Methods, systems, and procedures for identifying cultural relics in museum collections

Abstract

This application relates to the field of cultural relic identification technology, and in particular to a method, system, and program product for identifying the identity of museum-collected cultural relics. The method includes: acquiring multimodal data of the museum-collected cultural relics; preprocessing the multimodal data to generate a multidimensional feature vector; generating fingerprint features of the museum-collected cultural relics based on the multidimensional feature vector; constructing a decision tree based on the fingerprint features; and identifying the identity of the museum-collected cultural relics based on the decision tree. This solves the problems of low identification efficiency, poor accuracy, and low standardization in traditional methods of identifying museum-collected cultural relics.

Description

Technical Field

[0001] This application relates to the field of cultural relic identification technology, and in particular to a method, system and program product for identifying the identity of cultural relics in museum collections. Background Technology

[0002] Identifying the artifacts in a museum's collection is a core aspect of artifact protection and management, and its accuracy directly affects the security and authenticity of the artifacts.

[0003] In related technologies, during the circulation of cultural relics, different business scenarios have significantly different requirements for identification accuracy and efficiency. If a single feature identification method is used, the complex surface of cultural relics is easily affected by factors such as changes in ambient light. If a multimodal feature fusion method is used, there is a lack of systematic feature screening and optimization mechanisms. Summary of the Invention

[0004] This application provides a method, system, and program product for identifying cultural relics in museum collections, in order to solve the problems of low efficiency, poor accuracy, and poor standardization in the identification methods of cultural relics in museum collections in related technologies.

[0005] The first aspect of this application provides a method for identifying the identity of a museum collection, comprising the following steps: acquiring multimodal data of the museum collection; preprocessing the multimodal data and generating a multidimensional feature vector based on the preprocessed data; generating fingerprint features of the museum collection based on the multidimensional feature vector; constructing a decision tree based on the fingerprint features; and identifying the identity of the museum collection based on the decision tree.

[0006] Optionally, the multimodal data includes multiple types of data, such as three-dimensional geometric data, surface texture data, spectral reflectance data, and environmental data.

[0007] Optionally, the preprocessing method includes at least one of data registration and alignment processing, noise filtering processing, and normalization processing.

[0008] Optionally, generating fingerprint features of museum artifacts based on multidimensional feature vectors includes: dividing the surface of the museum artifacts into multiple local regions; extracting geometric features, texture features, and spectral features of each local region from the multidimensional feature vectors; calculating the information entropy of the corresponding local region based on the geometric features, texture features, and spectral features; normalizing the information entropy of each local region to obtain the standard entropy value of the local region; and generating fingerprint features of the museum artifacts based on the standard entropy value of the local region.

[0009] Optionally, the formula for calculating information entropy is:

[0010]

[0011] Where X represents a random variable, H(X) represents the information entropy of the random variable X, and p(x)i ) represents the i-th value of the random variable X, and log represents the logarithmic operation.

[0012] Optionally, a decision tree is constructed based on fingerprint point features, including: identifying the standard entropy value of the fingerprint point features; calculating the importance level of the fingerprint point features based on the standard entropy value; determining the root node based on the importance level, determining the internal nodes based on the value range of the fingerprint point features, determining the leaf nodes based on the classification result of the fingerprint point features; and constructing the decision tree based on the root node, internal nodes, and leaf nodes.

[0013] Optionally, before identifying the identity of the museum artifacts based on the decision tree, the process may include: obtaining the cost complexity and Gini coefficient of the decision tree; pruning the decision tree based on the indicated cost complexity; and optimizing the number of fingerprint features in the decision tree based on the Gini coefficient.

[0014] Optionally, identifying the identity of a museum artifact based on a decision tree also includes: obtaining the artifact's circulation type; determining the artifact's security level based on the circulation type; determining the target number of fingerprint features required for artifact identification based on the security level; and identifying the artifact based on the target number of fingerprint features in the decision tree.

[0015] A second aspect of this application provides a system for identifying cultural relics in a museum collection, comprising: an acquisition module for acquiring multimodal data of the cultural relics in the museum collection; a preprocessing module for preprocessing the multimodal data and generating a multidimensional feature vector based on the preprocessed data; and a generation module for generating fingerprint features of the cultural relics in the museum collection based on the multidimensional feature vector, constructing a decision tree based on the fingerprint features, and identifying the cultural relics in the museum collection based on the decision tree.

[0016] Optionally, the multimodal data includes multiple types of data, such as three-dimensional geometric data, surface texture data, spectral reflectance data, and environmental data.

[0017] Optionally, the preprocessing method includes at least one of data registration and alignment processing, noise filtering processing, and normalization processing.

[0018] Optionally, the generation module is further used to: divide the surface of the museum artifact into multiple local regions; extract the geometric features, texture features and spectral features of each local region from the multidimensional feature vector, calculate the information entropy of the corresponding local region based on the geometric features, texture features and spectral features; normalize the information entropy of each local region to obtain the standard entropy value of the local region, and generate the fingerprint point features of the museum artifact based on the standard entropy value of the local region.

[0019] Optionally, the formula for calculating information entropy is:

[0020]

[0021] Where X represents a random variable, H(X) represents the information entropy of the random variable X, and p(x) i ) represents the i-th value of the random variable X, and log represents the logarithmic operation.

[0022] Optionally, the generation module is further used to: identify the standard entropy value of fingerprint point features; calculate the importance level of fingerprint point features based on the standard entropy value; determine the root node based on the importance level, determine the internal nodes based on the value range of fingerprint point features, determine the leaf nodes based on the classification results of fingerprint point features; and construct a decision tree based on the root node, internal nodes, and leaf nodes.

[0023] Optionally, the museum artifact identification system also includes: an optimization module, used to obtain the cost complexity and Gini coefficient of the decision tree before identifying the museum artifact based on the decision tree; to prune the decision tree according to the indicated cost complexity; and to optimize the number of fingerprint features of the decision tree based on the Gini coefficient.

[0024] Optionally, the generation module is further used to: obtain the cultural relic circulation type of the collection; determine the security level of the collection based on the cultural relic circulation type; determine the target number of fingerprint point features required for the identification of the collection based on the security level; and identify the identity of the collection based on the target number of fingerprint point features in the decision tree.

