Computer three-dimensional modeling-based cultural relic restoration decision method and system

Abstract

The application discloses a cultural relic repair decision method and system based on computer three-dimensional modeling, relates to the technical field of cultural relic repair, and comprises the following steps: establishing a unified time baseline, generating a phase reference field, introducing a double-frequency comb interference structure to reconstruct a humidity refractive field and locking a zero-phase anchor point, and forming initial constraint conditions of a scanning path; under the initial constraint conditions, a Schlieren heterodyne tomography method is used to obtain an air flow phase gradient, a nonlinear diffraction kernel is calculated, and a refractive residual spectrum is output, which is used for modeling correction. The application realizes real-time compensation by combining phase sheath and humidity control through interference measurement and tomographic identification of humidity disturbance, and improves the three-dimensional modeling precision and the scientific nature of the repair decision by combining mineralization mechanism playback solidification process and using neural field modeling to fuse cracks and material characteristics.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cultural relic restoration, in particular to a cultural relic restoration decision-making method and system based on computer three-dimensional modeling. BACKGROUND

[0002] The "cultural relic restoration decision-making based on computer three-dimensional modeling" refers to that in the cultural relic restoration work, the three-dimensional digital modeling technology is used to comprehensively collect and model the morphology, structure, disease distribution and environmental impact factors of cultural relics, and then virtual simulation and multi-scheme comparison are performed to assist the restorers in scientifically selecting between different restoration materials and methods, so as to improve the feasibility and accuracy of the restoration scheme. In combination with the above material content, this method can introduce the experimental data such as physical properties, weather resistance, color difference compatibility of green restoration materials such as microbial reinforced mortar and active biological mud into the three-dimensional model, and through the visual simulation and quantitative evaluation of the crack filling and pottery bonding effect, the strength, durability and appearance coordination after restoration are predicted in advance, so as to avoid the secondary damage caused by insufficient experience or improper material selection in traditional restoration. In other words, it is an innovative idea of combining the digital virtual decision-making platform with new biological restoration materials, so as to gradually change the cultural relic protection work from experience dependence to scientific quantification and intelligent decision-making.

[0003] The prior art has the following disadvantages: In the prior art, three-dimensional modeling for grottoes or pottery mainly relies on optical scanning means to collect point cloud data on the target surface. However, in the presence of humid air flow, the scanning beam is easily affected by the combined action of humidity gradient and air disturbance, and nonlinear diffraction occurs. This phenomenon causes spatial distortion and geometric distortion of local point cloud data, so that the three-dimensional model reconstructed cannot accurately restore the real trend and depth characteristics of the cracks, resulting in misjudgment of the restoration decision in the crack filling path and the complementary direction. Further in the construction stage, the stress transmission path of the restoration material filling area will deviate from the actual stress state, and under the action of long-term environmental load and disease accumulation, in extreme cases, it may even cause large-area secondary peeling or overall collapse of the cultural relic structure, which seriously affects the safety and restoration reliability of the cultural relic body.

[0004] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide a cultural relic restoration decision-making method and system based on computer three-dimensional modeling to solve the problems in the background.

[0006] In order to achieve the above object, the present application provides the following technical scheme: a cultural relic restoration decision-making method based on computer three-dimensional modeling, comprising the following steps:

[0007] A unified time baseline is established, a phase reference field is generated, a double-frequency comb interference structure is introduced to reconstruct a humidity refractive field and lock a zero-phase anchor point, and initial constraint conditions of a scanning path are formed;

[0008] Under the initial constraint conditions, a Schlieren heterodyne tomography method is used to obtain an air flow phase gradient, a nonlinear diffraction kernel is calculated, and a refractive residual spectrum is output for modeling correction;

[0009] Based on the refractive residual spectrum, bidirectional staggered scanning is performed, and a sub-threshold random phase traction is introduced to disperse artifacts and invert light path disturbance trajectories to generate compensation constraint conditions;

[0010] Under the compensation constraint conditions, a neural field driven by physical prior is constructed, the crack topological structure and material microstructure are mapped to a stress distribution field, and the geometric features are calculated in combination with the refractive residual spectrum;

[0011] Based on the geometric features, the restoration material solidification process is replayed in the neural field, a microbial mineralization dynamics mechanism is introduced, a risk potential field is generated, and a phase-aligned complementary trajectory is calculated;

[0012] Under the complementary trajectory constraint, a time-reversal optical flow regulation mechanism is enabled, a phase conjugate sheath is constructed using a programmable acousto-optic super surface, and a micro-mist micro-spraying array is linked to adjust the humidity gradient, so that the imaging process is stably operated in real time, and the closed-loop control process is completed.

[0013] Preferably, the initial constraint condition forming step is as follows:

[0014] On the basis of establishing a unified time baseline, a phase reference field is generated, and a double-frequency comb interference structure is introduced in the phase reference field;

[0015] After introducing the double-frequency comb interference structure, the refractive index change caused by humidity in the air is extracted through the light frequency comb interference pattern, and the humidity refractive field is reconstructed;

[0016] In the humidity refractive field, a spatial point with the smallest phase fluctuation is selected as a zero-phase anchor point, and the initial constraint conditions of the scanning path are constructed with the zero-phase anchor point as a spatial reference;

[0017] When constructing the initial constraint conditions of the scanning path, the continuous region of the phase gradient in the humidity refractive field is taken as a propagation channel, and a three-dimensional optical scanning trajectory set with the smallest phase distortion is established.

[0018] Preferably, the refractive residual spectrum output step is as follows:

[0019] Based on the initial constraints of the scanning path, Schlirren heterodyne tomography was used to observe the phase gradient of the airflow from multiple perspectives, obtain synchronous laser spot trajectory image data and record environmental disturbance parameters.

[0020] Based on the acquisition of multi-frequency, different-angle observation images, a three-dimensional phase gradient vector field is constructed, and the phase gradient information is compared and corrected with a preset phase reference field to obtain full-space disturbance distribution data;

[0021] Based on the obtained phase gradient vector field, a nonlinear diffraction propagation kernel is constructed, and a spatial response model of beam propagation is established through path integral and forward propagation analysis.

[0022] Based on the nonlinear diffraction propagation kernel, a refractive residual spectrum is generated through phase difference comparison and frequency domain analysis, which serves as the observation basis for subsequent geometric modeling and error correction.

[0023] Preferably, the steps for generating compensation constraints are as follows:

[0024] After obtaining the refractive residual spectrum, a bidirectional staggered scan is performed, with equidistant laser scans conducted along the forward and reverse directions of the optical path. By comparing the differences between the forward and reverse scan images, the directional sensitivity perturbation behavior and reverse phase fluctuation information are extracted.

[0025] While performing bidirectional scanning, subthreshold-level random phase perturbation is introduced. By controlling the phase perturbation during the laser beam propagation process, minute changes are induced in the interferometric image, and latent perturbation features and narrowband artifact distribution are identified.

[0026] After completing bidirectional scanning and introducing phase perturbation, the optical path perturbation trajectory is inverted based on the difference in spot position and perturbation response intensity between the previous and subsequent images. A three-dimensional phase correction matrix containing spatial coordinates, phase offset, and time information is constructed to form compensation constraints for geometric reconstruction.

[0027] Preferably, the geometric feature calculation process is as follows:

[0028] After obtaining the compensation constraint matrix, a neural field with physical prior characteristics is constructed. A quaternary point cloud set containing phase error, coordinate position, perturbation direction and time label is used as the modeling basis, and the material parameters of the cultural relic substrate and the refraction change data of the environmental medium are injected.

[0029] By embedding the topological structure of cracks and the microstructural response properties of repair materials into a neural field, and by extracting crack paths, structural parameters and material physical behavior functions, the material response relationship and stress transmission path between spatial nodes are established.

[0030] By combining the constructed neural field and the refractive residual spectrum, a weighted mapping is performed according to the spatial location and the intensity of the disturbance to generate a disturbance-induced stress field. Under the drive of the tensor field, the three-dimensional structural deformation prediction and the reconstruction of the true geometric features are completed.

