Artwork identification system based on Internet of Things technology

Abstract

The invention relates to the technical field of the Internet of Things, and discloses an artwork identification system based on the Internet of Things technology. The system comprises a data acquisition module, an image processing module, a data analysis module and an Internet of Things transmission module, wherein the data acquisition module acquires multi-dimensional data of artworks through a multi-modal sensor and a high-resolution imaging device, and the multi-modal sensor comprises a temperature and humidity sensor, an illumination sensor and a micro-vibration sensor; and the high-resolution imaging equipment comprises a multispectral camera and a laser scanner and is used for acquiring image data and three-dimensional surface structure data of the artwork. A non-linear partial differential equation model is adopted in an image noise removal unit, and a regularization detail enhancement item is introduced, so that the aim of enhancing image high-frequency features while noise suppression is realized, and the effect of improving the artwork image detail recovery capability is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of the Internet of Things, and particularly to an identification system for artworks based on Internet of Things technology. Background Art

[0002] With the growth of the demand for the protection, identification and trading of artworks, the identification system for artworks based on Internet of Things technology has become a research hotspot. Traditional artwork identification systems usually adopt a single image processing method to suppress noise and extract features from artwork images. However, in the face of complex artwork texture features, diverse art styles, and potential surface damage problems, traditional systems cannot balance noise removal, detail enhancement, and edge feature extraction, resulting in loss of detail information and insufficient identification accuracy.

[0003] Meanwhile, in terms of artwork damage detection and style classification, the existing technologies mainly rely on manual analysis by experienced experts, lacking automated and intelligent means. Traditional methods are inefficient, costly, and limited by expert experience, and cannot accurately identify complex damages and subtle style differences. In addition, existing damage detection systems can mostly provide classification results of the damage degree, and cannot intuitively display the specific location and scope of the damage, bringing difficulties to the repair and protection of artworks.

[0004] To solve the above problems, in recent years, artificial intelligence technology has been gradually introduced into the field of artwork identification. For example, the application of convolutional neural networks in image classification has made certain progress, but there are still limitations in simultaneously capturing global features and local details. On the other hand, generative adversarial networks are used for image generation, but their application in artwork damage detection is not yet mature, and their potential in visualizing damaged areas has not been fully exploited.

[0005] In addition, the use of multi-modal data fusion technology has gradually received attention. For example, by combining ultraviolet and infrared light images, details of artworks that cannot be identified in traditional visible light images are discovered. However, the current systems still lack targeted optimization in multi-modal image fusion and three-dimensional surface feature capture.

[0006] In summary, the existing technologies have the following technical problems in the image processing, style classification, and damage detection of artworks:

[0007] The noise suppression method is likely to cause the loss of high-frequency details of the image, affecting the presentation of the authenticity of artworks;

[0008] Edge feature extraction cannot comprehensively capture complex textures and multi-directional features;

[0009] The style classification technology fails to simultaneously focus on global features and local details, resulting in a reduction in classification accuracy;

[0010] The lack of intuitive visualization in damage detection results makes it impossible to locate and quantify the damage range.

[0011] Therefore, those skilled in the art provide an identification system for artworks based on Internet of Things technology to solve the above-mentioned problems. Summary of the Invention

[0012] Aiming at the deficiencies of the prior art, the present invention provides an identification system for artworks based on Internet of Things technology to solve the problems raised in the above background technology.

[0013] To achieve the above objectives, the present invention is realized through the following technical solutions: An identification system for artworks based on Internet of Things technology, the system includes a data acquisition module, an image processing module, a data analysis module, and an Internet of Things transmission module:

[0014] The data acquisition module collects multi-dimensional data of artworks through multi-modal sensors and high-resolution imaging devices. The multi-modal sensors include temperature and humidity sensors, light sensors, and micro-vibration sensors for collecting environmental parameters. The high-resolution imaging devices include multi-spectral cameras and laser scanners for collecting image data and three-dimensional surface structure data of artworks. The collected data is preliminarily processed by the local processing unit in the data acquisition module and then sent to the image processing module;