[0025] A third aspect of this application provides a computer program product, including a computer program or instructions, which, when executed, implement the method for identifying cultural relics in a collection as described in the above embodiments.

[0026] Therefore, this application has at least the following beneficial effects:

[0027] This application's embodiments acquire multimodal data of museum collections, preprocess this data, generate corresponding multidimensional feature vectors based on the preprocessed multimodal data, and then extract fingerprint features from the museum collections based on these multidimensional feature vectors. A decision tree model is then constructed using these fingerprint features as the core foundation. Finally, the constructed decision tree enables accurate identification of the museum collections. Through collaborative analysis of multimodal features, the reliability of artifact identification is improved, achieving automation and intelligence in the identification process. This significantly reduces the labor and time costs of artifact preservation. The system can automatically adjust the identification intensity and dynamically optimize the identification strategy according to actual application scenarios, ensuring the security of museum artifact management, improving the efficiency of artifact management, and providing the most suitable identification solutions for different application scenarios. Therefore, it solves the problems of low efficiency, poor accuracy, and low standardization in traditional museum artifact identification methods.

[0028] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0029] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0030] Figure 1 This is a flowchart illustrating a method for identifying cultural relics in a museum collection, according to an embodiment of this application.

[0031] Figure 2 This is a technical flowchart of a method for identifying cultural relics in a museum collection, provided according to an embodiment of this application.

[0032] Figure 3 This is a schematic diagram of a system for identifying cultural relics in a museum collection, provided according to an embodiment of this application. Detailed Implementation

[0033] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0034] The method, system, and program product for identifying museum artifacts according to embodiments of this application are described below with reference to the accompanying drawings. Addressing the problems of low efficiency, poor accuracy, and low standardization in traditional museum artifact identification methods mentioned in the background, this application provides a method for identifying museum artifacts. In this method, multimodal data of the museum artifacts are acquired; the multimodal data is preprocessed to generate a multidimensional feature vector; fingerprint features of the museum artifacts are generated based on the multidimensional feature vector; a decision tree is constructed based on the fingerprint features; and the identity of the museum artifacts is identified based on the decision tree. This solves the problems of low efficiency, poor accuracy, and low standardization in traditional museum artifact identification methods.

[0035] Specifically, Figure 1 This is a flowchart illustrating a method for identifying cultural relics in a museum collection, as provided in an embodiment of this application.

[0036] like Figure 1 As shown, the method for identifying the artifacts in this collection includes the following steps:

[0037] In step S101, multimodal data of the museum's collection of cultural relics are acquired.

[0038] Multimodal data is a collection of information about the same cultural relic from different dimensions obtained through various information collection methods.

[0039] It is understood that by acquiring multimodal data of museum collections, the embodiments of this application can perform quantitative analysis of the composition of museum collections, restore the truth of the cultural relics, reflect the authenticity, integrity and safety of cultural relics from multiple dimensions, and improve the detection accuracy.

[0040] Specifically, a structured light 3D scanner was used to acquire high-precision point cloud data of the artifact's surface, achieving a resolution of 0.05 mm, and capturing geometric features such as surface curvature and normal vector distribution. A high-resolution digital camera was used to acquire multi-angle RGB (Red, Green, Blue) images of the artifact's surface, equipped with a ring-shaped LED (Light Emitting Diode) light source to eliminate shadow effects. Eight images were captured at each acquisition point, covering the entire surface area of ​​the artifact. A hyperspectral imaging system with a spectral range of 400-2500 nm and a spectral resolution of 3 nm was used to acquire spectral reflectance data of the artifact's surface. Simultaneously, environmental parameters such as temperature, humidity, and light intensity were collected to establish a data acquisition environment profile. During the scanning process, the ambient temperature was maintained at 20±2℃ and the humidity at 45%±5% to ensure data acquisition stability.

[0041] Furthermore, in embodiments of this application, the multimodal data includes multiple types of three-dimensional geometric data, surface texture data, spectral reflectance data, and environmental data.

[0042] Among them, three-dimensional geometric data is quantitative data that records the shape, size, and positional relationship of cultural relics in three-dimensional space; surface texture data is quantitative data that records the microscopic concavity and convexity features and texture details of the surface of cultural relics; spectral reflectance data is data that records the proportion of light intensity reflected by different wavelengths on the surface of cultural relics; and environmental data is quantitative data that records the key parameters of the environment in which cultural relics are stored / exhibited.

[0043] It is understandable that multimodal data includes multiple data such as three-dimensional geometric data, surface texture data, spectral reflectance data and environmental data. It can form a multimodal data system for cultural relics from aspects such as the three-dimensional truth of cultural relics, the subtle texture of cultural relics, the material composition of cultural relics, and the preservation / exhibition conditions of cultural relics, so as to realize the complete preservation and accurate interpretation of cultural relic information.

[0044] Specifically, three-dimensional geometric data is generated by using a laser emitter to emit laser beams onto the surface of the artifact, and by receiving the time difference of the reflected light, the spatial coordinates of each laser point are calculated. This process is repeated point by point across the entire artifact to generate point cloud data containing millions of points, which is used to restore the three-dimensional truth of the artifact.

[0045] Surface texture data includes microscopic bump parameters, texture pattern features, and texture correlation data. High-resolution texture scanners can capture micron-level fine features on the surface of cultural relics and fully extract texture data related to the craftsmanship and materials of the relics. High-resolution texture scanners used for the digitization of cultural relics typically have a basic imaging resolution of more than 10 million pixels. These high-resolution devices are sufficient to capture ultra-fine features such as hair-thin lines and micron-level features on the surface of cultural relics, ensuring that the texture of the digitized cultural relics is highly consistent with the real object. The surface of the cultural relics is scanned area by area to generate texture maps and restore the fine texture of the cultural relics.

[0046] Spectral reflectance data can cover a wavelength range from 400 nanometers to 2500 nanometers, with each wavelength corresponding to a reflectance value. A hyperspectral imaging system is used to decompose continuous wavelengths of light into individual wavelengths, which are then used to illuminate the surface of the artifact in sequence. The intensity of the reflected light is received by a detector, and the reflectance of each wavelength is calculated to accurately analyze the material composition of the artifact.

