Cultural relic non-contact splicing restoration method and device based on three-dimensional point cloud model data

Abstract

The invention discloses a cultural relic non-contact splicing repair method and device based on three-dimensional point cloud model data, and relates to the technical field of cultural relic repair. The method comprises a cultural relic three-dimensional point cloud data acquisition and preprocessing step, a cultural relic digital twin construction and virtual environment mapping step, a multi-parameter virtual splicing simulation step, a repaired structure stability prediction step and a repair scheme comparison and optimal screening step. According to the method, the digital twins of the cultural relics are constructed, various splicing sequences and repairing parameters are simulated in a virtual environment, multiple sets of preliminary splicing schemes are generated, a scoring system is constructed by using an analytic hierarchy process, comprehensive comparison and optimal screening are performed on all the sets of schemes, and repairing personnel are assisted to quickly determine the optimal repairing scheme; parameters are dynamically adjusted through a response surface method and a particle swarm optimization algorithm, dynamic optimization of a repairing scheme is achieved, and the repairing efficiency and quality are further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cultural relic restoration, in particular to a cultural relic non-contact splicing restoration method and device based on three-dimensional point cloud model data. BACKGROUND

[0002] With the continuous development of cultural relic protection and restoration technology, how to efficiently and accurately restore damaged cultural relics has become a key challenge in the industry. Traditional cultural relic restoration methods rely heavily on the experience and manual operation of restorers. For complex damaged cultural relics, especially those with numerous fragments and different shapes, traditional methods are not sufficient.

[0003] Traditional cultural relic restoration techniques have the following shortcomings: first, traditional methods rely on the personal experience and skill level of restorers, resulting in a high degree of subjectivity and uncertainty in the restoration results, making it difficult to ensure the stability and consistency of the restoration quality. Second, for complex damaged cultural relics, traditional manual splicing is not only time-consuming and labor-intensive, but also prone to causing secondary damage to the cultural relics during the splicing process. In addition, traditional methods are difficult to accurately predict and evaluate the structural stability of restored cultural relics, increasing the risk of restoration work. Especially when dealing with valuable cultural relics, any minor mistake can result in irreparable loss.

[0004] In view of the shortcomings of traditional cultural relic restoration techniques, the present application has developed a cultural relic non-contact splicing restoration method and device based on three-dimensional point cloud model data, which is particularly important. SUMMARY

[0005] The present application aims to overcome the shortcomings of the prior art and provides a cultural relic non-contact splicing restoration method and device based on three-dimensional point cloud model data. It can obtain accurate three-dimensional data of cultural relics through non-contact three-dimensional scanning equipment, and use a computer for virtual splicing and simulation, effectively avoiding the shortcomings of traditional manual splicing. In addition, by combining deep learning algorithms and improved finite element analysis techniques, the structural stability of restored cultural relics can be accurately predicted and evaluated, greatly improving the accuracy and safety of cultural relic restoration. Furthermore, this method uses the analytic hierarchy process and optimization algorithm to comprehensively compare and dynamically optimize the restoration scheme, further improving the restoration efficiency and quality.

[0006] To solve the above technical problems, the present application provides the following technical solutions: on the one hand, a cultural relic non-contact splicing restoration method based on three-dimensional point cloud model data, the specific steps of which are as follows:

[0007] The cultural relic three-dimensional point cloud data collection and preprocessing step: a non-contact three-dimensional scanning device is used to scan the cultural relic body and each fragment at full angle to obtain three-dimensional point cloud original data containing the surface texture, contour characteristics and size parameters of the cultural relic; an improved statistical filtering algorithm is used for denoising, an improved ICP algorithm is used for registration, and an adaptive voxel downsampling algorithm based on feature retention is used for simplification to obtain a standardized three-dimensional point cloud model; at the same time, a feature point detection algorithm based on deep learning is used to extract the edge feature points, texture feature points and hole feature points of each fragment, and the confidence of the feature points is calculated;

[0008] The cultural relic digital twin construction and virtual environment mapping step: based on the preprocessed standardized three-dimensional point cloud model, the historical background data and material attribute data of the cultural relic are fused, a multi-scale modeling method is used to construct a digital twin that is 1:1 mapped with the cultural relic entity and has the ability to simulate physical properties, and the mapping relationship of the physical property parameters is established; a virtual environment consistent with the actual repair scene is built, the digital twin is imported, a multi-feature fusion-based matching algorithm is used to calculate the relative position relationship between the fragments, a spatial coordinate system is constructed, and the virtual mapping of the actual splicing and repair process is completed;

[0009] Multi-parameter virtual splicing simulation step: in the virtual environment, at least two different splicing sequence parameters are set by using a fragment importance sorting-based method, at least two different repair parameters are determined according to the compressive strength of the cultural relic material, including splicing pressure threshold, transition region size and curvature parameter, virtual splicing operation is performed for each parameter combination, key data such as position deviation and surface fit degree in the splicing process are recorded, and multiple preliminary splicing schemes are generated;

[0010] Post-repair structure stability prediction step: based on an improved finite element analysis algorithm, the time-varying characteristics and environmental coupling effects of the cultural relic material are introduced, the structure stability of the digital twin corresponding to each virtual splicing scheme is analyzed, stress distribution and strain distribution data are calculated, and the load-bearing capacity and environmental interference resistance of the repaired cultural relic are predicted;

[0011] Repair scheme comparison and optimal selection step: based on the virtual splicing simulation data and structure stability prediction data, an analytic hierarchy process is used to construct a scoring system to generate a comprehensive comparison report of multiple schemes; a dynamic optimization step of the repair scheme can be selectively executed, and the parameters are dynamically adjusted through a response surface method and a particle swarm optimization algorithm to assist repair personnel in selecting the optimal repair scheme.

