A digital editing system based on intelligent review of content of teaching aids

Abstract

The application provides a kind of digital editing system based on intelligent review of teaching aid book content, it is related to content review technical field, the application is integrated dynamic digital watermark engine and blockchain two-way binding technology through multi-modal copyright protection module, realize reversible watermark generation and real-time tamper monitoring, combine three-dimensional version atlas and intelligent contract traceability, solve the copyright tracking problem;Based on the operation conversion algorithm and dynamic permission control of cloud collaborative review module, realize multi-person real-time collaborative editing and conflict detection;Through adaptive image enhancement module and hybrid segmentation network, significantly improve the text segmentation accuracy in complex scene, comprehensively solve the copyright protection, collaborative editing and image processing problems in the review of teaching aid books.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of content review, in particular to a digital editing system based on intelligent review of teaching aid book content. BACKGROUND

[0002] The digital editing system is a multimedia content processing platform based on computer technology, integrating data collection, storage, editing and output functions of text, images, audio, video and other data. Its core technologies include non-linear editing, cloud computing storage, artificial intelligence assisted processing (such as automatic labeling, intelligent editing) and cross-terminal collaboration. The system improves efficiency through algorithm optimization (such as encoding compression, real-time rendering), gradually replacing traditional physical medium editing methods, and becomes an industry core tool for teaching aid book content review.

[0003] In the prior art, the digital editing system based on teaching aid book content enhancement with publication number CN115953785B, relates to the field of image processing, specifically to a digital editing system based on teaching aid book content enhancement. The system includes a data acquisition module, a text segmentation module and a teaching aid resource retrieval module. The data acquisition module acquires teaching plan images. The text segmentation module obtains a projection histogram from the teaching plan image. The possible degree of each point as a segmentation point is obtained from the projection histogram, and then a plurality of first candidate segmentation points are obtained. The segmentation point compliance of each first candidate segmentation point is calculated. Then a plurality of second candidate segmentation points are obtained, and the probability of each second candidate segmentation point is calculated. The accurate segmentation point is obtained according to the probability of each second candidate segmentation point. All standard texts of the teaching plan image are obtained according to the accurate segmentation point. The teaching aid resource retrieval module retrieves the corresponding teaching aid materials using the keywords in all standard texts of the teaching plan image, thereby improving the accuracy of teaching plan image text segmentation.

[0004] However, the above technical solution still has the following technical defects when applied to the teaching aid book content review industry:

[0005] 1. Incomplete copyright protection and content traceability mechanism:

[0006] The above solution does not integrate digital watermarking, copyright blockchain and other technologies, making it difficult to trace the source of teaching aid materials or prevent unauthorized tampering, which poses a risk in education publishing compliance review;

[0007] 2. Lack of real-time collaboration and version management functions:

[0008] The above solution focuses on single-machine or local processing and does not introduce cloud-based collaborative editing, real-time annotation or version tracking functions, making it difficult to meet the needs of teacher team collaboration in revising teaching aid materials in modern education scenarios.

[0009] 3. The text segmentation accuracy is greatly affected by the image quality:

[0010] The above scheme relies on binary image and projection histogram for text segmentation. If the teaching image has blur, tilt, shadow or complex background interference (such as handwritten notes, table lines, etc.), it may lead to misjudgment of segmentation points (such as continuous characters are incorrectly split or connected characters are not separated), affecting the accuracy of subsequent OCR recognition.

[0011] The above information disclosed in the background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0012] The purpose of the present application is to provide a digital editing system based on intelligent review of the content of teaching aids, in order to solve the problems raised in the background art.

[0013] To achieve the above purpose, the present application provides the following technical solutions:

[0014] A digital editing system based on intelligent review of the content of teaching aids, comprising a multi-modal copyright protection module, a cloud-based collaborative review module, an adaptive image enhancement module and an intelligent semantic verification module;

[0015] Multi-modal copyright protection module: for generating reversible watermark images, automatically adjusting the watermark strength according to the content type, and monitoring tampering behavior in real time and triggering traceability alarm;

[0016] Cloud-based collaborative review module: establishes a data channel with the multi-modal copyright protection module, and uses an operation conversion algorithm to realize real-time collaborative editing by multiple people;

[0017] By constructing a three-dimensional version graph, the record content revision history and watermark change track of the reversible watermark image are compared, and when a modification conflict is identified, the reversible watermark image is marked as a pending image;

[0018] For the remaining reversible watermark images that do not have conflicts, mark them as compliant images, and dynamically control their watermark visibility and editing range based on role permissions;

[0019] Adaptive image enhancement module: first receives the pending images marked by the cloud-based collaborative review module, and uses a cascade processing of a degradation model detector and a conditional GAN to enhance the image quality in real time;

[0020] Then use a hybrid segmentation network combined with a dynamic projection histogram to segment the text area of the pending image;

[0021] An intelligent semantic verification module is configured to receive all reversible watermark images, and perform content compliance review, watermark authorization verification and semantic analysis, and automatically generate alternative solutions for illegal content.

[0022] Further, the multi-modal copyright protection module includes an image generation unit.

[0023] The image generation unit converts the text, images and formulas in the supplementary textbook content into reversible watermark images with copyright identifiers through a dynamic digital watermark engine, and extracts the peak signal-to-noise ratio aa, structural similarity index ab and normalized correlation coefficient ac from the comparison between the reversible watermark images and the pixel matrix of the original content in the supplementary textbook content.

[0024] Further, the multi-modal copyright protection module further includes a tampering monitoring unit.

[0025] The tampering monitoring unit is configured to calculate and evaluate the watermark distortion degree Asy in real time to monitor tampering behavior and trigger a traceability alarm.