[0047] Environmental data can include temperature and humidity data, light data, pollutant databases, air pressure and vibration intensity data in the storage room / exhibition hall, etc. By installing temperature and humidity sensors, light sensors, and pollutant sensors in key locations in the artifact storage room and exhibition hall, data can be collected in real time. The risk of damage to artifacts can be predicted through environmental data, and environmental standards can be customized for artifacts of different materials to optimize preservation / exhibition conditions.

[0048] This application embodiment can perform quantitative analysis of the composition of museum collections by acquiring multimodal data of museum collections, restore the truth of the cultural relics, reflect the authenticity, integrity and safety of cultural relics from multiple dimensions, and improve the detection accuracy.

[0049] In step S102, the multimodal data is preprocessed, and a multidimensional feature vector is generated based on the preprocessed data.

[0050] Preprocessing involves a series of operations to remove impurities, unify the format, fill in missing information, and standardize the original multimodal data. Multidimensional feature vectors are digital vectors formed by extracting key feature indicators from each of the preprocessed multimodal data and then combining them.

[0051] It is understood that the embodiments of this application preprocess multimodal data and generate multidimensional feature vectors, transforming a large amount of complex raw data into concise and computable core information assets, providing accurate data support for cultural relic identification, protection early warning, classification management, etc.

[0052] Specifically, the original 3D point cloud data may contain air points around the cultural relic, dust reflection points, or some missing points due to occlusion. During preprocessing, invalid points are first removed, and then the missing points are filled in using an algorithm. Finally, the point cloud data is converted into a mesh model with uniform precision to ensure that the 3D data format of different cultural relics is consistent.

[0053] The original texture image may have uneven lighting or slight blemishes. During preprocessing, the image brightness / contrast is adjusted, geometric and radiometric corrections are performed to eliminate the effects of lens distortion and uneven lighting, blemishes are removed using algorithms, and the texture image is cropped to a uniform resolution to ensure clear texture details and standardized dimensions.

[0054] The raw spectral data may contain abnormal reflectance values ​​due to ambient light interference. These abnormal values ​​will be removed during preprocessing. The Savitzky-Golay filtering algorithm will be used to smooth the spectral curve, retain the characteristic bands, and then the spectral data of different cultural relics will be unified to the same wavelength range and the reflectance values ​​will be calibrated.

[0055] Raw environmental data may contain abrupt values ​​or missing data due to sensor malfunctions. During preprocessing, abrupt outliers are removed, missing values ​​are filled with normal data from adjacent time periods, and the data is then standardized to ensure the continuity and consistency of environmental data.

[0056] Furthermore, in the embodiments of this application, the preprocessing methods include at least one of data registration and alignment processing, noise filtering processing, and normalization processing.

[0057] Among them, data registration is the process of associating cultural relic data collected from different sources, modalities, or perspectives with the same three-dimensional coordinate system; alignment is the process of adjusting the position, time, or order of multiple sets of scattered data under the same modality; noise filtering is the process of removing or weakening invalid interference information in the original data; and standardization is the process of unifying cultural relic data collected from different devices, at different times, and in different batches to the same format, unit, range, or precision.

[0058] It is understood that the embodiments of this application may use at least one of the following preprocessing methods—data registration and alignment, noise filtering, and standardization—to preprocess multimodal data, which can transform the original data into high-quality data that can be fused and computed, providing key support for subsequent applications such as multidimensional feature vector generation and cultural relic identification.

[0059] Specifically, data registration and alignment are performed by coarse alignment using marker points. This coarse alignment is achieved through automatic feature point extraction. The algorithm identifies three marker points on the surface of the artifact, either pre-defined or automatically extracted. These three marker points must meet the geometric condition of not being on the same straight line and being distributed as evenly as possible on the artifact's surface to construct a stable spatial triangular reference. Using these marker points as a reference, the texture image is roughly converted into a point cloud model. An improved ICP (Iterative ClosestPoint) algorithm is then used to achieve precise registration between the point cloud data and the texture image, with a registration error of less than 0.1 mm. The pixel positions corresponding to these feature points are found in the texture image, and the positional deviation between the two is calculated. The ICP algorithm automatically fine-tunes the position of the texture image based on the deviation of all feature points. After each iteration, the deviation gradually decreases until the deviation for three consecutive iterations is less than 0.1 mm, at which point the algorithm stops. Through feature-level fusion, geometric, texture, and spectral features are integrated into a unified multidimensional feature vector.

[0060] Anisotropic diffusion filtering is used to remove point cloud noise. Based on the point cloud density, noise distribution characteristics, and target feature retention requirements, parameters such as the number of neighboring points, diffusion threshold, and iteration count are set. For each filtered point, its surrounding neighboring points are extracted according to the set number to construct a local point cloud neighborhood structure. The average distance and arrangement direction consistency between the current point and its neighboring points in the neighborhood are calculated. If both meet the diffusion threshold requirements, it is determined to be a feature point. If it exceeds the threshold and the arrangement is irregular, it is determined to be a potential noise point. The coordinate position of the feature point remains unchanged. For the noise point, the average coordinate of all its neighboring points in the neighborhood is calculated, and the coordinate of the noise point is adjusted to the average coordinate.

[0061] The noise filtering process uses nonlocal mean filtering to process image noise. The original image is standardized in pixel format and color space, and grayscale / brightness thresholds are set to remove extreme outliers exceeding the normal grayscale range. Based on noise types such as grain noise and dark spot noise, resolution, and detail preservation requirements, parameters such as search window size, similarity window size, similarity threshold, and weighting coefficients are set. All pixel blocks in the image are traversed in row / column order, with each pixel block as the target pixel block. Similarity search and calculation are initiated, extracting all candidate pixel blocks within the set search window. By calculating the grayscale difference and texture consistency between the target and candidate pixel blocks, a set of pixel blocks meeting the similarity threshold is selected. For the selected similar pixel blocks, weights are assigned based on their similarity to the target pixel block. A weighted average algorithm is then used to calculate the grayscale mean of these similar pixel blocks, which is used as the new grayscale value of the target pixel block, replacing the original pixel block's grayscale value. This achieves image noise processing using nonlocal mean filtering.

