processed one by one. The category data represents the result type corresponding to the feature to be processed, and the preset tag categories include top secret type, confidential type, secret type, and non-confidential type. The tag category represents the type that is matched first when matching the preset tag types.

[0064] Specifically, obtaining the tag category according to the category data set and the preset tag category includes the following steps: sequentially obtaining data of the preset tag type and using it as the comparison category, and matching the comparison category in the category data set, if matched, using the comparison category as the tag category, and ending the matching. If not matched, sequentially obtaining the next data of the preset tag type until all the data of the preset tag type are matched.

[0065] It should be noted here that if there is marking information for the feature to be processed, the marking information is identified, and the feature to be identified of the marking information and the corresponding position information are obtained for identification, and the above-obtained feature to be identified is skipped in the feature set to be processed, thereby reducing the recognition of the entire image to be identified.

[0066] It should be noted here that for the extraction of marked information, for the areas where the identification information such as "secret level" and "confidentiality level" of the paper archives are located, we first use the target detection model based on the attention mechanism to accurately locate them, and then use CNN to extract the text's character shape, size and other features, as well as text alignment, font color and other format features.

[0067] In one embodiment, key feature comparison is performed on the features to be processed in sequence to obtain a corresponding category data set, including the following steps: S321, determining whether the feature to be processed is a processable feature.

[0068] S322: If the feature to be processed is a processable feature, a key feature comparison is performed based on the feature to be processed to obtain category data corresponding to the feature to be processed.

[0069] S323: If the feature to be processed is not of a processable type, the feature to be processed is segmented to obtain corresponding branch features, and key feature comparison is performed on the branch features to obtain corresponding branch result types.

[0070] S324, marking the branch result type according to the preset confidentiality level to obtain category data corresponding to the feature to be processed.

[0071] Specifically, the processable feature represents a feature that can be compared with key features, and judging whether the feature to be processed is a processable feature includes the following steps: S321-1, obtaining a feature identifier based on the feature to be processed, and determining whether the feature identifier is unique.

[0072] S321-2: If the feature identifier is unique, determine that the feature to be processed is a processable feature.

[0073] S321-3, if the feature identifier is not unique, it is determined that the feature to be processed is not a processable feature.

[0074] The feature identifier represents the type of the feature to be processed, and the feature identifier can be a text identifier, a seal image, etc.

[0075] In step S323-step S324, when the feature to be processed is not of a processable type, the feature to be processed is segmented to obtain the corresponding branch feature, and the key feature comparison is performed on the branch features to obtain the corresponding branch result type. The branch result type is marked according to the preset confidentiality level to obtain the category data corresponding to the feature to be processed.

[0076] For the case where there are multiple feature identifiers in the same feature to be processed, such as both printed text and seals, and different levels of confidentiality, a multi-branch network structure is used to perform independent feature extraction and classification judgment on different types of feature areas. Then, the classification results of each branch are comprehensively analyzed through a fusion strategy and processed separately according to the classification attributes. The marking box only marks the feature area of ​​the current classification attribute, and the hyperlink marking point only marks the image features and classification standards of the classification attribute it is in, ensuring the accuracy and interpretability of the classification results.

[0077] This embodiment determines the type of feature identification. When the feature identification only has printed text, it can be directly detected and identified, and the confidential type of the feature to be processed can be automatically identified, for example, the marking information of the feature to be processed can be automatically obtained. When the image file with a confidentiality seal on the feature identification is detected and identified, the confidential type of the feature to be processed can be automatically identified. When the feature identification is handwritten text, the handwritten text image file in the digitized result can be detected and identified, the confidential type of the feature to be processed can be automatically identified, and the confidential files with confidentiality can be assisted in stripping. In addition, for some paper archives with confidential information, the confidential information includes identifications such as secret level and confidentiality degree, and the confidentiality degree includes confidential and top secret. When the identified confidential information is a secret level, the confidential type of the paper archive is determined to be a non-confidential type. When the identified confidential information is a secret level and confidentiality degree, the confidential type of the paper archive is determined to be a confidential situation.

[0078] It should be noted here that confidential situations can be determined based on the degree of confidentiality, so I will not go into details here.