[0026] The watermark distortion degree Asy is obtained by non-dimensional processing and fitting of the peak signal-to-noise ratio aa, the structural similarity index ab and the normalized correlation coefficient ac, and the specific calculation formula is as follows:

[0027]

[0028] In the formula, Asy∈(0,1], and the value tends to 1, indicating smaller distortion.

[0029] Based on the supplementary image samples containing text, formulas and images, the watermark distortion degree Asy distribution under normal editing and malicious tampering is counted to determine the classification boundary. After aligning with the industry standard, the first watermark distortion threshold A1 and the second watermark distortion threshold A2 are preset, and the first watermark distortion threshold A1 is greater than the second watermark distortion threshold A2.

[0030] The first watermark distortion threshold A1 and the second watermark distortion threshold A2 are compared and evaluated with the watermark distortion degree Asy to monitor tampering behavior and trigger a traceability alarm. The specific evaluation content is as follows:

[0031] If the watermark distortion degree Asy is greater than the first watermark distortion threshold A1, it is determined to be a legal operation, the blockchain audit log is updated, and it is marked as "compliant version";

[0032] If the second watermark distortion threshold A2 is less than the watermark distortion degree Asy and the first watermark distortion threshold A1, it indicates that there is non-malicious editing, including compression and format conversion; at the same time, a first alarm trigger manual review request is issued.

[0033] Watermark distortion Asy < second watermark distortion threshold A2, determine as malicious tampering; At this time, the second alarm is issued, and the document editing permission is automatically frozen.

[0034] Further, the cloud collaborative review module includes a collaborative editing unit.

[0035] The collaborative editing unit is configured to establish a data channel with the multi-modal copyright protection module through a secure API interface, enabling bidirectional transmission of reversible watermark images and editing instructions. A collaborative editing algorithm based on operation adjustment and merging is used to sequentially reconstruct and resolve conflicts of editing operations of multiple users. A coordination mechanism adjusts and synchronizes editing operations of different users based on operation timestamps and context-dependent rules. The specific adjustment method is as follows:

[0036] The order of operations is determined by comparing operation timestamps, and in the case of location conflicts or logical conflicts in operation content, adjustment strategies such as operation transfer, insertion redirection, or position offset are dynamically executed.

[0037] Further, the cloud collaborative review module further includes a version conflict unit.

[0038] The version conflict unit is based on three types of data: version timeline of reversible watermark images, editing operation sequence, and watermark change trajectory. It uses a graph database or chain structure to construct a three-dimensional version map of reversible watermark images, with each editing behavior as a node and editing order and dependency as edges. Through automatic capture of content modification logs and watermark parameter change logs during each image editing, the three-dimensional version map is associated and mapped for tracking and recording content and watermark change trajectories.

[0039] By analyzing the specific editing operation type and position change in the content modification log, the content difference between versions is calculated to obtain the content change distance Ba. By analyzing the embedding position, intensity, or feature parameters in the watermark parameter change log, the offset degree of the watermark trajectory is calculated to obtain the watermark offset intensity Bb.

[0040] Then, the content change distance Ba and the watermark offset intensity Bb are extracted and dimensionless processed, and the following formula is used to calculate the version difference degree Bcy to identify whether there is a modification conflict:

[0041]

[0042] The preset version difference threshold B is compared and evaluated with the version difference degree Bcy, and the specific evaluation content is as follows:

[0043] If the version difference Bcy is less than or equal to the version difference threshold B, it is considered that the content change of the version is within a reasonable consistent range with the watermark change, no modification conflict is identified, and the version is automatically fused;

[0044] If the version difference Bcy is greater than the version difference threshold B, it is considered that there is an abnormal difference between the content change of the version and the watermark change, including a modification conflict that cannot be reconciled, at this time, the reversible watermark image corresponding to the version is marked as a "to-be-processed image" and a conflict warning is triggered.

[0045] Further, the cloud collaborative review module further includes a compliance control unit;

[0046] The compliance control unit is configured to mark images without conflicts as compliant; and through a pre-set role permission configuration rule, different user roles are assigned corresponding watermark visibility levels and editable area parameters, when a user accesses or edits a reversible watermark image, the permission configuration data is automatically called to perform visibility processing on the watermark information in the image, including masking, blurring or complete display, while limiting the operable image area boundary, only allowing the user to perform editing operations within the authorized range, and operation requests beyond the range will be prohibited.

[0047] Further, the adaptive image enhancement module includes an image enhancement unit;

[0048] The image enhancement unit combines the version difference Bcy to prioritize image processing; the greater the version difference Bcy, the more significant the difference between versions, the higher the image processing priority, and enhancement processing is performed first;

[0049] The image enhancement unit internally integrates a degradation model detector for identifying the degradation type of the input image; the degradation type includes blurring, noise and compression artifacts, and the degradation intensity parameter is extracted as an enhancement condition according to the detection result;

[0050] The adaptive image enhancement module uses the content change distance Ba and the watermark offset strength Bb to model the degradation scene in multiple dimensions;

[0051] Subsequently, the content change distance Ba and the watermark offset strength Bb are input as a condition vector into a conditional generative adversarial network to perform personalized enhancement operations on the image.

[0052] Further, the adaptive image enhancement module further includes a text version segmentation unit;

[0053] Based on the enhanced to-be-processed image, first, the to-be-processed image is subjected to feature extraction and preliminary region division through a hybrid segmentation network to construct a candidate region set; in the process, based on the region feature map extracted by the hybrid segmentation network, the texture complexity fa, the edge continuity fb and the morphological consistency fc of the candidate region are quantitatively collected through a texture analysis operator, an edge detection operator and a morphological operator respectively;

[0054] Subsequently, after the texture complexity fa, the edge continuity fb and the morphological consistency fc are extracted and subjected to dimensionless processing, the region text confidence coefficient Fbz is calculated and acquired to quantify the credibility of each candidate region as a text region; the specific calculation formula is as follows:

[0055]

[0056] In the formula, 0≤region text confidence coefficient Fbz≤1;

[0057] When the region text confidence coefficient Fbz tends to 0, it indicates that the region text feature conforms to the ideal feature;

[0058] When the region text confidence coefficient Fbz tends to 1, it indicates that the region text feature does not conform to the ideal feature;

[0059] Finally, in combination with the dynamic projection histogram algorithm, the candidate region whose region text feature conforms to the ideal feature is positioned and segmented, and a structured text region is extracted.