[0062] The standardization process involves Z-score standardization of geometric, texture, and spectral features. The original data of 3D geometric data, surface texture data, and spectral reflectance data are aggregated separately and split according to a single feature dimension. Extreme outliers in each dimension are removed, and missing values ​​are filled to ensure data integrity for each feature dimension. The mean and standard deviation are calculated separately for each feature dimension. For each feature value of each sample, the standardized value is calculated according to the following logic: the original value of the sample in a certain feature dimension is subtracted from the mean of that dimension, and then divided by the standard deviation of that dimension to obtain the Z-score standardized value of the sample in that feature dimension. All the transformed feature dimension data are then integrated and correlated by sample to form a standardized feature set that eliminates the influence of dimensions, which is used for the subsequent construction of multidimensional feature vectors.

[0063] This application embodiment preprocesses multimodal data and generates multidimensional feature vectors, transforming a large amount of complex raw data into concise and computable core information assets, providing accurate data support for cultural relic identification, protection early warning, classification management, etc.

[0064] In step S103, fingerprint features of the museum artifacts are generated based on the multidimensional feature vector, a decision tree is constructed based on the fingerprint features, and the identity of the museum artifacts is identified based on the decision tree.

[0065] Among them, fingerprint point features are a subset of features extracted from multi-dimensional feature vectors that possess uniqueness and discriminative power; the decision tree is a hierarchical and progressive judgment model built based on fingerprint point features.

[0066] It is understood that the embodiments of this application identify the identity of museum artifacts by generating fingerprint features of museum artifacts through multi-dimensional feature vectors and constructing decision trees. High-discrimination indicators can be extracted from multi-dimensional feature vectors, and multi-feature cross-validation can be achieved through hierarchical and progressive logical judgment of decision trees. This provides clear data support for the identification conclusions, greatly reduces the false positive rate and false negative rate, and ensures the authenticity and accuracy of the identity of museum artifacts.

[0067] Specifically, from the multidimensional feature vector, indicators that are meaningless or have low distinguishability for identity judgment are removed, while indicators that can significantly distinguish cultural relics of different identities are retained. The selected high-discrimination features are sorted according to identification priority to form a set of features with strong discrimination power, namely fingerprint features.

[0068] Fingerprint features are used as training samples and sorted according to the principle of highest discriminative power. A judgment threshold is set for each fingerprint feature, and branches are split layer by layer. After the first layer judgment, for the samples under each branch, the next fingerprint feature with the highest discriminative power is selected, and the splitting continues until the samples of all branches correspond to a unique identity conclusion, and the decision tree is completed.

[0069] Furthermore, in the embodiments of this application, generating fingerprint features of museum artifacts based on multidimensional feature vectors includes: dividing the surface of the museum artifact into multiple local regions; extracting geometric features, texture features, and spectral features of each local region from the multidimensional feature vector; calculating the information entropy of the corresponding local region based on the geometric features, texture features, and spectral features; normalizing the information entropy of each local region to obtain the standard entropy value of the local region; and generating fingerprint features of the museum artifact based on the standard entropy value of the local region.

[0070] Among them, a local area is a small, independent region after the overall surface of the museum's collection is divided according to fixed rules; geometric features are quantitative indicators reflecting the spatial shape and size of each local area extracted from multi-dimensional feature vectors; texture features are quantitative indicators reflecting the surface texture of each local area extracted from multi-dimensional feature vectors; spectral features are quantitative indicators reflecting the reflection / absorption characteristics of each local area for different wavelengths of light extracted from multi-dimensional feature vectors; information entropy is a single quantitative value calculated based on the geometric, texture, and spectral features of a single local area through specific logic; normalization is a process of uniformly scaling the information entropy values ​​of all local areas; and the standard entropy value is the final quantitative value obtained after normalizing the information entropy of each local area.

[0071] It is understood that the embodiments of this application divide the surface of the cultural relics into multiple local areas, extract features from each local area and calculate the corresponding information entropy, normalize the information entropy to generate fingerprint features of the cultural relics, and use the unique digital identifier of the fingerprint features of the cultural relics to quickly complete identity verification, thereby improving the accuracy and applicability of the identification of cultural relics in modern museums.

[0072] Specifically, a fixed reference point on the surface of the cultural relic is selected as the origin of the coordinate system to establish a three-dimensional coordinate system. The X and Y axes are parallel to the main plane of the surface of the cultural relic, and the Z axis is perpendicular to the surface. The coordinate unit is uniformly millimeters to ensure that each position on the surface of the cultural relic has a unique coordinate. Along the X and Y axes, a uniform rectangular grid is divided on the surface of the cultural relic with a step size of 10 mm, dividing the surface of the cultural relic into several local areas. Each 10 mm × 10 mm grid is a candidate fingerprint point, and its coordinate range is associated with the corresponding three-dimensional / texture / spectral data position.

[0073] Geometric, textural, and spectral features are extracted from local areas of the museum's collection. Geometric features such as surface curvature, rate of change of normal vector, and shape index are extracted; textural features such as contrast, energy, entropy, and homogeneity of gray-level co-occurrence matrix are extracted; and spectral features such as reflectance of feature bands and spectral absorption depth are extracted. For each feature dimension, the normalized values ​​are arranged into a one-dimensional list in a fixed order. Continuous normalized values ​​are converted into statistical interval frequencies, and frequencies are converted into probabilities. The final entropy value is calculated through probability distribution to quantify feature complexity.

[0074] From the information entropy of all local regions in the collection of cultural relics to be identified, the global minimum and global maximum values ​​are extracted. The original entropy value of the region is subtracted from the global minimum value, and then divided by the difference between the global maximum and global minimum values. The result is mapped to the 0-1 interval, which is the standard entropy value. It is ensured that all standard entropy values ​​are within the 0-1 interval. If a value is found to be outside the range, it is corrected to 0 or 1. The standard entropy values ​​of all regions are organized into an ordered entropy value list according to the fixed numbering order of the local regions. A discrimination threshold is set, and the entropy values ​​of high discrimination regions with variance ≥ the threshold are retained, while the entropy values ​​of low discrimination regions are removed. A fixed feature length is set, and the entropy values ​​after filtering / supplementation are arranged in ascending order of region number by truncating high discrimination features or supplementing suboptimal low discrimination features. They are combined into a fixed-length one-dimensional numerical sequence to verify the uniqueness and completeness of the features. The basic information of the cultural relics is associated and stored as structured data to generate the final fingerprint feature.