[0079] In one embodiment, after obtaining the feature identifier according to the feature to be processed, the following steps are also included: S410, determining whether the feature identifier corresponding to the feature to be processed is a text identifier.

[0080] S420: If the feature identifier is a text identifier, a plurality of feature words to be identified are obtained based on the feature to be processed and the feature identification technology.

[0081] S430, screening the feature words to be identified in a preset database, and determining whether corresponding comparison feature words can be obtained in the preset database.

[0082] S440: If the corresponding comparison feature word can be obtained in the preset database, the feature word to be identified is screened in the feature set to be processed to obtain a screening position.

[0083] S450, obtaining a corresponding comparison result according to the comparison feature word, and marking the comparison result for the feature to be processed based on the screening position to obtain marking information.

[0084] Among them, the feature word to be identified represents the feature word obtained based on the feature to be processed. The feature word is a key feature word. Several groups of feature words are stored in the preset database. The feature word corresponds to a corresponding confidentiality level. The screening position represents the position of the feature word in the image to be identified.

[0085] Specifically, the feature words to be identified are screened in the feature set to be processed, the screening positions with the same confidentiality level are obtained, and the feature words to be identified at the screening positions are marked according to the comparison results to obtain the marking information.

[0086] The implementation principle is: Based on the preset digital database, a set of pictures to be identified corresponding to the archives to be managed is obtained, and the pictures to be identified in the set of pictures to be identified are divided into regions and marked to obtain the marked regions to be identified. The images to be identified are obtained according to the marked regions to be identified, and the images to be identified are sequentially input into the preset training model to obtain the set of features to be processed corresponding to the images to be identified. The features to be processed in the set of features to be processed are classified and identified sequentially to obtain the marking information corresponding to the features to be processed. The images to be identified are marked with confidentiality based on the result type to obtain confidentiality marked images, and the archives to be managed are managed and stored based on the confidentiality marked images.

[0087] The embodiment of the present application also discloses an image recognition archive management system.

[0088] like Figure 6As shown, the archive management system for image recognition includes an image acquisition module 10, a feature extraction module 20, a classification and identification module 30, and a tag generation module 40. The image acquisition module 10 obtains a set of pictures to be identified corresponding to the archive to be managed based on a preset digital database, divides and marks the pictures to be identified in the set of pictures to be identified, so as to obtain a marked area to be identified. The feature extraction module 20 obtains the image to be identified based on the marked area to be identified, and sequentially inputs the image to be identified into a preset training model to obtain a set of features to be processed corresponding to the image to be identified, and the set of features to be processed includes at least one feature to be processed. The classification and identification module 30 sequentially classifies and identifies the features to be processed in the set of features to be processed, so as to obtain tag information corresponding to the image to be identified, and the tag information includes the result type. The tag generation module 40 marks the image to be identified with a confidentiality level based on the result type, so as to obtain a confidentiality-marked image, and manages and stores the archive to be managed based on the confidentiality-marked image.

[0089] The other functions performed in the above-mentioned image acquisition module 10, feature extraction module 20, classification and identification module 30 and label generation module 40 and the technical details of each function are the same or similar to the corresponding features in the image recognition archive management method described above, so they are not repeated here.

[0090] The confidentiality mark includes a hyperlink mark, and the mark information also includes image features and location information. The image recognition archive management system also includes a result output module 50 and an image preprocessing module. The result output module 50 is used to display the result type, image features and location information in a list form on the display screen for staff to view. The image preprocessing module allocates corresponding image preprocessing models according to different image types to achieve denoising, background removal and occlusion removal of feature areas, thereby improving image quality.

[0091] Specifically, the image acquisition module 10 is responsible for collecting images from the digitalization results of paper archives according to the set rules, corresponding to the function of calling the image acquisition script, ensuring that the acquired images are complete and meet the requirements of subsequent processing. The feature extraction module 20 is built based on CRNN and is responsible for extracting the features of various types of images to form an image feature set. The classification and identification module 30 is equipped with a Bi-LSTM model, which classifies and identifies images according to the image feature set and determines their confidentiality level. The tag generation module 40 completes image tagging, generates hyperlink tag points, and outputs a report containing auxiliary stripping information based on the result output module 50 to facilitate archive management operations.