[0031] Preferably, the process for calculating the replacement trajectory is as follows:

[0032] After completing geometric feature reconstruction and physical neural field construction, the material curing process is replayed based on spatial structure information. The activation sequence of injection nodes is controlled by time series, and volume changes and stress transmission behavior are recorded.

[0033] Based on the solidification process, a microbial-induced mineralization reaction mechanism is introduced. The induction path is activated according to the environmental parameters of the material nodes to simulate the crystal deposition process and enhance the structural continuity within the cracks.

[0034] After the simulation was completed, a three-dimensional spatial risk potential field was constructed based on the stress concentration degree, solidification shrinkage rate and mineralization failure probability, and potential failure areas were identified by the spatial voxel comprehensive risk coefficient.

[0035] By combining the risk potential field and geometric features, the starting point of the crack end is extracted, the replenishment path is extended along the low-risk direction, and a replenishment trajectory aligned with the phase of the geometric features is established based on the curing time and phase response value.

[0036] Preferably, under the constraint of the replenishment trajectory, the time-reversal optical flow control mechanism is activated, a phase conjugate sheath is constructed using a programmable acousto-optic metasurface, and the humidity gradient is adjusted in conjunction with the micro-mist and micro-spray array as follows:

[0037] Under the constraint of the supplementary trajectory, the time-reversal optical flow control mechanism is activated according to the time response state to construct a two-dimensional optical flow vector field, reconstruct the beam propagation trajectory and obtain the perturbation offset region.

[0038] A dynamic optical path mapping reference frame is constructed based on the optical flow vector field, and the beam propagation stability is spatially graded to form a three-dimensional stability heat map for the location of the disturbance region.

[0039] Based on the thermal map results, programmable acousto-optic metasurfaces are deployed at the edge of the scanning path to suppress disturbances and construct a propagation sheath through spatial phase conjugation.

[0040] Based on the propagation sheath, it is linked with the peripheral micro-mist and micro-spray array to control the spray parameters according to the disturbance level and dynamically adjust the humidity gradient to maintain the stability of the air refractive index.

[0041] The stable imaging process is completed through the synergistic effect of humidity control and phase sheath, and all control results are fed back to the modeling and decision-making process to form a closed-loop control logic.

[0042] The computer-based 3D modeling-based decision-making system for cultural relic restoration includes a phase reference establishment module, a refraction field acquisition module, a disturbance compensation module, a neural field modeling module, a restoration path deduction module, and an imaging stabilization module.

[0043] The phase reference establishment module establishes a unified time baseline, generates a phase reference field, introduces a dual-frequency comb interference structure to reconstruct the humidity refraction field and lock the zero-phase anchor point, forming the initial constraint conditions for the scanning path.

[0044] The refraction field acquisition module, under initial constraints, uses the Schlirren heterodyne tomography method to obtain the airflow phase gradient, calculates the nonlinear diffraction nucleus, and outputs the refraction residual spectrum for modeling and correction.

[0045] The disturbance compensation module performs bidirectional interleaved scanning based on the refractive residual spectrum and introduces subthreshold random phase traction to break up artifacts and invert the optical path disturbance trajectory to generate compensation constraints.

[0046] The neural field modeling module constructs a physical prior-driven neural field under compensated constraints, maps the crack topology and material microstructure to the stress distribution field, and combines the refractive residual spectrum to solve the geometric features.

[0047] The repair path deduction module replays the solidification process of the repair material in the neural field based on geometric features, introduces the microbial mineralization dynamics mechanism, generates a risk potential field, and solves the phase-aligned matching trajectory.

[0048] The imaging stabilization module, under the constraint of the supplementary trajectory, enables the time-reversal optical flow control mechanism, uses a programmable acousto-optic metasurface to construct a phase conjugate sheath, and links the micro-mist and micro-spray array to adjust the humidity gradient.

[0049] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0050] This invention introduces dual-frequency comb interferometry and Schlielen tomography to accurately identify optical nonlinear disturbances caused by humidity in the scanning path. Through phase conjugate sheathing and humidity gradient linkage control, real-time steady-state compensation is achieved at the scanning execution level. Furthermore, by leveraging neural field modeling based on physical priors, the crack structure and material microscopic behavior are effectively integrated to form a geometric model that more closely approximates the actual structural state. Finally, by combining microbial mineralization mechanisms and visual playback of the solidification process, a multi-scale mapping from material behavior prediction to refined path control is completed. Overall, this method not only overcomes the technical bottleneck of model distortion caused by environmental interference in existing technologies but also achieves synchronous coordination between intelligent material decision-making and structural restoration planning, fundamentally improving the scientific rigor, accuracy, and feasibility of cultural relic restoration decisions. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0052] Figure 1 This is a flowchart of the method for making decisions on cultural relic restoration based on computer 3D modeling, as described in this invention.

[0053] Figure 2 This is a schematic diagram of the modules of the cultural relic restoration decision system based on computer 3D modeling of the present invention. Detailed Implementation

[0054] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0055] This invention provides, for example Figure 1 The method for making decisions on cultural relic restoration based on computer 3D modeling, as shown, includes the following steps:

[0056] A unified time baseline is established, a phase reference field is generated, a dual-frequency comb interference structure is introduced into the phase reference field, the humidity refraction field is reconstructed and the zero-phase anchor point is locked to form the initial constraint conditions for the optical scanning path.

[0057] To address the nonlinear diffraction interference caused by humid air disturbance in the 3D modeling of cultural relics, a phase constraint mechanism based on a time baseline is proposed. The specific implementation process is as follows:

[0058] A unified time baseline is established to form a global reference framework for various optical and environmental measurement data throughout the modeling process. In practice, a highly stable rubidium atomic clock is selected as the time reference source, and the time signal is distributed to various on-site devices used for data acquisition, laser emission, image reception, and environmental monitoring via fiber optic synchronization links. During this process, real-time data on changes in disturbances such as temperature, relative humidity, and air velocity in the on-site environment are collected, and each type of disturbance information is timestamped to form a multivariate time synchronization sequence. This time-series data is not stored in isolation but is uniformly mapped to a continuous linear time space at the nanosecond level through a set of multidimensional time interpolation algorithms, forming a high-resolution time baseline covering the entire scanning cycle. This time baseline serves not only as a data synchronization tool but also as a reference for subsequent phase drift tracking and interference error correction, achieving modeling consistency in the time dimension. Compared with traditional methods that rely solely on frame sequence numbering for time synchronization, this approach significantly improves time accuracy and physical disturbance response, making it particularly suitable for environments like cultural relic sites where disturbances are frequent but high precision requirements are paramount.

[0059] In practice, the system uses a unified time reference clock, and each environmental sensor automatically attaches a time stamp during each sampling. Each piece of raw data is recorded as a time stamp + data category + measured value, and stored in chronological order according to the time stamp. Under the same time stamp, disturbance data of different categories are aggregated through a data bus to form a record set corresponding to multiple disturbance quantities at a single time point. In this way, over the entire scanning cycle, a multivariate time synchronization sequence is obtained, arranged chronologically, with each time point containing multiple disturbance quantities such as temperature, relative humidity, and air velocity.

[0060] To map the aforementioned sequences uniformly into a continuous linear time space with nanosecond-level steps, the system pre-defines a target time axis covering the entire scanning cycle, which is sequentially divided with nanoseconds as the minimum time interval. For each time point on the target time axis, the system automatically finds the two immediately preceding and following actual sampling time points in the original time synchronization sequence and interpolates and estimates the perturbation at the current time point based on the measurement values ​​of these two sampling points. When a perturbation does not change significantly in adjacent time intervals, the most recent sampled value can be directly used for filling. Through the above interpolation process, corresponding multivariate perturbation data can be generated for each nanosecond-level time point on the time axis. All time points arranged sequentially constitute a high-resolution time baseline covering the entire scanning cycle.

