A certificate anti-counterfeiting method and certificate anti-counterfeiting system

Abstract

The present invention relates to the field of information security technology, and specifically to a certificate anti-counterfeiting method and a certificate anti-counterfeiting system, comprising the following steps: using Internet of Things technology and data crawling methods to collect multimodal information including certificate images, texts, and voices. In the present invention, the quality and integrity of the data are ensured through comprehensive and in-depth data collection and preprocessing. Through deep learning and natural language processing algorithms, image features and text features are extracted from the original data efficiently and accurately. These features are integrated by combining feature fusion and adversarial generative networks to further improve the accuracy and depth of detection. By applying blockchain and cloud computing technologies, a certificate anti-counterfeiting management system with high security and ease of use is established. This method has higher execution efficiency and accuracy in data processing, feature extraction, anti-counterfeiting detection, and system construction, and can better meet the needs of certificate anti-counterfeiting in real life.

Description

Technical Field

[0001] The present invention relates to the field of information security technology, and in particular to a certificate anti-counterfeiting method and a certificate anti-counterfeiting system. Background Art

[0002] Information security technology is a discipline that studies how to protect information systems and data resources from threats such as unauthorized access, use, disclosure, destruction, interference, and tampering. Certificate anti-counterfeiting methods, among other things, refer to a series of technical means and measures used to prevent the forgery, tampering, duplication, or misappropriation of certificates. Their purpose is to protect the authenticity and integrity of certificates and prevent the risks and losses associated with malicious use or misappropriation. To achieve this goal, certificate anti-counterfeiting methods typically utilize physical anti-counterfeiting technologies (such as specialized materials and printing techniques), digital anti-counterfeiting technologies (such as cryptography and data encryption), biometric identification technologies (such as fingerprint and facial recognition), and remote query and verification technologies (providing online verification services via the internet). By combining these means and technologies, certificate anti-counterfeiting methods enhance the security and credibility of certificates and reduce the risks of forgery and misappropriation.

[0003] Among the existing methods for anti-counterfeiting of certificates and licenses, most are based on fixed patterns and rules, such as static feature detection and regular matching of manual rules. This method may not be able to accurately detect some complex and clever counterfeiting behaviors, and it cannot identify fake certificates that rely on manual rules and are specially designed for anti-counterfeiting behaviors. Existing methods are often too dependent on a single technology or method in anti-counterfeiting technology, such as relying solely on image recognition or OCR technology, which to a certain extent limits their adaptability and accuracy when facing complex situations. Many existing methods lack effective data security and privacy protection measures, which makes certificate data at risk of being stolen and tampered with during transmission and storage. Moreover, the data storage methods of these methods are often relatively scattered, lacking unified and efficient data management, making certificate management more difficult and unable to meet the needs of the big data era. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a certificate anti-counterfeiting method and certificate anti-counterfeiting system.

[0005] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solution: a certificate anti-counterfeiting method, comprising the following steps:

[0006] S1: Use IoT technology and data crawling methods to collect multimodal information including license images, text, and voice. Generate a multimodal dataset through data cleaning algorithms including missing value filling and outlier processing.

[0007] S2: Based on the multimodal dataset, using a convolutional neural network and a natural language processing algorithm to perform feature extraction and generate a feature matrix, wherein the feature matrix includes image features and text features;

[0008] S3: using the feature matrix to perform primary anti-counterfeiting detection through a support vector machine classifier to generate a primary anti-counterfeiting label set, wherein the primary anti-counterfeiting label set includes primary anti-counterfeiting labels;

[0009] S4: fusing information of the primary anti-counterfeiting label set with the feature matrix, using a feature fusion algorithm specifically an ensemble learning algorithm to improve the accuracy of anti-counterfeiting detection, and generating a comprehensive anti-counterfeiting label set;

[0010] S5: Based on the comprehensive anti-counterfeiting label set, deep learning and generative adversarial networks are used to generate anti-counterfeiting images of certificates and licenses with high anti-counterfeiting features, and a deep anti-counterfeiting image library is integrated;

[0011] S6: Based on the deep anti-counterfeiting image library and comprehensive anti-counterfeiting label set, a comprehensive certificate and license anti-counterfeiting management system is constructed using blockchain technology and cloud computing.

[0012] As a further solution of the present invention, the Internet of Things technology and data crawling methods are used to collect multimodal information including certificate images, text, and voice. The steps of generating a multimodal dataset are specifically as follows:

[0013] S101: Use IoT sensors to collect image information of licenses and certificates, and use infrared sensors to capture infrared images to generate infrared image datasets.

[0014] S102: Based on the infrared image dataset, extract text information in the image using OCR technology to generate a preliminary text dataset;

[0015] S103: Using speech recognition technology, convert the collected speech information to generate a speech-to-text dataset;

[0016] S104: Integrate the preliminary text dataset and the speech text dataset, apply a data cleaning algorithm to fill missing values ​​and process outliers, and generate a multimodal dataset.

[0017] As a further solution of the present invention, based on the multimodal dataset, a convolutional neural network and a natural language processing algorithm are used to extract features and generate a feature matrix, wherein the feature matrix includes image features and text features. Specifically, the steps are as follows:

[0018] S201: extracting features from image information in the multimodal dataset using a convolutional neural network to generate an image feature vector;

[0019] S202: Based on the multimodal dataset, extract features from the text information using a natural language processing algorithm to generate a text feature vector;

[0020] S203: applying a speech feature extraction algorithm to the speech data in the multimodal dataset to generate a speech feature vector;

[0021] S204: Integrate the image feature vector, text feature vector, and speech feature vector to generate a feature matrix.

