Cultural relic digital fingerprint identification method and system based on geometric texture and spectral information

Abstract

The invention discloses a cultural relic digital fingerprint identification method based on geometric texture and spectral information, and the method comprises the steps: obtaining multi-source data, carrying out the texture mapping, and forming a three-dimensional grid model with texture data; curvature values and topological features of all vertexes of the three-dimensional grid model are calculated, feature fusion is carried out on the basis of local saliency adjustment feature weights, and fingerprint point location screening is completed; based on the screened fingerprint point locations, a significant plane is constructed through a bit plane, and texture features of the cultural relics are extracted based on an improved histogram and a gray-level co-occurrence matrix; extracting characteristic wave bands of the cultural relics based on the screened fingerprint point locations in combination with the hyperspectral data; and multi-modal fusion identification is carried out based on features of geometry, texture and spectral information. According to the method, the geometric structure, the texture features and the hyperspectral information are fused, multi-mode collaborative cultural relic identity high-precision identification is achieved, and the method has the advantages of being high in accuracy, high in robustness, good in interpretability and capable of adapting to complex environments.

Description

Technical Field

[0001] The present invention relates to the technical field of digital fingerprint identification of cultural relics, and in particular to a feature extraction technology and a cultural relic identity identification method that integrate geometric texture and hyperspectral data. Background Art

[0002] As important historical and cultural resources and carriers of the five-thousand-year Chinese civilization, cultural relics carry rich cultural information and historical memories. They are not only artworks but also witnesses to the development process of human civilization. With the increasing frequency of cultural exchanges, the lending and exhibition of cultural relics in museums have become important forms and dissemination channels for exchanges and cooperation between museums and for promoting cultural diversity. However, the process of lending and returning cultural relics usually involves cross-regional and cross-national movements, and the management chain is complex and diverse. Due to insufficient supervision, imperfect identification technology, and the complexity of the external environment, it is easy to have situations such as counterfeit cultural relics or genuine ones being swapped, which affects the safety and protection of cultural relics. Currently, in the process of lending and returning cultural relics in many museums, the identification of the identity of cultural relics still mainly relies on traditional empirical identification and paper records, and these means are often difficult to cope with the challenges of high-quality fakes and technical swapping. In addition, cultural relics may be affected by the environment during the lending process, and the changes that occur when they are returned increase the difficulty of identification. How to ensure the accuracy of the identity identification of cultural relics during the lending and returning process through modern technical means has become a key issue in the management of museum cultural relics and an important topic that urgently needs to be solved in the current cultural relic protection and safety management.

[0003] Current cultural relic identity identification methods usually focus on single-modal means such as geometric structure analysis, texture feature matching, or hyperspectral data mining. However, in practical applications, each of these three methods has limitations: geometric features lack stability under complex surface deformations or perspective changes, texture information is easily distorted by light and wear, and although hyperspectral analysis has the advantage of material differentiation, it has a high computational cost and is easily affected by noise. It can be seen that the lack of a multi-modal fusion strategy makes it difficult for existing methods to take into account spatial structure, surface texture, and material composition, and there is still a large room for improvement in the overall recognition accuracy and robustness. Summary of the Invention

[0004] To overcome the deficiencies of the above-mentioned prior art, the present invention provides a method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information. By integrating geometric structure, texture features, and hyperspectral information, it realizes high-precision identification of cultural relic identity with multi-modal collaboration, and has the advantages of high accuracy, strong robustness, good interpretability, and adaptability to complex environments.

[0005] According to one aspect of the specification of the present invention, there is provided a method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information, including: Obtain multi-source data and perform texture mapping to form a three-dimensional mesh model with texture data; Calculate the curvature values and topological features of each vertex of the three-dimensional mesh model, perform feature fusion by adjusting feature weights based on local saliency, and complete the screening of fingerprint points; Based on the screened fingerprint points, construct a significant plane through bit planes, and extract the texture features of the cultural relics based on the improved histogram and gray-level co-occurrence matrix; Based on the screened fingerprint points, combine hyperspectral data to extract the characteristic bands of the cultural relics; Perform multi-modal fusion identification based on the features of geometry, texture, and spectral information.

[0006] As a further technical solution, obtaining multi-source data and performing texture mapping also includes: Collect the high-density geometric grid of the cultural relics to form a three-dimensional mesh model; Collect high-definition texture maps of the cultural relics through multi-angle photography; After processing the high-definition texture maps, map them to the three-dimensional mesh model to form a three-dimensional mesh model with texture data.