[0060] Further, the intelligent semantic verification module includes a semantic analysis unit;

[0061] The semantic analysis unit is configured to perform text region segmentation on the input compliant image, also adopts a combination of a hybrid segmentation network and a dynamic projection histogram, and extracts a structured text region; then, based on a natural language processing model, the text region of the enhanced to-be-processed image and the compliant image is subjected to content compliance review, including sensitive word detection, knowledge error checking and semantic compliance analysis; finally, a compliance judgment result is output, and the position and type of the illegal content are marked.

[0062] Further, the intelligent semantic verification module further includes an authorization verification unit;

[0063] The authorization verification unit is configured to verify the image watermark authorization state, trigger a traceability alarm by comparing the watermark change track and the version graph record, identify unauthorized tampering behavior, and generate a replacement proposal, including recommending compliant material replacement, marking traceability correction suggestions and triggering an artificial review process.

[0064] Compared with the prior art, the beneficial effects of the present application are: through the integration of a dynamic digital watermark engine and a two-way binding technology of a blockchain in a multi-modal copyright protection module, reversible watermark generation and tampering monitoring of teaching aid content are realized; combined with three-dimensional version atlas recording watermark change track and intelligent contract traceability report, the reference source tracking and unauthorized tampering risk are effectively solved.

[0065] According to the operation conversion algorithm of the cloud collaborative review module and the three-dimensional version atlas construction, the present application supports real-time collaborative editing and automatic conflict detection of multiple people; through the dynamic configuration of role permissions of the compliance control unit, including watermark visibility grading and editing area boundary locking, the safety and compliance of the collaboration process are ensured.

[0066] The present application also significantly improves the text segmentation robustness in complex scenes including handwritten annotations and table line interference through the use of a degradation model detector of an adaptive image enhancement module and conditional GAN enhancement, combined with the region text confidence coefficient Fbz quantization of a hybrid segmentation network and dynamic projection histogram cross-validation. BRIEF DESCRIPTION OF DRAWINGS

[0067] Figure 1 It is the structural framework schematic diagram of the whole system of the present application. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application is further described in detail below in combination with specific embodiments.

[0069] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present application should be understood as the usual meaning understood by those skilled in the art to which the present application belongs. The "first", "second" and similar words used in the present application do not represent any order, quantity or importance, but are only used to distinguish different components. "Include" or "contain" and similar words mean that the elements or objects before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connected" or "connected" and similar words are not limited to physical or mechanical connection, but can include electrical connection, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to represent the relative positional relationship, when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0070] Example one:

[0071] Please refer to Figure 1 A digital editing system based on intelligent review of teaching aid book content, characterized by: comprising a multi-modal copyright protection module, a cloud collaborative review module, an adaptive image enhancement module and an intelligent semantic verification module;

[0072] Multi-modal copyright protection module: for generating reversible watermark images, while automatically adjusting the watermark intensity according to the content type, and monitoring tampering behavior in real time and triggering traceability alarm;

[0073] Cloud collaborative review module: establish data channel with multi-modal copyright protection module, and realize multi-person real-time collaborative editing by operation conversion algorithm;

[0074] By constructing a three-dimensional version graph, the record content revision history and watermark change track of the reversible watermark image are compared, and when a modification conflict is identified, the reversible watermark image is marked as a to-be-processed image;

[0075] For the rest of the reversible watermark images without conflict, mark them as compliant images, and dynamically control their watermark visibility and editing range based on role permissions;

[0076] Adaptive image enhancement module: first receive the to-be-processed images marked by the cloud collaborative review module, and perform real-time enhancement on the image quality through the cascade processing of the degradation model detector and the conditional GAN;

[0077] Then use a hybrid segmentation network combined with a dynamic projection histogram to segment the text area of the to-be-processed image;

[0078] Intelligent semantic verification module, for receiving all reversible watermark images, and performing content compliance review, watermark authorization verification and semantic analysis, and automatically generating alternative solutions for illegal content.

[0079] In this embodiment, the multi-module cooperation realizes the intelligent management of the whole process of teaching aid content: the multi-modal copyright protection module realizes content traceability protection by generating reversible watermark images and automatically adjusting the watermark intensity, and ensures the safety of copyright by monitoring tampering behavior in real time; The cloud collaborative review module supports multi-person real-time collaborative editing by using the operation conversion algorithm, compares the content revision history and watermark change track through the three-dimensional version graph, accurately identifies the modification conflict and marks the to-be-processed image, and dynamically controls the watermark visibility and editing range of the compliant image based on role permissions; The adaptive image enhancement module improves the quality of the to-be-processed image through the cascade processing of the degradation model detector and the conditional GAN, and realizes accurate text area segmentation combined with the hybrid segmentation network and the dynamic projection histogram; The intelligent semantic verification module performs content compliance review, watermark authorization verification and semantic analysis on the enhanced image and the compliant image, and automatically generates alternative solutions for illegal content; The modules work together to form a complete closed loop from copyright protection, collaborative editing, image enhancement to semantic verification, which significantly improves the efficiency of teaching aid content review and the quality of publication.