[0022] As a further solution of the present invention, the feature matrix is ​​used to perform primary anti-counterfeiting detection through a support vector machine classifier to generate a primary anti-counterfeiting label set, wherein the primary anti-counterfeiting label set includes the following steps:

[0023] S301: Based on the feature matrix, a support vector machine is used for training to generate an SVM model;

[0024] S302: Using the SVM model to perform classification prediction on the feature matrix to generate a primary anti-counterfeiting prediction result;

[0025] S303: calibrating the primary anti-counterfeiting prediction result, optimizing the classification effect using decision boundary adjustment technology, and generating an optimized primary anti-counterfeiting result;

[0026] S304: Generate a primary anti-counterfeiting label set for each data point label according to the optimized primary anti-counterfeiting result.

[0027] As a further solution of the present invention, the primary anti-counterfeiting label set is fused with the feature matrix, and a feature fusion algorithm specifically an ensemble learning algorithm is used to improve the accuracy of anti-counterfeiting detection. The steps of generating a comprehensive anti-counterfeiting label set are specifically as follows:

[0028] S401: Based on the primary anti-counterfeiting label set, use random forest to perform feature importance evaluation to obtain a feature importance vector;

[0029] S402: performing feature selection on the feature matrix using the feature importance vector to generate an optimized feature matrix;

[0030] S403: Based on the optimized feature matrix, using a boosting algorithm to perform ensemble learning to generate an ensemble learning model;

[0031] S404: Utilize the integrated learning model to perform anti-counterfeiting detection and generate a comprehensive anti-counterfeiting label set.

[0032] As a further solution of the present invention, based on the comprehensive anti-counterfeiting label set, deep learning and generative adversarial networks are used to generate anti-counterfeiting images of certificates and licenses with high anti-counterfeiting features. The steps of integrating the deep anti-counterfeiting image library are specifically as follows:

[0033] S501: Generate preliminary anti-counterfeiting images using the generator part of the generative adversarial network to obtain a preliminary anti-counterfeiting image set;

[0034] S502: Evaluate the preliminary anti-counterfeiting image set using a discriminator of an anti-generation network to generate an evaluation result;

[0035] S503: Based on the generation evaluation result, fine-tune and optimize the generator of the generative adversarial network to obtain an optimized generator;

[0036] S504: Based on the optimized generator, the optimized generator is used to regenerate the anti-counterfeiting image to generate a deep anti-counterfeiting image library.

[0037] As a further solution of the present invention, based on the deep anti-counterfeiting image library and comprehensive anti-counterfeiting label set, the steps of constructing a comprehensive certificate and license anti-counterfeiting management system using blockchain technology and cloud computing are as follows:

[0038] S601: Based on the deep anti-counterfeiting image library, a unique hash value is generated for each image using blockchain technology and recorded in the blockchain to obtain an image hash value chain;

[0039] S602: Using the image hash value chain in combination with a hash tree algorithm, a query or verification operation is completed to generate a hash tree data structure;

[0040] S603: Based on the hash tree data structure, design and deploy distributed nodes on the cloud computing platform to establish a distributed verification node system to process license verification requests in parallel;

[0041] S604: Integrate the distributed verification node system, and use API gateway and load balancing technology to ensure system stability and build a comprehensive certificate anti-counterfeiting management system.

[0042] A certificate anti-counterfeiting system is used to execute the above-mentioned certificate anti-counterfeiting method. The certificate anti-counterfeiting system is composed of a data collection module, a feature extraction module, a primary anti-counterfeiting module, a feature fusion module, an advanced anti-counterfeiting module, an anti-counterfeiting image generation module, and an anti-counterfeiting management module.

[0043] As a further solution of the present invention, the data collection module uses IoT sensors and data crawling algorithms to collect multimodal information, including images, text, and voice, to generate a multimodal raw data set;

[0044] The feature extraction module uses a convolutional neural network and a natural language processing algorithm to extract features based on the multimodal original data set and generate a multimodal feature matrix;

[0045] The primary anti-counterfeiting module uses a support vector machine to perform preliminary classification based on a multimodal feature matrix, generates primary anti-counterfeiting labels, and establishes a primary anti-counterfeiting label set;

[0046] The feature fusion module uses an ensemble learning algorithm to perform feature fusion based on the primary anti-counterfeiting label set and the multimodal feature matrix to generate a comprehensive anti-counterfeiting label set;

[0047] The advanced anti-counterfeiting module performs deep learning and adversarial generative network training and optimization based on a comprehensive anti-counterfeiting tag set to generate a deep anti-counterfeiting image library;

[0048] The anti-counterfeiting image generation module generates highly anti-counterfeiting certificate images using an image generation algorithm based on a deep anti-counterfeiting image library to generate a final anti-counterfeiting image set;

[0049] The anti-counterfeiting management module uses blockchain technology and cloud computing to perform comprehensive certificate anti-counterfeiting management based on the final anti-counterfeiting image set and the comprehensive anti-counterfeiting label set, and generates a comprehensive certificate anti-counterfeiting management system.

[0050] As a further solution of the present invention, the data collection module includes an image collection submodule, a text collection submodule, and a voice collection submodule.

[0051] The feature extraction module includes an image feature extraction submodule, a text feature extraction submodule, and a speech feature extraction submodule.

[0052] The primary anti-counterfeiting module includes an SVM training submodule, a prediction submodule, and a calibration submodule.

[0053] The feature fusion module includes feature importance evaluation submodule, feature selection submodule, and ensemble learning submodule.