[0012] Further, in the artifact three-dimensional point cloud data acquisition and preprocessing step, the improved statistical filtering algorithm is used for denoising processing. First, the distance deviation of each point and its neighborhood points is calculated, and a weight factor is introduced to realize adaptive denoising according to the material characteristics of the artifact. The weight factor is determined by the balance coefficient, the ratio of the local distance standard deviation to the global mean, and the ratio of the local area elastic modulus to the global mean. The balance coefficient is in the range of 0.6-0.8, and the high value is taken for hard artifacts according to the hardness adjustment of the artifact material. When the calculated noise weight exceeds the threshold value in the range of 1.2-1.5, which is set by the repair expert according to the precious degree of the artifact, the noise point is determined and removed. The improved ICP algorithm is used in the registration process, and the adaptive voxel downsampling algorithm based on feature reservation is used for simplification processing to ensure that the reservation rate of the key feature points of the artifact is not less than 95%.

[0013] Further, in the artifact three-dimensional point cloud data acquisition and preprocessing step, the improved statistical filtering algorithm is used for denoising processing. First, the distance deviation of each point and its neighborhood points is calculated, and a weight factor is introduced to realize adaptive denoising according to the material characteristics of the artifact. The weight factor is determined by the balance coefficient, the ratio of the local distance standard deviation to the global mean, and the ratio of the local area elastic modulus to the global mean. The balance coefficient is in the range of 0.6-0.8, and the high value is taken for hard artifacts according to the hardness adjustment of the artifact material. When the calculated noise weight exceeds the threshold value in the range of 1.2-1.5, which is set by the repair expert according to the precious degree of the artifact, the noise point is determined and removed. The improved ICP algorithm is used in the registration process, and the adaptive voxel downsampling algorithm based on feature reservation is used for simplification processing to ensure that the reservation rate of the key feature points of the artifact is not less than 95%. where C p is the confidence of feature point p, η is the weight coefficient, and the value is 0.6, which is determined by model training. N p is the number of successful matches of feature point p in the training set, N total is the total number of training, I p is the information entropy of feature point p, reflecting the complexity of the surrounding point cloud, and I max is the maximum information entropy value. When C p >0.7, it is determined as a high-confidence feature point for fragment matching; otherwise, it is a low-confidence feature point, which is only used as an auxiliary reference.

[0014] Further, in the artifact three-dimensional point cloud data acquisition and preprocessing step, the improved statistical filtering algorithm is used for denoising processing. First, the distance deviation of each point and its neighborhood points is calculated, and a weight factor is introduced to realize adaptive denoising according to the material characteristics of the artifact. The weight factor is determined by the balance coefficient, the ratio of the local distance standard deviation to the global mean, and the ratio of the local area elastic modulus to the global mean. The balance coefficient is in the range of 0.6-0.8, and the high value is taken for hard artifacts according to the hardness adjustment of the artifact material. When the calculated noise weight exceeds the threshold value in the range of 1.2-1.5, which is set by the repair expert according to the precious degree of the artifact, the noise point is determined and removed. The improved ICP algorithm is used in the registration process, and the adaptive voxel downsampling algorithm based on feature reservation is used for simplification processing to ensure that the reservation rate of the key feature points of the artifact is not less than 95%. where d ij is the matching distance between the i-th feature point and the j-th feature point, β1, β2, β3 are the weights of geometric features, texture features and topological features, respectively, and the values are 0.4, 0.3 and 0.3, respectively, which are obtained by training 300 groups of artifact fragment samples. g i , t i , p irespectively are the geometric feature vector, the texture feature vector and the topological feature vector of the i th feature point, g j , t j , p j respectively are the geometric feature vector, the texture feature vector and the topological feature vector of the j th feature point, when d ij < θ, θ is a matching threshold, which is dynamically adjusted according to the complexity of the cultural relic surface, and the range is 0.01-0.05mm, and it is determined as a matching point pair, and then a spatial coordinate system of the fragment is constructed.

[0015] Further, in the cultural relic digital twin construction and virtual environment mapping step, a multiscale modeling method is used to construct the digital twin, the cultural relic is divided into a macro area and a micro area, the macro area uses a continuous medium model, and the micro area uses a discrete element model, cross-scale simulation is realized through a coupling algorithm, and the mapping relationship of the physical property parameters is established through the following formula: M virtual = κ·M real +(1-κ)·M sim , wherein M virtual is a physical property parameter of the digital twin, M real is a measured physical property parameter of the cultural relic entity, and M sim is a physical property parameter obtained based on a similar material experiment simulation, κ is a weight coefficient, and the value is 0.7-0.9, which is determined according to the reliability of the measured data, and the higher the value is, the more reliable the measured data is. The digital twin constructed through the method can accurately simulate the physical response of the cultural relic under different stress states, and provides an accurate model basis for virtual splicing and structure stability prediction.