[0007] As a further technical solution, processing the high-definition texture maps also includes: Extract the main body area of the cultural relics from multi-view images, eliminate the texture matching errors caused by background interference, and perform three-dimensional reconstruction to generate dense point clouds and meshes; Adopt a two-stage matching strategy of rough matching and fine matching to obtain the transformation matrix; Based on the transformation matrix, accurately map the texture map to the surface of the three-dimensional mesh model.

[0008] As a further technical solution, the screening of fingerprint points also includes: Construct an information entropy model through the included angle of neighborhood normal vectors, perform curvature analysis based on information entropy adaptive neighborhood to describe the local geometric complexity of the cultural relics; Perform topological structure optimization, calculate the multi-scale clustering coefficient, and ensure the rationality of the spatial distribution of points; Fuse geometric and topological information, dynamically adjust the contribution weights, and screen out the fingerprint points of the cultural relics.

[0009] As a further technical solution, after obtaining the screened fingerprint points, it also includes: Map the screened fingerprint points to the two-dimensional texture map through the collinearity equation to obtain the two-dimensional image coordinates of the fingerprint points; Perform bit-plane slicing on the image to form 8 binary images, calculate the variance, kurtosis, and average entropy values of different bit planes, and based on visual discrimination and the established three-plane evaluation indexes, select the characteristic plane and the original image to extract texture features together; Based on the Lab color space, the Gaussian mixture model is used to classify cultural relic images to obtain a clustering histogram; Based on the gray-level co-occurrence matrix, the relative positions of pixel pairs in the image are analyzed from multiple relative directions to extract the texture features of cultural relics; Based on the two complementary texture feature values of the calculated color histogram and gray-level co-occurrence matrix, standardization and normalization processing are performed and integrated into a unified feature vector.

[0010] As a further technical solution, after obtaining the selected fingerprint points, it further includes: Collecting hyperspectral data and performing radiometric correction and data matching; Using competitive adaptive reweighting to perform preliminary selection of characteristic bands; Using the genetic algorithm to globally optimize the band selection to obtain the final characteristic bands.

[0011] As a further technical solution, multi-modal fusion identification based on the characteristics of geometry, texture, and spectral information further includes: Unifying the encoding of local features of the three types of modalities; According to the attention mechanism, automatically learning the importance weights of features in different regions of different modalities, and realizing dynamic perception and adaptive weighting of the geometric structure stability, texture detail expressiveness, and spectral material difference; Performing multi-modal feature fusion identification based on cosine similarity.

[0012] According to one aspect of the specification of the present invention, there is provided a cultural relic digital fingerprint identification system based on geometric texture and spectral information, including: A data acquisition module for acquiring multi-source data and performing texture mapping to form a three-dimensional grid model with texture data; A fingerprint point selection module for calculating the curvature values and topological features of each vertex of the three-dimensional grid model, adjusting the feature weights based on local saliency for feature fusion, and completing the selection of fingerprint points; A texture feature extraction module for, based on the selected fingerprint points, constructing a significant plane through bit planes and extracting the texture features of cultural relics based on an improved histogram and gray-level co-occurrence matrix; A hyperspectral feature extraction module for, based on the selected fingerprint points, combining hyperspectral data to extract the characteristic bands of cultural relics; A feature fusion module for performing multi-modal fusion identification based on the characteristics of geometry, texture, and spectral information.

[0013] According to one aspect of the specification of the present invention, there is provided an artifact digital fingerprint identification device based on geometric texture and spectral information, including a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the artifact digital fingerprint identification method based on geometric texture and spectral information.

[0014] According to one aspect of the specification of the present invention, there is provided a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the artifact digital fingerprint identification method based on geometric texture and spectral information.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1) The present invention combines structured light scanning and photogrammetry to generate a three-dimensional model of the artifact with high-precision geometric reconstruction and high-definition texture mapping. From the two perspectives of adaptive curvature based on information entropy and topological features based on the atlas, the geometric features of the collected artifact model are calculated. Based on the geometric features, stable and representative artifact fingerprint points are selected and identified. An effective solution is provided for the local details and overall structural characteristics of the artifact, laying a solid foundation for subsequent artifact identification and protection.