[0080] Embodiment 2

[0081] The multi-modal copyright protection module includes an image generation unit;

[0082] The image generation unit converts the text, images and formulas in the supplementary book content into reversible watermark images with copyright identifiers through a dynamic digital watermarking engine, and extracts the peak signal-to-noise ratio aa, structural similarity index ab and normalized correlation coefficient ac from the comparison between the reversible watermark images and the pixel matrix of the original content in the supplementary book content.

[0083] The multi-modal copyright protection module further comprises a tampering monitoring unit;

[0084] The tampering monitoring unit is used for real-time calculation of watermark distortion degree Asy and evaluation, so as to monitor the tampering behavior and trigger the traceability alarm at the same time;

[0085] The watermark distortion degree Asy is obtained by non-dimensional processing and fitting of the peak signal-to-noise ratio aa, the structural similarity index ab and the normalized correlation coefficient ac, and the specific calculation formula is as follows:

[0086]

[0087] In the formula, Asy∈(0,1], and the value tends to 1, indicating smaller distortion;

[0088] Based on the supplementary image samples containing text, formulas and images, the watermark distortion degree Asy distribution under normal editing and malicious tampering is counted to determine the classification boundary; then after aligning with the industry standard, the first watermark distortion threshold A1 and the second watermark distortion threshold A2 are preset, and the first watermark distortion threshold A1 is greater than the second watermark distortion threshold A2;

[0089] The first watermark distortion threshold A1 and the second watermark distortion threshold A2 are compared and evaluated with the watermark distortion degree Asy to monitor the tampering behavior and trigger the traceability alarm; the specific evaluation content is as follows:

[0090] The watermark distortion degree Asy is greater than or equal to the first watermark distortion threshold A1, which is determined as a legal operation, the blockchain audit log is updated, and is marked as "compliant version";

[0091] The second watermark distortion threshold A2 is less than or equal to the watermark distortion degree Asy and less than the first watermark distortion threshold A1, indicating that there is non-malicious editing, including compression and format conversion; at the same time, a first alarm trigger manual review request is sent out;

[0092] The watermark distortion degree Asy is less than the second watermark distortion threshold A2, which is determined as malicious tampering; at this time, a second alarm is sent out, and the document editing permission is automatically frozen.

[0093] In this embodiment, the content of the teaching aid books is obtained by scanning paper teaching materials or importing electronic documents to obtain original materials, which are converted by the image generation unit using a dynamic digital watermarking engine: the text content is embedded with frequency domain watermarking based on discrete cosine transform (DCT), the image content is embedded with transparent watermarking using a wavelet transform (DWT) and singular value decomposition (SVD) combined algorithm, and the formula content is losslessly embedded by LaTeX vector watermarking technology to form reversible watermarking images with copyright identifiers; After the content classifier automatically identifies the material type, the watermark strength adjustment process applies the preset strength strategy: text uses 8dB-12dB PSNR reference value, image maintains 15dB-20dB visual fidelity, and formula class fixes 6bit-8bit quantization depth;

[0094] In the blockchain binding stage, the system calls the smart contract to generate a SHA-256 hash value containing the timestamp and author information, writes the hash value into the EXIF metadata of the watermark image and the transaction record into the blockchain distributed ledger, and reversely embeds the transaction ID returned by the blockchain into the image alpha channel, realizing two-way binding; In this process, the system calculates the peak signal-to-noise ratio aa, structural similarity index ab, and normalized correlation coefficient ac of the watermark image and the original material in real time to ensure that the watermark quality meets the standard before completing the final output;

[0095] In the calculation process of watermark distortion Asy, the peak signal-to-noise ratio aa and the structural similarity index ab are obtained by comparing the original content matrix and the watermark image matrix; the normalized correlation coefficient ac is obtained by extracting the matching degree of the watermark feature and the preset template; The parameter calculation result is written into the blockchain metadata as the baseline value for subsequent tamper monitoring;

[0096] In the process of presetting the watermark distortion threshold, 10000+ pages of teaching aid materials are collected to build a sample library, including 60% of text, 25% of formula, and 15% of image, each type of sample contains normal editing including format adjustment and content revision, and malicious tampering including content replacement and watermark destruction;

[0097] Secondly, in the process of determining the classification boundary, the watermark distortion Asy value is divided into three intervals by K-means clustering analysis: normal editing central interval [0, 0.35], transition interval (0.35, 0.65], and malicious tampering central interval (0.65, 1], and the midpoint 0.5 of the transition interval is taken as the initial classification boundary;

[0098] The process of aligning with industry standards is as follows: according to the requirements of ISO / IEC 29190 standard, the threshold is calibrated to A1=0.82 and A2=0.65 to ensure the balance between standard compliance and actual detection needs;

[0099] In the evaluation process of watermark distortion Asy, when there is non-malicious editing, the risk area is marked in the three-dimensional version map, and the risk operation of export and printing is limited; when it is determined to be malicious tampering, after the document editing permission is automatically frozen, it is rolled back to the latest version with Asy≥A1, and a traceability report is generated by calling the smart contract to locate the tamperer.

[0100] Embodiment 3

[0101] The cloud collaborative review module includes a collaborative editing unit;

[0102] The collaborative editing unit is configured to establish a data channel with the multi-modal copyright protection module through a secure API interface, so that the reversible watermark image and the editing instruction can be bidirectionally transmitted; wherein a collaborative editing algorithm based on operation adjustment and merging is adopted to sequentially reconstruct and resolve conflicts of editing operations of multiple users; and a coordination mechanism is adopted to adjust and synchronize the editing operations of different users based on operation time stamps and context dependency rules, and the adjustment manner is as follows:

[0103] The order of each operation is determined by comparing the operation time stamps, and the adjustment strategies of operation transfer, insertion redirection, or position offset are dynamically executed for the operations with position conflicts or logical conflicts in the content based on the context dependency rules.