[0092] It should be noted here that CRNN is a deep learning technology that combines convolutional neural networks (CNN) and recurrent neural networks (RNN). The CNN part uses the convolution kernel of the convolution layer to slide in the data space dimension to achieve efficient extraction of local features of the input data. The input data can be two-dimensional images and multi-dimensional signals. After stacking layers, multi-scale features from basic edges to complex structures can be mined. The pooling layer downsamples the convolution results, reduces the data dimension, retains key features, and enhances the robustness of the model to data transformation. The RNN part, especially when using long short-term memory networks (LSTM) or gated recurrent units (GRU), effectively handles time series dependencies through gating mechanisms. At each time step, the hidden state is updated based on the current input and the hidden state at the previous moment to achieve continuous memory and transmission of sequence information. Bi-LSTM, or bidirectional long short-term memory network, is a deep learning model that performs well in sequence data processing. It can effectively capture forward and reverse context information.

[0093] The embodiment of the present application also discloses an electronic device.

[0094] Reference Figure 7 The electronic device includes a processor 61 and a memory 62 coupled to each other, and the memory 62 stores a computer program 63 that can be run on the processor 61.

[0095] When the computer program 63 is executed by the processor 61 , the file management method of image recognition is implemented.

[0096] The memory 62 may be a ROM or other type of static storage device that can store static information and instructions, a random access memory 62, or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory 62, a read-only optical disc or other optical disc storage, an optical disc storage (including a compact disc, a laser disc, an optical disc, a digital versatile disc, a Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 62 may be an internal storage unit in some embodiments.

[0097] The processor 61 may be a central processing unit 61, a general purpose processor 61, a digital signal processor 61, an application specific integrated circuit, a field programmable gate array or other programmable logic device, a transistor logic device, a hardware component or any combination thereof, for running the program code stored in the memory 62 or processing data.

[0098] The processor 61 and the memory 62 are connected via a bus. The bus may include a path to transmit information between the above components. The bus may be a peripheral component interconnect standard bus or an extended industrial standard structure bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0099] Figure 7 Only an electronic device having a memory 62, a processor 61, and a bus is shown, and those skilled in the art can understand that Figure 7 The structure shown does not constitute a limitation on the electronic device, which can be a bus structure or a star structure. The electronic device can also include more or fewer components than shown in the figure, or combine certain components, or deploy different components. Other existing or future electronic devices may be applicable, should also be included in the scope of protection, and are included here by reference.

[0100] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the instructions of the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise clearly stated in this document, the execution of these steps is not strictly limited in order and can be performed in other orders.

[0101] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A file management method for image recognition, characterized in that: The following steps are involved: Based on a preset digital database, a set of pictures to be identified corresponding to the archives to be managed is obtained, and the pictures to be identified in the set of pictures to be identified are divided into regions and marked to obtain marked regions to be identified; Acquire an image to be identified according to the marked area to be identified, and sequentially input the image to be identified into a preset training model to obtain a set of features to be processed corresponding to the image to be identified, wherein the set of features to be processed includes at least one feature to be processed; Classifying and identifying the features to be processed in the set of features to be processed in sequence to obtain tag information corresponding to the features to be processed, wherein the tag information includes a result type; The image to be identified is marked with a confidentiality level based on the result type to obtain a confidentiality level marked image, and the archive to be managed is managed and stored based on the confidentiality level marked image.

2. The image recognition file management method according to claim 1, characterized in that: The result type includes at least one item of data, and the step of sequentially classifying and identifying the features to be processed in the set of features to be processed to obtain tag information corresponding to the features to be processed includes the following steps: Determine whether there is marking information for the feature to be processed; If there is no marking information for the feature to be processed, key feature comparison is performed on the feature to be processed in sequence to obtain a corresponding category data set, wherein the category data set includes at least one category data, and the category data corresponds to the feature to be processed one by one; Determining whether the category data in the category data set are the same; If the category data in the category data set are the same, taking one of the category data as the result type corresponding to the image to be identified; If the category data in the category data set are different, a tag category is obtained according to the category data set and a preset tag category, and the tag category is used as the result type of the image to be identified.