[0016] Further, in the multi-parameter virtual splicing simulation step, a method based on fragment importance ranking is used to set the splicing sequence parameter, and the fragment importance is calculated through the following formula: I k = γ1·S k +γ2·F k +γ3·C k , wherein I k is an importance index of the k th fragment, γ1, γ2 and γ3 are respectively an area weight, a feature richness weight and an integrity weight of the fragment, and the values are respectively 0.3, 0.4 and 0.3, which are determined by expert scoring method, S k is an area proportion of the k th fragment relative to the total area of the cultural relic, F k is a feature point number proportion of the k th fragment, and C k is a completeness coefficient of the k th fragment, and the value is 0-1, the higher the value is, the more complete the edge is, and the splicing pressure threshold in the repair parameter is determined according to the compressive strength of the cultural relic material, and the calculation formula is: P th =k·σ c / n, wherein k is a safety factor, taking a value of 0.3-0.5, determined according to the brittleness degree of the cultural relics, σ c is the compressive strength of the material of the cultural relics, and n is the number of splicing contact points.

[0017] Further, in the structure stability prediction step after repair, stress distribution and strain distribution data are calculated based on an improved finite element analysis algorithm, time-varying characteristics and environmental coupling effects of the material of the cultural relics are introduced, and a correction formula for stress calculation is: wherein σ ij (t, T, H) is a stress value considering the influences of time t, temperature T and humidity H, t is time, T is temperature, H is humidity, is an initial stress value, δ(t) is a time attenuation coefficient, obtained based on cultural relic aging experiment data, ∈(T) is a temperature influence coefficient, and ζ(H) is a humidity influence coefficient, and the stress distribution data calculated by the formula are more in line with the stress state of the cultural relics in the actual preservation environment, and the accuracy of the structure stability prediction is improved.

[0018] Further, in the repair scheme comparison and optimal screening step, an analytic hierarchy process is used to construct an evaluation system to generate a comprehensive comparison report, and a total score formula is: S = λ1·S p + λ2·S v + λ3·S s , wherein S is the comprehensive score of the scheme, λ1, λ2 and λ3 are precision weight, visual effect weight and structure safety weight respectively, taking values of 0.3, 0.2 and 0.5 respectively, determined through repair expert questionnaire survey, S p is the precision score, calculated based on position deviation and fitting degree, with a full score of 100, S v is the visual effect score, calculated based on surface flatness and texture coherence, with a full score of 100, and S s is the structure safety score, calculated based on the safety factor, with a full score of 100, the schemes are ranked according to the comprehensive score, the scheme with the highest score is the optimal repair scheme, and the parameter sensitivity analysis of each scheme is also included in the report, providing a reference for scheme adjustment for repair personnel.

[0019] Further, the repair scheme dynamic optimization step in the repair scheme comparison and optimal screening step specifically includes: based on the parameters of the optimal scheme, a function relationship between repair effect and parameters is constructed by a response surface method, a particle swarm optimization algorithm is used to dynamically adjust the parameters, and an optimization objective function is: min J = μ1·e p + μ2·(1-e v ) + μ3·(1-e s), wherein J is an optimization target value, minJ is a minimum value of the target function, μ1, μ2, μ3 are weights of the accuracy error, the visual effect and the structural safety respectively, and the values are 0.2, 0.2 and 0.6 respectively, e p is the accuracy error rate, e v is the visual effect evaluation index, e s is the structural safety evaluation index, through iterative optimization, the target function J reaches the minimum value, and a more optimal repair parameter combination is obtained, and the scientificity and reliability of the repair scheme are further improved. The dynamic optimization step can be selectively executed according to the needs of repair personnel and is suitable for precious cultural relics with extremely high repair accuracy requirements.

[0020] In another aspect, a cultural relic non-contact splicing repair device based on three-dimensional point cloud model data, the device comprises: a data acquisition and preprocessing module, a digital twin construction and environment mapping module, a multi-parameter virtual splicing module, a structural stability analysis module and a scheme comparison optimization and storage module;

[0021] The data acquisition and preprocessing module: integrates a non-contact three-dimensional scanning device and a data processing unit, is used for full-angle scanning of cultural relics and each fragment to obtain three-dimensional point cloud original data, and performs denoising, registration and simplification preprocessing to generate a standardized three-dimensional point cloud model, and extracts feature points of each fragment and calculates feature point confidence;

[0022] The digital twin construction and environment mapping module: is connected with the data acquisition and preprocessing module, is used for fusing the standardized three-dimensional point cloud model, cultural relic historical background data and material property data, constructing a 1:1 digital twin body with physical property simulation capability, simultaneously building a virtual environment consistent with the actual repair scene, establishing spatial position correlation between fragments through a feature matching algorithm and completing virtual mapping of the repair process;

[0023] The multi-parameter virtual splicing module: is connected with the digital twin construction and environment mapping module, is used for setting a splicing sequence parameter based on fragment importance sorting, determining a repair parameter according to the material characteristics of cultural relics, and performing virtual splicing simulation on different parameter combinations and recording position deviation and surface fitting degree key data;

[0024] The structural stability analysis module: is connected with the multi-parameter virtual splicing module, is used for introducing material time-varying characteristics and environmental coupling effects by using an improved finite element analysis algorithm, calculating stress distribution and strain distribution of the digital twin body after virtual splicing, and predicting the load-bearing capacity and environmental interference resistance of the repaired cultural relics;

[0025] The scheme comparison optimization and storage module is connected with the multi-parameter virtual splicing module and the structure stability analysis module respectively, is used for generating a scheme comparison report based on an analytic hierarchy process to construct a comprehensive scoring system, performs dynamic parameter optimization through a response surface method and a particle swarm optimization algorithm, and stores whole-process data to assist repair personnel in screening an optimal scheme.