[0016] 2) The present invention introduces a bit-plane hierarchical processing method in texture feature extraction, splitting the grayscale image into multiple bit planes, especially highlighting the key description ability of the high-order bit planes for the object shape and texture structure, thereby enhancing the expressiveness of texture features under weak differences. At the same time, combining the grayscale histogram and grayscale co-occurrence matrix of the local area, texture information is extracted from the two levels of statistical distribution and spatial dependence relationship respectively, realizing a fine description of the surface pattern and structural changes of the artifact, effectively improving the discrimination and robustness of texture features, and providing a stable and reliable texture basis for identity identification.

[0017] 3) The present invention introduces two feature selection methods, CARS (Competitive Adaptive Reweighted Sampling) and GA (Genetic Algorithm), in hyperspectral data processing, effectively screening out the significant bands that are most discriminative for artifact identity, avoiding redundant information interference and the curse of dimensionality problem.

[0018] 4) The present invention studies the design of a feature fusion network based on the attention mechanism for the feature differences of different objects, automatically allocating weights on the basis of ensuring the independent expression of various features, and finally constructing a fused digital fingerprint, providing double support in theory and practice for the multi-modal fusion identification of artifact digital fingerprints.

[0019] 5) By integrating geometric structures, texture features, and hyperspectral information, the present invention achieves high-precision identification of the identity of cultural relics through multi-modal collaboration, and has the advantages of high accuracy, strong robustness, good interpretability, and adaptability to complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0021] Figure 1 It is a schematic diagram of the technical route of the embodiment of the present invention.

[0022] Figure 2 It is a schematic diagram of the scanner parameters of the embodiment of the present invention.

[0023] Figure 3 It is a schematic diagram of the camera parameters of the embodiment of the present invention.

[0024] Figure 4 It is a schematic diagram of the removal of background noise in the embodiment of the present invention.

[0025] Figure 5 It is a schematic diagram of the 3D reconstruction in the embodiment of the present invention.

[0026] Figure 6 It is a schematic diagram of the generation of a high-definition 3D cultural relic model in the embodiment of the present invention.

[0027] Figure 7 It is a schematic diagram of the mapping results of some fingerprint points in the embodiment of the present invention.

[0028] Figure 8 It is a schematic diagram of the bit plane in the embodiment of the present invention.

[0029] Figure 9 It is a GMM clustering histogram in the embodiment of the present invention.

[0030] Figure 10 It is a schematic diagram of the GLCM features in the embodiment of the present invention.

[0031] Figure 11 It is a schematic diagram of the hyperspectral data acquisition and denoising process in the embodiment of the present invention.

[0032] Figure 12 It is a spectral curve fitting diagram in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] The present invention discloses a method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information, which can perform multimodal matching and stacking on cultural relics and realize high-precision identity identification of cultural relics. The present invention focuses on the accurate identification and comparison of digital fingerprints of cultural relics, constructs a multimodal information fusion route based on geometric, texture and hyperspectral information, and systematically solves the difference and fusion problems of different types of features. In terms of geometric feature extraction, the normal vector information entropy is first introduced to calculate the curvature of each vertex of the three-dimensional model, and from a morphological perspective, a topological map is constructed and the multi-scale clustering coefficient is calculated to capture the stable structural information of the cultural relics. The feature weight is adjusted based on local significance to finally realize the screening and identification of geometric fingerprint points. Based on high-resolution texture mapping, the texture feature information of the cultural relics is obtained from a statistical perspective around the fingerprint points, and the digital fingerprint point identification based on texture information is further completed by comparing the microscopic details in the texture neighborhood. The present invention uses a competitive adaptive reweighted equal band selection method to retain the most significant feature bands of the cultural relics, and realizes the detail comparison of the cultural relics to be compared and the reference cultural relics at the spectral level. In order to achieve the coordinated expression of different types of data, the method of the present invention realizes dynamic perception and adaptive weighting of geometric structure stability, texture detail expression and spectral material differences, ultimately improving the accuracy and generalization ability of cultural relics fingerprint identification.