[0075] Furthermore, in the embodiments of this application, the formula for calculating information entropy is:

[0076]

[0077] Where X represents a random variable, H(X) represents the information entropy of the random variable X, and p(x) i ) represents the i-th value of the random variable X, and log represents the logarithmic operation.

[0078] It is understood that the embodiments of this application use the Shannon information entropy formula to calculate the information entropy of a local area by combining geometric, textural, and spectral features, thereby quantifying the uncertainty and complexity of the features. Based on the calculated texture entropy and spectral entropy of the cultural relic area, the identity of the cultural relic in the collection can be automatically identified, thereby improving the accuracy and applicability of modern museum cultural relic management.

[0079] Specifically, Shannon entropy is a core indicator in information theory, used to measure the degree of uncertainty of all possible states of a feature. For texture features in a local area, it represents the probability of different texture patterns appearing in that area. The larger the texture feature calculated by substituting it into the formula, the more texture patterns there are and the more dispersed their distribution; the smaller the entropy value, the more uniform the texture.

[0080] For example, the geometric feature probability distribution of damaged areas of cultural relics is more dispersed, corresponding to higher entropy values; the geometric shape of intact areas is more regular, with lower entropy values. Damaged / damaged areas can be quickly located through entropy value differences; diseases on the surface of cultural relics will make local texture features more chaotic, and the corresponding texture entropy value is significantly higher than the uniform texture of intact areas; the spectral features of pollutants on cultural relics vary greatly, and the spectral entropy values ​​of polluted areas are more dispersed, accurately distinguishing the original material area from the polluted / repaired area.

[0081] Furthermore, in the embodiments of this application, constructing a decision tree based on fingerprint point features includes: identifying the standard entropy value of the fingerprint point features; calculating the importance level of the fingerprint point features based on the standard entropy value; determining the root node based on the importance level, determining the internal nodes based on the value range of the fingerprint point features, determining the leaf nodes based on the classification result of the fingerprint point features; and constructing a decision tree based on the root node, internal nodes, and leaf nodes.

[0082] Among them, importance level is an indicator that measures the degree of influence of a single fingerprint feature on the fingerprint classification / recognition result; the root node is the first classification node of the decision tree; the internal nodes are the branch judgment nodes in the middle of the decision tree; and the leaf nodes are the final result nodes of the decision tree.

[0083] It is understood that the embodiments of this application determine the standard entropy value of fingerprint point features through sample data statistics. Based on the standard entropy value, the distinguishability of each feature to fingerprint classification is measured to calculate the importance level. Then, based on the importance level, the feature with the most significant impact on classification is selected as the root node of the decision tree. The internal nodes are divided according to the value range of the fingerprint point features to gradually narrow the classification range. The classification result of the fingerprint point features is used as the leaf node to complete the construction of the decision tree. This allows the decision tree to accurately focus on key fingerprint features. By relying on the objective reference of the standard entropy value, the subjective bias of feature evaluation is avoided, effectively improving the accuracy of fingerprint point classification and recognition. The construction logic is clear and interpretable, and efficient classification and recognition of fingerprint points are achieved.

[0084] Specifically, the C4.5 algorithm is used to construct the decision tree, with information entropy as the splitting criterion. The C4.5 algorithm uses information gain ratio to select the optimal splitting attribute, penalizing attributes with a large number of values, making the decision tree's splitting logic more reasonable and suitable for processing multimodal feature data of cultural relics. The construction of the decision tree uses information entropy as the splitting criterion, determining the node splitting order by calculating the information gain ratio of each feature. The surface curvature feature is selected as the root node due to its significant discriminative power in representing the geometric shape of cultural relics; this selection is based on its good representation ability of the complexity of the surface of cultural relics. The construction of internal nodes comprehensively considers multidimensional information such as texture complexity and spectral features, ensuring that each node can distinguish different categories of fingerprint points to the greatest extent through a recursive feature evaluation method. The information gain ratio of each feature is calculated, and the feature with the largest information gain ratio is selected as the root node. The calculated information gain ratio of the surface curvature feature is 0.85, significantly higher than other features, therefore it is selected as the root node of the decision tree. The splitting threshold of the root node is a curvature value of 0.35, which divides the samples into high-complexity regions with curvature greater than 0.35 and low-complexity regions with curvature less than or equal to 0.35.

[0085] Furthermore, in the embodiments of this application, before identifying the identity of the museum artifacts based on the decision tree, the method further includes: obtaining the cost complexity and Gini coefficient of the decision tree; pruning the decision tree according to the indicated cost complexity; and optimizing the number of fingerprint features of the decision tree based on the Gini coefficient.

[0086] Among them, cost complexity is a comprehensive index that simultaneously relates the complexity of the decision tree structure and the error in identifying cultural relics; the Gini coefficient is an index that measures the classification purity of the cultural relic sample set corresponding to fingerprint features; and the number of fingerprint features is the number of fingerprint features that participate in the construction of the cultural relic identification decision tree.

[0087] It is understood that the embodiments of this application first obtain the cost complexity and Gini coefficient of the decision tree, and then prune the decision tree according to the cost complexity. At the same time, the number of fingerprint point features used by the decision tree is optimized by using the Gini coefficient. This can eliminate redundant and complex branches in the decision tree, and select fingerprint point features that are more suitable for the needs of distinguishing cultural relics. This effectively avoids the problem of the decision tree overfitting cultural relic samples, and makes the structure and feature configuration of the decision tree more in line with the actual identification scenario. It reduces the computational cost and interference of redundant features when identifying cultural relics, and improves the operating efficiency and stability of the decision tree in the identification of cultural relics.

[0088] Specifically, the fingerprint point selection process employs a cost complexity pruning strategy for optimization. Through systematic pruning, redundant nodes that contribute little to the overall classification are removed, while key branches with significant discriminative power are retained. During the selection process, particular attention is paid to the feature distribution of the region represented by each leaf node, prioritizing points with high information entropy and good feature stability as candidate fingerprint points. The selected fingerprint points not only possess high identification value but also maintain good stability under different environmental conditions, ensuring the reliability of the cultural relic identification system. By optimizing the decision tree through cost complexity pruning, high-discriminative fingerprint points are ultimately selected, with an average information entropy of 0.85, demonstrating high identification value.