[0015] After receiving the environmental parameters and artwork image data transmitted by the data acquisition module, the image processing module analyzes the image by constructing the relationship between spatial pixel values, degrades the noise part of the image using the change of pixel gradient, generates a pixel value distribution map based on spatial coordinates, obtains optimized image data and relevant environmental information, and transmits it to the data analysis module;

[0016] After receiving the structured image file and environmental data file transmitted by the image processing module, the data analysis module extracts features from the image data according to the predefined artwork classification and feature tags, extracts the edge and texture features of the image layer by layer, combines the environmental parameters to complete the artwork style classification and damage detection, generates an analysis result data file containing the classification result and damage features, and returns the file together with the structured image file to the Internet of Things transmission module;

[0017] After receiving the analysis result data file and structured image file from the data analysis module, the Internet of Things transmission module packages the data into a transmission file, encrypts the data using a distributed encryption method, uploads the file to the cloud platform in real time through the Internet of Things transmission protocol, and caches the encrypted file locally for users to access and retrieve relevant results.

[0018] Preferably, the multispectral camera in the data acquisition module is used to acquire images of the artworks under visible light, infrared light, and ultraviolet light. Among them, the ultraviolet light image is used to detect potential cracks and corrosion marks on the surface of the artworks, and the infrared light image is used to detect the internal structure and material properties of the artworks;

[0019] The laser scanner in the data acquisition module is used to acquire the three-dimensional structure data of the surface of the artworks. The height map of the surface of the artworks is established based on the intensity of the laser signal reflected by the laser scanner. The height map satisfies the following relationship:

[0020]

[0021] where Z(x, y) is the height at the coordinate point (x, y), c is the speed of light constant, and t is the round-trip time of the laser signal.

[0022] Preferably, the image processing module includes the following units:

[0023] The image noise removal unit is used to receive the image data of the artworks transmitted by the data acquisition module, and remove the noise part of the image through a nonlinear partial differential equation model to generate smoothed image data. The specific expression of the partial differential equation is:

[0024]

[0025] where u(x, y, t) is the pixel value of the image at time t, is the gradient of the image, is the diffusion coefficient of the image, λ is the regularization coefficient, and R(u) is the detail enhancement term;

[0026] The detail enhancement unit receives the smoothed image data output from the image noise removal unit, and enhances the high-frequency details of the image using the locally weighted regression method to generate image data with high texture clarity. The weighting function of the locally weighted regression is:

[0027]

[0028] where w(x ′ , y ′ ) is the weighting coefficient of the pixel (x ′ , y ′ ), u(x ′ , y ′ ) is the pixel value of the neighboring pixel (x ′ , y ′ ), u(x, y) is the pixel value of the central pixel (x, y), and σ is the parameter controlling the weighting degree;

[0029] The edge enhancement unit receives the high-detail clarity image output by the detail enhancement unit and enhances the edges of the image using the Laplace operator. The expression of the Laplace operator is as follows:

[0030]

[0031] where Δu(x, y) is the second derivative of the image at the pixel point (x, y), and u(x, y) is the pixel value of the image at the pixel point (x, y).

[0032] Preferably, the diffusion coefficient of the nonlinear partial differential equation model in the image noise removal unit is defined as:

[0033]

[0034] where is the magnitude of the pixel gradient of the image, and K is the diffusion threshold;

[0035] If , enhance diffusion to remove noise;

[0036] If , weaken diffusion to retain edge features.

[0037] Preferably, the local weighted regression method in the detail enhancement unit combines multi-scale processing, processes images of different scales by constructing a Gaussian pyramid, and refines details layer by layer. The pixel weighting function for each layer is:

[0038]

[0039] where w s (x ′ , y ′ ) is the weighting coefficient of the pixel (x ′ , y ′ ) in the s-th layer,

[0040] u s (x, y) is the pixel value of the central pixel (x, y) in the s-th layer,

[0041] u s (x ′ , y ′ ) is the pixel value of the central pixel (x ′ , y ′ ) in the s-th layer,

[0042] σ s is the weight control parameter of the s-th layer.

[0043] Preferably, the Laplacian operator used by the edge enhancement unit is combined with multi-directional gradient calculation to enhance edge features in different directions. The specific expression is:

[0044]

[0045] where N is the number of gradient directions, is the second-order derivative of the image in the i-th direction, and i represents the direction number.