[0054] The advanced anti-counterfeiting module includes a generator training submodule, a discriminator training submodule, and an optimization submodule.

[0055] The anti-counterfeiting image generation module includes an image generation submodule, an image evaluation submodule, and an image optimization submodule.

[0056] The anti-counterfeiting management module includes a blockchain record sub-module, a hash tree construction sub-module, a distributed node deployment sub-module, and a system integration sub-module.

[0057] Compared with the prior art, the advantages and positive effects of the present invention are:

[0058] In this invention, comprehensive and in-depth data collection and preprocessing ensure data quality and integrity. Deep learning and natural language processing algorithms are used to efficiently and accurately extract image and text features from raw data. These features are integrated using feature fusion and a generative adversarial network, further improving detection accuracy and depth. By applying blockchain and cloud computing technologies, a highly secure and user-friendly certificate and license anti-counterfeiting management system is established. This method boasts higher efficiency and accuracy in data processing, feature extraction, anti-counterfeiting detection, and system construction, better meeting real-life requirements for certificate and license anti-counterfeiting. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0060] Figure 2 This is a detailed flow chart of S1 of the present invention;

[0061] Figure 3 This is a detailed flow chart of S2 of the present invention;

[0062] Figure 4 This is a detailed flow chart of S3 of the present invention;

[0063] Figure 5 This is a detailed flow chart of S4 of the present invention;

[0064] Figure 6 This is a detailed flow chart of S5 of the present invention;

[0065] Figure 7 This is a detailed flow chart of S6 of the present invention;

[0066] Figure 8 is a system flow chart of the present invention;

[0067] Figure 9 This is a system block diagram of the present invention. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0069] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0070] Example 1

[0071] See also Figure 1 The present invention provides a technical solution: a certificate anti-counterfeiting method, comprising the following steps:

[0072] S1: Use IoT technology and data crawling methods to collect multimodal information including license images, text, and voice. Generate a multimodal dataset through data cleaning algorithms including missing value filling and outlier processing.

[0073] S2: Based on the multimodal dataset, convolutional neural networks and natural language processing algorithms are used to extract features and generate a feature matrix, which includes image features and text features;

[0074] S3: Using the feature matrix, a support vector machine classifier is used to perform primary anti-counterfeiting detection to generate a primary anti-counterfeiting label set, which includes primary anti-counterfeiting labels;

[0075] S4: Fusing the primary anti-counterfeiting label set with the feature matrix, using a feature fusion algorithm specifically an ensemble learning algorithm to improve the accuracy of anti-counterfeiting detection, and generating a comprehensive anti-counterfeiting label set;

[0076] S5: Based on a comprehensive anti-counterfeiting label set, deep learning and generative adversarial networks are used to generate highly anti-counterfeiting images of certificates and licenses, integrating a deep anti-counterfeiting image library.

[0077] S6: Based on a deep anti-counterfeiting image library and a comprehensive anti-counterfeiting label set, a comprehensive certificate and license anti-counterfeiting management system is built using blockchain technology and cloud computing.

[0078] This method leverages IoT technology and data crawling methods, combined with the collection and processing of multimodal information, to provide a comprehensive data foundation for certificates and licenses. This multimodal dataset and feature extraction algorithm accurately capture the image and text features of certificates and licenses, improving the accuracy of anti-counterfeiting detection. The generation of primary and comprehensive anti-counterfeiting label sets, based on feature matrices and fusion algorithms, further enhances the accuracy and reliability of anti-counterfeiting detection. By applying methods such as ensemble learning and fusing the results of multiple classifiers, it can better determine the authenticity of certificates and licenses. Deep learning and generative adversarial networks are used to generate certificate and license anti-counterfeiting images with highly robust anti-counterfeiting features, helping to improve the difficulty of anti-counterfeiting technology and reduce the likelihood of counterfeiting. The establishment of a deep anti-counterfeiting image library provides a rich image resource for subsequent anti-counterfeiting detection and management. The certificate and license anti-counterfeiting management system, based on blockchain technology and cloud computing, ensures the authenticity and security of certificates and improves the efficiency and security of certificate management. The decentralized nature of blockchain and the elasticity and scalability of cloud computing ensure data credibility and system stability.

[0079] See also Figure 2 , using IoT technology and data crawling methods to collect multimodal information including license images, text, and voice. Through data cleaning algorithms including missing value filling and outlier processing, the steps to generate a multimodal dataset are as follows:

[0080] S101: Use IoT sensors to collect image information of licenses and certificates, and use infrared sensors to capture infrared images to generate infrared image datasets.

[0081] S102: Based on the infrared image dataset, use OCR technology to extract text information in the image to generate a preliminary text dataset;

[0082] S103: Using speech recognition technology, convert the collected speech information to generate a speech-to-text dataset;

[0083] S104: Integrate the preliminary text dataset and the speech text dataset, use the data cleaning algorithm to fill in missing values ​​and process outliers, and generate a multimodal dataset.

[0084] First, this method uses IoT technology and data crawling methods to comprehensively collect multimodal information about certificates. By acquiring information in multiple forms, including images, text, and voice, the characteristics and content of the certificates are more comprehensively reflected, providing a more comprehensive data foundation. Second, a data cleaning algorithm is used to fill missing values ​​and process outliers in the multimodal dataset, improving the quality and integrity of the dataset. This eliminates gaps or unreasonable values ​​in the dataset, ensuring data accuracy and reliability, and providing reliable data support for subsequent anti-counterfeiting detection and management. Furthermore, this method integrates the characteristics of multimodal information and, through data cleaning and feature extraction, enhances the accuracy and reliability of anti-counterfeiting technology. The image, text, and voice features in the multimodal dataset are effectively extracted, enabling anti-counterfeiting detection to more comprehensively and comprehensively determine the authenticity of certificates.

