A cultural relic restoration scheme generation method, system and device

Abstract

The application provides a cultural relic restoration scheme generation method, system and device, and relates to the technical field of cultural relic digitization protection, and the scheme comprises the following steps: obtaining a full information model of a cultural relic to be restored; determining the characteristics of the cultural relic to be restored according to the full information model; comparing the characteristics of the cultural relic to be restored with a target reference model to determine the damage condition of the cultural relic to be restored; and generating a cultural relic restoration scheme based on the full information model and the damage condition. According to the scheme, multi-source data such as digital images, spectra and point cloud data are used to construct a high-precision full information model, so that the unified expression of geometry, texture and material is realized; multi-scale feature extraction and cross-modal fusion are combined to improve the modeling accuracy; a virtual restoration algorithm combining physical simulation and sample driving fusion is introduced to accurately identify damage and complete missing parts with high fidelity, so that a scientific and reliable cultural relic restoration scheme is generated.

Description

Technical Field

[0001] This invention relates to the field of digital preservation technology for cultural relics, and in particular to a method, system, and device for generating cultural relic restoration plans. Background Technology

[0002] Cultural relics are important carriers of human history and culture, possessing extremely high historical, artistic, and scientific value. However, due to the passage of time and the impact of the natural environment, many cultural relics face varying degrees of damage and weathering. Traditional methods of cultural relic restoration mainly rely on the experience of experts and manual operations, which are not only time-consuming and costly but may also cause secondary damage to the relics. Furthermore, traditional methods struggle to comprehensively and accurately record and preserve the multidimensional information of cultural relics, hindering their long-term protection and research.

[0003] With the development of digital technology, digital preservation of cultural relics has gradually become a research hotspot in the field. Digital technology enables high-precision 3D modeling, texture acquisition, and information recording of cultural relics, allowing for virtual restoration and display. However, most existing digital preservation technologies focus only on the acquisition and processing of single data sources, such as digital images or 3D laser scanning data, which cannot comprehensively and accurately reflect the true condition of the relics. Furthermore, limitations exist in data fusion and model construction, resulting in room for improvement in the accuracy and realism of virtual restoration results. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, and apparatus for generating cultural relic restoration plans, so as to solve at least one of the above-mentioned technical problems existing in the prior art.

[0005] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for generating a cultural relic restoration plan, comprising: acquiring a full-information model of the cultural relic to be restored; determining the characteristics of the cultural relic to be restored based on the full-information model; comparing the characteristics of the cultural relic to be restored with a target benchmark model to determine the damage status of the cultural relic to be restored; and generating a cultural relic restoration plan based on the full-information model and the damage status.

[0006] Optionally, obtaining a full-information model of the cultural relic to be restored includes: unifying the resolution scale of the three-dimensional spatial model and material feature information and then fusing the information to obtain a fused image model; unfolding the fused image model into a plane to obtain a texture map; mapping the texture map onto the surface of the three-dimensional geometric model based on the mapping relationship to obtain a three-dimensional texture geometric material model; the mapping relationship is obtained by performing coordinate transformation on the texture map and the three-dimensional geometric model; and rendering the three-dimensional texture geometric material model after consistency adjustment to obtain a full-information model.

[0007] Optionally, the three-dimensional spatial model is established as follows: acquiring digital image data of the cultural relic to be restored; determining the results of close-range photogrammetric image pairs and relative control measurement control points based on the digital image data; generating an initial spatial model based on the results of close-range photogrammetric image pairs and relative control measurement control points; mapping the digital image data as texture onto the surface of the initial spatial model to obtain the three-dimensional spatial model; establishing material characteristic information as follows: acquiring hyperspectral image data of the cultural relic to be restored; determining the endmember abundance map of the cultural relic to be restored after radiometric correction of the hyperspectral image data; determining material characteristic information based on the endmember abundance map; establishing the three-dimensional geometric model as follows: acquiring initial point cloud data of the cultural relic to be restored; preprocessing the initial point cloud data to obtain usable point cloud data; and using surface reconstruction based on the usable point cloud data to obtain the three-dimensional geometric model.

[0008] Optionally, based on the full information model, the characteristics of the cultural relic to be restored are determined, including: after color correction of the full information model, image segmentation is performed using superpixels to obtain initial segmentation results; multi-scale morphological gradient reconstruction and clustering are performed on the initial segmentation results to generate local spatial features and global color features; multi-scale feature extraction is performed on the full information model and a fusion strategy is adopted to generate fused features; and the local spatial features, global color features and fused features are used as the characteristics of the cultural relic to be restored.

[0009] Optionally, based on the characteristics of the cultural relic to be restored, a comparison is made with the target benchmark model to determine the damage status of the cultural relic to be restored, including: obtaining the benchmark features of the target benchmark model; after aligning the features of the cultural relic to be restored and the benchmark features, extracting the geometric features of each local area to obtain the local feature parameters of the cultural relic to be restored and the benchmark local feature parameters; monitoring the differences between the local feature parameters of the cultural relic to be restored and the benchmark local feature parameters to obtain the difference distribution data; and determining the damage status based on the difference distribution data.

[0010] Optionally, the damage status is determined based on the differential distribution data, including: clustering the differential distribution data to obtain the spatial distribution of the damaged parts; determining the damage level of any damaged part based on the number of differences in the spatial distribution; and taking the damaged part and damage level as the damage status.

[0011] Optionally, based on the full information model and the damage situation, a cultural relic restoration plan is generated, including: determining the mechanical parameter data of the cultural relic to be restored according to the full information model; discretizing and establishing mechanical and physical equations based on the full information model and the mechanical parameter data to obtain a physical simulation model; simulating the damage evolution process of the cultural relic to be restored using preset damage parameters based on the physical simulation model; and generating a cultural relic restoration plan according to the damage evolution process and preset restoration goals.

[0012] Optionally, based on the full information model and the damage situation, a cultural relic restoration plan is generated, including: determining the basic characteristics of the cultural relic to be restored according to the full information model; extracting the closest target sample from a pre-built sample library according to the basic characteristics; and using the target sample as a template to geometrically complete the damage situation and generate a cultural relic restoration plan.

[0013] Secondly, based on the same inventive concept, this application also provides a cultural relic restoration scheme generation system, specifically including: a model acquisition module for acquiring a full-information model of the cultural relic to be restored; a feature extraction module for determining the features of the cultural relic to be restored based on the full-information model; a damage determination module for comparing the features of the cultural relic to be restored with a target benchmark model to determine the damage status of the cultural relic to be restored; and a virtual restoration module for generating a cultural relic restoration scheme based on the full-information model and the damage status.

[0014] Thirdly, based on the same inventive concept, this application also provides a cultural relic restoration scheme generation device, specifically including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the cultural relic restoration scheme generation method as described above.

[0015] By employing the above technical solution, this invention achieves the following beneficial effects: By 3D registration and fusion of multi-source information from digital images, hyperspectral images, and point cloud data, a high-precision full-information model containing multi-dimensional information on geometry, texture, and material is constructed. Combined with technologies such as superpixel segmentation, multi-scale feature extraction, and cross-modal fusion, comprehensive feature acquisition of the geometric structure, surface texture, and spectral characteristics of cultural relics is realized. Furthermore, a difference monitoring mechanism based on a target benchmark model is introduced to accurately locate and quantify the damage range and level of cultural relics. Moreover, by utilizing both physical simulation and example-driven methods to generate restoration schemes, not only can the actual damage evolution process be simulated, but the geometric and textural completion of missing parts can also be achieved, thereby improving the accuracy, scientific validity, and visualization effect of the restoration schemes and meeting the practical needs of digital preservation and virtual restoration of cultural relics. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating the method for generating cultural relic restoration schemes provided in this embodiment of the invention;

[0018] Figure 2 Another flowchart illustrating the method for generating cultural relic restoration schemes provided in this embodiment of the invention;

[0019] Figure 3 This is a schematic diagram of the structure of the cultural relic restoration scheme generation system provided in an embodiment of the present invention;

[0020] Figure 4 This is another structural schematic diagram of the cultural relic restoration scheme generation system provided in the embodiments of the present invention;

[0021] Figure 5 This is a schematic diagram of the structure of the cultural relic restoration scheme generation device provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0024] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0025] In order to at least partially solve one or more of the above-mentioned problems and other potential problems, this disclosure proposes a method for generating cultural relic restoration schemes. The invention will be further explained below with reference to specific embodiments.

[0026] Example 1:

[0027] This disclosure provides a method for generating cultural relic restoration plans. Figure 1 This is a flowchart illustrating a method for generating a cultural relic restoration scheme according to an embodiment of this disclosure, as shown below. Figure 1 As shown, the method for generating this cultural relic restoration plan includes:

[0028] Step S101: Obtain a full information model of the cultural relic to be restored.

[0029] Step S102: Determine the characteristics of the cultural relic to be restored based on the full information model.

[0030] Step S103: Based on the characteristics of the cultural relic to be repaired, compare it with the target benchmark model to determine the damage status of the cultural relic to be repaired.

[0031] Step S104: Generate a cultural relic restoration plan based on the full information model and damage status.

[0032] Among them, the full information model refers to a three-dimensional digital model that can comprehensively describe the form, structure, material and surface condition of cultural relics by collecting and integrating them in multiple dimensions and in all aspects.

[0033] In this embodiment, high-precision external geometric data of the cultural relic can first be obtained using technologies such as 3D laser scanning and photogrammetry. Then, high-definition photography can be used to acquire surface texture and color information, combined with point spectral analysis and hyperspectral data analysis to obtain feature information. Finally, cross-modal data modeling technology can be used to fuse and model the aforementioned multi-source data, forming a unified digital model with geometric, spectral, texture, and material elements. The above is merely an illustrative example and does not constitute a limitation on all possible scenarios for obtaining a complete information model of the cultural relic to be restored; it is simply not exhaustive.