[0089] For each fingerprint feature of a cultural relic, iterate through all its value division methods, calculate the Gini index of the subset, and then calculate the weighted Gini coefficient of the feature, which is the number of subset samples. Record the minimum Gini coefficient corresponding to each fingerprint feature, which is the Gini index under the optimal division of the feature.

[0090] After each increase, the cost complexity of the entire tree is recalculated. The subtree with the smallest cost complexity gain is found, and the root node of the subtree is pruned into a leaf node. The category of the leaf node is set to the cultural relic identity category with the highest proportion in the subtree. This operation is repeated until the number of nodes in the tree drops to a reasonable range, or the classification error rate after pruning rises to a preset threshold. At this point, pruning is stopped and the current tree structure is retained.

[0091] All fingerprint features of cultural relics are sorted in ascending order according to the minimum Gini coefficient. The top k features are selected from the sorted features, and a simplified decision tree is reconstructed using these features. The classification accuracy is then tested on the cultural relic validation set. The k value corresponding to the highest accuracy on the validation set is found, and this value is used as the optimized number of fingerprint features.

[0092] Furthermore, in the embodiments of this application, identifying the identity of a museum artifact based on a decision tree further includes: obtaining the artifact's circulation type; determining the security level of the artifact based on the circulation type; determining the target number of fingerprint features required for identifying the artifact based on the security level; and identifying the artifact based on the target number of fingerprint features in the decision tree.

[0093] Among them, the cultural relic circulation type is the classification of the source, channel and usage scenario of the cultural relic in the collection before entering the museum and in the process of museum management; the security level is the classification of the risk and importance of cultural relics based on the cultural relic circulation type; the target number is the minimum number of fingerprint features used to identify the identity of cultural relics by decision tree, determined according to the security level.

[0094] It is understood that the embodiments of this application can first obtain the cultural relic circulation type, determine the security level of the cultural relic based on the circulation type, then determine the target number of fingerprint point features required for identification based on the security level, and finally complete the identification of the cultural relic based on the fingerprint point features corresponding to the target number in the decision tree. This can make the cultural relic identification process accurately match the circulation risk and importance of the cultural relic, and significantly improve the adaptability and accuracy of identification of different types of cultural relic.

[0095] Specifically, the types of cultural relics circulation in the museum's collection are internal circulation, exhibition outside the museum, and inventory in the warehouse. Among them, exhibition outside the museum is the highest security level, internal circulation is the medium security level, and inventory in the warehouse is the standard security level.

[0096] As shown in Table 1, the number of identification points is determined based on different types of cultural relics, materials, circulation operations, and security levels. The circulation type of the cultural relics in the collection is determined according to the current business scenario corresponding to the business type in the table. If a precious cultural relic is under a prohibited exit / exit management scenario, its in-museum circulation corresponds to a medium security type, and the target number of fingerprint feature identification points must account for more than 50%. Out-of-museum circulation corresponds to the highest security type, and the target number of fingerprint feature identification points must account for more than 85%. Warehouse inventory corresponds to a routine security type, and the target number of fingerprint feature identification points is 1. If a precious cultural relic is a first-class cultural relic, its in-museum circulation corresponds to a medium security type, and the target number of fingerprint feature identification points must account for more than 50%. Out-of-museum circulation corresponds to the highest security type, and the target number of fingerprint feature identification points must account for more than 1. 75%, the security type corresponding to the warehouse inventory is regular, and the target number of fingerprint points required for identification is 1; if the precious cultural relic is a second-level cultural relic, the security type corresponding to its circulation within the museum is medium, and the target number of fingerprint points required for identification needs to be greater than or equal to 20%, the security type corresponding to its circulation outside the museum is the highest, and the target number of fingerprint points required for identification needs to be greater than or equal to 50%, the security type corresponding to the warehouse inventory is regular, and the target number of fingerprint points required for identification is 1; if the precious cultural relic is a second-level cultural relic, the security type corresponding to its circulation within the museum is regular, and the target number of fingerprint points required for identification is 2, the security type corresponding to its circulation outside the museum is medium, and the target number of fingerprint points required for identification needs to be greater than or equal to 20%, the security type corresponding to the warehouse inventory is regular, and the target number of fingerprint points required for identification is 1.

[0097] Table 1. Location and Modal Requirements for Digital Fingerprint Extraction and Identification

[0098]

[0099] This application embodiment uses multi-dimensional feature vectors to generate fingerprint features of museum artifacts and constructs a decision tree to identify the identity of museum artifacts. It can extract high-discrimination indicators from multi-dimensional feature vectors and realize multi-feature cross-validation through hierarchical logical judgment of the decision tree. This provides clear data support for the identification conclusion, greatly reduces the false positive rate and false negative rate, and ensures the authenticity and accuracy of the identity of museum artifacts.

[0100] To better understand the solution of this application, the method or execution process for identifying museum artifacts is described below through a specific embodiment:

[0101] like Figure 2 The diagram shown is a technical flowchart of the method for identifying cultural relics in the museum's collection.

[0102] In step 1: Multimodal data acquisition of museum collection.

[0103] 1.1 Three-dimensional geometric data acquisition: High-precision point cloud data of the artifact surface was acquired using a structured light 3D scanner with a resolution of 0.05mm, and geometric features such as surface curvature and normal vector distribution were collected.

[0104] 1.2 Surface texture data acquisition: Multi-angle RGB images of the artifact surface are acquired using a high-resolution digital camera, and a ring light source is used to eliminate the influence of shadows.

[0105] 1.3 Spectral feature acquisition: Hyperspectral imaging system (spectral range 400-2500nm, spectral resolution ≤5nm) was used to acquire spectral reflectance data of the artifact surface.

[0106] 1.4 Environmental Parameter Recording: Collect environmental parameters such as temperature, humidity, and light intensity simultaneously to establish a data collection environment archive.

[0107] In step 2: multimodal data preprocessing and feature fusion.

[0108] 2.1 Data Registration and Alignment: The ICP algorithm is used to achieve accurate registration of point cloud data and texture images, with a registration error of less than 0.1 mm.

[0109] 2.2 Noise filtering: Anisotropic diffusion filtering is used to remove point cloud noise, and nonlocal mean filtering is used to process image noise.

[0110] 2.3 Feature Standardization: Geometric, textural, and spectral features are Z-score standardized to eliminate the influence of dimensions.