[0026] Compared with the prior art, the non-contact splicing and repairing method and device based on a three-dimensional point cloud model data have the following beneficial effects:

[0027] Firstly, the digital twin of cultural relics is constructed, and multiple splicing sequences and repairing parameters are simulated in a virtual environment to generate multiple preliminary splicing schemes, a scoring system is constructed by using an analytic hierarchy process, and each scheme is comprehensively compared and optimally screened to assist repair personnel in quickly determining an optimal repairing scheme.

[0028] Secondly, three-dimensional point cloud data of cultural relics and fragments are obtained by using a non-contact three-dimensional scanning device, and are preprocessed by using an improved statistical filtering algorithm, an ICP algorithm and an adaptive voxel downsampling algorithm based on feature reservation to effectively remove noise and reserve key features of cultural relics.

[0029] Other advantages, objects and features of the present application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following or can be learned from practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0031] Figure 1It is a kind of cultural relics non-contact splicing repair method flow operation chart based on three-dimensional point cloud model data;

[0032] Figure 2 It is a kind of cultural relics non-contact splicing repair device flow operation chart based on three-dimensional point cloud model data;

[0033] Figure 3 It is a kind of connection relationship schematic diagram of the cultural relics non-contact splicing repair device based on three-dimensional point cloud model data. DETAILED DESCRIPTION

[0034] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined object of the application, the specific embodiments, structures, features and effects according to the present application will be described in detail below in combination with the drawings and preferred embodiments.

[0035] Example one

[0036] The non-contact three-dimensional laser scanning equipment is adopted to perform full-angle scanning on the Tang Dynasty pottery figurine body and six scattered fragments, the equipment can avoid physical damage to the fragile pottery material of the pottery figurine to the greatest extent without contacting the surface of the pottery figurine, and accurately capture details such as the texture of the pottery figurine surface, the facial contour, the limb shape, and the edges, surface textures and hole regions of each fragment, so as to obtain three-dimensional point cloud original data containing the texture, contour features and size parameters of the pottery figurine surface, and provide comprehensive and accurate basic data support for subsequent repair work.

[0037] When the original data is preprocessed, firstly, an improved statistical filtering algorithm is adopted for denoising processing, the denoising logic is adjusted by combining the characteristics of the pottery figurine pottery material, the noise points generated in the scanning process due to environmental light interference and equipment errors are effectively removed, and the accuracy of subsequent point cloud registration and feature extraction is avoided. The influence of noise data; then, an improved ICP algorithm is adopted to register the point cloud data of the pottery figurine body and each fragment, the algorithm can significantly improve the alignment accuracy between different point cloud data, so that each part of the point cloud data is in a unified coordinate system, and lays a foundation for subsequent overall model construction; then, an adaptive voxel downsampling algorithm based on feature reservation is adopted to simplify the registered point cloud data, while greatly reducing the data amount and improving the subsequent processing efficiency, the key features such as the pottery figurine texture and face are accurately reserved, and the loss of important repair reference information caused by data simplification is prevented.

[0038] Finally, a feature point detection algorithm based on deep learning is adopted to extract the edge feature points, texture feature points and hole feature points of each fragment, these feature points are the core reference basis for subsequent fragment matching and splicing, which can help the repair personnel to quickly locate the corresponding relationship between the fragments; at the same time, the confidence of each feature point is calculated, the feature point confidence calculation formula is used in the calculation process, and the formula calculation is as follows: wherein C p is the confidence of the feature point p, η is a weight coefficient, N p is the number of successful matches of the feature point p in the training set, N total is the total number of training, I p is the information entropy of the feature point p, I max is the maximum information entropy value, through the formula, the reliability of the feature point can be accurately judged, and the invalid feature points with low confidence are removed, further improving the accuracy of subsequent splicing matching, and laying a solid data foundation for subsequent virtual splicing.

[0039] Based on the pre-processed standardized three-dimensional point cloud model of the pottery figurine, the historical background data and material attribute data of the Tang Dynasty pottery figurine are fused, and a digital twin entity which is 1:1 mapped with the pottery figurine entity is constructed by using a multi-scale modeling method. The multi-scale modeling method can accurately model the macro and micro regions of the pottery figurine respectively, and cross-scale simulation is realized through coupling algorithm, ensuring that the digital twin entity can not only present the overall shape of the pottery figurine, but also restore the fine features. At the same time, the mapping relationship of the physical attribute parameters of the pottery figurine is established, so that the digital twin entity has the ability to simulate the physical characteristics of the pottery figurine entity, providing real physical model support for subsequent structure stability analysis.