[0034] The characteristics of the cultural relic to be restored refer to the set of data that can characterize the uniqueness, structural properties, and condition of the cultural relic. For example, the characteristics of the cultural relic to be restored may include: geometric features, texture features, material features, and damage-related features, etc.

[0035] In this embodiment, the full-information model can first be preprocessed to facilitate subsequent feature extraction. For example, the preprocessing of the full-information model may involve using a 3D point cloud filtering algorithm to remove scanning error points and denoise the data; further, using texture correction methods to perform illumination equalization and color calibration on the artifact surface image data to avoid color distortion caused by uneven illumination; further still, using geometric reconstruction methods, such as Poisson reconstruction or Delaunay triangulation, to convert the scattered point cloud into a continuous mesh model; and even further, performing coordinate system transformation and processing on the model to align the model coordinate axes and unify the scale, facilitating comparison with subsequent benchmark models. Then, the preprocessed full-information model can be used for geometric feature extraction, texture feature extraction, material feature identification, and preliminary labeling of damaged areas. Finally, the extracted feature information can be classified and layered, data format standardized, and spatially indexed and associated with locations, organizing the feature information into structured data for easy subsequent retrieval and comparison. The above is merely an illustrative example and is not intended to limit all possible situations for determining the characteristics of the cultural relic to be restored; it is simply not an exhaustive list.

[0036] The target benchmark model is a standard model of the cultural relic used as a reference. This model is used as a comparison to identify the defects, deformations, and damage of the cultural relic to be restored. For example, the target benchmark model can be a digital model of the best-preserved example among similar cultural relics, a prototype model reconstructed through speculative restoration algorithms, or an original state model provided in historical archives.

[0037] The damage status refers to the specific description and quantitative information obtained by comparing the characteristics of the cultural relic to be restored with the target benchmark model, addressing issues such as damage, loss, and deformation. For example, the damage status can include the type, location, extent, degree, and manifestation of the damage, typically a digitized and visualized result, which can guide the generation of subsequent restoration plans.

[0038] In this embodiment, a geometric registration method is first used to precisely match and align the extracted features of the artifact to be restored with the features of the baseline model. A difference calculation algorithm is then used to analyze information such as geometric deformation, missing volume, and color differences. Based on the difference data, the damage is categorized into different types, and finally, a visual damage distribution map is output for reference in the restoration design. For example, damage categories may include smoke damage, cracks, scratches, stains, fading, efflorescence, hollow areas, and salt stains. The above is merely an illustrative example and does not constitute a limitation on all possible situations regarding the damage to the artifact to be restored; it is simply not exhaustive.

[0039] Among them, the cultural relic restoration plan refers to a digital and executable restoration plan formulated based on the current complete data information of the cultural relic, the results of damage assessment, and historical restoration experience.

[0040] In this embodiment, repair plans for different damage types can be generated first, based on the full information model and damage distribution results, combined with the complete morphological information of the material database, repair process library, and target benchmark model. This process can begin by classifying and matching each damaged area, and then determining feasible repair strategies based on corresponding material properties, processing techniques, mechanical and aesthetic requirements. Subsequently, a model reconstruction algorithm is used to simulate the repair effect in digital space, outputting a structured plan file containing repair location, method, parameters, material list, and construction steps. A virtual repair model is also generated as a preview of the effect, providing direct guidance for actual cultural relic restoration. The above is merely an illustrative example and does not constitute a limitation on all possible scenarios for generating cultural relic restoration plans; it is simply not exhaustive.

[0041] The technical solution of this disclosure achieves digitalization, precision, and visualization of cultural relic restoration by constructing a full-information model, extracting features, comparing and analyzing with a benchmark model, and generating a restoration plan. This reduces subjective human error, lowers the risk of direct contact with cultural relics, and improves the scientific rigor of damage identification and restoration plan development. Furthermore, it standardizes and archives the entire restoration process for long-term preservation and reference in subsequent restorations, thereby enhancing the efficiency and quality of cultural relic protection work.

[0042] In some embodiments, obtaining a full-information model of the cultural relic to be restored includes: unifying the resolution scale of the three-dimensional spatial model and material feature information and then fusing the information to obtain a fused image model; unfolding the fused image model into a plane to obtain a texture map; mapping the texture map onto the surface of the three-dimensional geometric model based on a mapping relationship to obtain a three-dimensional texture geometric material model; the mapping relationship is obtained by performing coordinate transformation on the texture map and the three-dimensional geometric model; and rendering the three-dimensional texture geometric material model after consistency adjustment to obtain a full-information model.

[0043] In this context, a three-dimensional spatial model refers to a spatially acquired model with texture information, which is essentially a geometric position plus surface color or texture mapping information. In this embodiment of the invention, the texture map of the three-dimensional spatial model can be image information obtained by texture finishing a digital image.

[0044] Material characteristic information refers to data describing the material properties of an object, which is often obtained through spectral imaging, point spectral sampling, etc. For example, material characteristic information may include spectral reflectance curves, aging degree, crystal structure of the material, etc.

[0045] In this embodiment, the material feature information can first be spatially resampled to make its pixel resolution consistent with the texture resolution of the 3D spatial model. For example, an upsampling method can be used to spatially resample the material feature information. Then, the coordinate systems of the 3D spatial model and the material feature information can be aligned using feature point matching or control point correction methods. Finally, under the premise of unifying their spatial coordinate systems and resolutions, the texture data of the 3D spatial model and the material feature information can be fused point-by-point on a vertex or face unit basis to obtain a fused image model that simultaneously contains spatial texture and material spectral information. For example, this fusion process can be performed using weighted averaging or multi-channel stacking. The above is merely an illustrative example and is not intended to limit all possible cases for obtaining the fused image model; it is simply not exhaustive.

[0046] In this embodiment, a mapping relationship between the surface vertices of the fused image model and the two-dimensional UV coordinates can first be established. Then, the three-dimensional fused image model is unfolded in a plane, that is, the three-dimensional fused image is mapped onto a two-dimensional plane to generate a texture map containing fused texture and material information. Exemplarily, this plane unfolding process can be implemented using UV unfolding or mesh-based parameterization methods. Finally, the texture map can be stitched and seamlessly spliced ​​to eliminate unfolding gaps and edge distortion, resulting in a texture material map. The above is only an illustrative example and is not intended to limit all possible cases of obtaining a texture material map; it is simply not exhaustive.

[0047] Among them, the three-dimensional geometric model refers to the object shape data obtained by three-dimensional scanning or modeling, which only contains geometric shapes such as point clouds and triangular meshes, and usually does not contain additional information such as color and material.

[0048] In this embodiment, rigid or non-rigid registration algorithms can be used to spatially align the fused image model and the 3D geometric model, calculating the spatial transformation matrix between them to place them in the same coordinate system. Subsequently, using this transformation matrix, the 3D coordinates of each surface point in the fused image model are transformed into the coordinate system of the 3D geometric model. Combined with the mapping information from the planar unfolding of the fused image model, the UV coordinates of its associated 2D texture map are accurately transferred to the corresponding surface position of the 3D geometric model, thereby establishing a mapping relationship between the 2D texture map and the 3D geometric model. The above is merely an illustrative example and is not intended to limit all possible scenarios for obtaining the mapping relationship; it is simply not exhaustive.

[0049] In this embodiment of the disclosure, for each top surface or face of the three-dimensional geometric model, based on the established mapping relationship between the two-dimensional texture material map and the three-dimensional geometric model, parameters such as texture and material information are sampled from the texture material map and mapped to the vertex or face attributes of the three-dimensional geometric model to obtain a three-dimensional texture geometric material model. The above is only an illustrative example and is not intended to limit all possible cases of obtaining a three-dimensional texture geometric material model; it is simply not exhaustive.

[0050] In this embodiment, a consistency check can be performed on the obtained 3D textured geometric material model. For example, the consistency check may include checking for holes or seams in the geometric mesh, checking for misalignment in the texture bonding, and checking for anomalies or abrupt changes in material values. Then, the 3D textured geometric material model is optimized, including smoothing seams and adjusting colors in the model texture, and unifying the range and units of material parameters. Further, texture maps or parameter files such as reflectivity, roughness, and refractive index required for physical rendering can be calculated based on the material properties. Then, the 3D textured geometric material model is loaded into the physical rendering engine, and rendering conditions such as lighting and camera settings are set to generate a realistic visualization result. Finally, an information-based cultural relic model with accurate shape, realistic appearance, and complete physical properties is obtained, i.e., a full-information model, which can be used for scientific analysis, digital display, and restoration simulation. The above is merely an illustrative example and is not intended to limit all possible scenarios for obtaining a full-information model; it is simply not exhaustive.

[0051] In this way, by fusing image data and hyperspectral data to obtain a three-dimensional fused image model, generating texture maps that centrally store color and material characteristics, using spatial registration and coordinate transformation to accurately map the texture maps onto a three-dimensional geometric model established from the point cloud data of cultural relics and assigning complete material information, and performing consistency adjustments and rendering, a fully informational digital model integrating shape, appearance texture, and material information is constructed. This ensures that geometric, texture, and material data correspond accurately in the same coordinate system, maximizing the restoration of the true form and texture of cultural relics. This not only improves the visualization authenticity and integrity of digital cultural relics, but also provides a high-precision and reproducible digital benchmark for subsequent damage analysis, restoration plan design, and long-term preservation.