[0061] Based on a time baseline, a dynamically responsive phase reference field is constructed to sense changes in the refractive properties of the air medium along the optical scanning path in real time. In practice, a continuous laser with a stable wavelength of 532 nm is used as the reference light source, coupled with a high-precision electrically controlled dual-axis scanning platform, to project the laser beam along a preset path onto the space in front of the target artifact surface. Simultaneously, a coherent detection structure records the phase change of the laser beam during its propagation in the air. Combined with the dynamic changes in the refractive index of the air medium, a full-field interferometric reconstruction algorithm is used to establish a reference field with phase-responsive properties in three-dimensional space. To improve the adaptability of the reference field to environmental changes, the temporal variation parameters of temperature, humidity, and airflow disturbance are integrated throughout the reference field generation process. A multivariate coupling model dynamically updates the phase state value at each spatial point, ensuring the reference field's reconfigurability and temporal consistency at different time points. After construction, this reference field does not exist as a static background image but is dynamically adjusted over time as a matrix, serving as the propagation reference path for all scanning beams, significantly improving path planning and positioning accuracy. This step not only establishes a phase environment background with real-time response capability, but also provides a stable foundation for the subsequent introduction of interference structures.

[0062] In practical implementation, the coherent detection structure continuously acquires multiple frames of interferometric images during the scanning process, each frame containing complete interference fringe information. The system processes these interferometric images frame by frame, calculating the phase value corresponding to each pixel using known phase shift settings or reference optical path configurations. Then, combined with pre-calibrated imaging geometric parameters, the phase value of each pixel is mapped to its corresponding spatial location. Through multi-view interferometric image reconstruction, phase attributes can be assigned to each voxel within a regular grid in three-dimensional space, thus forming a three-dimensional reference field covering the scanned area, where each spatial point stores a phase state value.

[0063] During the reference field update process, the system reads the corresponding temperature, humidity, and airflow disturbance data at each time step based on the aforementioned high-resolution time baseline, and establishes a correspondence between these data and the phase state values ​​in the spatial reference field. For example, a set of mapping rules between environmental parameters and phase state can be established for each voxel in the software. When a change in temperature or humidity is detected, the phase state value of that voxel at the current moment is adjusted according to this rule. After each time step, the system performs a comprehensive update of the phase state values ​​of all voxels and stores the reference field states at multiple consecutive time steps in chronological order, thus forming a dynamic matrix structure indexed by three-dimensional spatial coordinates and time. This dynamic matrix can reflect the evolution of the reference field in real time as time flows, providing basic data for subsequent optical path fitting and disturbance compensation.

[0064] Supported by a phase reference field, a dual-frequency comb interferometer structure is introduced to reconstruct the humidity refraction field with high sensitivity. Specifically, two sets of optical frequency comb laser sources with center frequencies of 195 THz and 200 THz are used, and spatially combined by a tunable laser beam combiner to form a group of interfering laser beams. This interferometer structure has a highly stable phase difference capability, and its interference pattern can accurately reflect the minute refractive index differences in the air caused by humidity variations. During operation, the interference light field is projected onto the air region in front of the artifact scanning path, and a high frame rate area array camera is used to synchronously acquire changes in the interference pattern. The acquisition time corresponding to each frame is recorded with a time baseline. Further, the phase distribution of the interference fringes is extracted using image processing algorithms, and Fourier analysis is performed on the fringe spacing at different positions to obtain the relative refractive index change curves of the air at each position. By combining the above curves with the known humidity-refractive index correspondence model, the humidity refraction field distribution map of the entire scanning area can be reconstructed. Compared to existing modeling methods using structured light or single-frequency laser beams, this dual-frequency comb interferometry structure offers higher phase resolution, maintaining high accuracy in spatial refraction measurements, especially in environments with high-frequency humidity disturbances or trace water vapor migration. This step solves the problem of traditional optical scanning struggling to acquire accurate spatial refraction information in complex humidity environments.

[0065] In one specific implementation of this invention, different locations can be understood as multiple analysis regions on the interference image, each analysis region consisting of several adjacent pixels. For example, the image can be divided by pixel rows or by pixel blocks of a fixed size, with each pixel row or pixel block representing a location. For each location, the system extracts one or more brightness variation sequences along the arrangement direction of the interference fringes and performs spectral analysis on these sequences to identify the repeating period and local spacing variation trends of the fringes. By analyzing the variation trend of the fringe spacing, a quantized curve of the relative change in optical path at that location can be obtained, further reflecting the relative change in the air refractive index at that location.

[0066] After obtaining the relative refractive index variation curves at various locations, the system calls a pre-established humidity-refractive index correspondence model. This model, based on experimental calibration or publicly available data, provides the variation law of refractive index under different humidity conditions. The system performs point-by-point transformation on the relative refractive index variation curve at each location, mapping each change on the curve to a humidity change, thereby obtaining humidity distribution information covering the entire image range. Combined with the geometric calibration parameters of interferometric imaging, the humidity information at each location in the two-dimensional image can be back-projected into three-dimensional space, mapped onto the corresponding spatial voxels, and then, through interpolation and tomographic reconstruction, each voxel is assigned a refractive index value corresponding to humidity in the three-dimensional grid, ultimately forming a humidity refractive field distribution map of the entire scanned area.

[0067] In the humidity refraction field, the most stable spatial phase point is selected as the zero-phase anchor point, and the initial constraints of the entire optical scanning path are established based on this anchor point. In practice, by traversing the reconstructed humidity refraction field data, the phase fluctuation amplitude of each spatial point in multiple consecutive time segments is analyzed, and the spatial node with the smallest standard deviation and the most stable center position is selected as the zero-phase anchor point. After selecting the anchor point, it is used as the three-dimensional spatial reference origin, and combined with the surrounding phase gradient information, a set of minimum disturbance propagation paths centered on the anchor point is constructed as the reference path for subsequent optical scanning. The path construction process uses a curved light trajectory simulation algorithm, fitting the spatial path based on the air refractive index distribution and the propagation trend in the phase reference field, obtaining a set of scanning trajectories that are physically feasible in space and have minimal phase distortion. Finally, the shortest path is selected as the initial scanning path, and subsequent geometric feature capture and material path simulation operations are performed on this path.

[0068] In practical implementation, the aforementioned humidity refraction field is first superimposed with the phase reference field to obtain a three-dimensional field describing the spatial distribution of air refractive index and the phase propagation trend. Then, within this three-dimensional field, starting from the scanning device position, the system progressively calculates the propagation direction and positional changes of light in space according to a preset exit angle and step size. At each step, the light direction is adjusted based on the refractive index distribution and phase change at the current position, thus obtaining a curved light trajectory. By changing the initial exit angle and the scanning endpoint position, multiple candidate light trajectories can be generated. The system records the cumulative amount of phase change and the degree of passage through highly disturbed regions along the path of each candidate light trajectory, and integrates these indicators into an evaluation value to measure the phase distortion degree of the trajectory. Finally, the system selects multiple candidate trajectories whose evaluation values ​​meet preset thresholds and simultaneously satisfy device motion constraints and scanning coverage requirements, forming a set of scanning trajectories that are physically feasible in space and have low phase distortion, for subsequent 3D modeling and image acquisition.

[0069] Under initial constraints, the Schlirren hetero-frequency tomography method is used to obtain the airflow phase gradient, calculate the nonlinear diffraction nucleus, and output the refractive residual spectrum to provide the accurate observations required for subsequent modeling and correction.