[0046] represents the value of the Laplacian operator of the image at the pixel point (x, y).

[0047] Preferably, the data analysis module includes:

[0048] A style classification unit that extracts multi-layer features of the art image using a convolutional neural network to complete style classification;

[0049] A damage detection unit that detects damaged areas in the image based on a pixel-by-pixel classification method, and combines the results of the detail enhancement and edge enhancement units to label the damaged parts.

[0050] Preferably, the training objective of the convolutional neural network is defined by the cross-entropy loss function as:

[0051]

[0052] where L is the loss value, y i is the true label of the i-th class, is the predicted probability of the i-th class, and M is the total number of classification categories.

[0053] Preferably, the convolutional neural network model of the style classification unit combines a self-attention mechanism to capture global features of the art image. The weight calculation formula of the self-attention mechanism is:

[0054]

[0055] where α ij is the attention weight between the i-th pixel and the j-th pixel, e ij is the similarity between pixels i and j, and R is the total number of pixels in the image.

[0056] Preferably, the IoT transmission module combines a distributed edge computing unit. The edge computing unit processes part of the data locally and caches the processing results. If the network is interrupted, it automatically delays uploading and automatically synchronizes to the cloud platform after the network resumes. The distributed edge computing unit encrypts the transmitted file in slices. The expression of the slice encryption algorithm is:

[0057] C i =E(Ki , M i ),

[0058] Among them, C i is the encrypted data shard, K i is the encryption key corresponding to each data shard, M i is the original data shard, and E is the encryption algorithm function.

[0059] The present invention provides an identification system for artworks based on Internet of Things technology. It has the following beneficial effects:

[0060] 1. By adopting a non-linear partial differential equation model in the image noise removal unit and introducing a regularization detail enhancement term, the present invention achieves the goal of enhancing the high-frequency features of the image while suppressing noise, and obtains the effect of improving the detail restoration ability of the artwork image.

[0061] 2. By combining the Laplacian operator with multi-directional gradient calculation in the edge enhancement unit, the present invention achieves the comprehensive capture and enhancement of edge features in different directions of the image, and obtains the effect of improving the expression ability of the artwork contour and complex morphological features.

[0062] 3. By adopting a convolutional neural network combined with a self-attention mechanism in the style classification unit, the present invention achieves the simultaneous capture of global features and local details of the artwork image, and obtains the effect of improving the classification accuracy of complex art styles.

[0063] 4. By introducing a generative adversarial network in the damage detection unit, generating a potential damage area prediction map and combining it with a discriminator for verification, the present invention achieves the intuitive visualization of the specific damage location and range of the artwork, and obtains the effect of enhancing the accuracy and visualization ability of damage detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 is the system framework diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0065] To enable those skilled in the art to understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0066] The following describes the present invention in detail with reference to the accompanying drawings:

[0067] Embodiment:

[0068] Please refer to the attached Figure 1, an embodiment of the present invention provides an identification system for artworks based on Internet of Things technology. The system includes a data acquisition module, an image processing module, a data analysis module, and an Internet of Things transmission module:

[0069] The data acquisition module collects multi-dimensional data of artworks through multi-modal sensors and high-resolution imaging devices. The multi-modal sensors include temperature and humidity sensors, light sensors, and micro-vibration sensors for collecting environmental parameters. The high-resolution imaging devices include multi-spectral cameras and laser scanners for collecting image data and three-dimensional surface structure data of artworks. The collected data is preliminarily processed by the local processing unit in the data acquisition module and then sent to the image processing module;

[0070] After receiving the environmental parameters and artwork image data transmitted by the data acquisition module, the image processing module analyzes the image by constructing the relationship between spatial pixel values, degrades the noise part of the image using the change of pixel gradient, generates a pixel value distribution map based on spatial coordinates, obtains optimized image data and relevant environmental information, and transmits it to the data analysis module;

[0071] After receiving the structured image file and environmental data file transmitted by the image processing module, the data analysis module extracts features from the image data according to predefined artwork classifications and feature tags, extracts the edges and texture features of the image layer by layer, combines the environmental parameters to complete artwork style classification and damage detection, generates an analysis result data file containing classification results and damage features, and returns the file together with the structured image file to the Internet of Things transmission module;

[0072] After receiving the analysis result data file and structured image file from the data analysis module, the Internet of Things transmission module packages the data into a transmission file, encrypts the data using a distributed encryption method, uploads the file to the cloud platform in real time through the Internet of Things transmission protocol, and caches the encrypted file locally for users to access and retrieve relevant results.