[0052] In some embodiments, the three-dimensional spatial model is established as follows: acquiring digital image data of the cultural relic to be restored; determining the results of close-range photogrammetric image pairs and relative control measurement control points based on the digital image data; generating an initial spatial model based on the results of close-range photogrammetric image pairs and relative control measurement control points; mapping the digital image data as texture onto the surface of the initial spatial model to obtain the three-dimensional spatial model; establishing material feature information as follows: acquiring hyperspectral image data of the cultural relic to be restored; determining the endmember abundance map of the cultural relic to be restored after radiometric correction of the hyperspectral image data; determining material feature information based on the endmember abundance map; establishing the three-dimensional geometric model as follows: acquiring initial point cloud data of the cultural relic to be restored; preprocessing the initial point cloud data to obtain usable point cloud data; and using surface reconstruction based on the usable point cloud data to obtain the three-dimensional geometric model.

[0053] Digital image data refers to a collection of high-resolution two-dimensional images acquired through digital cameras, industrial cameras, or other digital imaging devices. Each image records information such as color, texture, shape, and outline of the artifact's surface from a specific perspective.

[0054] In this embodiment, firstly, an imaging device that meets the accuracy requirements of close-range photogrammetry can be selected, and the lens distortion parameters are ensured to be calibrated. Then, the imaging device is used to surround the artifact, taking pictures from multiple heights and angles to ensure that the images cover the entire visible surface of the artifact. Specifically, before shooting, a reasonable shooting plan can be formulated based on the characteristics of the artifact and the display requirements. During the shooting process, attention should be paid to controlling factors such as lighting conditions, shooting angle, and distance to ensure the quality of the image data. Furthermore, all captured images are stored in the shooting order, retaining the original parameter information. Finally, the acquired image data is preprocessed to obtain digital image data of the artifact to be restored. For example, an X-Rite color chart can be used to perform color chart correction on the acquired images, or professional graphics software can be used to restore the true colors of the artifact. Professional photogrammetry software can also be used to preprocess the acquired images, such as denoising and distortion correction, to improve image quality. The above is only an illustrative example and is not intended to limit all possible situations for obtaining digital image data of the artifact to be restored; it is simply not exhaustive.

[0055] Among them, close-range photogrammetry image pairs refer to two or more images with a high degree of overlap taken from different positions, and their three-dimensional spatial geometric relationship can be deduced through stereo matching.

[0056] The relative control measurement control point results refer to the spatial position data of identifiable reference points in different images. These control points provide a benchmark for calculating the relative position and attitude between images. In this embodiment of the invention, the relative control measurement control point results may include the image coordinates, relative spatial coordinates, and matching relationships of the control points.

[0057] In this embodiment, image screening and pairing can be performed first. From the acquired image data, suitable image pairs for stereo reconstruction are selected based on overlap and shooting angle, serving as close-range photogrammetry image pairs. Then, stable feature points are extracted from each image pair using a conventional feature detection algorithm, and matched between paired images to obtain a set of corresponding point pairs. Subsequently, points with unique and significant features that can be accurately located in different images are selected from multiple image pairs as control points. Their pixel positions in each image are measured manually or automatically, and their relative coordinates are calculated. Based on the correspondence of control points in the image pairs, the relative camera pose and position parameters between images are calculated to obtain the relative control measurement control point results. The above is merely an illustrative example and is not intended to limit all possible cases for determining close-range photogrammetry image pairs and relative control measurement control point results; it is simply not exhaustive.

[0058] In this embodiment, the process of generating the initial spatial model can begin by combining the camera positions and orientations of each close-range photogrammetric image pair to construct the spatial structure of the entire imaging system. Then, a conventional stereo matching algorithm is used to perform pixel-level matching on the image pairs, calculating the 3D coordinates of the densely matched points to obtain the initial matching results. Next, missing points are filled in based on the initial matching results to generate dense 3D point cloud data covering the surface of the artifact. Finally, based on the point cloud data, the initial spatial model is obtained through surface triangulation or polygon mesh generation methods.

[0059] For example, the process of constructing the initial spatial model can also involve using close-range photogrammetric image pairs and relative control measurement control point results for interior orientation to determine the internal geometric parameters of each image, thereby eliminating nonlinear distortions in the optical imaging of the image itself. Then, using corresponding point pairs from the close-range photogrammetric image pairs combined with the results of the relative control measurement control points as a reference, exterior orientation is performed. The relative positional relationship of the images is calculated through spatial resection, and the known coordinates of the control points are used to unify the camera coordinate system with the actual coordinate system of the artifact. Bundle adjustment is then used to optimize the global camera exterior orientation element solution, improving overall consistency and ultimately determining the position and orientation of each image in the model coordinate system. Further, the polar geometric relationship between the two image pairs is calculated using the interior and exterior orientation results. The original images are then reprojected and sampled according to the epipolar equation, ensuring that the matching points in the image pair only have horizontal parallax. Epipolar-corrected image pairs are then output, ensuring that stereo matching converges more easily and is less prone to mismatches. This step utilizes epipolar resampling to resample the image, eliminating the vertical component in the parallax. This ensures that points of the same object are located in the same horizontal row within the matched image pair, simplifying subsequent stereo matching. Furthermore, feature point detection and matching can be performed on the epipolar-corrected image to establish an initial 3D sparse point cloud framework. Multi-view stereo matching algorithms or semi-global matching algorithms are then used to find the parallax value for each pixel, converting the image depth information into a dense point cloud. The local point clouds generated from different image pairs are then fused in the global coordinate system to obtain a high-density point cloud of the complete artifact surface. This step uses image matching and dense point cloud generation methods to recover the 3D spatial coordinates of each surface point from the image pair, obtaining a dense 3D point cloud to provide input data for modeling. Further, the normal information of each point is calculated to determine the surface orientation. Poisson reconstruction and progressive triangulation algorithms are used for surface reconstruction, converting the point cloud data into a continuous triangular mesh to form the overall surface. Finally, redundant patches are removed, holes are repaired, and the surface is smoothed to obtain the initial spatial model.

[0060] The above is merely an illustrative example and is not intended to limit all possible scenarios for generating the initial spatial model; it is simply not an exhaustive list.

[0061] In this embodiment, each surface point of the initial spatial model can first be projected onto the corresponding image based on the position and orientation of each digital image in the model coordinate system, thereby obtaining the texture information of the projected points in the image. Then, when the same model area is covered by multiple images, a stitching or fusion algorithm is used to synthesize a unified texture. Finally, the fused texture is mapped onto the model surface to obtain a three-dimensional spatial model with realistic colors. For example, the texture mapping process can use the outward orientation result to project the three-dimensional model surface onto the corresponding image, calculate the two-dimensional position of the surface points on the image, extract color information from the image, and apply the processed texture map to the three-dimensional model surface to obtain a textured three-dimensional spatial model. The above is merely an illustrative example and is not intended to limit all possible cases of obtaining a three-dimensional spatial model; it is simply not exhaustive.

[0062] Specifically, by utilizing the shooting equipment parameters and the surface geometry of the 3D model, an inverse transformation of the texture coordinates can be performed to eliminate distortion caused by shooting tilt. Then, color balancing and brightness adjustment are applied to the image to eliminate lighting differences. Finally, the model surface is unfolded onto a uniform plane to generate an orthophoto image without perspective distortion, which can be directly used for two-dimensional image display of cultural relics.

[0063] Hyperspectral image data refers to images that capture continuous spectral information across dozens to hundreds of narrow bands when photographing the same target area. Each pixel contains not only its two-dimensional spatial location but also a complete spectral curve, reflecting the object's reflectance characteristics at various wavelengths.

[0064] In this embodiment, when acquiring hyperspectral image data of the cultural relic to be restored, a hyperspectral imager with a suitable spectral range and resolution (such as a pushbroom, snapshot, or linear array scanner) can be selected first. During the imaging process, attention should be paid to controlling the ambient light source, such as selecting a stable artificial light source and eliminating interference from changes in natural light. Then, spatial and spectral cubic data of the surface of the cultural relic are acquired step by step according to the scanning method. Finally, the scanned data is saved in hyperspectral data format, and the acquisition metadata (such as wavelength, resolution, exposure, scanning speed, etc.) is recorded for subsequent analysis and retrieval. In particular, a portable ground object spectrometer can also be used simultaneously to acquire point spectral data of the cultural relic. Point spectral data can be used to verify the accuracy of the hyperspectral image analysis results and provide more detailed information for the spectral study of the cultural relic. The above is only an illustrative example and is not intended to limit all possible situations for acquiring hyperspectral image data of the cultural relic to be restored; it is simply not exhaustive.

[0065] Among them, the end-member is the spectral curve of the material with the purest spectral characteristics in the scene in hyperspectral analysis, representing a typical material.

[0066] An endmember abundance map, based on a spectral mixing model, indicates that each pixel may be composed of multiple endmembers mixed in different proportions. An endmember abundance map is a two-dimensional distribution map representing the proportion of a particular endmember in each pixel of an image.

[0067] In this embodiment, the acquired hyperspectral image can first undergo radiometric correction, that is, the pixel digital values ​​of the original hyperspectral data are converted into true physical reflectance values ​​to eliminate the influence of instrument and environmental factors on the spectral image data. Then, endmember extraction is performed using algorithms such as Pixel Purity Index (PPI) or Vertex Component Analysis (VCA) to automatically identify the pixels with the purest spectrum from the artifact's hyperspectral image data as endmember spectra. Further, a linear spectral mixing model can be used to calculate the proportion of each endmember in each pixel. Finally, a corresponding abundance distribution map is generated for each endmember, where the pixel value represents the proportion of that endmember at that location. The above is merely an illustrative example and does not constitute a limitation on all possible cases for determining the endmember abundance map of the artifact to be restored; it is simply not exhaustive.

[0068] Among them, material characteristic information refers to the distribution of physical and chemical material properties in different areas of the surface of cultural relics, including pigment types, mineral composition, distribution range and content ratio, etc.