[0040] A virtual environment consistent with the actual cultural relic restoration laboratory is built, and scene elements such as the spatial layout, lighting conditions and repair operation platform of the laboratory are accurately restored in the virtual environment. This highly restored virtual environment can enable the repair personnel to obtain similar operation experience in the virtual scene as in the actual repair, reducing the deviation between subsequent virtual splicing simulation and actual repair. After the constructed digital twin entity of the pottery figurine is imported into the virtual environment, a matching algorithm based on multi-feature fusion is used to calculate the relative position relationship between the fragments, and the formula is: wherein d ij is the matching distance between the i th feature point and the j th feature point, β1, β2 and β3 are the weights of geometric features, texture features and topological features respectively, g i , t i , p i are the geometric feature vector, texture feature vector and topological feature vector of the i th feature point respectively, g j , t j , p j are the geometric feature vector, texture feature vector and topological feature vector of the j th feature point respectively. This algorithm realizes accurate matching between fragments by constructing multi-dimensional description vectors of feature points, considering multiple feature dimensions comprehensively, and then constructs a unified spatial coordinate system to complete the virtual mapping of the actual splicing and repair process of the Tang Dynasty pottery figurine, providing a real and reliable environment and coordinate basis for subsequent virtual splicing simulation, and ensuring that the virtual splicing process is highly consistent with the actual repair process.

[0041] In the built virtual environment, the method based on the importance of the fragments is used to set the splicing sequence parameters, and the formula is calculated as follows: k = γ1·S k + γ2·F k + γ3·C k , wherein I k is the importance index of the kth fragment, γ1, γ2, γ3 are the area weight, feature richness weight and integrity weight of the fragment respectively, S k is the area proportion of the kth fragment, F k is the number of feature points of the kth fragment, C k is the integrity coefficient of the kth fragment, and the area size, feature richness and integrity of each fragment are considered in the sorting process to determine the importance of the fragments. This sorting method can preferentially select fragments with greater influence on the overall shape of the pottery figurine and richer feature information as the basis for splicing, thereby improving the splicing efficiency and accuracy. Accordingly, two different splicing sequences are set. The first one is based on the fragments with larger area and richer features, and the other one is based on the fragments with close connection to the body of the pottery figurine, and then the edge fragments are spliced. By comparing the effects of different splicing sequences, more reference dimensions are provided for subsequent selection of the optimal scheme.

[0042] According to the compressive strength of the pottery clay of the pottery figurine, two different repair parameters are determined, mainly considering the force-related parameters applied to the fragments during splicing. This parameter setting can simulate the effect of different operation forces on the pottery fragments during actual repair, avoid "damage" to the fragments during virtual splicing due to improper parameters, and ensure the rationality and safety of the virtual splicing process. For each combination of splicing sequence and repair parameter, virtual splicing operation is performed, and key data such as position deviation and surface fit degree of each fragment during splicing are recorded in real time. These data can intuitively reflect the splicing effect under different parameter combinations, provide quantitative basis for subsequent scheme evaluation, and finally generate two preliminary pottery splicing schemes according to these data.

[0043] An improved finite element analysis algorithm is used to analyze the structural stability of the digital twin of the pottery figurine corresponding to the two virtual splicing schemes, and the formula is as follows: , wherein σ ij (t,T,H) is the stress value considering the influence of time t, temperature T and humidity H, t is time, T is temperature, H is humidity, is the initial stress value, δ(t) is the time attenuation coefficient, ∈(T) is the temperature influence coefficient, ζ(H) is the humidity influence coefficient, compared with the traditional finite element analysis algorithm, the algorithm can more accurately simulate the mechanical properties of complex structures, and provide a more reliable analysis tool for the stability evaluation of pottery structure; The time-varying characteristics and environmental coupling effect of pottery clay material are introduced in the analysis process. The introduction of these factors can make the structure stability analysis closer to the actual preservation environment of pottery, and avoid the deviation of the analysis results from the actual situation due to the neglect of environmental and time factors.

[0044] The stress distribution and strain distribution data of the whole pottery and each splicing part are calculated by the algorithm. These data can clearly present the state of each part of the pottery under stress, helping repair personnel identify weak areas of splicing; According to these data, the bearing capacity and environmental interference resistance of the repaired pottery are predicted, which provides key basis for evaluating the feasibility and long-term reliability of the splicing scheme, and ensures that the selected scheme can meet the needs of long-term preservation and display of pottery.

[0045] According to the position deviation, surface fitting degree and other data recorded in the virtual splicing simulation process, as well as the stress distribution, strain distribution, bearing capacity and environmental interference resistance obtained by structure stability prediction, a comprehensive scoring system is constructed by using the analytic hierarchy process, and the formula is: S = λ1·S p + λ2·S v + λ3·S s , wherein S is the comprehensive score of the scheme, λ1, λ2 and λ3 are the precision weight, visual effect weight and structure safety weight respectively, S p is the precision score, S v is the visual effect score, and S s is the structure safety score. According to the comprehensive score, the schemes are sorted, and the scheme with the highest score is the optimal repair scheme. The analytic hierarchy process can systematically integrate multiple complex evaluation indexes, ensure the scientificity and objectivity of the scoring system, and avoid the one-sidedness caused by single index evaluation. The scoring system covers three dimensions of precision, visual effect and structure safety, and comprehensively covers the key evaluation elements of the repair scheme, generating a comprehensive comparison report of the two schemes, so that the repair personnel can intuitively understand the advantages and disadvantages of the two schemes.