[0073] Benefits of the data acquisition module: Through temperature and humidity sensors, light sensors, and micro-vibration sensors, it can collect environmental parameters closely related to the preservation environment of artworks in real time, helping to monitor the preservation conditions of artworks and prevent damage caused by environmental factors. The multi-spectral camera captures visible light, ultraviolet light, and infrared light images to comprehensively obtain detailed information of artworks. The laser scanner constructs three-dimensional surface structure data to highly accurately restore the surface features of artworks, providing multi-dimensional support for subsequent detail restoration and damage detection. The local processing unit preliminarily sorts and optimizes the collected data, reducing the amount of transmitted data and improving the overall response efficiency of the system.

[0074] Benefits of the image processing module: By analyzing the gradient changes of spatial pixel values, a non-linear processing method is used to degrade the noise part, retain the high-frequency details of the artwork, and avoid detail loss caused by traditional denoising methods. Generate a pixel value distribution map based on spatial coordinates, optimize the contrast and clarity of the artwork image, and provide high-quality input for subsequent analysis. Combine environmental parameters with image data to ensure that the generated optimized image can reflect the state changes of the artwork under different environmental conditions, and improve the authenticity and applicability of the overall data.

[0075] Benefits of the data analysis module: Extract the edge and texture features of the image layer by layer, comprehensively analyze the artwork in combination with environmental parameters, effectively identify the style and material characteristics of the artwork, and improve the accuracy of style classification. By analyzing the structured image file, accurately locate the damage position and scope of the artwork, form an analysis result containing damage characteristics, and provide a scientific basis for the restoration and protection of the artwork. Combine the artwork style classification with damage detection to provide a complete artwork status analysis report, which is convenient for users to comprehensively evaluate the history, preservation, and restoration value of the artwork.

[0076] Benefits of the IoT transmission module: Adopt a distributed encryption method to ensure the security of analysis results and image files during transmission, and avoid data leakage or tampering. Through the IoT transmission protocol, upload files to the cloud platform in real time, cache the encrypted files locally, and ensure the availability of data in case of network interruption. After the files are stored, users can access and retrieve relevant results through terminal devices at any time, enhancing the user experience and the usability of the system.

[0077] The multispectral camera in the data acquisition module is used to obtain images of the artwork under visible light, infrared light, and ultraviolet light. Among them, the ultraviolet light image is used to detect potential cracks and corrosion marks on the surface of the artwork, and the infrared light image is used to detect the internal structure and material characteristics of the artwork;

[0078] The laser scanner in the data acquisition module is used to obtain the three-dimensional structure data of the surface of the artwork. The height map of the surface of the artwork is established by the intensity of the laser signal reflected by the laser scanner. The height map satisfies the following relationship:

[0079]

[0080] Among them, Z(x, y) is the height at the coordinate point (x, y), c is the speed of light constant, and t is the round-trip time of the laser signal.

[0081] Advantages of the multispectral camera: Provide multi-dimensional information of visible light, ultraviolet light, and infrared light, can identify surface damage (cracks, corrosion) and internal characteristics (materials, structures), and enhance the comprehensiveness and reliability of the artwork image data.

[0082] Advantages of the laser scanner: With high-precision 3D reconstruction capabilities and support for the laser time reflection formula, it can capture the fine texture of the art surface, providing important 3D information support for the protection and digital preservation of artworks.

[0083] Comprehensive value: The collaborative work of the multispectral camera and the laser scanner ensures the comprehensiveness, accuracy, and multi-dimensional support of artwork data acquisition, providing a solid technical foundation for subsequent image processing, analysis, and artwork restoration, significantly enhancing the overall performance and practical value of the system.