[0104] The cloud collaborative review module further includes a version conflict unit;

[0105] The version conflict unit is configured to use a graph database or a chain structure to construct a three-dimensional version map of the reversible watermark image, taking three types of data, i.e., version time axis of the reversible watermark image, editing operation sequence, and watermark change track, as the basis, taking each editing behavior as a node, and taking editing order and dependency relationship as edges; wherein the content modification log and the watermark parameter change log during each image editing are automatically captured, and are associated and mapped in the three-dimensional version map, so as to track and record the content and watermark change track;

[0106] The content difference between versions is calculated to obtain the content change distance Ba by analyzing the specific editing operation type and position change in the content modification log; and the offset degree of the watermark track is calculated to obtain the watermark offset strength Bb by analyzing the embedding position, strength, or feature parameter in the watermark parameter change log;

[0107] Then, the content change distance Ba and the watermark offset strength Bb are extracted and dimensionless processed, and the version difference degree Bcy is calculated by the following formula to identify whether there is a modification conflict:

[0108]

[0109] A preset version difference threshold B is compared with the version difference degree Bcy, and specific evaluation contents are as follows:

[0110] If the version difference degree Bcy is less than or equal to the version difference threshold B, it is considered that the content change of the version and the watermark change are kept within a reasonable consistent range, no modification conflict is identified, and the version is automatically fused;

[0111] If the version difference degree Bcy is greater than the version difference threshold B, it is considered that the content change of the version and the watermark change have abnormal differences, including an unresolvable modification conflict, at this time, the reversible watermark image corresponding to the version is marked as a “to-be-processed image” and a conflict warning is triggered.

[0112] The cloud collaborative review module further includes a compliance control unit;

[0113] The compliance control unit is configured to mark images without conflicts as compliant; at the same time, through a preset role permission configuration rule, different user roles are assigned corresponding watermark visibility levels and editable area parameters, when a user accesses or edits a reversible watermark image, the permission configuration data is automatically called to perform visibility processing on the watermark information in the image, including masking, blurring or complete display, while limiting the operable image area boundary, only allowing the user to perform editing operations within the authorized range, and operation requests beyond the range will be prohibited.

[0114] In this embodiment, the cloud collaborative review module realizes efficient and secure teaching aid content management through multi-unit collaboration: the collaborative editing unit realizes bidirectional transmission of reversible watermark images and editing instructions through a secure API interface, adopts a collaborative editing algorithm based on operation adjustment and merging, and solves multi-user editing conflicts by combining operation time stamps and context dependency rules;

[0115] When the collaborative editing unit detects a position conflict caused by multiple users modifying the same paragraph at the same time, it automatically performs operation transfer, including moving the latter operation to an adjacent available position; when encountering a logical conflict caused by formula editing and text description mismatch, it triggers an insertion redirection, including adding a conflict marker for manual confirmation;

[0116] For layout conflicts such as misaligned graphic-text layout, position offset is implemented, including dynamically adjusting the spacing between elements according to priority, and through these three dynamic adjustment strategies, the integrity and consistency of multi-user editing are ensured;

[0117] The version conflict unit constructs a three-dimensional version graph, calculates the version difference degree Bcy through the content change distance Ba and the watermark offset strength Bb, and marks the image as to-be-processed and triggers a warning when the version difference degree Bcy is greater than the threshold B;

[0118] The compliance control unit dynamically controls the watermark visibility level and the editable area parameter based on the role permission configuration rule; each unit cooperates to ensure that the editing process is traceable through Ba and Bb quantitative change difference, version consistency is ensured through Bcy evaluation of version difference, operation range is ensured through role limitation, and a complete working closed loop from conflict detection to permission management is formed, effectively improving the safety and efficiency of collaborative editing of teaching aid content;

[0119] The role permission configuration rule assigns the permission level of the user according to the responsibility of the user in the collaborative review process; it contains two parts: one is the watermark visibility level, which is used to define the visibility of the embedded watermark in the image, including complete display, blur processing or complete shielding; the second is the editable area parameter, which is used to define the boundary of the image area that the user can operate. The permission configuration rule is uniformly formulated by the system administrator or the policy setting module according to the user roles of reviewers, editors and operators, and can be dynamically adjusted according to the project requirements;

[0120] When the user accesses or edits the image, the system will automatically match the corresponding permission configuration to realize the visual control of the watermark information and the area operation limitation, and ensure the safety of the image content, the order of the review process and the compliance of the editing operation;

[0121] The purpose of dynamically executing the adjustment strategy is to realize the automatic fusion and synchronization of operations while maintaining the editing intentions of each user;

[0122] When the content change distance Ba ranges from 0 to 1, 0 represents no change and 1 represents complete reconstruction;

[0123] When the watermark offset strength Bb ranges from 0 to 1, 0 represents no offset and 1 represents complete failure of the watermark;

[0124] The specific calculation formula of the content change distance Ba and the watermark offset strength Bb is as follows:

[0125]

[0126] In the formula, Dc represents the content modification difference, including the text edit distance and the image feature vector difference; L represents the total length of the original content for normalization; Dw represents the total offset value of the watermark parameters of the current version and the previous version; P represents the total quantity of the watermark parameters, including the position, strength and normalized dimension sum of the features.

[0127] Embodiment 4

[0128] The adaptive image enhancement module includes an image enhancement unit;

[0129] The image enhancement unit combines the version difference Bcy to prioritize image processing; the greater the version difference Bcy, the more significant the difference between versions, the higher the image processing priority, at which time the enhancement processing is prioritized;

[0130] The image enhancement unit internally integrates a degradation model detector for identifying the degradation type of the input image; the degradation type includes blur, noise and compression artifacts, and the degradation intensity parameter is extracted as an enhancement condition according to the detection result;

[0131] The adaptive image enhancement module uses the content variation distance Ba and the watermark offset strength Bb to model the degradation scene in multiple dimensions;

[0132] Subsequently, the content variation distance Ba and the watermark offset strength Bb are input as a condition vector into a conditional generative adversarial network to perform personalized enhancement operations on the image.