[0069] In this embodiment, endmember extraction is performed first. The extracted endmembers are matched with pure endmember spectral curves from existing spectral libraries or laboratory spectral measurements to determine the type or main component of the endmembers. Then, abundance estimation is performed, statistically analyzing the abundance vector of each pixel to identify the dominant endmember category, or retaining the proportion information of multiple endmembers. Next, based on the matched endmember categories, the abundance image of the corresponding region is retrieved, and the material category and its content data at each location are recorded to generate a two-dimensional material distribution map. Finally, the material feature information is bound to its corresponding spatial coordinates to obtain material feature information that can be fused with a three-dimensional spatial model.

[0070] For example, VCA can be used to extract endmembers from the radiometrically corrected spectral data to determine the characteristic spectra of different substances in the artifact. Then, non-negative constrained least squares is used to calculate the abundance of various endmembers in each pixel of the radiometrically corrected spectral image to improve the inversion accuracy and obtain an endmember abundance map. Finally, the endmember abundance map is interpreted, and material characteristic information of the artifact, such as pigment composition, disease distribution, or material characteristics, is extracted based on the obtained endmember abundance map. In particular, a point spectrometer can be used to measure high-precision spectra at specific points on-site, and the spectral curve of the "pure" sample measured by the point spectrometer can be compared with the endmember spectral curve extracted from the image to verify its accuracy. At the same time, the high-precision details of the point spectra can also supplement and correct the endmember library, providing more detailed information for the spectral research of artifacts.

[0071] The above is merely an illustrative example and is not intended to limit all possible scenarios for determining material characteristics; it is simply not an exhaustive list.

[0072] Point cloud data is a set of three-dimensional coordinate points obtained through three-dimensional scanning or photogrammetry. The coordinates of the points are usually represented by (X,Y,Z), and each point represents a small spatial location on the surface of the cultural relic.

[0073] In this embodiment, scanning devices such as laser scanners, structured light scanners, stereo vision systems, or multi-view photogrammetry can be used. During the scanning process, the scanner parameters and scanning path can be reasonably set according to the actual situation such as the size and shape of the cultural relic. The scanning device can collect surface data of the cultural relic from multiple angles to cover all visible areas, ensuring that the acquired point cloud data has high accuracy and completeness. Finally, the point clouds from different scanning positions are aligned to the same coordinate system through the scanner's built-in registration function or calibration plate to obtain the initial point cloud data. The above is only an illustrative example and is not intended to limit all possible situations for obtaining the initial point cloud data of the cultural relic to be restored; it is simply not exhaustive.

[0074] In this embodiment, the acquired initial point cloud data can be preprocessed. The preprocessing process may include: noise removal, i.e., using statistical filtering, radius filtering, or other methods to remove isolated points or points with scanning errors; defect repair, i.e., locally resampling or adding points to areas with holes appearing during scanning; coordinate projection and alignment, i.e., translating and scaling the point cloud to a standard coordinate system range to facilitate subsequent algorithm processing; density homogenization, i.e., reducing the number of points in areas with excessively high density and interpolating points to add points in areas with insufficient density; and boundary detection and clipping, i.e., deleting background points or irrelevant points introduced during scanning. After these preprocessing steps, usable point cloud data can be obtained. The above is merely an illustrative example and does not constitute a limitation on all possible situations for obtaining usable point cloud data; it is simply not exhaustive.

[0075] Among them, surface reconstruction is the process of connecting discrete point cloud data into a continuous three-dimensional surface mesh.

[0076] In this embodiment, surface normal information can first be estimated for each point in the available point cloud data to determine the surface orientation. Then, surface reconstruction can be performed using algorithms such as Poisson reconstruction and progressive triangulation to convert the point cloud data into a continuous triangular mesh, forming the overall surface. Finally, mesh smoothing, redundant face removal, and topology repair can be performed to eliminate noise and improve mesh renderability, ultimately obtaining a three-dimensional geometric model. Exemplarily, the process of constructing a three-dimensional geometric model can also involve first using a point cloud simplification algorithm to significantly reduce the data volume while retaining key geometric features, thereby improving subsequent computational efficiency. Subsequently, surface reconstruction is performed using the simplified point cloud data, converting the discrete point cloud data into a continuous triangular mesh surface, thus constructing a preliminary three-dimensional geometric model of the artifact. Specifically, during the modeling process, for the unique features and structures of cultural relics such as the sharp patterns of bronzes, the smooth curved surfaces of porcelains, or the undulating brushstrokes of murals, techniques such as manual editing, feature line extraction, and local mesh optimization can be used to finely adjust and repair the model. This eliminates voids caused by scanning blind spots and flaws caused by smoothing noise, and ensures that the model's geometry and size proportions closely match the real cultural relic, ultimately resulting in a three-dimensional geometric model of the relic. The above is merely an illustrative example and does not represent all possible scenarios for obtaining a three-dimensional geometric model; it is simply not exhaustive.

[0077] In this way, by integrating geometric modeling, photogrammetry, and hyperspectral analysis technologies, high-precision integrated acquisition of the morphology and material information of cultural relics is achieved. It can first accurately construct a three-dimensional geometric model from the initial point cloud through preprocessing and surface reconstruction, and then use digital images to generate a three-dimensional spatial model with realistic textures. At the same time, it combines hyperspectral imaging to extract and analyze the spatial distribution and proportion of different materials, organically combining geometric shape, surface texture, and material characteristics. This not only comprehensively and realistically reproduces the appearance and material composition of cultural relics, but also provides accurate, quantitative, and reusable data support for digital restoration, scientific research, virtual display, and protective treatment.

[0078] In some embodiments, determining the features of the cultural relic to be restored based on the full information model includes: performing color correction on the full information model, then using superpixels to segment the image to obtain an initial segmentation result; performing multi-scale morphological gradient reconstruction and clustering on the initial segmentation result to generate local spatial features and global color features; performing multi-scale feature extraction on the full information model and adopting a fusion strategy to generate fused features; and using the local spatial features, global color features, and fused features as the features of the cultural relic to be restored.

[0079] A superpixel is an image pre-segmentation unit that divides an image into sets of pixels that are adjacent in color, texture, and spatial location, making each superpixel region relatively consistent in color and texture. This reduces the complexity of subsequent processing while preserving object boundary information.

[0080] In this embodiment, a reference color chart or standard whiteboard model can be established first to correct the color of the texture image of the full-information model, eliminating differences in shooting lighting or sensor response. Then, using superpixel algorithms such as Simple Linear Iterative Clustering (SLIC), the color-corrected texture image is divided into multiple spatial regions based on color features to obtain the initial segmentation result. Using superpixels instead of single-pixel processing can reduce computational load, improve feature extraction efficiency, and obtain more consistent color and texture regions while preserving the object's edge structure, providing more stable data blocks for subsequent multi-scale feature analysis. The above is only an illustrative example and is not intended to limit all possible cases for obtaining the initial segmentation result; it is simply not exhaustive.

[0081] In this embodiment, morphological gradients can first be calculated in the boundary regions of the initial segmentation results using multiple structuring elements of different scales (such as 3×3, 5×5, 7×7, etc.). The two-dimensional gradient magnitude maps calculated at each scale are then integrated to obtain a multi-scale morphological gradient map set. Next, using the original gradient map as a label map, noise is smoothed while preserving the main boundary structure, and morphological reconstruction is performed on the two-dimensional gradient magnitude maps at each scale to obtain a multi-scale morphological gradient reconstruction result set. Further, each superpixel region can be located using the initial segmentation results. For each superpixel region, local spatial feature indices such as average gradient, variance, information entropy, and edge strength are calculated from the morphological gradient reconstruction results at each scale. The results are then integrated to obtain local spatial features. Further still, based on the initial segmentation results and the texture image of the full-information model, the color mean, color standard deviation, and color histogram distribution can be statistically analyzed in each superpixel region. These are then integrated to obtain global color features. Finally, a semi-supervised learning algorithm can be used for clustering and grouping, merging the local spatial features and global color features into a multi-dimensional feature vector to obtain the category identifier for each superpixel. The above is merely an illustrative example and is not intended to limit all possible cases of generating local spatial features and global color features; it is simply not an exhaustive list.

[0082] In this embodiment, the process of generating fused features can first be based on the full-information model, calculating the curvature of model vertices, normal vector variability, surface concavity, etc., at different scales, and integrating them to obtain multi-scale geometric features. Then, based on the spectral data in the material feature information of the full-information model, calculating the mean reflectance, spectral angle, band ratio, etc., under different band combinations and neighborhood sizes, and integrating them to obtain multi-scale spectral features. Further, based on the texture image of the full-information model, calculating the contrast, energy, homogeneity, etc., of the Local Binary Pattern (LBP) and Gray Level Co-occurrence Matrix (GLCM) at different scales, and integrating them to obtain multi-scale texture features. Further still, the multi-scale geometric features, multi-scale spectral features, and multi-scale texture features are aligned according to the position index of the full-information model and merged into a unified high-dimensional vector. Finally, the fusion result is normalized to eliminate differences in feature value ranges, resulting in fused features. The above is merely an illustrative example and does not constitute a limitation on all possible cases for generating fused features; it is simply not exhaustive.

[0083] In this embodiment, the process of obtaining the features of the cultural relic to be restored can be as follows: First, the local spatial features, global color features, and fused features are normalized. Then, the three types of features are concatenated into a comprehensive feature vector according to the spatial location index of the full information model, ensuring that each feature vector corresponds to a unique spatial coordinate position or superpixel. Finally, the comprehensive features are bound to the spatial coordinate mapping of the full information model and stored as a data structure that can be directly called by the subsequent classification and restoration decision system, generating a feature set of the cultural relic to be restored. In particular, this feature set of the cultural relic to be restored is a dataset that integrates local spatial features, global color features, and multimodal fused features, and is associated with spatial coordinates. The above is only an illustrative example and is not intended to limit all possible cases of obtaining the features of the cultural relic to be restored; it is simply not exhaustive here.

