AI poster rendering method based on quantum signature

By unifying data formatting and quantum password verification in AI poster rendering technology, the problems of style sheet redundancy and low rendering efficiency are solved, efficient and intelligent AI poster rendering and content management are achieved, preventing quantum attacks and providing more efficient , flexible rendering solution.

CN119048661APending Publication Date: 2024-11-29GANSU WANWEI INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411246977.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-06
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The existing AI poster rendering technology has problems such as redundancy and low rendering efficiency of style sheet files. The database storage and data retrieval process takes a long time, which affects the rendering effect. It cannot effectively improve the efficiency of asset intelligent management and is difficult to provide efficient and flexible multi-level Rendering solution.

Method used

By uniformly converting data in different formats into RTF format during the data acquisition stage, performing nine-grid segmentation and bit operations, combining HMI technology to generate AI posters, and using quantum cryptography for digital signature verification, preventing quantum attacks and ensuring the integrity and efficiency of the rendering process .

Benefits of technology

It improves the efficiency and quality of AI poster rendering, enhances the intelligence of data management, prevents quantum attacks during the rendering process, and achieves more efficient and flexible content production.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048661A_ABST
    Figure CN119048661A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of graphic processing, in particular to an AI poster rendering method based on quantum signature. In a data acquisition stage, creative text data including data sources and the like sent by an AI poster, an APP, a user terminal and a three-party interface are collected through anchor points deployed by a terminal; according to different formats of data, such as a picture jpeg, a video mp4, an audio and an advertisement, the data is uniformly converted into metadata of an RTF document, and associated data is used as meta-process data and is also stored in three data pools of an analog signal, an application program and text data separated from a data lake in the RTF document format. The problems that when a structured document renders a style, style sheet file redundancy and rendering efficiency reduction exist are solved, and the rendering effect is improved due to the fact that the time of a database storage data calling process is long.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of graphics processing technology, and in particular to a method for AI poster rendering based on quantum signature. Background Art

[0002] Artificial Intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. Artificial Intelligence, also known as intelligent machinery or machine intelligence, refers to the intelligence displayed by machines made by humans. Usually, artificial intelligence refers to the technology that presents human intelligence through ordinary computer programs. Through the progress of medicine, neuroscience, robotics and statistics, some predictions believe that countless human occupations will gradually be replaced by artificial intelligence.

[0003] With the popularization of AI and the improvement of AI creation quality, large models combined with poster works will be generated by AI assistance, which will become a future trend. However, AI rendering style has problems such as redundant style sheet files and reduced rendering efficiency. How to make AI poster rendering more efficient is a technical problem that needs to be solved urgently. Summary of the invention

[0004] Based on the shortcomings of the existing technology, this solution mainly solves the following two technical problems: First, this solution will solve the problems of redundant style sheet files and reduced rendering efficiency when rendering structured documents, as well as the long database storage data retrieval process affecting rendering effects. This will improve the efficiency of intelligent asset management.

[0005] Secondly, this solution will solve the complex structural problem of multi-level rendering of documents. Technologies such as digital images, audio, video and advertising in digital assets can provide more efficient and flexible content production methods for cultural creativity, making the produced content more diverse and vivid.

[0006] Step 1. In the data collection stage, creative text data including AI posters, APP, user terminals and data sources sent by three-party interfaces are collected through anchor points deployed on the terminal. According to different data formats such as picture jpeg, video mp4, audio mp3\wav and advertisement collected character strings or binary data, they are uniformly converted into metadata of RTF format documents and the associated data is stored as meta-process data in the three data pools of analog signals, applications and text data separated from the data lake in RTF format. Secondly, the image area in the AI ​​poster is divided into nine-square grids and a data set is generated, and an area identifier is generated at the same time. The data set with the area identifier is mapped to a bit array and converted into RTF format and stored in the miscellaneous data pool of the data lake. It is retrieved from the miscellaneous data pool when performing bit operations, and the creative text data and the scene to which it belongs are used as dynamic parameters through intelligent analysis. The large and small model training is combined with HMI technology to control the rendering to generate the AI ​​poster program to generate the final poster effect, thereby completing the secondary training of the large model; Step 2: Then, the bit array in the debris data pool uses quantum cryptography to replace the digital signature to generate a key for the data set of the area identifier and verify it. Finally, if the distance is greater than NormBound during the verification process, the verification fails; otherwise, the verification passes, indicating that the signature is valid. This completes the entire process of AI poster rendering and prevents rendering failures caused by possible quantum attacks during the process.