[0133] The adaptive image enhancement module also includes a text version segmentation unit;

[0134] Based on the enhanced image to be processed, first, a hybrid segmentation network is used to extract features and preliminarily divide the regions of the image to be processed, and a candidate region set is constructed; during this process, based on the region feature map extracted by the hybrid segmentation network, the texture complexity fa, edge continuity fb and morphological consistency fc of the candidate regions are quantitatively collected through a texture analysis operator, an edge detection operator and a morphological operator, respectively.

[0135] Subsequently, after the texture complexity fa, edge continuity fb and morphological consistency fc are extracted and dimensionless processed, the region text confidence coefficient Fbz is calculated and obtained, quantifying the credibility of each candidate region as a text region; the specific calculation formula is as follows:

[0136]

[0137] In the formula, 0≤region text confidence coefficient Fbz≤1;

[0138] When the region text confidence coefficient Fbz approaches 0, it indicates that the region text features meet the ideal features;

[0139] When the region text confidence coefficient Fbz approaches 1, it indicates that the region text features do not meet the ideal features;

[0140] Finally, combined with the dynamic projection histogram algorithm, the candidate regions whose text features meet the ideal features are located and segmented, and the structured text regions are extracted.

[0141] In this embodiment, the adaptive image enhancement module realizes intelligent optimization processing of the teaching aid image through the cooperation of the double units: the image enhancement unit identifies the quality degradation type of the image to be processed through the degradation model detector and drives the conditional GAN model for targeted enhancement, significantly improving the image definition, contrast and detail performance; the text version segmentation unit extracts features and divides regions based on the enhanced image using a hybrid segmentation network, calculates the region text confidence coefficient Fbz through the quantitative texture complexity fa, edge continuity fb and morphological consistency fc, accurately locates and segments the qualified text region through the dynamic projection histogram; this module undertakes the core functions of image quality repair and text structuring in the system, ensures the accuracy of text extraction through quantitative evaluation, provides high-quality image data basis for subsequent copyright review and semantic analysis, and forms a complete processing link from image enhancement to text segmentation;

[0142] By balancing and quantifying multiple key features including the dimensionless processed texture complexity fa, edge continuity fb and morphological consistency fc, and overall evaluation, the dimensional differences between different features can be effectively eliminated, ensuring the fairness and comparability of the evaluation results;

[0143] At the same time, the influence of abnormal features on Fbz is highlighted by using the root mean square form, which is particularly suitable for judging the text credibility of the region after multi-dimensional feature fusion, thereby improving the accuracy and robustness of text region extraction and subsequent recognition;

[0144]

[0145] In the formula, σ T represents the texture response standard deviation of the current region; σT ref represents the texture standard deviation of the ideal text region, and the reference value is 1;

[0146] E cont represents the number of continuous edge pixels, E all represents the number of all detected edge pixels, and the value range is 0 < fb < 1;

[0147] D shape represents the morphological difference degree of the current region and the template, D max represents the maximum possible morphological difference value, and the value range is 0 < fc < 1;

[0148] And the specific value range of the region text confidence coefficient Fbz is as follows:

[0149] 1. The determination standard of Fbz approaching to 0, i.e. high confidence:

[0150] Numerical range: Fbz ∈ [0, 0.3);

[0151] Technical features: the intersection over union IoU of the text region segmentation result and the labeled true value is greater than or equal to 0.85;

[0152] Data performance: OCR recognition accuracy is greater than or equal to 98% and character over-segmentation rate is less than 2%;

[0153] System response: directly enter the subsequent processing flow without manual review;

[0154] 2, the determination criterion of Fbz approaching to 1, that is, low confidence:

[0155] Numerical range: Fbz∈(0.7, 1];

[0156] Technical features: there is obvious fracture or adhesion in the text region, obvious fracture means the number of connected domains is greater than 1.5 times the expected value, and adhesion means the character width is greater than 2 times the average width;

[0157] Data performance: OCR recognition accuracy is less than 70% or pseudo-labeled confidence is less than 0.6;

[0158] System response: trigger the re-segmentation mechanism and record the abnormal event;

[0159] 3, transition interval processing, that is, Fbz∈[0.3, 0.7]:

[0160] Start the mixed verification process: simultaneously call the CNN classifier and the projection histogram for cross-validation;

[0161] Dynamic adjustment strategy: when Fbz∈(0.5, 0.7] for 3 consecutive frames, automatically increase the sensitivity parameter of the segmentation network by 15%;

[0162] The determination threshold can be dynamically calibrated according to the application scene through the ROC curve, and in critical fields such as medical text, it can be tightened to Fbz>0.5 to trigger an alarm;

[0163] In addition, when fa=fb=fc=1, it means Fbz=0, which means that the texture, edge and morphology of the region completely meet the ideal text characteristics, and at this time the text credibility is the highest.

[0164] Embodiment 5

[0165] The intelligent semantic verification module includes a semantic analysis unit;

[0166] The semantic analysis unit is used for text region segmentation of the input compliant image, also uses the mixed segmentation network combined with the dynamic projection histogram, extracts the structured text region; then constructs and based on the natural language processing model, performs content compliance review on the text region of the enhanced image to be processed and the compliant image, including sensitive word detection, knowledge error checking and semantic compliance analysis; finally, output the compliance determination result and mark the position and type of the illegal content.

[0167] The intelligent semantic verification module further comprises an authorization verification unit;

[0168] The authorization verification unit is configured to verify the image watermark authorization state, trigger a traceability alarm by comparing the watermark change track and the version graph record, identify unauthorized tampering behavior, and generate a replacement proposal, including recommending a compliant material replacement, a traceability correction suggestion, and triggering a manual review process.