[0084] Thus, by sequentially performing color correction and superpixel segmentation, multi-scale morphological gradient reconstruction and clustering, multi-modal and multi-scale feature fusion and feature integration based on a full-information model, a high-precision comprehensive feature representation for the cultural relic to be restored is formed. This process not only preserves the detailed boundaries of the cultural relic while extracting multi-dimensional information that includes local spatial structure, global color distribution, and spectral material characteristics, but also achieves accurate correspondence and fusion of multi-source features through spatial location indexing. This allows for a comprehensive, accurate, and quantifiable characterization of the cultural relic's morphological and material features, providing high-quality and robust data support for subsequent damage identification, disease analysis, and restoration decisions.

[0085] In some embodiments, the damage status of the cultural relic to be restored is determined by comparing it with a target benchmark model based on the characteristics of the cultural relic to be restored, including: obtaining the benchmark features of the target benchmark model; aligning the features of the cultural relic to be restored and the benchmark features, and extracting the geometric features of each local region to obtain the local feature parameters of the cultural relic to be restored and the benchmark local feature parameters; monitoring the differences between the local feature parameters of the cultural relic to be restored and the benchmark local feature parameters to obtain difference distribution data; and determining the damage status based on the difference distribution data.

[0086] The target benchmark model refers to a three-dimensional full-information model that can represent the ideal, complete or standard state of a cultural relic. It includes geometric shape, surface texture, color information, material characteristics and spectral characteristics, and is consistent with the cultural relic to be restored in terms of spatial scale, data structure and feature dimensions. It is used as a comparison reference to detect damage and defects.

[0087] Among them, the benchmark features refer to the multi-dimensional feature set extracted from the target benchmark model and used for comparison and judgment with the cultural relic to be restored.

[0088] In this embodiment, multi-scale geometric features, multi-scale spectral features, multi-scale texture features, local spatial features, and global color features, etc., of the target reference model can be extracted first from the target reference model using the same method and scale as the cultural relic to be restored. All extracted features are then used as reference features. The above is merely an illustrative example and is not intended to limit all possible cases of obtaining the reference features of the target reference model; it is simply not exhaustive.

[0089] Specifically, the target benchmark model is a fully informational model that represents the ideal state or standard form of a cultural relic, and it can be constructed using the following methods: First, it can be generated by collecting geometric data and texture images from historical archives, along with material information from historical records, based on archived information such as 3D scan data, high-resolution photographs, and hyperspectral measurement data of the cultural relic during its undamaged or early conservation stages. Registration and fusion techniques can then be used to generate a complete target benchmark model. Second, when a historical model of the target cultural relic is lacking, its original appearance can be inferred from similar artifacts. High-precision data from similar types, specifications, or batches of artifacts can be collected, and a sample that is as close as possible to the target relic in terms of "health" or "good preservation" can be selected. This sample can then undergo full-process 3D modeling and material acquisition to construct a fully informational model, serving as the target benchmark model. Third, it can be generated by collecting the 3D form, texture, and material of existing cultural relics based on fragments, local morphology, and art history and craftsmanship. Experts can then infer missing areas based on craftsmanship, structural symmetry, and artistic design. Finally, 3D modeling software can be used to manually fill in the missing parts, and material and texture information can be assigned to the filled parts to obtain the target benchmark model. The above is merely an illustrative example and is not intended to limit the scope of all possible scenarios for obtaining the target benchmark model; it is simply not an exhaustive list.

[0090] In this embodiment, the features of the artifact to be restored and the reference features can first be preliminarily matched according to the spatial location index in the full information model. When there is an overall deformation or pose difference between the two models, spatial transformation can be performed using Iterative Closest Point (ICP) or feature point-based registration methods. At a single spatial index position or superpixel region corresponding to the two features, geometric features are extracted from the features of the artifact to be restored and the reference features, respectively, to obtain local feature parameters of the artifact to be restored and local feature parameters of the reference. For example, the extracted geometric features may include curvature, normal vector change, boundary line length, surface roughness, etc. The above is merely an illustrative example and does not constitute a limitation on all possible cases for obtaining local feature parameters of the artifact to be restored and local feature parameters of the reference; it is simply not exhaustive.

[0091] Among them, difference monitoring refers to comparing the parameters of the cultural relic to be restored at the corresponding location with the benchmark parameters to detect the degree of difference between them, thereby quantifying the local damage.

[0092] In this embodiment, the geometric feature difference value of each spatial index position or superpixel region corresponding to the two features can first be calculated using the local feature parameters of the artifact to be restored and the reference local feature parameters. Then, all geometric feature difference values ​​are mapped back to spatial positions to generate difference distribution data. For example, the generated difference distribution data can be a three-dimensional difference heatmap. The above is only an illustrative example and is not intended to limit all possible cases of obtaining difference distribution data; it is simply not exhaustive.

[0093] Among them, the damage status refers to the descriptive result of a comprehensive judgment on the type, location and extent of damage to the surface of cultural relics based on differential distribution data.

[0094] In this embodiment, a difference threshold can be preset; for example, if the curvature change exceeds a certain specific value, it is judged as "damage". First, the spatial location is divided and labeled as "normal" and "damaged" areas according to the preset difference threshold, resulting in a preliminary damage distribution label map. Then, the difference feature pattern can be matched with a predefined damage type library to obtain a damage type label map containing spatial location and damage type. Finally, based on the damage type label map and the difference value intensity, a severity score (such as difference percentage, difference magnitude, etc.) can be calculated according to the damage type to obtain a complete damage situation including damage location, damage type, and damage degree. The above is only an illustrative example and is not intended to limit all possible situations in determining the damage situation; it is simply not exhaustive.

[0095] Specifically, the predefined damage type library can be a reference database established in the early stages through the collection, analysis, and classification of a large number of known damage cases of cultural relics. The acquisition process typically includes: collecting actual damage samples of various types of cultural relics (such as missing parts, cracks, deformation, contamination, etc.); extracting corresponding geometric, color, and material difference features using a full-information model; and having cultural relic restoration experts annotate and classify these features, ultimately forming a mapping table of damage types and corresponding feature patterns. This table is used to automatically match the difference features of unknown cultural relics and determine the damage type during the detection phase. The above is merely an illustrative example and does not represent all possible scenarios for the predefined damage type library; it is simply not exhaustive.

[0096] Thus, by acquiring the multidimensional benchmark features of the target benchmark model and performing spatial alignment and local geometric feature analysis with the features of the cultural relic to be restored, accurate detection and visualization of differences are achieved. Finally, the type, location, and extent of damage are determined by combining the difference feature patterns. This process not only ensures the spatial accuracy and comparability of the detection results but also quantifies various damage features, providing a complete, scientific, and repeatable decision-making basis for automated diagnosis, restoration plan formulation, and restoration priority ranking.

[0097] In some embodiments, determining the damage status based on differential distribution data includes: clustering the differential distribution data to obtain the spatial distribution of damaged parts; determining the damage level of any damaged part based on the number of differences in the spatial distribution; and taking the damaged part and damage level as the damage status.

[0098] In this embodiment, the difference values ​​and their corresponding spatial coordinates can be read first based on the difference distribution data to filter out obvious noise points. Then, a clustering algorithm suitable for three-dimensional spatial data (such as a hierarchical clustering algorithm based on Euclidean distance) can be selected to divide the difference points into several clusters according to the spatial proximity of the points. Each cluster represents a potential damaged part, resulting in preliminary damaged part grouping. Finally, the spatial range (such as minimum bounding rectangle, volume, and location center point) of each damaged part group is calculated and added to the spatial distribution information of the damaged parts to obtain damaged part information with spatial distribution description. The above is only an illustrative example and is not intended to limit all possible cases of obtaining the spatial distribution of damaged parts; it is simply not exhaustive.

[0099] In this embodiment, the number of differential points and the average difference magnitude within each damaged area can be counted first. The statistical values ​​are then compared with a threshold, and the damage level is classified according to a set rule. The above is merely an illustrative example and is not intended to limit all possible scenarios for determining the damage level of any damaged area; it is simply not exhaustive.

[0100] In this embodiment, data fusion can be performed on any identical damaged area to integrate its spatial location, extent, statistical differences, and damage level into a complete record as the damage status of that damaged area. Finally, the damage status data of all damaged areas can be stored in a database or exported as a report for visualization and repair decision-making. The above is merely an illustrative example and is not intended to limit the extent of all possible damage scenarios; it is simply not exhaustive.

[0101] Thus, by starting with differential distribution data, the scattered differential points are first spatially clustered to obtain the distribution of damaged parts. Then, the number and magnitude of the differences are used for classification, thereby obtaining a structured damage situation that includes both location, extent, and severity. This process not only improves the readability and operability of the detection results, but also elevates the operation from single-point comparison to a qualitative and quantitative analysis combining parts level, providing a scientific, intuitive, and repeatable basis for repair decisions.

[0102] In some embodiments, a cultural relic restoration plan is generated based on a full-information model and damage conditions, including: determining the mechanical parameter data of the cultural relic to be restored according to the full-information model; discretizing and establishing mechanical and physical equations based on the full-information model and mechanical parameter data to obtain a physical simulation model; simulating the damage evolution process of the cultural relic to be restored using preset damage parameters based on the physical simulation model; and generating a cultural relic restoration plan according to the damage evolution process and preset restoration goals.

[0103] Among them, the mechanical parameter data of the cultural relic to be restored refers to the set of quantitative data reflecting the properties of the cultural relic materials under stress. These parameters determine the mechanical response of the cultural relic under external force, temperature and humidity changes and other conditions, and are the basis for constructing a physical simulation model.