[0007] Beneficial effects of the invention: Technical innovation 1. In the data collection stage, creative text data including AI posters, APP, user terminals and data sources sent by three-party interfaces are collected through anchor points deployed by terminals, and the metadata of RTF format documents is uniformly converted according to different data formats such as pictures (jpeg), videos (mp4), audio (mp3\wav) and advertisements (collected strings or binary data), and the associated data is also stored in the RTF format document format as meta-process data in the three data pools (analog signals, applications and text data) separated from the data lake. Secondly, in the AI ​​poster, the image area is divided into nine squares and a data set is generated, and a region identifier is generated at the same time. The data set with the region identifier is mapped to a bit array and converted into RTF format and stored in the sundry data pool of the data lake. When performing bit operations, it is retrieved from the sundry data pool, and the creative text data and the scene to which it belongs are analyzed by combining the HMI intelligent analysis through secondary training of the size model, and then rendered to generate an AI poster. Technical innovation 2. Then, the bit array in the sundry data pool uses quantum cryptography to replace the digital signature to generate a key for the data set of the region identifier and verify it. Finally, during the verification process, if the distance is greater than NormBound, the verification fails; otherwise, the verification passes, indicating that the signature is valid. This completes the entire process of AI poster rendering and prevents rendering failures caused by quantum attacks that may occur during the process. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A schematic diagram of the data lake structure of the present invention; Figure 2 It is a schematic diagram of the rendering effect of the present invention. DETAILED DESCRIPTION

[0009] The present invention adopts the following technical solution: Step 1: In the data collection stage, the anchor points deployed by the terminal are used to collect creative text data including AI posters, APP, user terminals and data sources sent from the three-party interface. According to different data formats such as pictures (jpeg), videos (mp4), audio (mp3\wav) and advertisements (collected strings or binary data), they are uniformly converted into metadata of RTF format documents and the associated data is stored in the three data pools (analog signals, applications and text data) separated from the data lake as meta-process data in RTF format. Secondly, in the AI ​​poster, the image area is divided into nine squares and a data set is generated, and a region identifier is generated at the same time. The data set with the region identifier is mapped to a bit array and converted into RTF format and stored in the sundry data pool of the data lake. When performing bit operations, it is retrieved from the sundry data pool, and the creative text data and the scene to which it belongs are used as dynamic parameters through intelligent analysis. The large and small model training is combined with HMI (Human Machine Interface) technology to control the rendering and generate the AI ​​poster program to generate the final poster effect, thereby completing the secondary training of the large model.

[0010] S1. First, in the data collection stage, creative text data including AI posters, APP, user terminals and data sources sent by three-party interfaces are collected through anchor points deployed on the terminal. According to different data formats such as pictures (jpeg), videos (mp4), audio (mp3\wav) and advertisements (collected strings or binary data), they are uniformly converted into metadata of RTF format documents and the associated data is stored as meta-process data in RTF format documents in the three data pools (analog signals, applications and text data) separated from the data lake.

[0011] Analog signal data pool: Continuously similar collected data in RTF format documents, which has little analytical value due to its small volatility.

[0012] Application data pool: data related to core business in RTF format documents.

[0013] Text data pool: data in RTF format documents transferred in the form of text files.

[0014] Key words: Metadata: description of data (records, indexes, key values, and the relationship between different data attributes, etc.).

[0015] Meta-process data: more valuable for analysis than collected data, usually contains richer information (records, date, location, responsible person, recording equipment, other ancillary information) 1. Text data: uploaded text files (word, execl, etc.).

[0016] 2. Data collection characteristics: continuous and similar data collection 3. Application data: data transmitted between businesses, AI poster data, e-commerce advertising and promotional data.

[0017] 4. Miscellaneous data pool: It is used to store temporary data sets and release storage space after the results of the operations are completed.