[0034] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In addition, the technical features in the various embodiments or single embodiments provided by the present invention are arbitrarily combined with each other to form a new technical solution. This combination is not restricted by the sequence of steps and / or the structural composition mode, but must be based on the ability of ordinary technicians in this field to achieve. When the combination of technical solutions is contradictory or cannot be achieved, it should be considered that this combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0035] The embodiment of the present invention proposes a method for identifying cultural relics digital fingerprints based on geometric texture and spectral information, which includes the steps of data collection, geometric alignment, texture feature extraction, hyperspectral feature selection and feature fusion, wherein the cultural relics data are preprocessed by geometric alignment to extract key feature points; texture features are extracted by bit plane stratification, gray level co-occurrence matrix and histogram method; significant band information is screened by CARS and GA methods, and geometric, texture and spectral information are fused, and finally the identity of the cultural relics is identified by similarity measurement.

[0036] The embodiment of the present invention proposes a method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information, and its technical route is as follows: Figure 1 As shown, the following steps are included: Step 1) Acquire multi-source data and perform texture mapping, including 3D mesh model acquisition, texture data acquisition, and texture mapping of the 3D model.

[0037] This step mainly involves the generation of a 3D model with high-definition textures. During the data acquisition phase, a spiral progressive scanning path is preset to increase the overlap rate of adjacent scanning bands and ensure the integrity of complex surface coverage. The JMStudio software provided by JiMuYida is used for processing to complete the denoising optimization, data fusion and texture mapping of the original point cloud model of the cultural relic. At the same time, high-definition images of the cultural relic from multiple perspectives are systematically shot in the same environment to build evenly distributed photography stations to ensure the acquisition of texture information within a high dynamic range.

[0038] The following sub-steps can be used for implementation: Step 1.1, use the Mole handheld medium format 3D scanner (such as Figure 2 As shown in the figure, the 3D mesh model is collected to restore the cultural relics with high precision; Step 1.2, use a Sony Alpha 7 high-resolution SLR camera (such as Figure 3 As shown in the figure), the cultural relics are photographed from multiple angles to obtain high-definition texture maps; Step 1.3, texture mapping, first extract the main area of the cultural relic from the multi-view image to eliminate the texture matching error caused by background interference, such as Figure 4 As shown. Secondly, 3D reconstruction is performed to generate dense point clouds and meshes, as shown Figure 5 and 6 Then, a two-stage matching strategy of coarse matching and fine matching is adopted to obtain the transformation matrix. Finally, the texture map is accurately mapped to the scanned model surface. The results of some fingerprint point mapping are shown in Figure 7 shown.

[0039] Specifically, the background of the texture map is removed, and the standardized processing flow of Agisoft Metashape Professional is used to generate dense point clouds and triangular meshes, and the internal and external parameters of each image are obtained. In order to achieve spatial alignment between the photogrammetric model and the high-precision geometric scan model, the two models are mainly coarsely and finely registered so that they are in the same coordinate system. Based on the rigid body transformation matrix T solved in the registration stage, the multi-view texture images are fused to the surface of the high-precision scanned model.

[0040] Step 2) Fingerprint point identification method based on multi-scale geometric features and topological optimization, including calculating the curvature values and topological features of each vertex of the three-dimensional model of the cultural relic, adjusting the feature weights based on local saliency for feature fusion, and completing the screening of fingerprint points. First, in this step, an information entropy model is constructed through the angle between neighborhood normal vectors, and curvature analysis is carried out based on the information entropy adaptive neighborhood to characterize the local geometric complexity of the cultural relic. Then, topological structure optimization is performed, and the multi-scale clustering coefficient is calculated to ensure the rationality of the spatial distribution of points and avoid excessive concentration of points. Then, by integrating geometric and topological information, the contribution weights are dynamically adjusted to screen out the fingerprint points of the cultural relic. Finally, the stability and consistency of the fingerprint points are analyzed, and the cultural relic is identified based on geometric features.

[0041] Specific implementation can be achieved by the following sub-steps: Step 2.1, calculate the normal of each vertex of the three-dimensional model, use the histogram representation of the angle difference between neighborhood normal vectors to represent the distribution of normal vectors, calculate the entropy value based on its probability, and use the calculated normal vector information entropy as a weight factor to dynamically adjust the neighborhood radius of curvature calculation according to the surface complexity of different regions; Step 2.2, represent the three-dimensional grid one-to-one as a topological graph, introduce morphological optimization to process the topological graph, and calculate the multi-scale clustering coefficient to identify the points with relatively complex topological features in the cultural relic model; Step 2.3, normalize the curvature feature values and topological feature values, analyze each vertex on the three-dimensional grid model of the cultural relic, measure its relative importance in the local area, calculate the local saliency, and calculate the weight parameters of the features of each grid vertex to form the final fusion feature; Step 2.4, screen the fingerprint points and perform geometric identification of the fingerprint points by calculating the stability of the multi-scale geometric fusion feature calculation method proposed from two perspectives of local consistency analysis and repeated scan data evaluation under different acquisition conditions.