[0104] In this embodiment, material and spectral features are first used to segment different material regions on the model and mark the corresponding damaged areas. Then, the material labels are matched with a material mechanics database to read the initial values ​​of the corresponding mechanical parameters. For areas with measured data, these values ​​are replaced with the measured values. Finally, each mechanical parameter is mapped to the corresponding mesh cell or vertex of the full-information model, forming spatially distributed parameter data, thus obtaining the mechanical parameter data of the artifact to be restored. The above is merely an illustrative example and does not constitute a limitation on all possible scenarios for determining the mechanical parameter data of the artifact to be restored; it is simply not exhaustive.

[0105] Among them, the physical simulation model is a discretized numerical model that combines three-dimensional geometry, material mechanical properties and physical equations, and is used to calculate the stress, strain and damage evolution process of cultural relics under different conditions.

[0106] In this embodiment, the full-information model can first be divided into finite element meshes or voxel meshes, allowing each element to inherit corresponding mechanical parameters, resulting in a discretized model of the cultural relic to be restored with mechanical parameters. Then, appropriate equations (such as elasticity equations, thermo-elastic coupling equations, fracture mechanics equations, etc.) can be selected based on different physical processes, and boundary conditions (such as support location, external force application, environmental temperature and humidity changes, etc.) can be defined to obtain a physical simulation model containing geometric meshes, material parameters, boundary conditions, and equation descriptions. The above is merely an illustrative example and is not intended to limit all possible scenarios for obtaining the physical simulation model; it is simply not exhaustive.

[0107] Damage parameters are physical quantities used to describe the occurrence and spread of damage. In this embodiment, damage parameters are mathematical variables used to quantify the damage to cultural relics.

[0108] Among them, damage evolution refers to the process by which damage to cultural relics expands and changes over time and space under the influence of external forces, environment and other factors.

[0109] In this embodiment, initial damage parameters can first be assigned to the corresponding parts in the physical simulation model according to the damage level. Then, physical equations are solved under set loads or environmental conditions to track crack initiation and propagation, material yielding and failure processes. Finally, through this simulation solution process, the damage evolution process of the cultural relic under different environmental factors is simulated, and the trend of disease development is predicted. The above is only an illustrative example and is not intended to limit all possible situations in simulating the damage evolution process of the cultural relic to be restored; it is simply not exhaustive.

[0110] In this embodiment, the damage to certain parts can first be analyzed to determine which areas, under future trends, will significantly impact the structural integrity or appearance of the cultural relic, thus establishing the scope and priority of repair. Then, based on an existing database of repair techniques, and considering the cultural relic's value, style, compatibility of repair materials with the relic's material, and pre-defined repair goals, the most suitable repair method is selected for each part based on its material, damage type, and evolutionary trend. Finally, the repair plans for each part are integrated to obtain the final repair plan. The above is merely an illustrative example and does not represent a limit to all possible scenarios for generating a cultural relic repair plan; it is simply not exhaustive.

[0111] Specifically, the restoration plan can be overlaid into the physical simulation model, and numerical calculations can be performed under set boundary conditions to dynamically simulate the real-time changes of physical quantities such as stress, strain, and displacement during the restoration process. After the simulation is completed, the simulation results are used to evaluate the restoration effect and determine whether the structural strength, stability, and appearance of the cultural relic meet the preset restoration goals. This allows for restoration verification and plan optimization in a virtual environment.

[0112] Thus, starting from a fully informational model, the material mechanical parameters of the cultural relic are accurately extracted, and a physical simulation model combining geometric and physical equations is constructed. Subsequently, damage parameters are used to simulate the future damage evolution process, and finally, a visualized solution is generated based on the restoration goals. This process deeply integrates the digital information of the cultural relic with mechanical simulation, realizing an integrated closed loop from damage detection, structural prediction, and restoration decision-making. This not only improves the scientific rigor, relevance, and foresight of the restoration plan but also reduces the risk of physical intervention in the cultural relic.

[0113] In some embodiments, a cultural relic restoration plan is generated based on a full information model and the damage condition, including: determining the basic characteristics of the cultural relic to be restored according to the full information model; extracting the closest target sample from a pre-built sample library according to the basic characteristics; and using the target sample as a template to geometrically complete the damage condition and generate a cultural relic restoration plan.

[0114] In this embodiment, geometric features, texture features, color features, and size and proportion information can be extracted from the full-information model first. Then, the different features are normalized so that they can be compared in a unified metric space. Finally, the processed feature data is used as the basic features of the cultural relic to be restored. The above is merely an illustrative example and does not constitute a limitation on all possible scenarios for determining the basic features of the cultural relic to be restored; it is simply not exhaustive.

[0115] In this embodiment, similar intact cultural relic samples can be collected in advance, and digital modeling and feature extraction can be performed on them to construct a sample library containing full-information models and feature vectors of different cultural relic categories. Then, distance metric algorithms (such as cosine similarity, Euclidean distance, or weighted feature distance) can be used to calculate the similarity between the cultural relic to be restored and each sample in the sample library. The sample with the highest similarity is selected as the target sample, and its full-information model data is obtained. The above is only an illustrative example and is not intended to limit all possible cases for determining the closest target sample; it is simply not exhaustive.

[0116] Geometric completion refers to using the complete morphological information of the target sample to reconstruct the three-dimensional shape of the missing or damaged parts of the cultural relic to be restored, so that the model can be restored to a state as close as possible to the original appearance.

[0117] In this embodiment, a global registration algorithm (such as ICP or rigid transformation based on feature points) can be used to align the target sample to the coordinate system of the artifact to be restored, unifying the coordinate systems of the two sets of models. Then, the damaged areas on the full-information model of the artifact to be restored are marked according to the damage location, and the corresponding geometric regions are found in the target sample. Next, the corresponding geometric data in the target sample is copied and deformed to ensure seamless splicing onto the artifact to be restored. Finally, the restoration plan can be optimized, such as by smoothing boundaries, unifying the mesh to avoid geometric abrupt changes and cracks, and fine-tuning textures and colors to ensure the artifact appears natural and harmonious after restoration. The data from the restoration process is then organized to obtain the final restoration plan. The above is merely an illustrative example and does not represent all possible scenarios for generating artifact restoration plans; it is simply not exhaustive.

[0118] Thus, by transforming the full-information model into standardized features, and combining it with a pre-built sample library to retrieve the closest complete template, and then using this template to geometrically complete the damaged parts, a restoration plan with a high degree of consistency in form and style can be quickly generated. This method reduces subjective assumptions in the restoration process, improves automation and efficiency, and ensures that the restoration plan is both scientific and conforms to the aesthetic characteristics of the original appearance of the cultural relic.

[0119] Example 2:

[0120] In this embodiment, Figure 2This illustration shows another flowchart of the method for generating cultural relic restoration schemes provided in an embodiment of the present invention, such as... Figure 2 The following are included:

[0121] S211. Acquire or obtain digital image data of cultural relics. The image has high spatial resolution and low spectral resolution, and can be a regular digital photographic image or a multispectral image.

[0122] S212. Preprocess the digital image data. Preprocessing may include denoising, color correction, radiometric correction, and necessary image registration to obtain high-quality image data that meets the requirements of subsequent processing. A three-dimensional spatial model of the cultural relic can then be constructed based on the processed digital image data.

[0123] S213. Perform multi-scale morphological gradient reconstruction on digital image data. Based on the defined multi-scale morphological gradient reconstruction function, calculate the morphological gradients of the digital image at multiple scales, and fuse the gradient maps from each scale to obtain a superpixel image with more accurate contours and enhanced boundaries. This result can provide ideal initial segmentation size and boundary information for subsequent kernel-based improved fuzzy C-means clustering. For example, the multi-scale morphological gradient reconstruction function... It can include:

[0124]

[0125] In the formula, This is the original image. For tagged images, As a structural element, Indicates the scale of the smallest region. Indicates the scale of the largest region. This is a closing operation.

[0126] S214. Superpixel segmentation optimization of digital images. Based on the results of multi-scale morphological gradient reconstruction, superpixel segmentation is performed on the images, and the segmentation accuracy is improved by using a kernel-based improved fuzzy C-means clustering algorithm, which aggregates pixels of the same type into uniform regions, while reducing the number of colors and facilitating the fusion of local spatial information and global color features.

[0127] S215. Convert the superpixel segmentation results into labeled images. Assign a unique code to each superpixel region for subsequent multimodal feature extraction and clustering.

[0128] S216. Feature Image Extraction. Based on the labeled image, extract features such as color, local spatial distribution, and texture from the digital image at the superpixel scale, and generate corresponding feature images. These feature images intuitively represent the local spatial and global color features of the digital image and are used for subsequent feature mapping and alignment with the hyperspectral feature image.

[0129] S221. Collect hyperspectral image data of cultural relics. This image has high spectral resolution and low spatial resolution, and can reflect the spectral characteristics of the material of the cultural relics.

[0130] S222. Hyperspectral Image Data Preprocessing. Preprocessing operations such as radiometric correction, noise suppression, band selection, and geometric registration are performed on the acquired hyperspectral images to eliminate the influence of instrument and environmental factors and improve the accuracy and stability of spectral information.

[0131] S223a. Hyperspectral image data upsampling. The preprocessed hyperspectral image is upsampled to match the spatial resolution of the digital image, preparing it for subsequent feature mapping.

[0132] S223b. Using labeled images, spectral, spatial, and textural features are extracted from hyperspectral images according to superpixel regions, and a multi-scale superpixel nonlinear feature extraction method is combined to enhance the ability to distinguish different materials. Based on the processed hyperspectral images, endmember spectra are extracted using vertex component analysis to determine the characteristic spectra corresponding to different substances in the cultural relics; then, non-negative constrained least squares method is used for abundance inversion to obtain the endmember abundance map of each endmember in the image; based on the endmember abundance map, material feature information of the cultural relics is extracted and identified, such as pigment composition distribution, diseased areas, or specific material partitions.