[0070] To eliminate the nonlinear diffraction interference caused by air disturbance at the artifact site on optical modeling, after completing the initial constraints of the optical path, Schlielen heterodyne tomography was introduced to obtain the airflow phase gradient. This was then combined with inversion calculations to construct a nonlinear diffraction kernel, from which the refractive residual spectrum was generated. The specific implementation steps are as follows:

[0071] Under the constraints of a defined optical scanning path and phase anchor points, a multi-frequency interferometric observation platform was constructed to observe airflow disturbances, supporting high-precision acquisition of Schliren hetero-frequency tomography data. The platform was set up at five observation angles in front of the target area of ​​the artifact, positioned at 0°, 30°, 60°, 90°, and 120°, with each angle equipped with a pulsed laser and a high-speed area array imaging camera. The lasers used were high-stability multi-frequency laser pulse sources with wavelengths of 530 nm, 532 nm, 534 nm, 536 nm, and 538 nm, respectively. The energy output of each laser pulse was controlled within 8 millijoules, and the pulse duration within 5 nanoseconds. The laser pulses were triggered according to a unified time baseline to ensure complete synchronization of imaging time across all channels. The laser beams were projected into the air region near the artifact scanning path through an optical axis collimation system, where refraction interference with the airflow structure caused observable curvature changes in the laser propagation trajectory in the two-dimensional image. A high-speed camera captures the center trajectory of the laser spot in each frame of the image, with a sampling frame rate of 3000 frames per second, and simultaneously records the values ​​of temperature, humidity, and airflow velocity sensors at each imaging session. This set of multi-frequency, multi-angle image sequences provides spatiotemporally synchronized multi-dimensional observational data for subsequent 3D inversion.

[0072] Edge features are extracted frequency-wise and angle-wise from the acquired image sequences to extract the actual offset path of the light spot center trajectory in the pixel coordinate system. By constructing a unified image transformation matrix, image data from all perspectives are reprojected into a three-dimensional Cartesian coordinate system centered on the artifact scanning path. Using the wavelength of each frequency laser beam and the propagation distance of the laser beam in air as basic parameters, the actual phase offset of the light spot at each point is calculated. This calculation result, combined with optical measurement formulas and real-time acquired temperature and humidity field data, allows the derivation of the air refractive index distribution gradient at each spatial point. By performing spatial difference operations on the refractive index variation trends of all spatial points, a set of three-dimensional phase gradient vector fields is formed. This vector field details the optical path distortion trend caused by local humidity unevenness or air disturbance in the air and reflects the dynamic characteristics of disturbance changes on the time axis. To ensure reconstruction accuracy, the phase information of each image frame is compared with a pre-constructed phase reference field to correct errors caused by lens aberrations or angular deviations, thereby ensuring the full spatial consistency and temporal stability of the phase gradient. Compared with traditional single-frequency or single-plane Schlielen observation methods, this heterogeneous multi-view inversion method has higher spatial transparency and phase accuracy, and is especially suitable for cultural relics sites with complex structures and severe disturbances.

[0073] To construct a unified image transformation matrix, the system calibrates each camera and laser emitter after installation. During calibration, images of a calibration board with known geometry are acquired, and the imaging parameters and spatial relationships of each camera are calculated to determine the correspondence between pixel coordinates and scene 3D coordinates. Based on these calibration results, the system generates a set of transformation parameters for each camera to convert pixel coordinates to 3D coordinates. During scanning, the system reads the pixel coordinates of the light spot center in each frame and uses the corresponding transformation parameters to convert them into spatial points in the same 3D Cartesian coordinate system.

[0074] The three-dimensional Cartesian coordinate system is established centered on the artifact scanning path: First, the system fits the pre-defined or planned scanning path into a spatial curve. Then, the tangent direction of this curve at the starting position is used as one axis of the coordinate system, and two directions perpendicular to the tangent direction are used as the other two axes, thus determining the orientation of the three-dimensional coordinate system. The origin of the coordinate system can be selected at the starting point or the midpoint of the scanning path. In this way, all spatial points and light spot trajectories within the entire scanning area can be represented in a unified coordinate system centered on the scanning path.

[0075] After obtaining the three-dimensional coordinates of the laser beam trajectory, the system compares the actual propagation path length between the laser beam emitter and each spatial point with the ideal straight-line path length, treating the difference as the phase shift during propagation. Combining the operating wavelength of the laser beam at each frequency with real-time environmental data, the system can infer the phase shift of each spatial point relative to ideal propagation conditions. Subsequently, the system maps the phase shift data of all spatial points onto a three-dimensional mesh structure and calculates the difference in phase change between adjacent meshes to estimate the trend of refractive index variation with spatial position. By performing such difference calculations in three spatial directions, a vector representing the direction and rate of phase change can be obtained for each mesh point, thus forming a set of three-dimensional phase gradient vector fields that reflect the spatial distribution gradient of the air refractive index.

[0076] A nonlinear diffraction propagation kernel is constructed using the obtained three-dimensional airflow phase gradient vector field to accurately describe the propagation behavior of a laser scanning beam in a real air disturbance environment. Specifically, each node in the spatial vector field is considered a dynamic micro-disturbance source, and a refraction influence matrix is ​​constructed based on the direction and magnitude of its phase gradient change. Path integration is performed on this matrix to obtain the cumulative phase shift curve of the beam propagation along the actual scanning path. This cumulative phase shift data is compared with a preset ideal optical straight path to establish a disturbance correction mapping function. Then, through backpropagation analysis, the influence value of each disturbance node is mapped to the forward propagation space, realizing the forward propagation simulation of the disturbance effect. The resulting nonlinear diffraction kernel is a propagation response model distributed throughout space, containing information on the phase shift law, path deformation trend, and energy density change of the beam under each micro-disturbance condition in the air medium. This propagation kernel can be used to calculate the bending degree, coherence change, and focusing position drift of the beam under a given disturbance field, thus completely replacing the traditional method of path calculation using a fixed optical path or linear refraction compensation model, significantly improving the model's adaptability and predictive ability in complex environments.

[0077] In its implementation, the system first discretizes the aforementioned three-dimensional phase gradient vector field into a regular spatial grid, with each grid point recording the direction and magnitude of the phase change. Then, using the actual scanning path as a reference, the scanning path is divided into several equal-length path segments, and the positional relationship between each path segment and the surrounding spatial grid points is determined. The system assigns an influence coefficient to each spatial grid point and each path segment to represent the degree of influence of the phase gradient at that grid point on the beam propagation along the corresponding path segment. All influence coefficients are arranged according to the combination of grid points and path segments, forming a refraction influence matrix.

[0078] When calculating the cumulative phase shift of the beam propagating along the actual scanning path, the system sequentially traverses each path segment. For each segment, it summarizes the influence coefficients corresponding to that segment in the refraction influence matrix to obtain the phase shift of that segment. By summing the phase shifts of all path segments sequentially along the path, a cumulative phase shift curve describing the phase change of the beam throughout the entire scanning path can be obtained.

[0079] To establish the perturbation correction mapping function, the system pre-defines the ideal optical straight path and its theoretical phase distribution at various spatial locations. It then compares and analyzes the actual cumulative phase shift curve with the phase distribution of the ideal straight path to obtain the required correction amounts at different path locations. Based on these correction amounts, the system establishes a correspondence between the observed phase shift and the coordinate correction or phase compensation amount for each path location, storing this relationship in the form of a lookup table or interpolation function. During actual operation, when the system detects a phase shift of a specific magnitude at a certain path location, it can look up the corresponding compensation parameter through this mapping relationship and apply it to the point cloud coordinates or phase data to achieve perturbation correction and simulation of the forward propagation space.

[0080] Based on the constructed nonlinear diffraction propagation kernel, a refractive residual spectrum is derived and generated, serving as the primary observation basis for subsequent geometric reconstruction and error correction. In practice, the reference phase map generated by the ideal optical propagation path is first compared point-by-point with the actual phase map calculated using the nonlinear diffraction kernel. The phase shift at each spatial node is calculated and recorded as the original residual data. Subsequently, a three-dimensional Fourier transform and wavelet decomposition are performed on this residual data to obtain its intensity spectrum and frequency response curves at different frequency dimensions. A distribution map is then constructed according to the phase shift amplitude interval, forming a complete refractive residual spectrum atlas. This atlas not only reflects the spatial distribution density of perturbation effects but also reveals the locations of high-frequency perturbation regions, resonance path segments, and phase anomaly peaks, providing crucial input for subsequent optical path compensation, image reconstruction, and structural correction. This refractive residual spectrum is entirely derived from physical interference and real measurement data, possessing advantages such as strong observability, high spatial resolution, and fast time response. It is a crucial component in constructing a precise three-dimensional modeling process based on perturbation response.