[0111] 2.4 Multi-source data fusion: Different modal features are integrated into a unified multi-dimensional feature vector through feature-level fusion algorithms, and principal component analysis is used to reduce the feature dimensionality.

[0112] In step 3: Fingerprint point feature quantization based on information entropy.

[0113] 3.1 Local area division: The surface of the cultural relic is divided into several local areas (10mm×10mm), and each area is used as a candidate fingerprint point.

[0114] 3.2 Multimodal Feature Extraction: Extract the following features for each local region:

[0115] Geometric features: surface curvature, rate of change of normal vector, shape index

[0116] Texture features: contrast, energy, entropy, and homogeneity of the gray-level co-occurrence matrix.

[0117] Spectral characteristics: reflectance in characteristic bands, spectral absorption depth

[0118] 3.3 Information Entropy Calculation: The information entropy of each feature point is calculated using the Shannon entropy formula:

[0119]

[0120] in This represents the probability distribution of the occurrence of eigenvalues.

[0121] 3.4 Entropy normalization: The information entropy of all feature points is normalized by min-max to obtain the standard entropy value in the range of [0,1].

[0122] In step 4: fingerprint point decision tree construction and optimization.

[0123] 4.1 Feature Importance Assessment: The importance of each feature is assessed based on information gain.

[0124]

[0125] 4.2 Decision Tree Construction: The C4.5 algorithm is used to construct the decision tree, with information entropy as the splitting criterion.

[0126] The root node selects the feature with the largest information gain.

[0127] Internal nodes are divided according to the range of eigenvalues.

[0128] Leaf nodes represent the final fingerprint classification results.

[0129] 4.3 Decision Tree Pruning: Cost Complexity Pruning (CCP) is used to prevent overfitting and retain the most discriminative feature points.

[0130] 4.4 Determining the Number of Fingerprint Points: The optimal number of fingerprint points is determined based on the Gini coefficient to ensure a balance between identification accuracy and efficiency.

[0131] In step 5: the strategy for identifying cultural relics is dynamically generated.

[0132] 5.1 Business Scenario Analysis: Determine the security level based on the document flow type:

[0133] In-house circulation: Medium security level

[0134] Exhibitions outside the venue: Highest security level

[0135] Warehouse inventory: Standard security level

[0136] 5.2 Cultural Relics Grading Assessment: Combining the intrinsic value of the cultural relic (precious cultural relic, general cultural relic)

[0137] 5.3 Authentication Strategy Generation:

[0138] The number of identification points is determined based on the different types of cultural relics, materials, circulation business, and security levels.

[0139] Highest security level: Select the top 50% of fingerprints with high information entropy, and a confidence threshold ≥ 95%.

[0140] Medium security level: Select the top 20% of fingerprints with high information entropy and a confidence threshold ≥ 90%.

[0141] Standard security level: Select the fingerprint point with the highest information entropy.

[0142] 5.4 Dynamic adjustment mechanism: Based on real-time identification results, the fingerprint point selection strategy and confidence threshold are dynamically adjusted.

[0143] In step 6: Optimization and updating of identity authentication strategy.

[0144] 6.1 Performance Evaluation: Calculate indicators such as accuracy, false alarm rate, and recall rate.

[0145] 6.2 Model Update: Incremental learning is adopted to continuously optimize the decision tree model.

[0146] 6.3 Parameter Adjustment: Dynamically adjust feature weights and decision thresholds based on usage.

[0147] 6.4 Strategy Upgrade: Regularly update the feature library and algorithm model to maintain technological advancement.

[0148] In summary, the museum artifact identification system proposed in this application acquires multimodal data of museum artifacts, performs preprocessing on the acquired multimodal data, generates corresponding multidimensional feature vectors based on the preprocessed multimodal data, extracts fingerprint features of the museum artifacts based on the multidimensional feature vectors, constructs a decision tree model based on these fingerprint features, and finally achieves accurate identification of the museum artifacts using the constructed decision tree. Through multimodal feature collaborative analysis, the reliability of artifact identification is improved, and the identification process is automated and intelligent, significantly reducing the manpower and time costs of artifact preservation. The system can automatically adjust the identification intensity and dynamically optimize the identification strategy according to the actual application scenario, ensuring the security of artifact management in modern museums, improving the efficiency of artifact management, and providing the most suitable identification solution for different application scenarios.

[0149] Next, the system for identifying museum artifacts according to embodiments of this application is described with reference to the accompanying drawings.

[0150] Figure 3 This is a schematic diagram of the museum artifact identification system 300 according to an embodiment of this application.

[0151] like Figure 3 As shown, the artifact identification system 300 includes: an acquisition module 301, a preprocessing module 302, and a generation module 303.

[0152] The acquisition module 301 is used to acquire multimodal data of the museum’s collection of cultural relics; the preprocessing module 302 is used to preprocess the multimodal data and generate multidimensional feature vectors based on the preprocessed data; the generation module 303 is used to generate fingerprint point features of the museum’s collection of cultural relics based on the multidimensional feature vectors, construct a decision tree based on the fingerprint point features, and identify the identity of the museum’s collection of cultural relics based on the decision tree.

[0153] Furthermore, in embodiments of this application, the multimodal data includes multiple types of three-dimensional geometric data, surface texture data, spectral reflectance data, and environmental data.

[0154] Furthermore, in the embodiments of this application, the preprocessing methods include at least one of data registration and alignment processing, noise filtering processing, and normalization processing.

[0155] Furthermore, in the embodiments of this application, the generation module 303 is further used to: divide the surface of the museum artifact into multiple local regions; extract the geometric features, texture features and spectral features of each local region from the multidimensional feature vector, calculate the information entropy of the corresponding local region based on the geometric features, texture features and spectral features; normalize the information entropy of each local region to obtain the standard entropy value of the local region, and generate the fingerprint point features of the museum artifact based on the standard entropy value of the local region.

[0156] Furthermore, in the embodiments of this application, the formula for calculating information entropy is:

[0157]

[0158] Where X represents a random variable, H(X) represents the information entropy of the random variable X, and p(x) i ) represents the i-th value of the random variable X, and log represents the logarithmic operation.