[0085] See also Figure 3 Based on the multimodal dataset, convolutional neural network and natural language processing algorithm are used to extract features and generate feature matrix. The feature matrix includes image features and text features. The specific steps are as follows:

[0086] S201: Using a convolutional neural network to extract features from image information in a multimodal dataset and generate an image feature vector;

[0087] S202: Based on the multimodal dataset, a natural language processing algorithm is used to extract features from the text information and generate a text feature vector;

[0088] S203: applying a speech feature extraction algorithm to the speech data in the multimodal dataset to generate a speech feature vector;

[0089] S204: Integrate the image feature vector, the text feature vector, and the speech feature vector to generate a feature matrix.

[0090] First, feature extraction from image information in multimodal datasets using convolutional neural networks can capture visual features and patterns within the images. These image feature vectors provide rich image information, helping anti-counterfeiting detection systems accurately identify and compare the image features of documents and certificates, thereby improving the accuracy of anti-counterfeiting technology. Second, feature extraction based on the text information in multimodal datasets using natural language processing algorithms can extract important semantic and structural information from the text. These text feature vectors provide key features of the text content, helping anti-counterfeiting detection systems accurately analyze and identify the text information in documents and certificates. Furthermore, speech feature extraction from speech data in multimodal datasets can capture acoustic features of speech, such as spectrum, pitch, and speaking rate. These speech feature vectors provide important features of the speech information in documents and certificates, improving the ability of anti-counterfeiting technology to analyze and compare speech information. Combining image, text, and speech feature vectors to generate a feature matrix provides a comprehensive representation of multimodal information. This feature matrix helps anti-counterfeiting detection systems comprehensively consider multiple aspects of a document and certificate, improving the accuracy and reliability of authenticity assessments.

[0091] See also Figure 4 , using the feature matrix, the support vector machine classifier is used to perform primary anti-counterfeiting detection and generate a primary anti-counterfeiting label set. The primary anti-counterfeiting label set contains the following steps:

[0092] S301: Based on the feature matrix, use the support vector machine for training to generate an SVM model;

[0093] S302: Use the SVM model to perform classification prediction on the feature matrix to generate a primary anti-counterfeiting prediction result;

[0094] S303: Calibrate the primary anti-counterfeiting prediction result, optimize the classification effect using decision boundary adjustment technology, and generate an optimized primary anti-counterfeiting result;

[0095] S304: Generate a primary anti-counterfeiting label set for each data point label based on the optimized primary anti-counterfeiting result.

[0096] First, by using a feature matrix as input, this method can fully leverage features extracted from multimodal datasets, including images, text, and speech, to provide comprehensive information support. This helps improve the accuracy and robustness of primary anti-counterfeiting detection because it comprehensively considers multiple information sources. Second, using a support vector machine algorithm for classification training and prediction effectively learns and identifies patterns and regularities in certificates and licenses. Support vector machines have strong classification and generalization capabilities, enabling them to establish accurate decision boundaries in complex data spaces, thereby reliably determining the authenticity of certificates and licenses. Furthermore, calibration and optimization of the primary anti-counterfeiting prediction results can further improve classification accuracy and controllability. Calibration techniques can fine-tune the prediction results to better meet practical needs and anti-counterfeiting assessment requirements. Decision boundary adjustment techniques can adjust the classification boundaries to optimize classification performance based on specific scenarios and application requirements. Finally, the generated primary anti-counterfeiting label set provides a preliminary authenticity judgment for each data point. This label set can serve as the foundation for subsequent anti-counterfeiting work and provide an important reference for further development of advanced anti-counterfeiting technologies and management. This primary anti-counterfeiting label set can effectively screen out potentially counterfeit certificates and licenses, improving the efficiency and reliability of the anti-counterfeiting system.

[0097] See also Figure 5 , the primary anti-counterfeiting label set is fused with the feature matrix, and a feature fusion algorithm specifically for ensemble learning is used to improve the accuracy of anti-counterfeiting detection. The steps for generating a comprehensive anti-counterfeiting label set are as follows:

[0098] S401: Based on the primary anti-counterfeiting label set, use random forest to evaluate feature importance and obtain a feature importance vector;

[0099] S402: performing feature selection on the feature matrix using the feature importance vector to generate an optimized feature matrix;

[0100] S403: Based on the optimized feature matrix, a boosting algorithm is used to perform ensemble learning to generate an ensemble learning model;

[0101] S404: Utilize the integrated learning model to perform anti-counterfeiting detection and generate a comprehensive anti-counterfeiting label set.

[0102] First, by using random forests to assess feature importance, the importance of features in anti-counterfeiting detection can be determined. This helps eliminate features with little impact on anti-counterfeiting determination, reduces the interference of irrelevant information on anti-counterfeiting detection, and improves detection accuracy and efficiency. Second, by optimizing the feature matrix using feature importance vectors, the dimensionality of the feature space can be reduced while retaining the features most valuable for anti-counterfeiting determination. This reduces computational and storage overhead, accelerates the anti-counterfeiting detection process, and improves the accuracy and reliability of authenticity determination. Furthermore, by using a boosting algorithm for ensemble learning, the prediction results of multiple base classifiers can be combined to obtain more accurate and reliable classification results. Ensemble learning can effectively reduce overfitting and improve generalization, thereby enhancing the robustness and accuracy of anti-counterfeiting detection. Finally, the generated comprehensive anti-counterfeiting label set combines the results of the primary anti-counterfeiting label set and the ensemble learning results, providing a more comprehensive and reliable anti-counterfeiting determination. The comprehensive anti-counterfeiting label set accurately identifies the authenticity of each data point and provides a reliable basis for subsequent anti-counterfeiting decisions.