[0042] Step 3) Digital fingerprint point identification based on texture information, including constructing a significant plane based on bit planes, extracting texture features of the cultural relic based on an improved histogram and gray-level co-occurrence matrix, and performing fingerprint point identification.

[0043] This step mainly constructs a significant plane through bit plane layering, extracts texture features based on the method of an improved histogram and gray-level co-occurrence matrix, and further performs fingerprint point identification of the cultural relic based on texture features.

[0044] Specific implementation can be achieved by the following sub-steps: Step 3.1: For the fingerprint points selected based on the 3D scanning model, obtain the 2D image coordinates of the fingerprint points through steps such as coordinate system transformation, projection transformation, distortion correction, and obtaining 2D pixel coordinates. Step 3.2: Perform bit-plane slicing on the image to form 8 binary images. Calculate the average information entropy, variance, and kurtosis features of each plane image for the eight bit-planes of an image. Select the higher-order planes 5 - 8 and the original image together for texture feature extraction. The bit-planes are as Figure 8 shown.

[0045] Step 3.3: Based on the Lab color space, use the Gaussian Mixture Model (GMM) to classify the cultural relic images, and use the Expectation-Maximization algorithm for parameter estimation. The iterative process mainly includes two parts: the Expectation step (E-step) and the Maximization step (M-step). The GMM clustering histogram is as Figure 9 shown.

[0046] Step 3.4: Analyze the relative positions of pixel pairs in the image based on the Gray-Level Co-Occurrence Matrix (GLCM) from multiple relative directions including the horizontal direction, vertical direction, diagonal direction, and anti-diagonal direction, namely 0°, 90°, 45°, and 135°. Calculate the image texture features including Contrast, Homogeneity, Energy, and Correlation based on the gray-level co-occurrence matrices in the four directions. As Figure 10 shown.

[0047] Step 3.5: Based on the calculated two complementary texture feature values of the color histogram and the gray-level co-occurrence matrix, perform standardization and normalization processing on them, and integrate them into a unified feature vector for subsequent similarity calculation and analysis. Through this comprehensive feature vector, while ensuring the integrity and richness of the image texture information, it can effectively identify the fine texture of cultural relics.

[0048] Step 4) Digital fingerprint point identification based on spectral information, including hyperspectral data acquisition, radiometric correction and data matching, characteristic band selection, and spectral identification of the fingerprint points of cultural relics based on the hyperspectral characteristic bands of cultural relics.

[0049] Specific implementation can be achieved through the following sub-steps: Step 4.1, Use the Hyperspec® VNIR-SWIR Co-aligned airborne hyperspectral imager to collect data. The experimental data is collected in a darkroom to avoid the influence of light. At the same time, to ensure that the halogen lamps provide sufficient and uniform light sources, two halogen lamps are used in the research to provide uniform light for the cultural relics from both sides. Since the imager is equipped with a fixed-focus lens, to obtain high-quality and clear images, it is necessary to ensure that the imager is as close as possible to the surface of the cultural relic. To maximize the details of the image, the scanning angle of the instrument should be finely set according to the specific size of the cultural relic, and the cultural relic should be placed in the central area and kept at a similar distance from the edge of the scanned image to prevent large distortions at the image edge. The scanning angle range for each image collection is from 10° to -10°, a total of 20°, and the collection time is about 1 minute. To ensure that the data can fully cover the surface of the cultural relic, the research continuously adjusts the position and angle of the cultural relic during the collection process to ensure that all details of the cultural relic are fully collected.

[0050] Step 4.2, Collect the all-black image obtained by covering the lens cap and the standard whiteboard reflection image. Based on the original hyperspectral data Rraw, the dark current data Rdark, and the standard reflection whiteboard data Rwhite, calculate the reflectance. At the same time, use the cubic spline interpolation method to smooth the spectral data at the junction, which can effectively eliminate the offset difference between the visible light band and the near-infrared band at the junction. Hyperspectral data collection and denoising are as Figure 11 shown.