3. The image recognition file management method according to claim 2, characterized in that: The step of sequentially comparing the key features of the features to be processed to obtain a corresponding category data set includes the following steps: Determine whether the feature to be processed is a processable feature; If the feature to be processed is a processable feature, performing key feature comparison based on the feature to be processed to obtain category data corresponding to the feature to be processed; If the feature to be processed is not of a processable type, the feature to be processed is segmented to obtain corresponding branch features, and key feature comparison is performed on the branch features to obtain corresponding branch result types, and the branch result types are marked according to a preset confidentiality level to obtain category data corresponding to the feature to be processed.

4. The image recognition file management method according to claim 3, characterized in that: The step of determining whether the feature to be processed is a processable feature comprises the following steps: Acquire a feature identifier according to the feature to be processed, and determine whether the feature identifier is unique, wherein the feature identifier represents the type of the feature to be processed; If the feature identifier is unique, determining that the feature to be processed is a processable feature; If the feature identifier is not unique, it is determined that the feature to be processed is not a processable feature.

5. The image recognition file management method according to claim 1, characterized in that: The step of dividing and marking the to-be-recognized pictures in the to-be-recognized picture set to obtain the to-be-recognized marked areas comprises the following steps: Scanning the to-be-recognized pictures in sequence according to preset intervals based on preset scanning routes to obtain a plurality of sets of scanning area sets, wherein the scanning area sets include scanning areas and data types corresponding to the scanning areas; The scanning areas having the same data type and being continuous are taken as pending areas, and the pending areas are marked to obtain the marked areas to be identified.

6. The image recognition file management method according to claim 1, characterized in that: The step of obtaining the image to be identified based on the marked area to be identified comprises the following steps: Matching a corresponding preprocessing model in a preprocessing database according to the mark area to be identified, wherein the preprocessing database stores a plurality of groups of preprocessing models, and the preprocessing model is used to perform clear processing on the mark area to be identified; A set of images to be recognized is generated based on the preprocessing model, wherein the set of images to be recognized includes at least one image to be recognized.

7. The image recognition file management method according to claim 4, characterized in that: After obtaining the feature identifier according to the feature to be processed, the following steps are also included: Determine whether the feature identifier corresponding to the feature to be processed is a text identifier; If the feature identifier is a text identifier, a plurality of feature words to be identified are obtained based on the feature to be processed based on feature identification technology; Screening the feature words to be identified in a preset database, and determining whether corresponding comparison feature words can be obtained in the preset database; If the corresponding comparison feature word can be obtained in the preset database, the feature word to be identified is screened in the feature set to be processed to obtain a screening position, wherein the screening position represents the position of the comparison feature word in the image to be identified; A corresponding comparison result is obtained according to the comparison feature word, and the comparison result is used to mark the feature to be processed based on the screening position to obtain marking information.

8. An image recognition archive management system, characterized in that: include: An image acquisition module (10), the image acquisition module (10) acquiring a set of to-be-identified pictures corresponding to the to-be-managed archive based on a preset digital database, dividing and marking the to-be-identified pictures in the set of to-be-identified pictures into regions, so as to obtain a marked region to be identified; A feature extraction module (20), the feature extraction module (20) acquiring an image to be identified based on the marked area to be identified, and sequentially inputting the image to be identified into a preset training model to obtain a set of features to be processed corresponding to the image to be identified, the set of features to be processed comprising at least one feature to be processed; A classification and identification module (30), wherein the classification and identification module (30) sequentially classifies and identifies the features to be processed in the set of features to be processed, so as to obtain label information corresponding to the image to be identified, wherein the label information includes a result type; A marking generation module (40) is used to mark the image to be identified with a confidentiality level based on the result type to obtain a confidentiality level marked image, and to manage and store the file to be managed based on the confidentiality level marked image.

9. The image recognition archive management system according to claim 8, characterized in that: The confidentiality level mark comprises a hyperlink mark, the mark information further comprises image features and position information, and further comprises a result output module (50), wherein the result output module (50) is used to display the result type, image features and position information in a list form on a display screen.

10. An electronic device, characterized in that: The electronic device comprises a processor (61) and a memory (62) coupled to each other, wherein the memory (62) stores a computer program (63) that can be run on the processor (61); When the computer program (63) is executed by the processor (61), the image recognition archive management method according to any one of claims 1 to 7 is implemented.

Images