[0133] For example, suppose the original data contains A dataset of samples, in which For the first One original sample, For the first One original sample, multi-scale superpixel nonlinear feature extraction algorithm The formula can include:

[0134]

[0135] Among them, the clustering prototype in the algorithm This can be represented as a weighted sum of mapped data. Its iterative formula can include:

[0136]

[0137]

[0138]

[0139] In the formula, Indicates color level, 1≤ ≤ , The number of clusters for the superpixel image. and All are positive integers; This represents the number of samples in the dataset. For the first The number of pixels in each region; For mapping functions; and They are the first The first original sample and the first Each original sample is mapped to the pixels measured in kernel space; Color levels For the first Fuzzy membership degree of each cluster center; As a weighting factor; These are Gaussian radial basis functions; Represents a pixel; This represents the Gaussian radial basis function corresponding to a pixel. Represents the transpose of a matrix; Indicates after the first The data obtained after the next iteration.

[0140] S224. Establish a mapping relationship between digital image data and hyperspectral data. Based on material feature information, upsampled hyperspectral image data, and a 3D spatial model, spatial registration and coordinate transformation are used to unify the spatial coordinate system of the material feature information and the 3D spatial model, thereby establishing a mapping relationship between the two and providing a mapping foundation for subsequent cross-modal fusion.

[0141] S225. Perform cross-modal information fusion on digital image data and hyperspectral data. Based on the mapping relationship, the texture features of the 3D spatial model and the material feature information of the hyperspectral image can be fully extracted and fused through an iterative method. Then, appropriate strategies are adopted according to the importance and correlation of features to perform cross-modal information fusion and obtain a fused image model.

[0142] S226. Unwrap the fused image model using UV mapping to generate a texture map.

[0143] S231. Collect three-dimensional point cloud data of cultural relics and obtain geometric morphology information of the surface of cultural relics through laser scanning or structured light scanning.

[0144] S232. Preprocess the original point cloud, such as denoising, simplifying redundant points, repairing holes, and unifying the coordinate system, to ensure the point cloud quality required for model reconstruction.

[0145] S233. Based on the preprocessed point cloud data, surface reconstruction is performed to generate a triangular mesh geometric model, resulting in a three-dimensional geometric model without texture information.

[0146] S234. Establish the mapping relationship between two-dimensional texture maps and three-dimensional geometric models.

[0147] S235. Texture mapping based on mapping relationships. The texture material map is mapped onto the surface of the 3D geometric model, achieving precise matching between image texture, material feature information, and model geometry, generating a 3D textured geometric material model. Based on the 3D textured geometric material model, texture mapping is refined and color difference correction is performed using professional 3D modeling software to ensure the model accurately reproduces the details of the cultural relic, obtaining a complete information model of the cultural relic.

[0148] S236. Generate a cultural relic restoration plan and virtually recreate the cultural relic. Based on the full-information model of the cultural relic and all the extracted cultural relic feature information, compare and analyze the damage of the cultural relic with the target benchmark model, and generate a cultural relic restoration plan based on the full-information model and damage condition. Finally, based on the cultural relic restoration plan, use restoration methods based on physical models or sample texture completion to recreate a virtual restored 3D cultural relic model with historical rationality and visual realism.

[0149] It should be understood that Figure 2 The schematic diagrams shown are merely illustrative and not limiting, and are scalable; those skilled in the art can use them as a basis. Figure 2 Even with various obvious changes and / or substitutions to the examples, the resulting technical solutions still fall within the scope of this disclosure.

[0150] Example 3:

[0151] In this embodiment, a digital management platform can be constructed to support the integrated release and hosting of multi-source data. Various data related to cultural relics, such as digital images, spectral data, point cloud data, and 3D models, can be uniformly managed and stored. Through the platform, users can easily query, browse, and download relevant data, enabling data sharing and utilization. Furthermore, a cultural relic protection and restoration archive can be established on the platform, recording information such as the protection history, restoration process, and effect evaluation of the cultural relics. Users can view the protection and restoration status of cultural relics through the platform, understand their current state and development trends, and provide a basis for cultural relic protection decisions. Even further, the constructed 3D model of the cultural relic can be integrated with the real-world environment, allowing users to experience the historical features and cultural connotations of the cultural relic immersively by wearing virtual reality devices. During the display, users can interact with the cultural relic, such as zooming in, zooming out, and rotating, observing the details of the cultural relic from different angles, enhancing the user experience.

[0152] Example 4:

[0153] This disclosure provides a system for generating cultural relic restoration plans, such as... Figure 3As shown, the system may include: a model acquisition module 301, used to acquire a full-information model of the cultural relic to be restored; a feature extraction module 302, used to determine the features of the cultural relic to be restored based on the full-information model; a damage determination module 303, used to compare the features of the cultural relic to be restored with a target benchmark model to determine the damage status of the cultural relic to be restored; and a virtual restoration module 304, used to generate a cultural relic restoration plan based on the full-information model and the damage status.

[0154] In some embodiments, the model acquisition module 301 includes: a fusion image submodule, used to fuse information by unifying the resolution scale of the 3D spatial model and material feature information to obtain a fusion image model; an image unfolding submodule, used to unfold the fusion image model in a plane to obtain a texture map; a feature mapping submodule, used to map the texture map onto the surface of the 3D geometric model based on a mapping relationship to obtain a 3D texture geometric material model; the mapping relationship is obtained by performing coordinate transformation on the texture map and the 3D geometric model; and a model adjustment submodule, used to perform consistency adjustment on the 3D texture geometric material model before rendering to obtain a full-information model.

[0155] In some embodiments, the digital image modeling module ( Figure 3 (Not shown in the image), used to acquire digital image data of the cultural relic to be restored; based on the digital image data, determine the close-range photogrammetric image pairs and relative control measurement control points; based on the close-range photogrammetric image pairs and relative control measurement control points, generate an initial spatial model; the digital image data is mapped as texture onto the surface of the initial spatial model to obtain a three-dimensional spatial model. Material feature module ( Figure 3 (Not shown in the image), used to acquire hyperspectral image data of the cultural relic to be restored; after radiometric correction of the hyperspectral image data, endmember extraction is used to determine the endmember abundance map of the cultural relic to be restored; material feature information is determined based on the endmember abundance map. Point cloud modeling module ( Figure 3 (Not shown in the image) is used to obtain the initial point cloud data of the cultural relic to be restored; the initial point cloud data is preprocessed to obtain usable point cloud data; based on the usable point cloud data, a three-dimensional geometric model is obtained by surface reconstruction.

[0156] In some embodiments, the feature extraction module 302 includes: a superpixel segmentation submodule, used to perform image segmentation using superpixels after color correction of the full-information model to obtain an initial segmentation result; a reconstruction and clustering submodule, used to perform multi-scale morphological gradient reconstruction and clustering on the initial segmentation result to generate local spatial features and global color features; a fusion feature submodule, used to extract features from the full-information model at multiple scales and use a fusion strategy to generate fused features; and a feature generation submodule, used to use the local spatial features, global color features, and fused features as features of the cultural relic to be restored.

[0157] In some embodiments, the damage determination module 303 includes a baseline acquisition submodule, used to acquire baseline features of the target baseline model; a feature alignment submodule, used to align the features of the artifact to be restored and the baseline features, and then extract the geometric features of each local region to obtain local feature parameters of the artifact to be restored and local feature parameters of the baseline; a difference monitoring submodule, used to monitor the differences between the local feature parameters of the artifact to be restored and the local feature parameters of the baseline, and obtain difference distribution data; and a damage result submodule, used to determine the damage status based on the difference distribution data.

[0158] In some embodiments, the damage result submodule is used to determine the damage status based on differential distribution data, including: clustering the differential distribution data to obtain the spatial distribution of damaged parts; determining the damage level of any damaged part based on the number of differences in the spatial distribution; and taking the damaged part and damage level as the damage status.

[0159] In some embodiments, the virtual restoration module 304 includes: a mechanical parameter submodule, used to determine the mechanical parameter data of the cultural relic to be restored based on the full information model; a simulation model submodule, used to discretize and establish mechanical and physical equations based on the full information model and mechanical parameter data to obtain a physical simulation model; a damage evolution submodule, used to simulate the damage evolution process of the cultural relic to be restored using preset damage parameters based on the physical simulation model; and a scheme generation submodule, used to generate a cultural relic restoration scheme based on the damage evolution process and preset restoration goals.

[0160] In some embodiments, the virtual restoration module 304 further includes: a basic feature submodule, used to determine the basic features of the cultural relic to be restored based on the full information model; a sample extraction submodule, used to extract the closest target sample from a pre-built sample library based on the basic features; and a geometric completion submodule, used to perform geometric completion on the damage using the target sample as a template to generate a cultural relic restoration plan.

[0161] The specific functions and examples of each module and submodule of the system in this disclosure embodiment can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0162] Example 5:

[0163] Figure 4 This invention provides another structural schematic diagram of the cultural relic restoration scheme generation system, as shown in the embodiment of the invention. Figure 4 As shown, the system may include:

[0164] The data acquisition subsystem 410 includes: a digital image unit 411 for acquiring digital image data of cultural relics; a hyperspectral image unit 412 for acquiring hyperspectral image data; a point spectral unit 413 for acquiring point spectral image data, which can be used to verify the accuracy of hyperspectral image analysis results; and a point cloud data unit 414 for acquiring point cloud data and preprocessing the data.

[0165] The digital modeling subsystem 420 includes: a three-dimensional spatial model unit 421, used to construct a three-dimensional spatial model based on digital image data; an orthophoto unit 422, used to generate an orthophoto based on digital image data and the three-dimensional spatial model; a three-dimensional geometric model unit 423, used to construct a three-dimensional geometric model based on point cloud data; and a texture feature unit 424, used to construct a three-dimensional spatial model of the cultural relic based on digital image data and obtain the original texture features of the cultural relic.