[0051] Step 4.3, Use competitive adaptive reweighting to make a preliminary selection of the characteristic bands. Specifically, based on partial least squares regression (PLS), through its own iterative optimization mechanism, hundreds of spectral bands are gradually screened and compressed, and finally the characteristic bands with the highest information content are retained.

[0052] Step 4.4, Use the genetic algorithm to globally optimize the band selection. Utilize the genetic algorithm to make up for the deficiency that step 4.3 may fall into a local optimum, combine the global search ability of the genetic algorithm and the iterative optimization strategy of CARS, so that the finally selected characteristic bands are more stable and reliable.

[0053] Step 4.5, For the hyperspectral data of the cultural relic and the data to be compared, perform the selection and analysis of the characteristic bands at the same position to complete the spectral identification of the fingerprint points of the cultural relic. Specifically, extract the spectral reflectance information of each fingerprint point in each band, and construct the corresponding spectral curve, and perform a comparative analysis of the reference data and the data to be identified to complete the spectral identification of the fingerprint points of the cultural relic. Spectral curve fitting is as Figure 12 shown.

[0054] Step 5) Multimodal feature fusion, including the feature fusion and identification of geometric, texture, and spectral information.

[0055] In specific implementation, the optimization part can be achieved by the following sub-steps: Step 5.1, uniformly encode the local features of the three types of modalities; Step 5.2: Automatically learn the importance weights of features in different regions of different modalities based on the attention mechanism, and realize dynamic perception and adaptive weighting of geometric structure stability, texture detail expression, and spectral material differences; Step 5.3, complete the fusion identification of cultural relics digital fingerprints based on multimodal features.

[0056] In the embodiment of the present invention, through the geometric alignment and texture compensation mechanism, stable feature matching and comparison can be achieved even in complex environments such as lighting changes, local wear or viewing angle differences, thereby significantly improving the robustness of the system.

[0057] In the embodiment of the present invention, the introduction of hyperspectral data enables the method to have fine-grained material analysis capabilities, can identify extremely small surface material differences, and is suitable for high-precision scenarios such as counterfeit detection and microscopic restoration analysis.

[0058] In the embodiment of the present invention, the introduction of hyperspectral data enables the method to have fine-grained material analysis capabilities, can identify extremely small surface material differences, and is suitable for high-precision scenarios such as counterfeit detection and microscopic restoration analysis.

[0059] Through the above process, the present invention extracts the stable features of geometric texture and hyperspectral data on the basis of realizing high-precision three-dimensional reconstruction, image texture analysis and hyperspectral data processing, and optimizes the fusion method of cross-modal data, and finally constructs a unique and comparable digital fingerprint of cultural relics for identity identification.

[0060] In specific implementation, the method proposed by the present invention can be manually or automatically operated by those skilled in the art according to the technical solution of the present invention, and the digital fingerprint identification of cultural relics can be realized according to specific steps. The method includes steps such as feature point extraction, geometric alignment, texture analysis and hyperspectral data processing, through which high-precision identification of cultural relics is realized. The implementation of the method does not need to rely on a specific computer and can be completed by conventional equipment and tools.

[0061] In some possible embodiments, the present invention provides a method for identifying digital fingerprints of cultural relics, including the following steps: First, high-precision data collection is performed on the surface of the cultural relic through specific equipment to obtain geometric structures, texture features, and hyperspectral image data; Next, geometric alignment methods are used to preprocess the collected cultural relic data, extract key feature points, and achieve image registration; Then, based on texture feature analysis, including bit-plane stratification processing, gray-level co-occurrence matrix calculation, and histogram feature extraction, combined with significant band information in the hyperspectral data, feature fusion is performed; Finally, through similarity measurement and classification algorithms, digital fingerprint identification of cultural relics is carried out to achieve identity comparison and verification.

[0062] The implementation basis of each embodiment of the present invention is achieved through programmed processing by a device with processor functions. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present invention are encapsulated into various modules. Based on this actual situation, on the basis of the above embodiments, an embodiment of the present invention provides a system for identifying digital fingerprints of cultural relics based on geometric texture and spectral information, and this system is used to execute the method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information in the above method embodiments.