[0018] S2. Secondly, in the AI ​​poster, the data set with regional identifiers is mapped into a bit array and converted into RTF format and stored in the miscellaneous data pool of the data lake. When performing bit operations, it is retrieved from the miscellaneous data pool, and the AI ​​poster is generated by intelligently analyzing the creative text and rendering.

[0019] Intelligently analyze creative text and render to generate AI posters: 1. Obtain the nine-grid dataset trained in the AI ​​poster, and search the database for multiple scene definition images corresponding to the nine-grid dataset in the database; 2. Preprocessing the multiple database images by cleaning, cropping, size standardization, format conversion, etc.; 3. Based on the subject features, a generative model is used to generate multiple flat creative pictures containing the subject.

[0020] 4. Each graphic creative picture is split into explicit data and implicit data, and the indicators are analyzed through clustering.

[0021] Image explicit data: including data parameters for preprocessing such as image cleaning, cropping, size standardization, and format conversion.

[0022] Image implicit data: mainly refers to the style of the image, the scope of application scenarios, etc. These data implicitly reveal the matching and preferences of the rendered image assets to the scene grid.

[0023] Put the explicit data and implicit data into the cluster analysis index for operation to obtain the training data results with more typical AI poster features T, see Figure 2 .

[0024] Specific description: RMSSTD (Root - Mean - Square Standard Deviation): The comprehensive standard deviation of all variables in a group. The smaller the RMSSTD, the higher the similarity of individual objects in the group (cluster), and the better the clustering effect. The calculation formula is as follows: Among them, Si represents the sum of the standard deviations of the ith variable in each group, and p is the number of variables.

[0025] R-Square: The size of the difference between groups after clustering, that is, the proportion of variance in the original data that the clustering results can explain. The larger the R-Square, the higher the difference between groups (clusters), and the better the clustering effect. The calculation formula is as follows: Wherein, W represents the degree of difference within each group after clustering, B represents the degree of difference between groups after clustering, T represents the total degree of difference of all data objects after clustering, and T=W+B.

[0026] According to the idea of ​​clustering, a good clustering result should be in the range of R-Square E [0,11], and the closer Sanare is to 1, the better. This shows the difference between different groups, that is, the larger B is, the smaller the number of clusters within the same group (within the group). The smaller the difference between objects, that is, W, is, this is exactly what cluster analysis hopes to achieve. The calculation formula is as follows: Where p represents p indicators (variables), n represents n group members, Represents the overall mean.

[0027] S3, intelligent analysis of creative text; First, the creative text and the corresponding scenes in the miscellaneous data pool are used as dynamic parameters of the AI ​​poster and put into the big model training to obtain a data set to obtain the current intelligent analysis creative text data range.

[0028] Secondly, the creative texts and scenes that are matched with the features in the vector database are orderly taken as secondary dynamic parameters for secondary training, thus avoiding the possibility that the AI ​​poster effect may not be optimal due to inaccurate results of the first training of the large model.

[0029] Then, the training results of the small model are combined with HMI (Human Machine Interface) technology to control the rendering and generate the AI ​​poster program to generate the final poster effect, thereby completing the secondary training of the large model. This implementation method greatly reduces the migration ability of large model technology to small model technology.

[0030] Example 1: GPT is a text generation model. The GPT model trained by the above method can fully realize the AI ​​poster content generation with high accuracy. However, in order to further improve the stability of intelligent generation of multi-labels, that is, to ensure that the generated content must be the label content accumulated by humans for a long time, all labels of the AI ​​poster content are hashed through the produced tags. For example, "Creative Poster" is hashed to ['Creativity', 'Poster'], and then the hashed creative poster, that is, poster and creativity, are put into the label hash list. In the process of generating multi-labels for the Star Model, the probability of each hash candidate of the Star Model that is not in the label hash list is penalized by a coefficient, that is, multiplied by a coefficient of 0.1, that is, the generated hash does not belong to the label hash list, and is a feature label in the small model vector database.

[0031] Step 2: Then, the bit array in the debris data pool uses quantum cryptography to replace the digital signature to generate a key for the data set of the area identifier and verify it. Finally, if the distance is greater than NormBound during the verification process, the verification fails; otherwise, the verification passes, indicating that the signature is valid. This completes the entire process of AI poster rendering and prevents rendering failures caused by possible quantum attacks during the process.