[0084] The image processing module includes the following units:

[0085] The image noise removal unit is used to receive the artwork image data transmitted by the data acquisition module and remove the noise part of the image through a non-linear partial differential equation model to generate smoothed image data. The specific expression of the partial differential equation is:

[0086]

[0087] where u(x, y, t) is the pixel value of the image at time t, is the gradient of the image, is the diffusion coefficient of the image, λ is the regularization coefficient, and R(u) is the detail enhancement term;

[0088] The detail enhancement unit receives the smoothed image data output by the image noise removal unit and enhances the high-frequency details of the image using the locally weighted regression method to generate image data with high texture clarity. The weighting function of the locally weighted regression is:

[0089]

[0090] where w(x ′ , y ′ ) is the weighting coefficient of the pixel (x ′ , y ′ ), u(x ′ , y ′ ) is the pixel value of the neighboring pixel (x ′ , y ′ ), u(x, y) is the pixel value of the central pixel (x, y), and σ is the parameter controlling the weighting degree;

[0091] The edge enhancement unit receives the high-detail clarity image output by the detail enhancement unit and enhances the edges of the image using the Laplace operator. The expression of the Laplace operator is:

[0092]

[0093] Wherein, Δu(x, y) is the second-order derivative of the image at the pixel point (x, y), and u(x, y) is the pixel value of the image at the pixel point (x, y).

[0094] Advantages of the image noise removal unit: Accurately remove noise, retain key details, generate smooth and clear image data, and provide high-quality input for subsequent processing.

[0095] Advantages of the detail enhancement unit: Through the local weighted regression method, the high-frequency texture features of the artwork are restored, and the detail information of cracks and brushstrokes is refined to make the image realistic and visually expressive.

[0096] Advantages of the edge enhancement unit: Combined with detail enhancement results, it strengthens multi-directional edge features, highlights the contour structure of the artwork, and provides accurate edge data support for style classification and damage detection.

[0097] Overall synergy of modules: The processing logic is interconnected, from noise removal to detail enhancement to edge enhancement, gradually optimizing the image quality of artworks. The synergy of modules effectively solves the problem of noise and detail in traditional image processing methods, and provides image data foundation for subsequent analysis of artwork recognition.

[0098] Diffusion coefficient of the nonlinear partial differential equation model in the image noise removal unit Defined as:

[0099]

[0100] in, is the amplitude of the image pixel gradient, K is the diffusion threshold;

[0101] like When , diffusion is enhanced to remove noise;

[0102] like , the diffusion is weakened to preserve the edge features.

[0103] Adaptive noise removal: By dynamically adjusting the diffusion coefficient based on the gradient amplitude, the noise part of the image can be effectively removed while maintaining the clarity of edge and detail information, ensuring that the noise removal process does not affect important structures in the image.

[0104] Edge protection effect: The diffusion coefficient is weakened in the gradient area to avoid excessive smoothing of edge features, effectively retaining the outline and details of the artwork, and ensuring that the original features of the artwork in the image are preserved.

[0105] Efficient Denoising and Detail Preservation: This diffusion coefficient model achieves multi-scale noise removal effects by adaptively adjusting the diffusion intensity. The noise and the detailed parts of the image can be processed as needed in different regions to achieve the best noise removal effect.

[0106] Advantages of the Mathematical Model: The non-linear partial differential equation model can adaptively adjust the processing process to suit different image characteristics. Especially when removing noise, it can retain the important features of the image, making this method more accurate and efficient in the image processing of artworks.

[0107] Generally speaking, the image noise removal unit of the present invention, by adopting a non-linear partial differential equation model based on gradient magnitude and combining diffusion threshold adjustment, can efficiently remove noise while protecting and enhancing the key details and edge features in the image, providing input data for subsequent image analysis and processing.

[0108] The local weighted regression method in the detail enhancement unit, combined with multi-scale processing, processes images of different scales by constructing a Gaussian pyramid, refining details layer by layer. The pixel weighting function for each layer is:

[0109]

[0110] where w s (x ′ , y ′ ) is the weighting coefficient of the pixel (x ′ , y ′ ) in the s-th layer,

[0111] u s (x, y) is the pixel value of the central pixel (x, y) in the s-th layer,

[0112] u s (x ′ , y ′ ) is the pixel value of the central pixel (x ′ , y ′ ) in the s-th layer,

[0113] σ s is the weight control parameter of the s-th layer.