[0063] This system includes: a data collection module, which is used to obtain multi-source data and perform texture mapping to form a three-dimensional mesh model with texture data; a fingerprint point selection module, which is used to calculate the curvature values and topological features of each vertex of the three-dimensional mesh model, adjust the feature weights based on local saliency for feature fusion, and complete the selection of fingerprint points; a texture feature extraction module, which is used to construct a significant plane through bit planes based on the selected fingerprint points, and extract the texture features of cultural relics based on the improved histogram and gray-level co-occurrence matrix; a hyperspectral feature extraction module, which is used to combine hyperspectral data to extract the characteristic bands of cultural relics based on the selected fingerprint points; a feature fusion module, which is used to perform multi-modal fusion identification based on the features of geometry, texture, and spectral information.

[0064] The system for identifying digital fingerprints of cultural relics based on geometric texture and spectral information provided by the embodiment of the present invention faces the current situation that due to the lack of multi-modal fusion strategies, existing methods are difficult to take into account spatial structures, surface textures, and material compositions, and there is still a large room for improvement in the overall recognition accuracy and robustness. By using the aforementioned several modules, through the fusion of geometric structures, texture features, and hyperspectral information, high-precision identification of cultural relic identity with multi-modal collaboration is achieved, and it has the advantages of high accuracy, strong robustness, good interpretability, and adaptability to complex environments.

[0065] It should be noted that the system embodiments provided by the present invention are used not only to implement the methods in the above method embodiments, but also to implement the methods in other method embodiments provided by the present invention. The difference lies only in setting corresponding functional modules. The principle is basically the same as that of the above system embodiments provided by the present invention. As long as those skilled in the art, on the basis of the above system embodiments, refer to the specific technical solutions in other method embodiments, obtain corresponding technical means by combining technical features, and the technical solutions constituted by these technical means, and on the premise of ensuring the practicability of the technical solutions, improve the modules in the above system embodiments to obtain corresponding system-like embodiments for implementing the methods in other method-like embodiments.

[0066] Based on the same inventive concept as the foregoing embodiments, the embodiments of the present invention further provide a cultural relic digital fingerprint identification device based on geometric texture and spectral information, including a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the cultural relic digital fingerprint identification method based on geometric texture and spectral information.

[0067] Based on the same inventive concept as the foregoing embodiments, the embodiments of the present invention further provide a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the cultural relic digital fingerprint identification method based on geometric texture and spectral information as follows: Obtain multi-source data and perform texture mapping to form a three-dimensional mesh model with texture data; Calculate the curvature values and topological features of each vertex of the three-dimensional mesh model, and perform feature fusion by adjusting the feature weights based on local saliency to complete the screening of fingerprint points; Based on the screened fingerprint points, construct a significant plane through bit planes, and extract the texture features of the cultural relics based on the improved histogram and gray-level co-occurrence matrix; Based on the screened fingerprint points, extract the characteristic bands of the cultural relics in combination with hyperspectral data; Perform multi-modal fusion identification based on the characteristics of geometry, texture, and spectral information.

[0068] In summary, the present invention discloses a method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information, aiming to solve the limitations of single-modal methods in existing cultural relic identification technologies. This method collects the geometric structure, surface texture, and hyperspectral image data of cultural relics. In terms of geometric feature extraction, first, the normal vector information entropy is introduced to calculate the curvature of each vertex of the 3D model. From a morphological perspective, by constructing a topological graph, the multi-scale clustering coefficient is calculated to capture the stable structural information of the cultural relics. Based on local saliency, the feature weights are adjusted to finally achieve the screening and identification of geometric fingerprint points. Based on the high-resolution texture map, the texture feature information of the cultural relics is obtained from the feature vectors extracted from a statistical perspective, local feature descriptors, and deep learning around the fingerprint points. Further, the digital fingerprint points based on texture information are identified by comparing the microscopic details within the texture neighborhood. In view of the feature differences of different objects, a feature fusion network is designed based on the attention mechanism. On the basis of ensuring the independent expression of various features, weights are automatically assigned to finally construct the fused digital fingerprint, providing double support in theory and practice for the multi-modal fusion identification of digital fingerprints of cultural relics.