[0169] In this embodiment, the intelligent semantic verification module realizes intelligent review of teaching aid content through the cooperation of the two units: the semantic analysis unit extracts the compliant image text area using a hybrid segmentation network and a dynamic projection histogram, performs sensitive word detection, knowledge error checking, and semantic compliance analysis through a natural language processing model, accurately marks the location and type of illegal content, and the authorization verification unit identifies unauthorized tampering behavior by comparing the watermark change track and the version graph record, and generates a replacement proposal (including a compliant material replacement and a traceability correction suggestion). This module assumes the dual functions of content compliance review and copyright protection in the system, ensures the knowledge accuracy and copyright legality of teaching aid content through the cooperative work of semantic analysis and authorization verification, provides intelligent decision support for publication review, and forms a complete verification closed loop from content identification to infringement disposal.

[0170] It should be noted that all calculation formulas in this application file use regression analysis in machine learning algorithms, including but not limited to, to deeply analyze the collected relevant parameters, identify their natural trends and mutual relationships. Professional software such as Python's Scikit-learn library or R language is used to automatically generate mathematical models that match the data. Then, the model performance is objectively evaluated through cross-validation and other methods, and combined with continuous feedback and optimization to ensure that the created formula truly reflects the internal laws of the data, thereby ensuring its effectiveness and accuracy, and ensuring that the calculation process conforms to the constraints of natural laws, rather than being based on artificially set rules.

[0171] The technical solutions of the present application can be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), FLASH, hard disk or optical disk, etc., including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present application.

[0172] The logic and / or steps represented in the flow diagrams and / or otherwise described herein, for example, can be considered as a sequence of executable instructions, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For purposes of this specification, a "computer-readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be for example but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium include an electrical connection, hard-wired connection, fiber-optic cable, portable storage media, RAM, ROM, EEPROM, tape, magnetic disk, optical disk, optical fiber, and / or micro-wave transmission medium, and the like.

[0173] It should be noted that the above-mentioned embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalent replaced without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

[0174] It should be noted that the above-mentioned embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalent replaced without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A digital editing system based on intelligent review of content of a study aid book, characterized by: The multi-modal copyright protection module, the cloud collaborative review module, the adaptive image enhancement module and the intelligent semantic verification module are included. The multi-modal copyright protection module is used for generating reversible watermark images, automatically adjusting watermark intensity according to content types, monitoring tampering behaviors in real time and triggering traceability alarms. The cloud collaborative review module establishes a data channel with the multi-modal copyright protection module and realizes real-time collaborative editing of multiple people by using an operation conversion algorithm. By constructing a three-dimensional version graph, the record content revision history and the watermark change track of the reversible watermark image are compared, and when a modification conflict is identified, the reversible watermark image is marked as a to-be-processed image. For the remaining reversible watermark images without conflicts, they are marked as compliant images, and the watermark visibility and editing range are dynamically controlled based on role permissions. The adaptive image enhancement module includes an image enhancement unit, which first receives the to-be-processed images marked by the cloud collaborative review module, and performs real-time enhancement on the image quality through the cascade processing of the degradation model detector and the conditional GAN. Specifically, the image enhancement unit combines the version difference Bcy to prioritize image processing; the greater the version difference Bcy, the more significant the difference between versions, and the higher the image processing priority, which is then enhanced. The degradation model detector integrated in the image enhancement unit is used to identify the degradation type of the input image, including blur, noise and compression artifacts, and the degradation intensity parameter is extracted as the enhancement condition according to the detection result. The adaptive image enhancement module uses the content variation distance Ba and the watermark offset strength Bb to model the degradation scene in multiple dimensions.

2. The digital editing system based on intelligent review of the contents of a study aid book according to claim 1, characterized in that: Then, the content variation distance Ba and the watermark offset strength Bb are input as a conditional vector into the conditional generative adversarial network to perform personalized enhancement operations on the image. The intelligent semantic verification module is used to receive all reversible watermark images and perform content compliance review, watermark authorization verification and semantic analysis, and automatically generate alternative solutions for illegal content.

3. The digital editing system based on intelligent review of the contents of a study aid book according to claim 2, characterized in that: The multi-modal copyright protection module includes an image generation unit. The image generation unit converts the text, images and formulas in the supplementary textbook content into reversible watermark images with copyright identifiers through a dynamic digital watermarking engine, and extracts the peak signal-to-noise ratio aa, structural similarity index ab and normalized correlation coefficient ac from the comparison of the pixel matrix of the reversible watermark image and the original content in the supplementary textbook content. The multi-modal copyright protection module also includes a tampering monitoring unit. ； The tampering monitoring unit is used to calculate and evaluate the watermark distortion degree Asy in real time to monitor tampering behaviors and trigger traceability alarms. The watermark distortion degree Asy is obtained by dimensionless processing and fitting of the peak signal-to-noise ratio aa, the structural similarity index ab and the normalized correlation coefficient ac, and the specific calculation formula is as follows: In the formula, Asy∈(0,1], and the value tends to 1, indicating smaller distortion. Based on the teaching image samples containing text, formulas and images, the watermark distortion degree Asy distribution under normal editing and malicious tampering is counted to determine the classification boundary; then after aligning with the industry standard, the first watermark distortion threshold A1 and the second watermark distortion threshold A2 are preset, and the first watermark distortion threshold A1 is greater than the second watermark distortion threshold A2; The first watermark distortion threshold A1 and the second watermark distortion threshold A2 are compared and evaluated with the watermark distortion degree Asy to monitor the tampering behavior and trigger the traceability alarm; the specific evaluation content is as follows: The watermark distortion degree Asy is greater than or equal to the first watermark distortion threshold A1, which is determined as legal operation, and the blockchain audit log is updated and marked as "compliant version"; The second watermark distortion threshold A2 is less than or equal to the watermark distortion degree Asy and less than the first watermark distortion threshold A1, indicating that there is non-malicious editing, including compression and format conversion; at the same time, the first alarm trigger manual review request is sent out; The watermark distortion degree Asy is less than the second watermark distortion threshold A2, which is determined as malicious tampering; at this time, the second alarm is sent out, and the document editing permission is automatically frozen.