[0103] See also Figure 6 Based on the comprehensive anti-counterfeiting label set, deep learning and generative adversarial networks are used to generate certificate and license anti-counterfeiting images with high anti-counterfeiting features. The specific steps for integrating the deep anti-counterfeiting image library are as follows:

[0104] S501: Generate preliminary anti-counterfeiting images using the generator part of the generative adversarial network to obtain a preliminary anti-counterfeiting image set;

[0105] S502: Evaluate the preliminary anti-counterfeiting image set using the discriminator of the anti-generation network to generate an evaluation result;

[0106] S503: Based on the generated evaluation results, fine-tune and optimize the generator of the generative adversarial network to obtain an optimized generator;

[0107] S504: Based on the optimized generator, the anti-counterfeiting image is regenerated using the optimized generator to generate a deep anti-counterfeiting image library.

[0108] First, by using the generator portion of a generative adversarial network (GAN) to generate preliminary anti-counterfeiting images, the network can leverage the learned patterns and regularities to generate images with high security features. The GAN can learn the feature distribution of real images and generate realistic synthetic images, thereby creating recognizable and secure anti-counterfeiting images for certificates and licenses. Second, the GAN's discriminator is used to evaluate the preliminary set of anti-counterfeiting images, assessing the quality and security features of the generated images. The discriminator can identify differences between real and generated images, providing feedback on image quality and security features. This helps determine the rationality and realism of the images generated by the generator, providing guidance for subsequent optimization. Furthermore, based on the generated evaluation results, the GAN generator is fine-tuned and optimized to further improve the quality and security features of the generated images. Optimizing the generator based on the discriminator's feedback allows the generator to gradually approximate the feature distribution of real images, enhancing the realism and security performance of the generated images, thereby improving the quality of the deep anti-counterfeiting image library. Finally, based on the optimized generator, the optimized generator is used to regenerate security images to create a deep anti-counterfeiting image library. By using a fine-tuned and optimized generator to generate images, we can obtain a deep anti-counterfeiting image library with higher quality and stronger anti-counterfeiting features. This image library can be used for subsequent anti-counterfeiting verification and comparison, providing more accurate and reliable anti-counterfeiting identification.

[0109] See also Figure 7 Based on the deep anti-counterfeiting image library and comprehensive anti-counterfeiting label set, the steps for building a comprehensive certificate and license anti-counterfeiting management system using blockchain technology and cloud computing are as follows:

[0110] S601: Based on the deep anti-counterfeiting image library, a unique hash value is generated for each image using blockchain technology and recorded in the blockchain to obtain an image hash value chain;

[0111] S602: Using the image hash value chain in combination with the hash tree algorithm, a query or verification operation is completed to generate a hash tree data structure;

[0112] S603: Based on the hash tree data structure, design and deploy distributed nodes on the cloud computing platform, establish a distributed verification node system, and process license verification requests in parallel;

[0113] S604: Integrate the distributed verification node system, and use API gateway and load balancing technology to ensure system stability and build a comprehensive certificate and license anti-counterfeiting management system.

[0114] First, by using blockchain technology to generate a unique hash value for each image and recording it in the blockchain, the image's uniqueness and immutability are ensured. The distributed, decentralized, and immutable nature of blockchain makes the image hash chain highly secure and trustworthy, effectively preventing the forgery and tampering of certificate images. Second, combining the image hash chain with a hash tree algorithm can improve the efficiency of certificate query and verification. The hash tree data structure allows for rapid location and verification of the hash value of a specific image, eliminating the need to traverse the entire image database, significantly improving query and verification speed and efficiency. Furthermore, based on the hash tree data structure, designing and deploying distributed nodes on a cloud computing platform allows for the establishment of a highly available distributed verification node system. By parallelizing verification tasks across multiple nodes, the system's throughput and response speed are improved, enabling horizontal scalability to handle large-scale certificate verification requests. Finally, by integrating the distributed verification node system and leveraging API gateways and load balancing technologies to ensure system stability, a comprehensive certificate anti-counterfeiting management system can be constructed. This system can provide efficient and reliable certificate verification services, ensuring the authenticity and credibility of certificates and offering beneficial anti-counterfeiting support for specific application scenarios.

[0115] See also Figure 8 A certificate anti-counterfeiting system is used to execute the above-mentioned certificate anti-counterfeiting method. The certificate anti-counterfeiting system is composed of a data collection module, a feature extraction module, a primary anti-counterfeiting module, a feature fusion module, an advanced anti-counterfeiting module, an anti-counterfeiting image generation module, and an anti-counterfeiting management module.