[0081] Based on the refractive residual spectrum, a bidirectional interleaved scan is performed. Subthreshold random phase traction is introduced during the scan to break up narrowband artifacts and invert the optical path disturbance trajectory, thereby generating compensation constraints.

[0082] To achieve active compensation for optical path distortion caused by air disturbances, after extracting the refractive residual spectrum, a structured bidirectional interleaved scan is performed using the phase error distribution information reflected in the spectrum, and subthreshold phase perturbation is introduced to accurately invert the beam offset path. Finally, compensation constraints that can be used for modeling correction are generated. The specific implementation method is as follows:

[0083] Based on the spatial phase error distribution revealed in the refractive residual spectrum, bidirectional staggered scanning was implemented to construct redundant data channels and extract directional perturbation behavior. In practice, the original scanning path was first divided into equally spaced continuous sampling segments, each 50 mm long, and a complete scanning process was performed along both the forward and reverse directions of the path. A highly stable single-mode continuous laser beam with a wavelength of 532 nm was used as the scanning source. The laser beam's trajectory in space was controlled by an optomechanical scanning platform, with a scanning speed set at 1.5 mm / s, ensuring at least 20 laser sampling points per millimeter to meet spatial resolution requirements. During the forward scan, the laser beam started from the beginning of the scanning path and covered the entire area clockwise; during the reverse scan, the laser beam returned from the end point to the beginning point, maintaining the same path but with the opposite scanning direction. Both scans were triggered by a clock synchronization control unit at a unified time baseline to ensure consistent imaging time. During image acquisition, each frame of laser interferometry image was accompanied by recording environmental temperature and humidity, air velocity, laser power fluctuations, and other physical background parameters, forming a temporal information set that can be used for later comparison. By comparing the trends in spot offset, interference fringe morphology, and reflection peak variation in forward and reverse scan images, the direction-sensitive perturbation structure and asymmetric optical path bending behavior can be preliminarily identified. The key to this step is to use the perspective change introduced by path direction reversal to reveal the inverse phase fluctuations and residual errors that are difficult to capture in traditional unidirectional scanning.

[0084] During bidirectional scanning, subthreshold-level random phase perturbations are continuously introduced to disperse narrowband artifacts formed within the stable interference region and to make latent perturbation responses explicit. Specifically, an acousto-optic modulator is positioned along the laser beam emission path. A periodic, minute phase perturbation is superimposed on the laser beam propagation direction using a high-frequency oscillating voltage. The amplitude of this perturbation is kept within 5% of the main scanning phase change, insufficient to cause changes in image structure, but sufficient to trigger phase jumps and slight shifts in bright and dark fringes within the interference fringes. A random phase traction signal generator generates a non-periodic sinusoidal phase shift according to a pseudo-random sequence, which is transmitted to the laser beam front end in real time via the modulator, ensuring the laser beam carries weak phase perturbation characteristics when entering the scanning path region. When the laser beam passes through spatial segments with drastic local refractive index changes or frequent airflow disturbances, these perturbations are amplified due to phase superposition, forming an interference enhancement effect, which manifests as random flashes, brightness jump bands, or slightly blurred interference areas in the image. By comparing permeated scanned images with undisturbed standard images pixel by pixel and superimposing the perturbation input parameters recorded in each frame, a quantitative assessment of the sensitivity to local perturbation response and an inverse modeling of the perturbation-induced mechanism can be achieved. Unlike traditional methods that rely on image processing algorithms to identify artifacts later, this method completes the feedforward intervention of artifact interference sources during the data acquisition stage, possessing advantages in active identification and high-resolution perception. It is particularly suitable for pre-modeling detection in environments with complex artifact structures and unstable airflow.

[0085] Based on the bidirectional interleaved scanning results and subthreshold perturbation response data, high-precision inversion calculations of optical path perturbation trajectories are performed, and spatial compensation constraints used in the modeling process are generated accordingly. In the operation, a three-dimensional spatial reference mesh model is first constructed, mapping the image data obtained from the forward and backward scans to each node position in this mesh, with the node spacing controlled within 1 mm to cover the entire scanning path. For each mesh node, the directional perturbation offset vector is obtained by calculating the three-dimensional coordinate difference between the spot position in the forward image and the corresponding spot position in the backward image. Arranging all perturbation vectors in the scanning order along the time axis forms a complete set of spatial perturbation trajectories. Subsequently, the perturbation trajectories are correlated and calibrated with the phase error information in the refraction residual spectrum, and a phase compensation factor is introduced at each node to construct a three-dimensional spatial phase correction matrix. This correction matrix contains both the position offset vector and the phase difference compensation factor, and its data format is a triplet, consisting of spatial coordinates, phase offset, and timestamp. In 3D modeling, this matrix serves as an important input parameter for point cloud position correction and phase field resampling during the geometric reconstruction stage. It can effectively eliminate geometric offset, contour curvature, and local feature mismatch caused by air disturbances.

[0086] Under the condition of compensation constraint, a neural field based on physical prior is constructed to map the crack topology and the microstructure of the repair material to the stress distribution field, and the geometric features that approximate the real shape are obtained by combining the refractive residual spectrum.

[0087] To reconstruct the high-precision geometric structure of cultural relics in a complex and disturbed environment, after obtaining the compensation constraint matrix, a neural field model with a physical response mechanism is constructed. The crack topology and the structure of the repair material are injected into it, and coupled solution is performed by combining the refractive residual spectrum. The specific process is as follows:

[0088] Based on the constructed three-dimensional compensation constraint matrix, a neural field with physical prior characteristics is established to support subsequent mechanical response modeling and geometric structure evolution prediction. At the start of the operation, each spatial coordinate node in the compensation constraint matrix is ​​densely sampled, forming a set of four-element point clouds containing phase error, coordinate position, perturbation direction, and time label. This point cloud set not only records the original scanning deviation of each point in space but also includes reflection intensity and phase response information obtained from interferometry. Based on this, following the continuous medium assumption in physical mechanics, macroscopic material parameters of the artifact substrate are introduced, including elastic modulus, shear modulus, anisotropy coefficient, thermal expansion coefficient, and their response factor under humidity changes. Simultaneously, the refractive index variation range of environmental media such as air under different temperature and humidity conditions is considered and bound to the mapping relationship of the compensation point cloud, forming a priori physical information database integrating spatial perturbation, material response, and media refraction dynamics. During the modeling process, the prior information is attached in tensor form to the connection relationships between each point cloud node, ensuring that the topology of the neural field not only follows spatial proximity relationships but is also controlled by real physical parameters. Unlike existing 3D modeling methods that build neural networks based solely on geometric structures or grayscale images, this neural field possesses the ability to respond realistically to external physical disturbances, enabling subsequent simulation processes to follow the actual stress transmission logic rather than data fitting trends.

[0089] By embedding fracture topological features and the microstructural response properties of the repair material into the neural field structure, a structure-material joint mechanical mapping is formed. Fracture topological information is derived from a set of fracture boundaries extracted from a high-density point cloud based on depth gradient mutation and edge convolution, and its path, branch nodes, and tip evolution trends are tracked in three-dimensional space. Each fracture segment records its length, average opening width, sidewall roughness, principal stress direction angle, and spatial rotation vector, serving as the basis for topological encoding. This encoding is not only a geometric representation but also used to predict the coupling mode between the fracture and the external stress field. Simultaneously, physical experiments are performed on specific biomineral materials used for repair, extracting their pore structure, crystal morphology, grain orientation, hardening shrinkage rate, and interfacial bonding strength variation curves at different hydration stages. The filling interface is then structurally fitted using a microscale scanner, thereby constructing a set of material response functions within the neural field. This function set is indexed by material type and associated with nodes in the filling region, ensuring that each part of the field not only identifies its spatial location but also possesses a clear definition of its material properties. When coupling stress propagation paths between nodes, not only is the geometrically shortest path considered, but also material interface impedance, porosity-induced hysteresis, and microcrack propagation threshold are introduced to ensure that stress transmission laws are consistent with physical laws. This type of structure-material fusion not only surpasses the approximate modeling methods of existing simulation methods that rely solely on a single elastic constant, but also provides fine-grained control over material behavior, making it particularly suitable for cultural relic scenarios with strong heterogeneity and complex damage paths.