[0046] Subsequently, the dynamic optimization step of the repair scheme is selectively executed. Based on the preliminary selected optimal scheme parameters, the function relationship between the pottery repair effect and the parameters is constructed by using the response surface method. This method can accurately fit the relationship between the parameters and the effect, and provide a clear direction for parameter adjustment. Then, the particle swarm optimization algorithm is used to dynamically adjust the parameters, and the objective function is: min J = μ1·e p + μ2·(1-e v ) + μ3·(1-e s), where J is the optimization target value, minJ is the minimum value of the objective function, μ1, μ2, μ3 are the weights of precision error, visual effect and structural safety respectively, e p is the precision error rate, e v is the visual effect evaluation index, e s is the structural safety evaluation index, the particle swarm optimization algorithm can efficiently search for the optimal parameter combination, further improving the precision, visual effect and structural safety of the repair scheme; at the same time, the whole process stores the full-process data, which can not only be used for subsequent scheme tracing and review, but also provide data reference for similar pottery figurine repair, ultimately assisting repair personnel to screen out the optimal splicing repair scheme for the Tang Dynasty pottery figurine, ensuring that the repaired pottery figurine not only conforms to the historical original appearance, but also has good structural stability.

[0047] Example two

[0048] A non-contact three-dimensional optical scanning device is used to scan the Song Dynasty celadon bowl body and 8 broken fragments at full angle. During scanning, the glaze texture on the surface of the celadon bowl, the bowl mouth contour, the bowl body curvature, and the edge shape, glaze texture and damage hole of each fragment are focused on, to ensure that complete three-dimensional point cloud original data containing the surface texture, contour features and size parameters of the celadon bowl are obtained, especially without missing key details such as fine glaze cracks on the edge of the fragments.

[0049] In the preprocessing stage, an improved statistical filtering algorithm is used for denoising, and the denoising process is optimized in combination with the material properties of the celadon glaze and porcelain body to remove noise generated by device vibration and environmental dust during scanning; then an improved ICP algorithm is used to register the point cloud data of the celadon bowl body and each fragment to eliminate positional deviations between different scanning areas; then an adaptive voxel downsampling algorithm based on feature preservation is used to simplify the point cloud data, which not only reduces the data complexity, but also completely preserves important features such as the glaze texture and bowl mouth curvature of the celadon bowl.

[0050] Finally, a feature point detection algorithm based on deep learning is used to extract the edge feature points, texture feature points and hole feature points of each fragment, and a feature point confidence calculation formula is used to calculate the confidence of each feature point, and reliable feature points are selected for subsequent splicing and matching.

[0051] Based on the preprocessed standardized three-dimensional point cloud model of the celadon bowl, the historical background data and material property data of the Song Dynasty celadon bowl are fused, and a digital twin that maps 1:1 with the celadon bowl entity is constructed using a multi-scale modeling method.

[0052] During modeling, the macro and micro regions of the celadon bowl are modeled respectively, cross-scale simulation is realized through coupling algorithm, and the mapping relationship of the physical property parameters of the celadon bowl is established, so that the digital twin can accurately simulate the physical performance of the celadon bowl entity.

[0053] A virtual environment consistent with the actual cultural relic restoration workshop is built, and the workbench layout, lighting conditions, and protective equipment placement of the workshop are restored. The Song dynasty celadon bowl digital twin is imported into the virtual environment, a matching algorithm based on multi-feature fusion is used to calculate the relative position relationship between each fragment, a multi-dimensional feature point description vector is constructed to achieve accurate matching, and then a spatial coordinate system is constructed to complete the virtual mapping of the actual splicing and restoration process of the Song dynasty celadon bowl.

[0054] In the virtual environment, a method based on fragment importance ranking is used to set the splicing sequence parameters. The area proportion, feature richness, and integrity of each fragment are considered to determine the importance of the fragments. Two splicing sequences are set: one starts from the bottom fragments, gradually splices the body fragments, and finally splices the mouth fragments; the other splices the fragments with larger connection area first, and then splices the fragments with smaller connection area.

[0055] According to the compressive strength of the celadon bowl body material, two different restoration parameters are determined, mainly focusing on controlling the related parameters of the force acting on the fragments during splicing to prevent "porcelain body damage" during virtual splicing.

[0056] For each combination of splicing sequence and restoration parameter, virtual splicing operations are carried out, and key data such as position deviation and surface fit degree during splicing are recorded in real time to form two preliminary celadon bowl splicing schemes.

[0057] An improved finite element analysis algorithm is used to analyze the structural stability of the celadon bowl digital twin corresponding to the two virtual splicing schemes. In the analysis, the time-varying characteristics and environmental coupling effects of the celadon bowl body material are introduced.

[0058] The stress distribution and strain distribution data of the entire celadon bowl and each splicing part after splicing are calculated by the algorithm. Based on these data, the load-bearing capacity and environmental interference resistance of the restored celadon bowl are predicted, providing a basis for evaluating the rationality of the splicing scheme.

[0059] Based on the position deviation and surface fit degree data from virtual splicing simulation, and the stress distribution, strain distribution, load-bearing capacity, and environmental interference resistance data from structural stability prediction, a comprehensive scoring system is constructed using the analytic hierarchy process. The two schemes are scored from three dimensions of precision, visual effect, and structural safety, and a comprehensive comparison report is generated.