[0116] The data collection module uses IoT sensors and data crawling algorithms to collect multimodal information, including images, text, and voice, to generate multimodal raw data sets;

[0117] The feature extraction module uses convolutional neural networks and natural language processing algorithms to extract features based on the multimodal original data set and generate a multimodal feature matrix;

[0118] The primary anti-counterfeiting module uses a support vector machine to perform preliminary classification based on the multimodal feature matrix, generate primary anti-counterfeiting labels, and establish a primary anti-counterfeiting label set;

[0119] The feature fusion module uses an ensemble learning algorithm to perform feature fusion based on the primary anti-counterfeiting label set and the multimodal feature matrix to generate a comprehensive anti-counterfeiting label set;

[0120] The advanced anti-counterfeiting module conducts deep learning and adversarial generative network training and optimization based on a comprehensive anti-counterfeiting label set to generate a deep anti-counterfeiting image library;

[0121] The anti-counterfeiting image generation module uses an image generation algorithm based on the deep anti-counterfeiting image library to generate highly anti-counterfeiting certificate images and generate the final anti-counterfeiting image set;

[0122] The anti-counterfeiting management module uses blockchain technology and cloud computing to conduct comprehensive certificate and license anti-counterfeiting management based on the final anti-counterfeiting image set and the comprehensive anti-counterfeiting label set, generating a comprehensive certificate and license anti-counterfeiting management system.

[0123] First, the data collection module enriches the system's data sources by collecting multimodal information, including images, text, and voice. This provides diverse information features and improves the accuracy and reliability of certificate authenticity determination. Second, the feature extraction module utilizes convolutional neural networks and natural language processing algorithms to extract features from the multimodal raw data, generating a multimodal feature matrix. This feature extraction method captures key features of the certificate data, providing more discriminative input features for subsequent anti-counterfeiting determination. The primary anti-counterfeiting module uses a support vector machine to perform preliminary classification, generate primary anti-counterfeiting labels, and establish a primary anti-counterfeiting label set. The primary anti-counterfeiting module's judgment enables a basic assessment of the certificate, providing a reference for subsequent anti-counterfeiting decisions. The feature fusion module, based on the primary anti-counterfeiting results and multimodal features, employs an ensemble learning algorithm to fuse features and generate a comprehensive anti-counterfeiting label set. This feature fusion improves the accuracy and robustness of anti-counterfeiting determinations, resulting in more reliable comprehensive anti-counterfeiting labels. The advanced anti-counterfeiting module utilizes deep learning and a generative adversarial network for training and optimization, generating a deep anti-counterfeiting image library with highly reliable anti-counterfeiting features. This image library generates realistic anti-counterfeiting images, enhancing their authenticity and credibility. The anti-counterfeiting image generation module, based on the deep anti-counterfeiting image library, uses image generation algorithms to generate the final set of anti-counterfeiting images. These anti-counterfeiting images possess highly anti-counterfeiting characteristics and can be used for certificate verification and comparison, providing reliable anti-counterfeiting identification and determination. Through the anti-counterfeiting management module, the system leverages blockchain technology and cloud computing to build a comprehensive certificate anti-counterfeiting management system, providing secure and reliable image identification, verification, and management capabilities, ensuring the reliability and stability of the anti-counterfeiting system.

[0124] See also Figure 9 ,The data collection module includes image collection submodule, text collection submodule, and voice collection submodule.

[0125] The feature extraction module includes an image feature extraction submodule, a text feature extraction submodule, and a speech feature extraction submodule.

[0126] The primary anti-counterfeiting module includes an SVM training submodule, a prediction submodule, and a calibration submodule.

[0127] The feature fusion module includes feature importance evaluation submodule, feature selection submodule, and ensemble learning submodule.

[0128] The advanced anti-counterfeiting module includes a generator training submodule, a discriminator training submodule, and an optimization submodule.

[0129] The anti-counterfeiting image generation module includes an image generation submodule, an image evaluation submodule, and an image optimization submodule.

[0130] The anti-counterfeiting management module includes a blockchain record sub-module, a hash tree construction sub-module, a distributed node deployment sub-module, and a system integration sub-module.

[0131] First, the data collection module, which includes image, text, and voice collection submodules, can collect multimodal information related to certificates from multiple data sources. This diverse data collection provides more comprehensive and multi-faceted certificate information, enhancing the anti-counterfeiting system's ability to determine certificate authenticity.

[0132] Secondly, the feature extraction module uses image, text, and speech feature extraction submodules to extract key features from different data modalities. This feature extraction can capture the important characteristics of the certificate data in different modalities, providing more discriminative input features for the subsequent anti-counterfeiting module, thereby improving the accuracy and robustness of the anti-counterfeiting system.

[0133] The primary anti-counterfeiting module provides preliminary classification and determination capabilities through the SVM training submodule, prediction submodule, and calibration submodule. This module can quickly and preliminarily determine the authenticity of documents and licenses, and the calibration submodule optimizes and calibrates the classification model, improving the accuracy of primary anti-counterfeiting.

[0134] The feature fusion module utilizes a feature importance assessment submodule, a feature selection submodule, and an ensemble learning submodule to fuse and integrate information from different feature modalities. By comprehensively considering the importance of each feature modality and selecting discriminative features, the anti-counterfeiting system's discriminative ability and robustness are improved.

[0135] The Advanced Anti-Counterfeiting Module trains and optimizes deep learning and generative adversarial networks through generator training, discriminator training, and optimization submodules. Generator and discriminator training generates images with highly anti-counterfeiting features, while the optimization submodule improves the quality and realism of generated images.

[0136] The anti-counterfeiting image generation module, which includes image generation, image evaluation, and image optimization submodules, generates anti-counterfeiting images based on the deep anti-counterfeiting image library generated by the advanced anti-counterfeiting module. By evaluating and optimizing the generated images, the quality and anti-counterfeiting characteristics of the anti-counterfeiting images can be improved, enhancing the accuracy and credibility of anti-counterfeiting verification.

[0137] Finally, the anti-counterfeiting management module includes a blockchain recording submodule, a hash tree construction submodule, a distributed node deployment submodule, and a system integration submodule. These submodules leverage blockchain technology and cloud computing to build a comprehensive certificate and license anti-counterfeiting management system, ensuring the security and credibility of certificate information and providing distributed verification and management capabilities.