[0159] Furthermore, in the embodiments of this application, the generation module 303 is further used to: identify the standard entropy value of fingerprint point features; calculate the importance level of fingerprint point features based on the standard entropy value; determine the root node based on the importance level, determine the internal nodes based on the value range of fingerprint point features, determine the leaf nodes based on the classification result of fingerprint point features; and construct a decision tree based on the root node, internal nodes, and leaf nodes.

[0160] Furthermore, in the embodiments of this application, the museum artifact identification system 300 further includes: an optimization module, used to obtain the cost complexity and Gini coefficient of the decision tree before identifying the museum artifact based on the decision tree; to prune the decision tree according to the indicated cost complexity; and to optimize the number of fingerprint point features of the decision tree according to the Gini coefficient.

[0161] Furthermore, in the embodiments of this application, the generation module 303 is further used to: obtain the cultural relic circulation type of the collection; determine the security level of the collection based on the cultural relic circulation type; determine the target number of fingerprint point features required for the identification of the collection based on the security level; and identify the identity of the collection based on the target number of fingerprint point features in the decision tree.

[0162] It should be noted that the foregoing explanation of the embodiment of the method for identifying the identity of cultural relics in the collection also applies to the system for identifying the identity of cultural relics in the collection in this embodiment, and will not be repeated here.

[0163] In summary, the museum artifact identification system proposed in this application acquires multimodal data of museum artifacts, performs preprocessing on the acquired multimodal data, generates corresponding multidimensional feature vectors based on the preprocessed multimodal data, extracts fingerprint features of the museum artifacts based on the multidimensional feature vectors, constructs a decision tree model based on these fingerprint features, and finally achieves accurate identification of the museum artifacts using the constructed decision tree. Through multimodal feature collaborative analysis, the reliability of artifact identification is improved, and the identification process is automated and intelligent, significantly reducing the manpower and time costs of artifact preservation. The system can automatically adjust the identification intensity and dynamically optimize the identification strategy according to the actual application scenario, ensuring the security of artifact management in modern museums, improving the efficiency of artifact management, and providing the most suitable identification solution for different application scenarios.

[0164] This application also provides a computer program product, including a computer program or instructions, which, when executed, implement the above-mentioned method for identifying the identity of cultural relics in museum collections.

[0165] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0166] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0167] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0168] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0169] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

Claims

1. A method for identifying the identity of cultural relics in a museum collection, characterized in that, Includes the following steps: Acquire multimodal data of museum collections; The multimodal data is preprocessed, and a multidimensional feature vector is generated based on the preprocessed data; The fingerprint features of the cultural relics in the collection are generated based on the multidimensional feature vector, a decision tree is constructed based on the fingerprint features, and the identity of the cultural relics in the collection is identified based on the decision tree. The step of generating fingerprint features of the museum artifacts based on the multidimensional feature vector includes: dividing the surface of the museum artifacts into multiple local regions; extracting geometric features, texture features, and spectral features of each local region from the multidimensional feature vector; calculating the information entropy of the corresponding local region based on the geometric features, texture features, and spectral features; normalizing the information entropy of each local region to obtain the standard entropy value of the local region; and generating fingerprint features of the museum artifacts based on the standard entropy value of the local region. The step of constructing a decision tree based on the fingerprint point features includes: identifying the standard entropy value of the fingerprint point features; calculating the importance level of the fingerprint point features based on the standard entropy value; determining the root node based on the importance level, determining the internal nodes based on the value range of the fingerprint point features, and determining the leaf nodes based on the classification result of the fingerprint point features; and constructing the decision tree based on the root node, the internal nodes, and the leaf nodes. The step of identifying the artifacts in the collection based on the decision tree further includes: obtaining the circulation type of the artifacts in the collection; determining the security level of the artifacts in the collection based on the circulation type; determining the target number of fingerprint features required for identifying the artifacts in the collection based on the security level; and identifying the artifacts in the collection based on the target number of fingerprint features in the decision tree.

2. The method for identifying museum artifacts according to claim 1, characterized in that, The preprocessing methods include at least one of data registration and alignment processing, noise filtering processing, and normalization processing.

3. The method for identifying museum artifacts according to claim 1, characterized in that, The formula for calculating the information entropy is: Where X represents a random variable, H(X) represents the information entropy of the random variable X, and p(x) i ) represents the probability of the random variable X taking the i-th eigenvalue, and log represents the logarithmic operation.

4. The method for identifying museum artifacts according to claim 1, characterized in that, Before identifying the artifacts in the collection based on the decision tree, the process also includes: Obtain the cost complexity and Gini coefficient of the decision tree; The decision tree is pruned according to the cost complexity. The number of fingerprint features in the decision tree is optimized based on the Gini coefficient.

5. A system for identifying cultural relics in a museum collection, characterized in that, include: The acquisition module is used to acquire multimodal data of the museum's collection. The preprocessing module is used to preprocess the multimodal data and generate multidimensional feature vectors based on the preprocessed data. The generation module is used to generate fingerprint point features of the museum artifacts based on the multidimensional feature vector, construct a decision tree based on the fingerprint point features, and identify the identity of the museum artifacts based on the decision tree. The generation module is further configured to: divide the surface of the artifact into multiple local regions; extract the geometric features, texture features, and spectral features of each local region from the multidimensional feature vector; calculate the information entropy of the corresponding local region based on the geometric features, texture features, and spectral features; normalize the information entropy of each local region to obtain the standard entropy value of the local region; and generate the fingerprint feature of the artifact based on the standard entropy value of the local region. The generation module is further configured to: identify the standard entropy value of the fingerprint point feature; and calculate the importance level of the fingerprint point feature based on the standard entropy value. The root node is determined based on the importance level, the internal nodes are determined based on the value range of the fingerprint point features, and the leaf nodes are determined based on the classification results of the fingerprint point features. The decision tree is constructed based on the root node, the internal nodes, and the leaf nodes; The system for identifying cultural relics in a museum collection also includes: an identification module, used to obtain the circulation type of the cultural relics in the museum collection; determine the security level of the cultural relics in the museum collection based on the circulation type of the cultural relics; determine the target number of fingerprint point features required for the identification of the cultural relics in the museum collection based on the security level; and identify the identity of the cultural relics in the museum collection based on the target number of fingerprint point features in the decision tree.

6. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed, they implement the method for identifying cultural relics in a collection as described in any one of claims 1-4.

Images