[0032] The digital signature algorithm NTRUSign-251 based on lattice theory has the following specific process: (1) Generate a key. Select two polynomials on the ring R The number of 1s in the coefficients is , and calculate the public key based on f and g . Solve the polynomials F and G so that they satisfy the equation (2) Signature Process 2-1 Perform a hash transformation on the message m and convert it into a polynomial , where the polynomial are all polynomials over the ring R.

[0033] 2-2 Calculate the polynomials A, B, a, b on the ring R so that they satisfy: And it is required that the coefficients of A and a are greater than -q / 2 and less than q / 2.

[0034] 2-3 Calculate the polynomial s as follows: s is the signature of message m calculated using public key h.

[0035] Verification process: Hash transformation of message m is performed and converted into a polynomial , calculated from the signature to be verified s and the public key h: , If the distance is greater than NormBound, the verification fails; otherwise, the verification passes, indicating that the signature is valid.

[0036] Glossary Quantum entanglement, also translated as quantum entanglement, was proposed by Einstein, Podolsky, and Rosen in 1935. Quantum entanglement describes a "mysterious" connection between two or more entangled particles. Even if they are far apart and there is no medium between them, the behavior of one particle will affect the state of the other particle. If one of the particles is operated and its state changes, the other particle will also change accordingly. This cannot be explained by classical mechanics.

[0037] Rich Text Format (RTF) is a cross-platform, structured document format based on control tags. Multi-level dynamic rendering of rich text documents refers to the dynamic rendering of rich text document data existing in the database in the display layer. In the existing technology for rendering structured documents, when the document structure changes, the document control engine will shift and change each document element node. The existing technical solution has the following disadvantages: First, based on the rendering method of structured documents, when rendering documents, the style depends on the externally introduced style sheet file. If the document requires a lot of customized text size, color, and special characters, it will cause redundancy of the style sheet file and reduce rendering efficiency.

[0038] Secondly, in the existing solutions, digital images, audio, video and other technologies in digital assets can only be stored simply, and cannot provide more efficient and flexible content production assistance for cultural creativity.

[0039] Data Lake: A Data Lake is a central information repository that stores large amounts of raw data in its native or original format, which can be structured, unstructured, or semi-structured.

[0040] Quantum entanglement, also translated as quantum entanglement, was proposed by Einstein, Podolsky, and Rosen in 1935. Quantum entanglement describes a "mysterious" connection between two or more entangled particles. Even if they are far apart and there is no medium between them, the behavior of one particle will affect the state of the other particle. If one of the particles is operated and its state changes, the other particle will also change accordingly. This cannot be explained by classical mechanics.

Claims

1. A method for AI poster rendering based on quantum signature, characterized in that These include: Step 1: In the data collection stage, the anchor points deployed on the terminal are used to collect creative text data including AI posters, APPs, user terminals and data sources sent from the third-party interface. According to different data formats such as picture jpeg, video mp4, audio mp3\wav and advertisement collected character strings or binary data, they are uniformly converted into metadata of RTF format documents and the associated data is stored as meta-process data in the three data pools of analog signals, applications and text data separated from the data lake in RTF format. Secondly, the image area in the AI ​​poster is divided into nine-square grids and a data set is generated, and a region identifier is generated at the same time. The data set with the region identifier is mapped into a bit array and converted into RTF format and stored in the sundry data pool of the data lake. When performing bit operations, it is retrieved from the sundry data pool, and the creative text data and the scene to which it belongs are used as dynamic parameters through intelligent analysis. The large and small model training is combined with HMI technology to control the rendering to generate the AI ​​poster program to generate the final poster effect, thereby completing the secondary training of the large model. Step 2. Then, quantum cryptography is used to replace the digital signature in the bit array in the debris data pool to generate a key for the data set of the area identifier and verify it; finally, if the distance is greater than NormBound during the verification process, the verification fails; otherwise, the verification passes, indicating that the signature is valid; thereby completing the entire process of AI poster rendering and preventing rendering failures caused by possible quantum attacks during the process.