[0069] The terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information, characterized in that Including: Obtain multi-source data and perform texture mapping to form a three-dimensional mesh model with texture data; Calculate the curvature values and topological features of each vertex of the three-dimensional mesh model, adjust the feature weights based on local saliency for feature fusion, and complete the screening of fingerprint points; Based on the screened fingerprint points, construct a significant plane through bit planes, and extract the texture features of the cultural relics based on the improved histogram and gray-level co-occurrence matrix; Based on the screened fingerprint points, extract the characteristic bands of the cultural relics in combination with hyperspectral data; Perform multi-modal fusion discrimination based on the features of geometry, texture, and spectral information.

2. The method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information according to claim 1, wherein Obtain multi-source data and perform texture mapping, which also includes: Collect the high-density geometric grid of the cultural relics to form a three-dimensional mesh model; Collect the high-definition texture maps of the cultural relics through multi-angle photography; Process the high-definition texture maps and map them to the three-dimensional mesh model to form a three-dimensional mesh model with texture data.

3. The method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information according to claim 2, characterized in that, Processing the high-definition texture maps also includes: Extract the main body area of the cultural relics from multi-view images, eliminate the texture matching errors caused by background interference, and perform three-dimensional reconstruction to generate dense point clouds and meshes; Adopt a two-stage matching strategy of rough matching and fine matching to obtain the transformation matrix; Accurately map the texture map to the surface of the three-dimensional mesh model based on the transformation matrix.

4. The method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information according to claim 1, wherein The screening of fingerprint points also includes: Construct an information entropy model through the angle between neighborhood normal vectors, perform curvature analysis based on information entropy adaptive neighborhood to describe the local geometric complexity of cultural relics; Perform topological structure optimization, calculate the multi-scale clustering coefficient to ensure the rationality of the spatial distribution of points; Fuse geometric and topological information, dynamically adjust the contribution weights, and screen out the fingerprint points of cultural relics.

5. The method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information according to claim 4, wherein After obtaining the screened fingerprint points, it also includes: Map the screened fingerprint points to the two-dimensional texture map through the collinearity equation to obtain the two-dimensional image coordinates of the fingerprint points; Perform bit-plane layering on the image to form 8 binary images, calculate the variance, kurtosis, and average entropy values of different bit planes, and select the characteristic plane together with the original image for texture feature extraction based on visual discrimination and the established three-plane evaluation indexes; Based on the Lab color space, use the Gaussian mixture model to classify the cultural relics image to obtain the clustering histogram; Analyze the relative positions of pixel pairs in the image from multiple relative directions based on the gray-level co-occurrence matrix to extract the texture features of cultural relics; Based on the two complementary texture feature values of the calculated color histogram and gray-level co-occurrence matrix, perform standardization and normalization processing, and integrate them into a unified feature vector.

6. The method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information according to claim 4, characterized in that, After obtaining the screened fingerprint points, it also includes: Collect hyperspectral data and perform radiometric correction and data matching; Use competitive adaptive reweighting to perform preliminary selection of characteristic bands; Use the genetic algorithm to globally optimize the band selection to obtain the final characteristic bands.

7. The method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information according to claim 1, characterized in that, Performing multi-modal fusion discrimination based on the features of geometry, texture, and spectral information also includes: Unify the encoding of local features of the three types of modalities; According to the attention mechanism, automatically learn the importance weights of different regional features of different modalities, and realize the dynamic perception and adaptive weighting of the geometric structure stability, texture detail expressiveness, and spectral material difference. Multi-modal feature fusion and identification based on cosine similarity.

8. A cultural relic digital fingerprint identification system based on geometric texture and spectral information, characterized in that, Including: A data acquisition module for obtaining multi-source data and performing texture mapping to form a three-dimensional grid model with texture data; A fingerprint point selection module for calculating the curvature values and topological features of each vertex of the three-dimensional grid model, adjusting the feature weights based on local saliency for feature fusion, and completing the selection of fingerprint points; A texture feature extraction module for constructing a significant plane through bit planes based on the selected fingerprint points, and extracting the texture features of the cultural relics based on the improved histogram and gray-level co-occurrence matrix; A hyperspectral feature extraction module for combining hyperspectral data to extract the characteristic bands of the cultural relics based on the selected fingerprint points; A feature fusion module for performing multi-modal fusion and identification based on the features of geometric, texture, and spectral information.

9. A cultural relic digital fingerprint identification device based on geometric texture and spectral information, characterized in that, Including a memory and a processor, the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the method for identifying digital fingerprints of cultural relics based on geometric texture and spectral information according to any one of claims 1 to 7.

Images