[0138] Working principle:

[0139] IoT sensors and data scraping algorithms are used to collect multimodal information from licenses, including images, text, and voice. This involves using infrared sensors to capture infrared images, using optical character recognition (OCR) to extract text from images, and using speech recognition to convert voice information. Data cleaning algorithms are then used to fill missing values ​​and address outliers in the collected data, generating a multimodal raw dataset.

[0140] Based on the multimodal raw dataset, we use convolutional neural networks and natural language processing algorithms for feature extraction. We use convolutional neural networks to extract image feature vectors for images, natural language processing algorithms to extract text feature vectors for text, and speech feature extraction algorithms to extract speech feature vectors for speech. The resulting image, text, and speech feature vectors are integrated into a feature matrix.

[0141] A support vector machine classifier is used to perform preliminary anti-counterfeiting checks on the feature matrix. First, a support vector machine is trained to generate an SVM model. The trained SVM model is then used to perform classification predictions on the feature matrix, generating preliminary anti-counterfeiting prediction results. These preliminary anti-counterfeiting prediction results are calibrated, and decision boundary adjustment techniques are used to optimize the classification results, generating optimized preliminary anti-counterfeiting results. A preliminary anti-counterfeiting label set is generated for each data point label.

[0142] The primary anti-counterfeiting label set is fused with the feature matrix, and an ensemble learning feature fusion algorithm is used to improve anti-counterfeiting detection accuracy. Feature importance is assessed using a random forest algorithm to obtain a feature importance vector. The feature importance vector is used to select features from the feature matrix to generate an optimized feature matrix. Based on this optimized feature matrix, an ensemble learning algorithm is used for anti-counterfeiting detection to generate a comprehensive anti-counterfeiting label set.

[0143] Based on a comprehensive set of anti-counterfeiting labels, deep learning and a generative adversarial network are used to generate highly anti-counterfeiting images for certificates and licenses. A preliminary set of anti-counterfeiting images is generated using the generator portion of the generative adversarial network. This set of images is then evaluated using the discriminator to generate evaluation results. Based on the evaluation results, the generator is fine-tuned and optimized to produce an optimized generator. This optimized generator is then used to regenerate anti-counterfeiting images, creating a deep anti-counterfeiting image library.

[0144] Based on a deep anti-counterfeiting image library and a comprehensive set of anti-counterfeiting labels, a comprehensive certificate and license anti-counterfeiting management system is constructed using blockchain technology and cloud computing. Blockchain technology generates a unique hash value for each image and records it in the blockchain, forming an image hash value chain. A hash tree algorithm is used to perform query or verification operations, generating a hash tree data structure. Distributed nodes are designed and deployed on a cloud computing platform, establishing a distributed verification node system to process certificate and license verification requests in parallel. This distributed verification node system is integrated through an API gateway and load balancing technology to form a comprehensive certificate and license anti-counterfeiting management system.

[0145] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A certificate anti-counterfeiting method, characterized in that: The following steps are involved: Use IoT technology and data crawling methods to collect multimodal information including license images, text, and voice, and generate multimodal data sets through data cleaning algorithms including missing value filling and outlier processing; Based on the multimodal data set, a convolutional neural network and a natural language processing algorithm are used to extract features and generate a feature matrix, wherein the feature matrix includes image features and text features; Using the feature matrix, a primary anti-counterfeiting detection is performed by a support vector machine classifier to generate a primary anti-counterfeiting label set, wherein the primary anti-counterfeiting label set includes primary anti-counterfeiting labels; The primary anti-counterfeiting label set is subjected to information fusion with the feature matrix, and a feature fusion algorithm specifically an ensemble learning algorithm is used to improve the accuracy of anti-counterfeiting detection, thereby generating a comprehensive anti-counterfeiting label set; The steps of fusing the primary anti-counterfeiting label set with the feature matrix, using a feature fusion algorithm specifically an ensemble learning algorithm to improve the accuracy of anti-counterfeiting detection, and generating a comprehensive anti-counterfeiting label set are specifically as follows: Based on the primary anti-counterfeiting label set, using random forest to evaluate feature importance and obtain a feature importance vector; Using the feature importance vector to perform feature selection on the feature matrix to generate an optimized feature matrix; Based on the optimized feature matrix, using a boosting algorithm to perform ensemble learning to generate an ensemble learning model; Use ensemble learning models to conduct anti-counterfeiting detection and generate a comprehensive anti-counterfeiting label set; Based on the comprehensive anti-counterfeiting label set, deep learning and generative adversarial networks are used to generate certificate anti-counterfeiting images with high anti-counterfeiting features, and a deep anti-counterfeiting image library is integrated; Based on the deep anti-counterfeiting image library and comprehensive anti-counterfeiting label set, a comprehensive certificate anti-counterfeiting management system is constructed using blockchain technology and cloud computing.

2. The certificate anti-counterfeiting method according to claim 1, characterized in that: Use IoT technology and data crawling methods to collect multimodal information including license images, text, and voice. Use data cleaning algorithms including missing value filling and outlier processing to generate a multimodal data set. The specific steps are as follows: Use IoT sensors to collect image information of certificates and use infrared sensors to collect infrared images to generate infrared image data sets; Based on the infrared image dataset, extract text information in the image using OCR technology to generate a preliminary text dataset; Use speech recognition technology to convert the collected speech information and generate a speech-to-text data set; The preliminary text dataset and the speech text dataset are integrated, and a data cleaning algorithm is used to fill in missing values ​​and process outliers to generate a multimodal dataset.