[0090] In the constructed neural field, combined with the previously obtained refractive residual spectrum, a joint solution operation is performed to restore the three-dimensional geometric structure model that most closely resembles the real state. First, the refractive residual spectrum is segmented and discretized according to its frequency distribution range, extracting the high-frequency perturbation-dominant region, the mid-frequency transition region, and the low-frequency stable region, and mapping the corresponding data to coordinate nodes in the neural field. During the mapping process, a two-factor weighting strategy of position registration and perturbation intensity matching is adopted, that is, not only establishing a connection between the refractive spectrum and the node based on spatial location, but also performing a secondary correction based on the historical perturbation intensity at that location. Next, a perturbation-induced stress field is constructed based on the mapping results, and the perturbation experienced by each node is transformed into a predicted value of local structural displacement through a material interface function. After superimposing the force transmission response between nodes, a spatial deformation tensor field is formed. Under the guidance of the tensor field, the topological orientation of the crack path, the generation and termination mechanism of micro-branches are dynamically adjusted, while the actual distribution direction of the repair material in the pore structure and the interface bonding morphology are also adjusted. The final output three-dimensional geometric model consists of tens of thousands of nodes that have been perturbation corrected, material response modulated and topology feedback adjusted. It not only restores the potential deformation regions that were not identified during the scan, but also preserves the microstructure evolution process caused by changes in the physical environment.

[0091] Based on geometric features, the solidification process of the repair material is replayed in the neural field. The microbial mineralization dynamics mechanism is introduced to generate the risk potential field of the repair filling path, and the replacement trajectory aligned with the geometric features is further calculated.

[0092] After completing the geometric structure reconstruction and physical neural field construction, in order to simulate the behavioral evolution of the repair material in real application scenarios, it is necessary to dynamically replay the material solidification process based on the restored spatial structure. At the same time, a microbial-induced mineralization reaction mechanism is introduced to construct a risk potential field, and finally, the replenishment trajectory synchronized with the geometric features is derived. The specific implementation method is as follows:

[0093] Based on the constructed physical neural field and geometric feature model, the curing process of the repair material is dynamically replayed to reconstruct its spatiotemporal evolution path under actual injection scenarios. This process begins with crack filling and is time-series-based according to the material construction logic. Each injection node is assigned a start-up time, curing cycle, volume shrinkage function, and interfacial permeation resistance value. Based on thermal analysis experiments and scanning electron microscopy observations, the microscopic change parameters of the repair material at each stage under air exposure conditions are obtained, including the micelle formation rate in the early stage of gelation, the grain growth direction distribution in the middle stage, and the surface shrinkage ratio in the final stage. These parameters are injected into each material node in the neural field, enabling the node to possess the ability to evolve its structural state over time. During the simulation replay, the filling path nodes are activated layer by layer according to a time-stepping mechanism, recording the volume change rate and density change of each layer, and calculating its stress transmission effect on adjacent nodes in real time. Through the spatial superposition of continuous time series, the complete process of the material transitioning from a high-fluidity state in the early stage of injection to a condensation transformation state in the middle stage of curing, until the formation of the final hard shell layer, is reconstructed, thereby evaluating the changes in internal stress field and boundary deformation trends at each stage.

[0094] Building upon the material solidification and replay process, a microbial mineralization reaction mechanism is further introduced to enhance the structural integration and micromechanical stability of the filling process. Specifically, a *Bacillus subtilis* strain with strong adaptability to alkaline environments is selected. This strain can efficiently catalyze calcium carbonate formation in environments with calcium ion concentrations above 1.2 mmol / L and pH values ​​between 8.0 and 8.8. The calcium carbonate induction rate is cultivated and quantified under laboratory conditions, and various kinetic parameters, including reaction delay period, initial nucleation rate, crystal growth rate, maximum crystal size, and crystallization plateau period, are extracted. In a neural field, each material node is parameter-mapped according to environmental factors (temperature, humidity, microfluidic velocity, pH value) matched to its spatial location to simulate the mineralization reaction behavior of that node under microbial activity. When a node is in a state of contraction stress concentration or an adjacent node has completed preliminary solidification, the induction reaction pathway is activated, and the crystal deposition rate is derived by combining nucleation density and crystallization rate. During the mineralization reaction, the system records the crystallization direction, coverage area, crystal integrity index, and interfacial bonding coefficient with the material matrix of the formed calcium carbonate crystals, gradually forming a stable mineralization network. Finally, the microcrystalline deposition structure distributed in a band along the inside of the crack was reconstructed in the model, which effectively enhanced the structural continuity and formed a bonding reinforcement band at the crack edge, thereby improving the stress resistance of the overall repair area.

[0095] After simulating the solidification process and mineralization behavior, a risk potential field based on spatially coupled stress is constructed to identify potential material failure regions and structurally vulnerable areas in the filling path. The construction of the risk potential field uses the stress concentration degree, extreme solidification shrinkage rate, and mineralization failure probability of each material node as core input variables, mapping them onto a three-dimensional structural space to form a potential energy distribution map. Within each spatial voxel, the comprehensive risk coefficient of the voxel is calculated by statistically analyzing the anomalous stress gradient, the proportion of solidification delay nodes, and the density of anomalous mineralization crystallization nodes contained in that region. Based on standardized scoring, the entire space is divided into high-risk, medium-risk, and low-risk zones. This potential field reflects the non-uniform stress concentration regions and potential microcrack propagation paths formed by the repair material within the three-dimensional structure. High-risk continuous zones are extracted using isosurface analysis techniques, identifying their directionality, distribution density, and the angular relationship between them and the crack principal axis in space, providing a quantitative basis for subsequent path avoidance, reinforcement design, and material adjustment strategies.

[0096] Based on the spatial coupling relationship between the constructed risk potential field and known geometric features, a replacement trajectory highly aligned with the structural features is derived. The specific calculation method is as follows: First, all unclosed fracture path endpoints are identified in the 3D geometric model as candidate starting points for the replacement path. Then, the direction of lowest energy is found in the risk potential field, and the path extension is executed using the tangent direction of the equipotential surface as the initial planning vector. During the path extension process, the material solidification time point and phase response value of each node covered by the extension segment are referenced in real time to ensure that the path is synchronized with the constructed geometric features in the time dimension. When the path encounters high-risk areas or mineralization failure areas during extension, the direction is adjusted using a path offset algorithm to bypass weak structures and maintain continuity. After the path is generated, its corresponding 3D coordinate sequence is constructed, and the material consistency, phase matching degree, and environmental response consistency with adjacent nodes are recorded at each point. The resulting replenishment trajectory not only satisfies the docking conditions in geometric space, but also ensures the consistency of the solidification state, material interface properties and microbial mineralization response, achieving triple synchronization of three-dimensional structure, temporal behavior and material dynamics, providing a basic path for material guidance, injection control and quality verification in the subsequent construction stage.

[0097] Under the constraint of the supplementary trajectory, the time reversal optical flow control mechanism is activated, and a phase conjugate protective sheath is constructed at the edge of the scanning path using a programmable acousto-optic metasurface. The humidity gradient is dynamically adjusted in conjunction with the micro-mist and micro-spray array to achieve real-time stable operation of the scanning imaging process and complete the closed-loop control process of cultural relic restoration modeling and decision-making.