[0114] Advantages of Multi-scale Feature Processing: The hierarchical processing of the Gaussian pyramid can enhance the global features and local details of artworks images, and then achieve layer-by-layer optimization from large contours to small textures, solving the problem that single-scale methods cannot simultaneously take into account details and global features.

[0115] Precision Enhancement of Locally Weighted Regression: The pixel weighting function of each layer is dynamically adjusted according to the similarity of pixel values, which can accurately restore the high-frequency details of artworks, avoid artifacts or unnaturalness, and enhance the visual effect of the image.

[0116] Significant Synergistic Enhancement Effect: The synergistic effect of multi-scale processing and locally weighted regression enables the accurate restoration of image details while maintaining the coherence and consistency of overall features, providing strong technical support for the expression of complex features of artworks.

[0117] Adaptation to Diverse Artwork Features: This method can flexibly adapt to artworks of different types, resolutions, and complexities, and is applicable to the digital processing and enhancement of various artworks.

[0118] Through the detail enhancement unit of the present invention combined with the multi-scale processing method, both the global characteristics and local details of artworks can be effectively enhanced, providing image data support for the identification, protection, and restoration of artworks.

[0119] The Laplace operator used in the edge enhancement unit combines with multi-directional gradient calculation to enhance edge features in different directions. The specific expression is:

[0120]

[0121] where N is the number of gradient directions, is the second-order derivative of the image in the i-th direction, and i represents the direction number.

[0122] represents the value of the Laplace operator at the pixel point (x, y) of the image.

[0123] Comprehensive Edge Feature Capture: The Laplace operator combined with multi-directional gradient calculation can comprehensively extract edge features in the artwork image by accumulating second-order derivatives in multiple directions, adapting to complex and irregularly shaped edges.

[0124] Improve the Precision of Edge Enhancement: The sensitivity of the second-order derivative to gradient changes ensures the accurate positioning of edge features, avoiding blurring and omission problems in traditional methods, and enhancing the contrast between the edge and the background.

[0125] Provide Support for Subsequent Modules: The edge-enhanced image provides clear input data for the style classification and damage detection modules, significantly improving the accuracy and reliability of subsequent processing.

[0126] Flexibility and Adaptability: The flexible adjustment of multi-directional gradients enables this method to adapt to the processing requirements of various artwork images, taking into account both efficiency and effect, and providing a highly scalable technical solution for the processing of complex images.

[0127] Through the edge enhancement unit of the present invention, the edge features of the art image are comprehensively and accurately enhanced, laying a solid foundation for achieving high-precision art identification and protection.

[0128] The data analysis module includes:

[0129] A style classification unit that extracts multi-layer features of the art image using a convolutional neural network to complete style classification;

[0130] A damage detection unit that detects damaged areas in the image based on a per-pixel classification method, and combines the results of the detail enhancement and edge enhancement units to mark the damaged parts.

[0131] The training objective of the convolutional neural network is defined by the cross-entropy loss function as:

[0132]

[0133] where L is the loss value, y i is the true label of the i-th class, is the predicted probability of the i-th class, and M is the total number of classification categories.

[0134] The convolutional neural network model of the style classification unit incorporates a self-attention mechanism to capture the global features of the art image. The weight calculation formula of the self-attention mechanism is:

[0135]

[0136] where α ij is the attention weight between the i-th pixel and the j-th pixel, e ij is the similarity between pixels i and j, and R is the total number of pixels in the image.

[0137] Advantages of the style classification unit: By combining a convolutional neural network and a self-attention mechanism, it achieves a balance between global features and local details, significantly improving the accuracy of complex art style classification.

[0138] Advantages of the damage detection unit: Through a per-pixel classification method, combined with the results of detail enhancement and edge enhancement, it accurately marks the damaged parts of the art, providing a scientific basis for restoration and protection.

[0139] Overall value of module collaboration: The combination of style classification and damage detection forms a complete art status analysis system, providing comprehensive support for the intelligent management and protection of art.