[0060] Then, a repair scheme dynamic optimization step is performed to preliminarily optimize the parameters of the scheme, to construct a function relationship between the repair effect of the celadon bowl and the parameters by using a response surface method, and to dynamically adjust the parameters by using a particle swarm optimization algorithm, so that the accuracy, visual effect and structural safety of the scheme are further improved, and finally the optimal splicing repair scheme of the Song Dynasty celadon bowl is screened out by the repair personnel.

[0061] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above with a preferred embodiment, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content to obtain equivalent embodiments with equivalent changes, without departing from the technical solution of the present application. Any modification, change, equivalent change and modification of the above embodiments, which are based on the technical essence of the present application, still belong to the scope of the technical solution of the present application.

Claims

1. A non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data, characterized in that, The specific steps of this method are as follows: The steps for acquiring and preprocessing 3D point cloud data of cultural relics are as follows: A non-contact 3D scanning device is used to scan the entire cultural relic and its fragments from all angles to obtain raw 3D point cloud data containing the surface texture, contour features, and dimensional parameters of the relic; an improved statistical filtering algorithm is used for noise reduction, an improved ICP algorithm is used for registration, and an adaptive voxel downsampling algorithm based on feature preservation is used for simplification to obtain a standardized 3D point cloud model; simultaneously, a deep learning-based feature point detection algorithm is used to extract edge feature points, texture feature points, and hole feature points of each fragment, and the confidence level of the feature points is calculated. The steps for constructing a digital twin of a cultural relic and mapping it to a virtual environment are as follows: Based on a preprocessed standardized 3D point cloud model, historical background data and material attribute data of the cultural relic are integrated. A multi-scale modeling method is used to construct a digital twin that is 1:1 mapped to the physical entity of the cultural relic and has the ability to simulate physical attributes. The mapping relationship of physical attribute parameters is established. A virtual environment consistent with the actual restoration scene is built, the digital twin is imported, and a matching algorithm based on multi-feature fusion is used to calculate the relative positional relationship between fragments. A spatial coordinate system is constructed to complete the virtual mapping of the actual splicing and restoration process. Multi-parameter virtual splicing simulation steps: In the virtual environment, at least two different splicing sequence parameters are set using a method based on the importance of fragments. At least two different repair parameters are determined according to the compressive strength of the cultural relic material. Virtual splicing operation is performed for each parameter combination. Key data such as position deviation and surface fit during the splicing process are recorded to generate multiple preliminary splicing schemes. Post-repair structural stability prediction steps: Based on the improved finite element analysis algorithm, the time-varying characteristics of the cultural relic material and the environmental coupling effect are introduced to conduct structural stability analysis on the digital twins corresponding to each set of virtual splicing schemes, calculate stress distribution and strain distribution data, and predict the load-bearing capacity and resistance to environmental interference of the cultural relic after restoration. Repair scheme comparison and optimal selection steps: Based on virtual splicing simulation data and structural stability prediction data, a scoring system is constructed using the analytic hierarchy process (AHP) to generate a comprehensive comparison report of multiple schemes; The system can selectively perform dynamic optimization steps for repair solutions, dynamically adjusting parameters using response surface methodology and particle swarm optimization algorithms to assist repair personnel in selecting the optimal repair solution.

2. The non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data according to claim 1, characterized in that, In the step of data acquisition and preprocessing of the three-dimensional point cloud of cultural relics, when using an improved statistical filtering algorithm for denoising, the distance deviation between each point and its neighboring points is first calculated. At the same time, the material characteristics of the cultural relics are introduced to introduce a weight factor to achieve adaptive denoising. The weight factor is determined by the balance coefficient, the ratio of the local distance standard deviation to the global mean, and the ratio of the local elastic modulus to the overall mean. The balance coefficient ranges from 0.6 to 0.8 and is adjusted according to the hardness of the cultural relic material, with a higher value for hard cultural relics. When the calculated noise weight exceeds the threshold, it is judged as a noise point and removed.

3. The non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data according to claim 1, characterized in that, The three-dimensional point cloud data acquisition and preprocessing steps for cultural relics employ a deep learning-based feature point detection algorithm to extract feature points from each fragment. Specifically, this includes: constructing a feature point detection model for cultural relic fragments. This model takes three-dimensional point cloud data as input, extracts local and global features of the point cloud through a convolutional neural network, and outputs the coordinates and type labels of the feature points. The confidence level of the feature points is calculated using the following formula: Where C p Let η be the confidence level of feature point p, η be the weighting coefficient, and N be the weighting coefficient. p N represents the number of successful matches of feature point p in the training set. total For the total number of training iterations, I p Let I be the information entropy of feature point p. max This represents the maximum information entropy value.

4. The non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data according to claim 1, characterized in that, In the process of constructing the digital twin of the cultural relic and mapping it to the virtual environment, a matching algorithm based on multi-feature fusion is used to calculate the relative positional relationship between fragments. This algorithm constructs a multi-dimensional description vector of feature points, including geometric features, texture features, and topological features, and performs feature matching through weighted Euclidean distance. The formula is as follows: Where d ij Let g be the matching distance between the i-th feature point and the j-th feature point, and let β1, β2, and β3 be the weights of the geometric feature, texture feature, and topological feature, respectively. i t i p i Let g be the geometric feature vector, texture feature vector, and topological feature vector of the i-th feature point, respectively. j t j p j These are the geometric feature vector, texture feature vector, and topological feature vector of the j-th feature point, respectively.