4. The digital editing system based on intelligent review of the contents of a study aid book according to claim 3, characterized in that: The cloud collaborative review module includes a collaborative editing unit; The collaborative editing unit is used to establish a data channel between the multi-modal copyright protection module through a secure API interface, enabling bidirectional transmission of reversible watermark images and editing instructions; wherein a collaborative editing algorithm based on operation adjustment and merging is used to sequentially reconstruct and resolve conflicts of editing operations of multiple users; and a coordination mechanism adjusts and synchronizes editing operations of different users based on operation time stamps and context dependency rules; the specific adjustment method is as follows: By comparing the operation time stamps to determine the sequence of operations, and combining the context dependency rules, for the case of location conflict or logical conflict in operation content, the adjustment strategies of operation transfer, insertion redirection or position offset are dynamically executed.

5. The digital editing system based on intelligent review of the contents of a study aid book according to claim 4, characterized in that: The cloud collaborative review module also includes a version conflict unit; The version conflict unit is based on three types of data: version time axis of reversible watermark image, editing operation sequence and watermark change trajectory, and uses a graph database or chain structure to build a three-dimensional version graph of reversible watermark image, taking each editing behavior as a node and editing sequence and dependency relationship as an edge; wherein the content modification log and watermark parameter change log at each image editing are automatically captured and associated mapped in the three-dimensional version graph to track and record the content and watermark change trajectory; By analyzing the specific editing operation type and position change in the content modification log, the content difference between versions is calculated to obtain the content change distance Ba; by analyzing the embedding position, intensity or feature parameter in the watermark parameter change log, the offset degree of the watermark trajectory is calculated to obtain the watermark offset intensity Bb; Then, after extracting the content change distance Ba and the watermark offset intensity Bb and performing dimensionless processing, the version difference degree Bcy is calculated by the following formula to identify whether there is a modification conflict: ； The preset version difference threshold B is compared and evaluated with the version difference degree Bcy, and the specific evaluation content is as follows: If the version difference Bcy is less than or equal to the version difference threshold B, it is considered that the content change of the version is within a reasonable consistent range with the watermark change, no modification conflict is identified, and the version is automatically fused; If the version difference Bcy is greater than the version difference threshold B, it is considered that there is an abnormal difference between the content change of the version and the watermark change, including an unresolvable modification conflict, the reversible watermark image corresponding to the version is marked as a "to-be-processed image", and a conflict warning is triggered.

6. The digital editing system based on intelligent review of the contents of a study aid book according to claim 5, characterized in that: The cloud collaborative review module further comprises a compliance control unit; The compliance control unit is configured to mark the image without conflict as compliant, and assign corresponding watermark visibility levels and editable area parameters to different user roles according to preset role permission configuration rules. When a user accesses or edits the reversible watermark image, the permission configuration data is automatically called to process the watermark information in the image, including hiding, blurring or fully displaying, and the operable image area boundary is limited to allow the user to perform editing operations within the authorized range, and operation requests beyond the range will be prohibited.

7. The digital editing system based on intelligent review of the contents of a study aid book according to claim 6, characterized in that: The adaptive image enhancement module further comprises a text version segmentation unit; Based on the enhanced to-be-processed image, the feature of the to-be-processed image is extracted and the initial region is divided through the mixed segmentation network to construct a candidate region set. In the process, the texture complexity fa, the edge continuity fb and the morphological consistency fc of the candidate region are quantitatively collected through the texture analysis operator, the edge detection operator and the morphological operator based on the region feature map extracted by the mixed segmentation network. Subsequently, after the texture complexity fa, the edge continuity fb and the morphological consistency fc are extracted and dimensionless processed, the region text confidence coefficient Fbz is calculated and obtained to quantify the credibility of each candidate region as a text region. The specific calculation formula is as follows: ； In the formula, 0≤region text confidence coefficient Fbz≤1; When the region text confidence coefficient Fbz tends to 0, it indicates that the region text feature meets the ideal feature; When the region text confidence coefficient Fbz tends to 1, it indicates that the region text feature does not meet the ideal feature; Finally, the candidate regions with ideal region text features are positioned and segmented by combining the dynamic projection histogram algorithm, and the structured text region is extracted.

8. The digital editing system based on intelligent review of the contents of a study aid book according to claim 1, characterized in that: The intelligent semantic verification module comprises a semantic analysis unit; The semantic analysis unit is configured to segment the text region of the input compliant image, also adopts the mixed segmentation network combined with the dynamic projection histogram to extract the structured text region, then constructs and based on the natural language processing model, performs content compliance review on the text region of the enhanced to-be-processed image and the compliant image, including sensitive word detection, knowledge error checking and semantic compliance analysis, and finally outputs the compliance judgment result and marks the position and type of the illegal content.

9. The digital editing system based on intelligent review of the contents of a study aid book according to claim 8, characterized in that: The intelligent semantic verification module further comprises an authorization verification unit; The authorization checking unit is configured to check the image watermark authorization state, trigger a traceability alarm by comparing the watermark change track and the version graph record, identify unauthorized tampering behavior, and generate a replacement proposal, including recommending a compliant material replacement, labeling a traceability correction suggestion, and triggering a manual review process.

Images