[0140] Enhancement effect of the self-attention mechanism: Introducing global information into the classification model makes the style classification results of art more accurate, especially when dealing with artworks with complex textures and details.

[0141] Through the data analysis module of the present invention, the style characteristics and damage information of artworks can be comprehensively and accurately extracted and analyzed, providing technical innovation and efficient solutions for the protection, identification, and restoration of artworks.

[0142] The Internet of Things transmission module is combined with a distributed edge computing unit. The edge computing unit processes part of the data locally and caches the processing results. If the network is interrupted, the upload is automatically delayed, and after the network is restored, it is automatically synchronized to the cloud platform. The distributed edge computing unit encrypts the transmitted file in slices. The expression of the slice encryption algorithm is:

[0143] C i =E(K i ,M i ),

[0144] where C i is the encrypted data slice, K i is the encryption key corresponding to each data slice, M i is the original data slice, and E is the encryption algorithm function.

[0145] Efficiency and real-time performance of edge computing: The edge computing unit processes part of the data locally, reducing the pressure on the cloud and improving the efficiency of data transmission and processing. At the same time, it can cache the results and automatically synchronize when the network is interrupted, ensuring the integrity and continuity of the data.

[0146] Security and reliability of slice encryption: The slice encryption mechanism encrypts each data slice independently, enhancing the security of the transmitted data.

[0147] Network adaptability and scalability: This module can adapt to complex network environments, support large-scale data transmission requirements through distributed deployment and multi-node collaboration, and provide a flexible technical foundation for future expansion.

[0148] Through the Internet of Things transmission module of the present invention combined with a distributed edge computing unit, the system can efficiently, securely, and reliably complete data transmission and synchronization, providing support for the real-time performance, stability, and scalability of the artwork identification system.

[0149] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A system for identifying artworks based on Internet of Things technology, characterized in that: The system includes a data acquisition module, an image processing module, a data analysis module and an Internet of Things transmission module: The data acquisition module collects multi-dimensional data of the artwork through multi-modal sensors and high-resolution imaging equipment. The multi-modal sensors include temperature and humidity sensors, light sensors and micro-vibration sensors for collecting environmental parameters. The high-resolution imaging equipment includes multi-spectral cameras and laser scanners for collecting image data and three-dimensional surface structure data of the artwork. The collected data is initially processed by the local processing unit in the data acquisition module and then sent to the image processing module. After receiving the environmental parameters and artwork image data transmitted by the data acquisition module, the image processing module analyzes the image by constructing the relationship between spatial pixel values, degrades the noise part of the image by using the change of pixel gradient, generates a pixel value distribution map based on spatial coordinates, obtains optimized image data and related environmental information, and transmits it to the data analysis module; After receiving the structured image file and environmental data file transmitted by the image processing module, the data analysis module performs feature extraction on the image data according to the predefined artwork classification and feature labels, extracts the edge and texture features of the image layer by layer, completes the artwork style classification and damage detection in combination with the environmental parameters, and generates an analysis result data file containing the classification results and damage features, and returns the file together with the structured image file to the Internet of Things transmission module; After receiving the analysis result data file and structured image file from the data analysis module, the Internet of Things transmission module packages the data into a transmission file, encrypts the data using a distributed encryption method, uploads the file to the cloud platform in real time through the Internet of Things transmission protocol, and caches the encrypted file locally for users to access and retrieve relevant results.

2. The artwork recognition system based on Internet of Things technology according to claim 1 is characterized in that: The multispectral camera in the data acquisition module is used to obtain artwork images under visible light, infrared light and ultraviolet light, wherein the ultraviolet light image is used to detect potential cracks and corrosion marks on the surface of the artwork, and the infrared light image is used to detect the internal structure and material characteristics of the artwork; The laser scanner in the data acquisition module is used to obtain the three-dimensional structural data of the artwork surface. The height map of the artwork surface is established by the intensity of the laser signal reflected by the laser scanner. The height map satisfies the following relationship: Where Z(x, y) is the height at the coordinate point (x, y), c is the speed of light constant, and t is the round-trip time of the laser signal.