2. According to the method of AI poster rendering based on quantum signature according to claim 1, it is characterized in that It includes the following: Step 1 is divided into: S1. First, in the data collection stage, the anchor points deployed on the terminal are used to collect creative text data including AI posters, APPs, user terminals and data sources sent from the third-party interface. According to different data formats such as picture jpeg, video mp4, audio mp3\wav and advertisement collection string or binary data, they are uniformly converted into metadata of RTF format documents and the associated data is stored as meta-process data in RTF format documents in the three data pools of analog signals, applications and text data separated from the data lake; Analog signal data pool: continuously similar collected data in RTF format documents, which has little analytical value due to its low volatility; Application data pool: data related to core business in RTF format documents; Text data pool: data in RTF format documents transferred in the form of text files; Metadata: a description of data records, indexes, key values, and the relationships between different data attributes; Meta-process data: It is more valuable for analysis than collected data and usually contains richer information; Text data: uploaded text files word, execl; Characteristics of collected data: continuously similar collected data; Application data: data transmitted between businesses, AI poster data, e-commerce advertising and promotion data; Miscellaneous data pool: storage of temporary computing data sets and release of storage space after computing results; S2. Secondly, in the AI ​​poster, the data set with the regional identifier is mapped into a bit array and then converted into RTF format and stored in the miscellaneous data pool of the data lake. When performing bit operations, it is retrieved from the miscellaneous data pool, and the AI ​​poster is generated by intelligently analyzing the creative text and rendering; Intelligently analyze creative text and render to generate AI posters:

1. Obtain the nine-grid dataset trained in the AI ​​poster, and search the database for multiple scene definition images corresponding to the nine-grid dataset in the database; 2. Preprocessing the multiple database images by cleaning, cropping, size standardization, format conversion, etc.; 3. Select the subject features and generate multiple flat creative pictures containing the subject through the generative model; 4. Each graphic creative picture is split into explicit data and implicit data, and the indicators are analyzed through clustering; Image explicit data: including data parameters for preprocessing such as image cleaning, cropping, size standardization, and format conversion; Image hidden data: mainly refers to the style and application scenario of the image; Put the explicit data and implicit data into the cluster analysis index for calculation to obtain the training data results with more typical AI poster features T; RMSSTD is used to express it, and the calculation formula is as follows: Among them, Si represents the sum of the standard deviations of the ith variable in each group, and p is the number of variables; R-Square: The size of the difference between groups after clustering, that is, the proportion of the variance of the original data explained by the clustering results. The larger the R-Square, the higher the difference between groups (clusters), and the better the clustering effect; The calculation formula is as follows: Where W represents the difference within each group after clustering, B represents the difference between groups after clustering, T represents the total difference of all data objects after clustering, and T=W+B; The calculation formula is as follows: Where p represents p indicators (variables), n represents n group members, represents the overall mean; S3, intelligent analysis of creative text First, the creative texts and scenes in the sundry data pool are used as dynamic parameters of AI posters and put into the big model training to obtain a data set, and the current intelligent analysis creative text data range is obtained; Secondly, the creative texts and scenes in the set and the vector database features that match the same features are used as secondary dynamic parameters for secondary training, which avoids the possibility that the AI ​​poster effect is not optimal due to inaccurate results of the first training of the large model; Then, the training results of the small model are combined with HMI technology to control the rendering and generate the AI ​​poster program to generate the final poster effect, thus completing the secondary training of the large model; The digital signature algorithm NTRUSign-251 based on lattice theory has the following specific process: Generate Keys Choose two polynomials over the ring R The number of 1s in the coefficients is , and calculate the public key based on f and g ; Solve the polynomials F and G so that they satisfy the equation Signature Process 2-1 Perform a hash transformation on the message m and convert it into a polynomial , where the polynomial are all polynomials over the ring R; 2-2 Calculate the polynomials A, B, a, b on the ring R so that they satisfy: And it is required that the coefficients of A and a are greater than -q / 2 and less than q / 2; 2-3 Calculate the polynomial s as follows: s is the signature of message m calculated using public key h; Verification Process Perform a hash transformation on the message m and convert it into a polynomial , calculated from the signature to be verified s and the public key h: , If the distance is greater than NormBound, the verification fails; otherwise, the verification passes, indicating that the signature is valid.