[0166] The model analysis subsystem 430 includes: an image preprocessing unit 431 for preprocessing digital image data, hyperspectral image data, and point cloud data; a superpixel segmentation optimization unit 432 for performing superpixel segmentation and optimization on the digital image data; and a hyperspectral image analysis unit 433 for dividing the hyperspectral image into multi-scale regions based on spectral spatial distribution characteristics, extracting spectral features, spatial features, and texture features within each region, and generating a hyperspectral feature image. Subsequently, endmember spectra are extracted from the hyperspectral feature image to determine the characteristic spectra of different substances. Then, the endmember abundance distribution is inverted using the non-negative constrained least squares method to obtain an endmember abundance map, and the corresponding material feature information is extracted. A cross-modal information interaction unit 434 receives the material feature information extracted from the 3D spatial model and the hyperspectral image. A resampling unit 435 performs upsampling operations on the preprocessed hyperspectral image to make its spatial resolution consistent with the digital image. A multi-scale feature mapping unit 436 performs multi-scale feature mapping based on the 3D spatial model, material feature information, and the resampled hyperspectral image, establishing a one-to-one mapping relationship. Feature reconstruction unit 437 is used to construct a full-information model of the cultural relic based on mapping relationships, the original texture of the relic, the 3D spatial model, and material feature information. Virtual restoration unit 438 is used to compare and analyze the damage of the cultural relic with the target benchmark model based on the full-information model and all extracted feature information of the cultural relic, and generate a cultural relic restoration plan based on the full-information model and damage condition. Finally, based on the cultural relic restoration plan, the full-information model of the cultural relic is restored and completed using a restoration method based on physical models or sample texture completion, reproducing a virtual restored 3D cultural relic model with historical rationality and visual realism.

[0167] The interactive display subsystem 440 includes: a cultural relic overview display unit 441, used to display the general information of the cultural relic; a historical restoration information display unit 442, used to establish and display information such as cultural relic protection and restoration archives, the history of cultural relic protection, restoration process, and effect evaluation; a historical and cultural information display unit 443, used to display historical and cultural information related to the cultural relic; and a special construction information display unit 444, used to display various data generated by the cultural relic data acquisition subsystem, digital modeling subsystem, and model analysis subsystem, such as digital images, spectral data, point cloud data, and 3D models. A virtual reality display unit 445 is used to integrate the constructed 3D model of the cultural relic with the real-world environment, allowing users to experience the historical features and cultural connotations of the cultural relic immersively by wearing virtual reality devices.

[0168] Example 6:

[0169] This embodiment provides a device for generating cultural relic restoration plans. Figure 5 A schematic block diagram of an apparatus 500 that can be used to implement embodiments of the present disclosure is shown.

[0170] like Figure 5 As shown, the device 500 includes a computing unit 501, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. The RAM 503 may also store various programs and data required for the operation of the device 500. The computing unit 501, the ROM 502, and the RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0171] Multiple components in device 500 are connected to I / O interface 505, including: input unit 506, such as keyboard, mouse, etc.; output unit 507, such as various types of displays, speakers, etc.; storage unit 508, such as disk, optical disk, etc.; and communication unit 509, such as network card, modem, wireless transceiver, etc. Communication unit 509 allows device 500 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0172] The computing unit 501 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. The computing unit 501 performs the various methods and processes described above, such as a method for generating an artifact restoration scheme. For example, in some embodiments, the multi-label feature selection method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or mounted on device 500 via read-only memory 502 and / or communication unit 509. When the computer program is loaded into random access memory 503 and executed by computing unit 501, one or more steps of the multi-label feature selection method described above can be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform a method for generating an artifact restoration scheme by any other suitable means (e.g., by means of firmware).

[0173] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating a cultural relic restoration plan, characterized in that, The method includes: A complete information model of the cultural relic to be restored is obtained; the complete information model is obtained based on a three-dimensional spatial model, material characteristic information, and a three-dimensional geometric model. The material feature information is specifically established in the following manner: Acquire hyperspectral image data of the cultural relic to be restored; After preprocessing the hyperspectral image data, an upsampling operation is performed. Using labeled images, spectral, spatial, and texture features are extracted from the hyperspectral images according to superpixel regions. A multi-scale superpixel nonlinear feature extraction algorithm is combined to enhance the ability to distinguish different materials. The vertex component analysis method is used to extract endmember spectra, and the characteristic spectra corresponding to different substances are determined. Then, the non-negative constrained least squares method is used to perform abundance inversion to obtain the endmember abundance map of each endmember in the image. The marked image is obtained by acquiring digital image data of the cultural relic to be restored, preprocessing it, then performing multi-scale morphological gradient reconstruction, then superpixel segmentation optimization, and finally converting it. The multi-scale superpixel nonlinear feature extraction algorithm Specifically, it includes: ,in, Indicates the weights of the mapped data. Indicates color level, 1≤ ≤ , The number of clusters in the superpixel image; This represents the number of samples in the dataset. For the first The number of pixels in each region; Color levels For the Fuzzy membership degree of each cluster center; As a weighting factor; For mapping functions; and They are the first The first original sample and the first Each original sample is mapped to the pixels measured in kernel space; The material characteristic information is determined based on the endmember abundance map; Based on the full information model, the characteristics of the cultural relic to be restored are determined; the characteristics of the cultural relic to be restored include at least local spatial characteristics, global color characteristics, and fusion characteristics; Based on the characteristics of the cultural relic to be restored, a comparison is made with a target benchmark model to determine the extent of damage to the cultural relic; specifically including: Obtain the baseline features of the target baseline model; After aligning the features of the cultural relic to be restored with the reference features, the geometric features of each local region are extracted to obtain the local feature parameters of the cultural relic to be restored and the reference local feature parameters. Difference monitoring was conducted between the local characteristic parameters of the cultural relic to be restored and the baseline local characteristic parameters to obtain difference distribution data; The damage condition is determined based on the differential distribution data; Based on the full information model and the damage situation, a cultural relic restoration plan is generated.

2. The method according to claim 1, characterized in that, The method for obtaining a full-information model of the cultural relic to be restored includes: After unifying the resolution scale of the three-dimensional spatial model and material feature information, information fusion is performed to obtain a fused image model. The fused image model is unfolded in a plane to obtain a texture map; Based on the mapping relationship, the texture material map is mapped onto the surface of the three-dimensional geometric model to obtain a three-dimensional texture geometric material model; the mapping relationship is obtained by performing coordinate transformation on the texture material map and the three-dimensional geometric model; After performing consistency adjustments on the three-dimensional texture geometry material model, it is rendered to obtain the full-information model.

3. The method according to claim 2, characterized in that, The three-dimensional spatial model is established in the following way: Obtain digital image data of the cultural relic to be restored; Based on the digital image data, determine the results of close-range photogrammetry image pairs and relative control measurement control points; An initial spatial model is generated based on the close-range photogrammetry image pairs and the results of the relative control measurement control points. The digital image data is mapped as a texture onto the surface of the initial spatial model to obtain the three-dimensional spatial model; The three-dimensional geometric model is established in the following way: Obtain the initial point cloud data of the cultural relic to be restored; The initial point cloud data is preprocessed to obtain usable point cloud data; Based on the available point cloud data, the three-dimensional geometric model is obtained by surface reconstruction.

4. The method according to claim 1, characterized in that, The process of determining the characteristics of the cultural relic to be restored based on the full information model includes: After color correction of the texture image of the full information model, image segmentation is performed using superpixels to obtain the initial segmentation result; The initial segmentation results are subjected to multi-scale morphological gradient reconstruction and clustering to generate local spatial features and global color features; Multi-scale feature extraction is performed on the full-information model, and a fusion strategy is adopted to generate fused features; The local spatial features, the global color features, and the fusion features are used as the features of the cultural relic to be restored.

5. The method according to claim 1, characterized in that, Determining the damage status based on the differential distribution data includes: Clustering the differentially distributed data yields the spatial distribution of the damaged areas; The damage level of any damaged part is determined based on the number of differences in spatial distribution. The damaged part and the damage level are defined as the damage condition.

6. The method according to claim 1, characterized in that, The process of generating a cultural relic restoration plan based on the full information model and the damage status includes: Based on the full information model, the mechanical parameters of the cultural relic to be restored are determined; Based on the full information model and the mechanical parameter data, the mechanical and physical equations are discretized and established to obtain a physical simulation model. Based on the physical simulation model, the damage evolution process of the cultural relic to be repaired is simulated using preset damage parameters. Based on the damage evolution process and the preset restoration goals, the cultural relic restoration plan is generated.

7. The method according to claim 1, characterized in that, The process of generating a cultural relic restoration plan based on the full information model and the damage status includes: Based on the aforementioned full-information model, the basic characteristics of the cultural relic to be restored are determined; Based on the aforementioned basic characteristics, the closest target sample is extracted from a pre-built sample library; Using the target example as a template, the damage is geometrically completed to generate the cultural relic restoration plan.

8. A system for generating cultural relic restoration schemes using the method described in any one of claims 1-7, characterized in that, The system includes: The model acquisition module is used to acquire a full-information model of the cultural relic to be restored; The feature extraction module is used to determine the features of the cultural relic to be restored based on the full information model. The damage determination module is used to determine the damage status of the cultural relic to be repaired by comparing it with a target benchmark model based on the characteristics of the cultural relic to be repaired. The virtual restoration module is used to generate a cultural relic restoration plan based on the full information model and the damage condition.

9. A device for generating cultural relic restoration plans, characterized in that, include: At least one processor; And a memory communicatively connected to at least one processor; wherein the memory stores instructions executable by at least one processor, the instructions being executed by at least one processor to enable at least one processor to perform the method of any one of claims 1-7.

Images