3. The artwork recognition system based on Internet of Things technology according to claim 1 is characterized in that: The image processing module includes the following units: The image noise removal unit is used to receive the artwork image data transmitted by the data acquisition module, and remove the noise part of the image through a nonlinear partial differential equation model to generate smoothed image data. The specific expression of the partial differential equation is: Among them, u(x, y, t) is the pixel value of the image at time t, is the gradient of the image, is the diffusion coefficient of the image, λ is the regularization coefficient, and R(u) is the detail enhancement term; The detail enhancement unit receives the smoothed image data output from the image noise removal unit, and enhances the high-frequency details of the image using a local weighted regression method to generate image data with high texture clarity. The weighting function of the local weighted regression is: Among them, w(x ′ ,y ′ ) is the pixel (x ′ ,y ′ ), u(x ′ ,y ′ ) is the neighborhood pixel (x ′ ,y ′ ), u(x, y) is the pixel value of the center pixel (x, y), and σ is a parameter that controls the degree of weighting; The edge enhancement unit receives the high-detail definition image output by the detail enhancement unit, and uses the Laplace operator to enhance the edge of the image. The expression of the Laplace operator is: Wherein, Δu(x, y) is the second-order derivative of the image at the pixel point (x, y), and u(x, y) is the pixel value of the image at the pixel point (x, y).

4. The artwork recognition system based on Internet of Things technology according to claim 3 is characterized in that: Diffusion coefficient of the nonlinear partial differential equation model in the image noise removal unit Defined as: in, is the amplitude of the image pixel gradient, K is the diffusion threshold; like When , diffusion is enhanced to remove noise; like , the diffusion is weakened to preserve the edge features.

5. The artwork recognition system based on Internet of Things technology according to claim 3 is characterized in that: The local weighted regression method in the detail enhancement unit is combined with multi-scale processing to process images of different scales by constructing a Gaussian pyramid to refine the details layer by layer. The pixel weighting function of each layer is: Among them, w s (x ′ ,y ′ ) is the pixel (x) in the sth layer ′ ,y ′ ), u s (x, y) is the pixel value of the center pixel (x, y) of the sth layer, u s (x ′ ,y ′ ) is the center pixel of the sth layer (x ′ ,y ′ ), σ s is the weight control parameter of the sth layer.

6. The artwork recognition system based on Internet of Things technology according to claim 3 is characterized in that: The Laplace operator used by the edge enhancement unit is combined with multi-directional gradient calculation to enhance edge features in different directions. The specific expression is: Where N is the number of gradient directions, is the second-order derivative of the image in the i-th direction, i represents the number of the direction, Represents the Laplacian value of the image at the pixel point (x, y).

7. The artwork recognition system based on Internet of Things technology according to claim 1 is characterized in that: The data analysis module includes: The style classification unit uses a convolutional neural network to extract multi-layer features of artwork images and complete style classification; The damage detection unit detects the damaged area in the image based on the pixel-by-pixel classification method, and marks the damaged area by combining the results of the detail enhancement and edge enhancement units.

8. The artwork recognition system based on Internet of Things technology according to claim 7 is characterized in that: The training objective of the convolutional neural network is defined by the cross entropy loss function: Among them, L is the loss value, y i is the true label of the i-th category, is the predicted probability of the i-th category, and M is the total number of classification categories.

9. The artwork recognition system based on Internet of Things technology according to claim 8, characterized in that: The convolutional neural network model of the style classification unit is combined with a self-attention mechanism to capture the global features of the artwork image. The weight calculation formula of the self-attention mechanism is: Among them, α ij is the attention weight between the i-th pixel and the j-th pixel, e ij is the similarity between pixels i and j, and R is the total number of pixels in the image.

10. The artwork recognition system based on Internet of Things technology according to claim 1, characterized in that: The IoT transmission module is combined with a distributed edge computing unit. The edge computing unit processes part of the data locally and caches the processing results. If the network is interrupted, the upload is automatically delayed and automatically synchronized to the cloud platform after the network is restored. The distributed edge computing unit performs fragment encryption on the transmission file. The expression of the fragment encryption algorithm is: C i =E(K i ,M i ), Among them, C i is the encrypted data fragment, K i is the encryption key corresponding to each data shard, M i is the original data shard, and E is the encryption algorithm function.

Images