5. The non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data according to claim 1, characterized in that, In the process of constructing the digital twin of the cultural relic and mapping it to the virtual environment, a multi-scale modeling method is used to construct the digital twin. The cultural relic is divided into macroscopic and microscopic regions. The macroscopic region adopts a continuous medium model, and the microscopic region adopts a discrete element model. Cross-scale simulation is achieved through a coupling algorithm. The mapping relationship of physical property parameters is established by the following formula: M virtual =κ·M real +(1-κ)·M sim M virtual M represents the physical property parameters of the digital twin. real M represents the measured physical property parameters of the cultural relic. sim κ represents the physical property parameters obtained from experimental simulations of similar materials, and κ is the weighting coefficient.

6. The non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data according to claim 1, characterized in that, In the multi-parameter virtual splicing simulation step, the splicing order parameters are set using a method based on fragment importance ranking. The fragment importance is calculated using the following formula: I k =γ1·S k +γ2·F k +γ3·C k , where I k Let S be the importance index of the k-th fragment, and let γ1, γ2, and γ3 be the fragment area weight, feature richness weight, and integrity weight, respectively. k F represents the area percentage of the k-th fragment. k C represents the proportion of feature points in the k-th fragment. k Let be the integrity coefficient of the k-th fragment.

7. The non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data according to claim 1, characterized in that, In the post-repair structural stability prediction step, stress and strain distribution data are calculated based on an improved finite element analysis algorithm. The time-varying characteristics of the artifact's material and the environmental coupling effect are incorporated. The corrected formula for stress calculation is as follows: Where σ ij (t,T,H) represents the stress value considering the effects of time t, temperature T, and humidity H, where t is time, T is temperature, and H is humidity. Let be the initial stress value, δ(t) be the time decay coefficient, ∈(T) be the temperature influence coefficient, and ζ(H) be the humidity influence coefficient.

8. The non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data according to claim 1, characterized in that, In the comparison and optimal selection step of the repair schemes, the analytic hierarchy process (AHP) is used to construct a scoring system to generate a comprehensive comparison report. The total scoring formula is: S = λ1·S p +λ2·S v +λ3·S s Where S is the overall score of the scheme, and λ1, λ2, and λ3 are the accuracy weight, visual effect weight, and structural safety weight, respectively. p For accuracy scoring, S v Rate the visual effects, S s To assess structural safety, the solutions are ranked based on their overall scores, with the solution scoring the highest being the optimal repair solution.

9. The non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data according to claim 1, characterized in that, The dynamic optimization step of the repair scheme comparison and optimal selection process specifically includes: based on the parameters of the optimal scheme, constructing a functional relationship between the repair effect and the parameters using the response surface methodology, and dynamically adjusting the parameters using a particle swarm optimization algorithm. The optimization objective function is: minJ = μ1·e p +μ2·(1-e v )+μ3·(1-e s ), where J is the optimization objective value, minJ is the minimum value of the objective function, μ1, μ2, and μ3 are the weights of accuracy error, visual effect, and structural safety, respectively, and e p For the accuracy error rate, e v e is a visual effect evaluation index. s This is a structural safety evaluation index.

10. A non-contact splicing and restoration device for cultural relics based on three-dimensional point cloud model data, applicable to the non-contact splicing and restoration method for cultural relics based on three-dimensional point cloud model data as described in any one of claims 1-9, characterized in that, The device includes: a data acquisition and preprocessing module, a digital twin construction and environment mapping module, a multi-parameter virtual stitching module, a structural stability analysis module, and a scheme comparison, optimization, and storage module. The data acquisition and preprocessing module integrates a non-contact 3D scanning device and a data processing unit to perform full-angle scanning of the artifact and its fragments to obtain raw 3D point cloud data. At the same time, it performs noise reduction, registration, and simplification preprocessing to generate a standardized 3D point cloud model, and extracts the feature points of each fragment and calculates the confidence level of the feature points. The digital twin construction and environment mapping module is connected to the data acquisition and preprocessing module. It is used to integrate standardized three-dimensional point cloud models, cultural relic historical background data and material attribute data to construct a 1:1 digital twin with physical attribute simulation capabilities. At the same time, it builds a virtual environment consistent with the actual restoration scene and establishes spatial position associations between fragments through feature matching algorithms to complete the virtual mapping of the restoration process. The multi-parameter virtual splicing module is connected to the digital twin construction and environment mapping module. It is used to set splicing order parameters based on the importance of fragments, determine restoration parameters according to the material characteristics of cultural relics, perform virtual splicing simulation for different parameter combinations, and record key data such as position deviation and surface fit. The structural stability analysis module is connected to the multi-parameter virtual splicing module. It is used to calculate the stress distribution and strain distribution of the virtual spliced ​​digital twin by adopting an improved finite element analysis algorithm and introducing the time-varying characteristics of materials and the coupling effect of the environment. It predicts the load-bearing capacity and resistance to environmental interference of the cultural relic after restoration. The scheme comparison, optimization, and storage module is connected to the multi-parameter virtual splicing module and the structural stability analysis module, respectively. It is used to generate a scheme comparison report by constructing a comprehensive scoring system based on the analytic hierarchy process, perform dynamic parameter optimization through response surface methodology and particle swarm optimization algorithm, and store full-process data to assist repair personnel in selecting the optimal scheme.

Images