3. The certificate anti-counterfeiting method according to claim 1, characterized in that: Based on the multimodal data set, a convolutional neural network and a natural language processing algorithm are used to extract features and generate a feature matrix, wherein the feature matrix includes image features and text features. Specifically, the steps are as follows: Using a convolutional neural network to extract features from the image information in the multimodal data set to generate an image feature vector; Based on the multimodal data set, a natural language processing algorithm is used to extract features from text information to generate a text feature vector; Using a speech feature extraction algorithm on the speech data in the multimodal data set to generate a speech feature vector; The image feature vector, text feature vector and speech feature vector are integrated to generate a feature matrix.

4. The certificate anti-counterfeiting method according to claim 1, characterized in that: The steps of using the feature matrix to perform primary anti-counterfeiting detection through a support vector machine classifier to generate a primary anti-counterfeiting label set, wherein the primary anti-counterfeiting label set includes the primary anti-counterfeiting labels are as follows: Based on the feature matrix, a support vector machine is used for training to generate an SVM model; Use the SVM model to classify and predict the feature matrix to generate a primary anti-counterfeiting prediction result; Calibrate the primary anti-counterfeiting prediction result, optimize the classification effect by using decision boundary adjustment technology, and generate an optimized primary anti-counterfeiting result; According to the optimized primary anti-counterfeiting result, a primary anti-counterfeiting label set is generated for each data point label.

5. The certificate anti-counterfeiting method according to claim 1, characterized in that: Based on the comprehensive anti-counterfeiting label set, deep learning and adversarial generative networks are used to generate certificate anti-counterfeiting images with high anti-counterfeiting features. The steps of integrating the deep anti-counterfeiting image library are as follows: The generator part of the adversarial generative network is used to generate preliminary anti-counterfeiting images to obtain a preliminary anti-counterfeiting image set; Using a discriminator of a generative adversarial network to evaluate the preliminary anti-counterfeiting image set and generate an evaluation result; Based on the generation evaluation result, fine-tuning and optimizing the generator of the generative adversarial network to obtain an optimized generator; Based on the optimized generator, the anti-counterfeiting image is regenerated using the optimized generator to generate a deep anti-counterfeiting image library.

6. The certificate anti-counterfeiting method according to claim 1, characterized in that: Based on the deep anti-counterfeiting image library and comprehensive anti-counterfeiting label set, the steps of building a comprehensive certificate anti-counterfeiting management system using blockchain technology and cloud computing are as follows: Based on the deep anti-counterfeiting image library, blockchain technology is used to generate a unique hash value for each image and record it in the blockchain to obtain an image hash value chain; Using the image hash value chain in combination with a hash tree algorithm, a query or verification operation is completed to generate a hash tree data structure; Based on the hash tree data structure, design and deploy distributed nodes on the cloud computing platform, establish a distributed verification node system, and process certificate verification requests in parallel; Integrate the distributed verification node system, use API gateway and load balancing technology to ensure system stability, and build a comprehensive certificate anti-counterfeiting management system.

7. A certificate anti-counterfeiting system, characterized in that: The certificate anti-counterfeiting system is used to execute the certificate anti-counterfeiting method according to any one of claims 1 to 6, and the certificate anti-counterfeiting system is composed of a data collection module, a feature extraction module, a primary anti-counterfeiting module, a feature fusion module, an advanced anti-counterfeiting module, an anti-counterfeiting image generation module, and an anti-counterfeiting management module; The data collection module uses IoT sensors and data crawling algorithms to collect multimodal information, including images, text, and voice, to generate a multimodal raw data set; The feature extraction module uses a convolutional neural network and a natural language processing algorithm to extract features based on the multimodal original data set to generate a multimodal feature matrix; The primary anti-counterfeiting module uses a support vector machine to perform preliminary classification based on a multimodal feature matrix, generates primary anti-counterfeiting labels, and establishes a primary anti-counterfeiting label set; The feature fusion module uses an integrated learning algorithm to perform feature fusion based on the primary anti-counterfeiting label set and the multimodal feature matrix to generate a comprehensive anti-counterfeiting label set; The advanced anti-counterfeiting module performs deep learning and adversarial generative network training and optimization based on the comprehensive anti-counterfeiting label set to generate a deep anti-counterfeiting image library; The anti-counterfeiting image generation module generates highly anti-counterfeiting certificate images using an image generation algorithm based on a deep anti-counterfeiting image library to generate a final anti-counterfeiting image set; The anti-counterfeiting management module uses blockchain technology and cloud computing to perform comprehensive certificate anti-counterfeiting management based on the final anti-counterfeiting image set and the comprehensive anti-counterfeiting label set, and generates a comprehensive certificate anti-counterfeiting management system.

8. The certificate anti-counterfeiting system according to claim 7, characterized in that: The data collection module includes an image collection submodule, a text collection submodule, and a voice collection submodule; The feature extraction module includes an image feature extraction submodule, a text feature extraction submodule, and a speech feature extraction submodule; The primary anti-counterfeiting module includes SVM training submodule, prediction submodule, and calibration submodule; The feature fusion module includes a feature importance assessment submodule, a feature selection submodule, and an integrated learning submodule; The advanced anti-counterfeiting module includes a generator training submodule, a discriminator training submodule, and an optimization submodule; The anti-counterfeiting image generation module includes an image generation submodule, an image evaluation submodule, and an image optimization submodule; The anti-counterfeiting management module includes blockchain record sub-module, hash tree construction sub-module, distributed node deployment sub-module, and system integration sub-module.

Images