[0098] To ensure the accurate execution of the compensation trajectory during actual imaging and material guidance, it is necessary to dynamically adjust the beam propagation path, stabilize the humidity environment, and construct an optical sheath to achieve real-time closed-loop feedback between modeling and decision-making. The specific steps of this process are as follows:

[0099] Based on the time response state of nodes in the compensated trajectory, a time-reversal optical flow control mechanism is initiated to monitor and predict the propagation changes of the beam in the scanning path in real time. This process requires capturing the propagation direction, phase perturbation amplitude, and energy distribution of the beam in different air density regions using a high-speed phase velocimetry device. In practice, multiple photosensitive sensing points are uniformly arranged within the spatial distribution range of the scanning path, and optical parameters are collected at each point at a frequency of 100,000 times per second. Based on the known time baseline and phase values ​​of the compensated path nodes, a two-dimensional optical flow vector field is constructed, and the actual propagation trajectory of the beam in the previous cycle is reconstructed through inverse time interpolation. After superimposing this trajectory with the predicted propagation path, the perturbation offset region and its rate of change can be obtained, thereby realizing dynamic optical flow prediction capability and providing data support for subsequent control measures.

[0100] Using optical flow prediction results, a dynamic optical path mapping reference frame is constructed to grade and evaluate the optical propagation stability of various spatial regions in the scanning path. During implementation, the space surrounding the compensation trajectory is divided into several equally spaced three-dimensional grids, with grid side lengths controlled between 0.5 mm and 1 mm. Each grid node records the beam energy density, phase stability, perturbation duration, and surrounding air refractive index gradient at the corresponding time. Through overlay analysis of three consecutive cycles, the relative stability score of each grid point in the time dimension is obtained. Based on the score, four stability levels are set, corresponding to completely stable, slightly perturbed, moderately perturbed, and strongly perturbed regions, respectively. This reference frame is updated in real time, and a three-dimensional stability heatmap is generated using graphical rendering. This map is used to guide the subsequent positioning of key control areas and improve the accuracy of the optical path optimization strategy.

[0101] Guided by a dynamic reference frame, a programmable acousto-optic metasurface is deployed around the compensation trajectory to construct a phase-conjugate sheath for suppressing disturbances. During execution, the optical path disturbance information obtained through a time-reversal mechanism is first mapped to the region to be controlled, and phase calculations are performed on the laser beam's reverse propagation direction in the disturbance hotspots. Based on this direction, the acousto-optic response characteristics of each nanoscale unit in the metasurface are modulated, ensuring that its output wavefront has a spatial phase conjugate relationship with the incident wavefront. This process utilizes acoustically controlled piezoelectric signals to drive the internal lattice arrangement of the acousto-optic material, responding to phase adjustment requirements within microseconds. The sheath's construction path covers the entire scanning range edge and pre-compensates for predicted high-incidence disturbance areas. Unlike traditional static optical protection methods, this metasurface possesses both spatial dynamic response and programmable control properties, enabling the formation of a stable propagation channel in complex humidity-disturbed fields, preventing laser speckle distortion, image blurring, or target distortion.

[0102] After the phase-conjugate protective sheath is constructed, a micro-mist / micro-spray array deployed around the scanning area is further coordinated to dynamically adjust the humidity gradient distribution and maintain the relative uniformity of the refractive index in the air medium. The micro-spray array consists of densely packed piezoelectric ceramic spray units, each capable of releasing uniform water mist particles with a diameter of less than 3 micrometers within 0.01 seconds. During execution, the array controls the spray direction, concentration, and duration in real time according to the disturbance intensity level marked in the optical path reference frame. In areas where the refractive index abnormally increases due to excessive humidity, the array activates a local negative pressure desiccant device to recover free moisture through high-speed adsorption; in areas with excessively low humidity, low-dose pulsed spraying is used to increase local air humidity. This spray control process employs a linear scanning synchronization mechanism to ensure that the spray rhythm is consistent with the scanning equipment's operating speed, forming a spatial-temporal dual-dimensional linkage. Ultimately, a high-quality propagation band with stable refractive index and low air disturbance is constructed along the scanning path, providing optimal environmental conditions for laser scanning and image acquisition.

[0103] With the combined action of phase conjugate sheathing and humidity gradient control, the system achieves real-time stable operation throughout the entire scanning imaging process, and feeds all control results back to the modeling and decision-making process, forming a complete closed-loop control logic. In the feedback mechanism, the system quantifies and analyzes the image sharpness, spatial consistency, phase stability indicators, and compensation path execution error after each scan, generating an environmental dynamic control response curve. This response curve is applied in real-time to parameter adjustments for the next scanning cycle, including acousto-optic wavefront shape, micro-jetting frequency, humidity adjustment amplitude, and scanning delay compensation. During material injection and image acquisition, the feedback data is also used to correct minor deviations in the repair path and guide the material injection rhythm to synchronize with the optical scanning rate, achieving high-precision spatial synchronization control.

[0104] This invention introduces dual-frequency comb interferometry and Schlielen tomography to accurately identify optical nonlinear disturbances caused by humidity in the scanning path. Through phase conjugate sheathing and humidity gradient linkage control, real-time steady-state compensation is achieved at the scanning execution level. Furthermore, by leveraging neural field modeling based on physical priors, the crack structure and material microscopic behavior are effectively integrated to form a geometric model that more closely approximates the actual structural state. Finally, by combining microbial mineralization mechanisms and visual playback of the solidification process, a multi-scale mapping from material behavior prediction to refined path control is completed. Overall, this method not only overcomes the technical bottleneck of model distortion caused by environmental interference in existing technologies but also achieves synchronous coordination between intelligent material decision-making and structural restoration planning, fundamentally improving the scientific rigor, accuracy, and feasibility of cultural relic restoration decisions.

[0105] This invention provides, for example Figure 2 The computer-based 3D modeling-based decision-making system for cultural relic restoration shown includes a phase reference establishment module, a refraction field acquisition module, a disturbance compensation module, a neural field modeling module, a restoration path deduction module, and an imaging stabilization module.

[0106] The phase reference establishment module establishes a unified time baseline, generates a phase reference field, introduces a dual-frequency comb interference structure to reconstruct the humidity refraction field and lock the zero-phase anchor point, forming the initial constraint conditions for the scanning path.

[0107] The refraction field acquisition module, under initial constraints, uses the Schlirren heterodyne tomography method to obtain the airflow phase gradient, calculates the nonlinear diffraction nucleus, and outputs the refraction residual spectrum for modeling and correction.

[0108] The disturbance compensation module performs bidirectional interleaved scanning based on the refractive residual spectrum and introduces subthreshold random phase traction to break up artifacts and invert the optical path disturbance trajectory to generate compensation constraints.

[0109] The neural field modeling module constructs a physical prior-driven neural field under compensated constraints, maps the crack topology and material microstructure to the stress distribution field, and combines the refractive residual spectrum to solve the geometric features.

[0110] The repair path deduction module replays the solidification process of the repair material in the neural field based on geometric features, introduces the microbial mineralization dynamics mechanism, generates a risk potential field, and solves the phase-aligned matching trajectory.

[0111] The imaging stabilization module, under the constraint of the supplementary trajectory, activates the time-reversal optical flow control mechanism, uses a programmable acousto-optic metasurface to construct a phase conjugate sheath, and links the micro-mist and micro-spray array to adjust the humidity gradient, thereby achieving real-time stable operation of the imaging process and completing the closed-loop control process.

[0112] The cultural relic restoration decision-making method based on computer 3D modeling provided in this invention is implemented through the above-mentioned cultural relic restoration decision-making system based on computer 3D modeling. For details of the specific methods and processes of the cultural relic restoration decision-making system based on computer 3D modeling, please refer to the above-mentioned embodiments of the cultural relic restoration decision-making method based on computer 3D modeling, which will not be repeated here.

[0113] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for making decisions on cultural relic restoration based on computer 3D modeling, characterized in that, Includes the following steps: A unified time baseline is established, a phase reference field is generated, a dual-frequency comb interference structure is introduced to reconstruct the humidity refraction field and lock the zero-phase anchor point, forming the initial constraint conditions for the scanning path. Under initial constraints, the Schlirren hetero-frequency tomography method is used to obtain the airflow phase gradient, calculate the nonlinear diffraction kernel, and output the refractive residual spectrum for modeling and correction. Bidirectional interleaved scanning is performed based on the refractive residual spectrum, and subthreshold random phase traction is introduced to disperse artifacts and invert the optical path disturbance trajectory to generate compensation constraints. Under the condition of compensation constraint, a physical prior-driven neural field is constructed to map the crack topology and material microstructure to the stress distribution field, and the geometric characteristics are solved by combining the refractive residual spectrum. Based on geometric features, the solidification process of the repair material is replayed in the neural field. The microbial mineralization dynamics mechanism is introduced to generate a risk potential field and solve the phase-aligned matching trajectory. Under the constraint of the supplementary trajectory, the time-reversal optical flow control mechanism is activated, a phase conjugate sheath is constructed using a programmable acousto-optic metasurface, and the humidity gradient is adjusted in conjunction with the micro-mist and micro-spray array.

2. The method for making decisions on cultural relic restoration based on computer three-dimensional modeling according to claim 1, characterized in that, The steps for forming initial constraints are as follows: Based on the establishment of a unified time baseline, a phase reference field is generated, and a dual-frequency comb interference structure is introduced into the phase reference field; After introducing a dual-frequency comb interference structure, the change in refractive index caused by humidity in the air is extracted by the optical frequency comb interference pattern, and the humidity refraction field is reconstructed. In the humidity refraction field, the spatial point with the smallest phase fluctuation is selected as the zero-phase anchor point, and the initial constraint conditions of the scanning path are constructed using the zero-phase anchor point as the spatial reference. When constructing the initial constraints of the scanning path, the continuous region of phase gradient in the humidity refraction field is used as the propagation channel to establish a set of three-dimensional optical scanning trajectories with minimum phase distortion.

3. The method for making decisions on cultural relic restoration based on computer three-dimensional modeling according to claim 2, characterized in that, The steps for outputting the refractive residual spectrum are as follows: Based on the initial constraints of the scanning path, Schlirren heterodyne tomography was used to observe the phase gradient of the airflow from multiple perspectives, obtain synchronous laser spot trajectory image data and record environmental disturbance parameters. Based on the acquisition of multi-frequency, different-angle observation images, a three-dimensional phase gradient vector field is constructed, and the phase gradient information is compared and corrected with a preset phase reference field to obtain full-space disturbance distribution data; Based on the obtained phase gradient vector field, a nonlinear diffraction propagation kernel is constructed, and a spatial response model of beam propagation is established through path integral and forward propagation analysis. Based on the nonlinear diffraction propagation kernel, the refractive residual spectrum is generated by phase difference comparison and frequency domain analysis.

4. The method for making decisions on cultural relic restoration based on computer three-dimensional modeling according to claim 3, characterized in that, The steps for generating compensation constraints are as follows: After obtaining the refractive residual spectrum, a bidirectional staggered scan is performed, with equidistant laser scans conducted along the forward and reverse directions of the optical path. By comparing the differences between the forward and reverse scan images, the directional sensitivity perturbation behavior and reverse phase fluctuation information are extracted. While performing bidirectional scanning, subthreshold-level random phase perturbation is introduced. By controlling the phase perturbation during the laser beam propagation process, minute changes are induced in the interferometric image, and latent perturbation features and narrowband artifact distribution are identified. After completing bidirectional scanning and introducing phase perturbation, the optical path perturbation trajectory is inverted based on the difference in spot position and perturbation response intensity between the previous and subsequent images, and a three-dimensional phase correction matrix is ​​constructed to form compensation constraints for geometric reconstruction.

5. The method for making decisions on cultural relic restoration based on computer three-dimensional modeling according to claim 4, characterized in that, The geometric feature calculation process is as follows: After obtaining the compensation constraint matrix, a neural field with physical prior characteristics is constructed. The quaternary point cloud set is used as the modeling basis, and the material parameters of the cultural relic substrate and the refraction change data of the environmental medium are injected. By embedding the topological structure of cracks and the microstructural response properties of repair materials into a neural field, and by extracting crack paths, structural parameters and material physical behavior functions, the material response relationship and stress transmission path between spatial nodes are established. By combining the constructed neural field and the refractive residual spectrum, a weighted mapping is performed according to the spatial location and the intensity of the disturbance to generate a disturbance-induced stress field. Under the drive of the tensor field, the three-dimensional structural deformation prediction and the reconstruction of the true geometric features are completed.

6. The method for making decisions on cultural relic restoration based on computer three-dimensional modeling according to claim 5, characterized in that, The process of calculating the replacement trajectory is as follows: After completing geometric feature reconstruction and physical neural field construction, the material curing process is replayed based on spatial structure information. The activation sequence of injection nodes is controlled by time series, and volume changes and stress transmission behavior are recorded. Based on the solidification process, a microbial-induced mineralization reaction mechanism is introduced. The induction path is activated according to the environmental parameters of the material nodes to simulate the crystal deposition process and enhance the structural continuity within the cracks. After the simulation was completed, a three-dimensional spatial risk potential field was constructed based on the stress concentration degree, solidification shrinkage rate and mineralization failure probability, and potential failure areas were identified by the spatial voxel comprehensive risk coefficient. By combining the risk potential field and geometric features, the starting point of the crack end is extracted, the replenishment path is extended along the low-risk direction, and a replenishment trajectory aligned with the phase of the geometric features is established based on the curing time and phase response value.

7. The method for making decisions on cultural relic restoration based on computer three-dimensional modeling according to claim 6, characterized in that, Under the constraint of the supplementary trajectory, the time-reversal optical flow control mechanism is activated, a phase conjugate sheath is constructed using a programmable acousto-optic metasurface, and the humidity gradient is adjusted in conjunction with a micro-mist and micro-spray array as follows: Under the constraint of the supplementary trajectory, the time-reversal optical flow control mechanism is activated according to the time response state to construct a two-dimensional optical flow vector field, reconstruct the beam propagation trajectory and obtain the perturbation offset region. A dynamic optical path mapping reference frame is constructed based on the optical flow vector field, and the beam propagation stability is spatially graded to form a three-dimensional stability heat map for the location of the disturbance region. Based on the thermal map results, programmable acousto-optic metasurfaces are deployed at the edge of the scanning path to suppress disturbances and construct a propagation sheath through spatial phase conjugation. Based on the propagation sheath, it is linked with the peripheral micro-mist and micro-spray array to control the spray parameters according to the disturbance level and dynamically adjust the humidity gradient to maintain the stability of the air refractive index. The system achieves stable imaging through the combined action of humidity control and phase sheath, and feeds all control results back to the modeling and decision-making process.

8. A computer-based 3D modeling-based decision-making system for cultural relic restoration, used to implement the computer-based 3D modeling-based decision-making method for cultural relic restoration as described in any one of claims 1-7, comprising a phase reference establishment module, a refraction field acquisition module, a disturbance compensation module, a neural field modeling module, a restoration path deduction module, and an imaging stabilization module: The phase reference establishment module establishes a unified time baseline, generates a phase reference field, introduces a dual-frequency comb interference structure to reconstruct the humidity refraction field and lock the zero-phase anchor point, forming the initial constraint conditions for the scanning path. The refraction field acquisition module, under initial constraints, uses the Schlirren heterodyne tomography method to obtain the airflow phase gradient, calculates the nonlinear diffraction nucleus, and outputs the refraction residual spectrum for modeling and correction. The disturbance compensation module performs bidirectional interleaved scanning based on the refractive residual spectrum and introduces subthreshold random phase traction to break up artifacts and invert the optical path disturbance trajectory to generate compensation constraints. The neural field modeling module constructs a physical prior-driven neural field under compensated constraints, maps the crack topology and material microstructure to the stress distribution field, and combines the refractive residual spectrum to solve the geometric features. The repair path deduction module replays the solidification process of the repair material in the neural field based on geometric features, introduces the microbial mineralization dynamics mechanism, generates a risk potential field, and solves the phase-aligned matching trajectory. The imaging stabilization module, under the constraint of the supplementary trajectory, enables the time-reversal optical flow control mechanism, uses a programmable acousto-optic metasurface to construct a phase conjugate sheath, and links the micro-mist and micro-spray array to adjust the humidity gradient.

Images