Virtual-reality fusion guide system and method for cultural and museum exhibition halls based on digital twins

Abstract

The present invention provides a virtual-reality fusion navigation system and method for cultural and museum exhibition halls based on digital twins, which belongs to the field of exhibition hall navigation technology. The system extracts data from cultural and museum exhibition halls, obtains composite data of cultural and museum exhibition halls and performs data preprocessing, generates composite data of cultural and museum exhibition halls and merges them for simulation, constructs a visitor sensory analysis model, a visitor behavior analysis model and a cultural relic status analysis model, constructs a cultural and museum exhibition hall navigation model based on the visitor sensory analysis model, the visitor behavior analysis model and the cultural relic status analysis model, performs model distillation to obtain a lightweight cultural and museum exhibition hall navigation model and deploys it to various terminals of the cultural and museum exhibition hall, generates visitor navigation decisions based on the lightweight cultural and museum exhibition hall navigation model, performs virtual-reality fusion navigation for the cultural and museum exhibition hall based on the visitor navigation decision, obtains navigation results, optimizes and adjusts the lightweight cultural and museum exhibition hall navigation model according to the navigation results, and provides visitors with cultural and museum exhibition hall navigation decisions that take into account both cultural relic safety protection and visitor visiting experience.

Description

Technical Field

[0001] The present invention relates to the field of exhibition hall guidance technology, and in particular to a virtual-reality fusion guidance system and method for cultural and museum exhibition halls based on digital twins. Background Art

[0002] Currently, smart museums are widely adopting sensor networks and digital guide systems, initially enabling basic functions such as visitor tracking and cultural relic environmental monitoring. However, existing technologies often focus on single-point capabilities, such as independently deploying AR guides or security monitoring, and lack mechanisms for coordinating multi-source data.

[0003] For example, the Chinese invention patent with patent number 202411979783.2 provides a virtual-reality display system for the Metaverse Museum. The system collects user information and cultural relic data through a data acquisition module, and also grasps user needs and makes personalized recommendations through personalized tour planning and a combination of virtual and real methods. However, there are still certain limitations in terms of balancing tourist needs and cultural relic safety needs, and in terms of the difficulty of deploying high-precision models on lightweight equipment.

[0004] On the one hand, there is often an opposition between the sensory needs of tourists and the safety needs of cultural relics. Often, due to the need to take the safety needs of cultural relics into consideration, the sensory needs of tourists cannot be met. On the other hand, the terminal computing power restricts the experience upgrade. The high-precision model requires a large amount of data memory capacity and cannot be deployed on lightweight devices such as AR glasses, forcing the functions to be compulsorily restricted, causing the immersive experience to stagnate at the primary stage. Summary of the Invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a virtual-reality integrated navigation method for a cultural and museum exhibition hall based on digital twins, the method comprising:

[0006] Extracting data from the cultural and museum exhibition hall to obtain composite data of the cultural and museum exhibition hall, performing data preprocessing on the composite data of the cultural and museum exhibition hall to generate a composite data set of the cultural and museum exhibition hall, wherein the composite data set of the cultural and museum exhibition hall includes a visitor sensory data set, a visitor behavior data set, and a cultural relic status data set;

[0007] Conduct simulations based on the complex data set of cultural and museum exhibition halls to build visitor sensory analysis models, visitor behavior analysis models, and cultural relic status analysis models;

[0008] A cultural and museum exhibition guide model is constructed based on the visitor sensory analysis model, visitor behavior analysis model, and cultural relic status analysis model, and a lightweight cultural and museum exhibition guide model is obtained through model distillation.

[0009] Deploy the lightweight museum guide model to each terminal in the museum, including the visitor user terminal, edge node terminal, and edge cloud terminal, and generate visitor guide decisions based on the lightweight museum guide model.

[0010] Based on the tourist guide decision, a virtual-reality integrated guide is conducted for the cultural and museum exhibition hall, a guide result is obtained, and a lightweight cultural and museum exhibition hall guide model is optimized and adjusted according to the guide result.

[0011] As a further solution of the present invention, a simulation is performed based on a composite data set of cultural and museum exhibition halls to construct a visitor sensory analysis model, a visitor behavior analysis model, and a cultural relic status analysis model, including:

[0012] Perform tourist sensory feedback simulation based on the tourist sensory data set, obtain the tourist sensory simulation results, analyze the tourist sensory simulation results through a neural network, generate tourist sensory feedback mapping rules, and build a tourist sensory analysis model based on the tourist sensory feedback mapping rules;

[0013] Among them, the tourist perception feedback mapping rule is expressed as the quantitative mapping relationship between the tourist's biometric response and the multimodal feedback parameters corresponding to the cultural and museum exhibition hall;

[0014] Based on the tourist behavior data set, tourist group behavior simulation is carried out to obtain the tourist group behavior simulation results, and the tourist group behavior simulation results are analyzed through a neural network to generate the tourist group behavior knowledge entropy change conduction rules, and a tourist behavior analysis model is constructed based on the tourist group behavior knowledge entropy change conduction rules;

[0015] Among them, the conduction rule of tourist group behavior knowledge entropy change is expressed as the functional relationship between the tourist group cognitive association strength and the generation of cross-space guided path;

[0016] Conduct a micro-damage risk transmission simulation on cultural relics based on the cultural relics status data set, obtain the simulation results of micro-damage risk transmission on cultural relics, analyze the simulation results of micro-damage risk transmission on cultural relics through neural networks, generate micro-damage risk transmission rules for cultural relics, and build a cultural relics status analysis model based on the micro-damage risk transmission rules for cultural relics;

[0017] Among them, the risk transmission rule of micro-damage to cultural relics is expressed as the risk state transmission path of micro-damage caused by changes in the micro-environment of cultural relics to the structure of cultural relics.

[0018] As a further solution of the present invention, a museum guide model is constructed based on the visitor sensory analysis model, the visitor behavior analysis model, and the cultural relic status analysis model, including:

[0019] The tourist cognitive feedback vector generated by the tourist sensory analysis model, the tourist guide path decision vector generated by the tourist behavior analysis model, and the risk parameter vector generated by the cultural relic status analysis model are temporally aligned and normalized, and the rules corresponding to each model are integrated to construct a guide decision rule system.

[0020] Among them, the tourist cognitive feedback vector is used to express the tourist cognitive state and interaction needs, the tourist guide path decision vector is used to express the tourist guide decision path and the tourist spatiotemporal distribution state, and the risk parameter vector is used to express the cultural relic damage risk and protection needs;

[0021] Based on the real-time scene status of the cultural and museum exhibition hall, the weight distribution coefficients of the tourist sensory analysis model, tourist behavior analysis model and cultural relic status analysis model are obtained. The weight ratio of the corresponding feature vectors of each model is dynamically adjusted according to the weight distribution coefficient to generate a joint feature vector of the cultural and museum exhibition hall.

[0022] As a further solution of the present invention, the rules corresponding to each model are integrated to construct a navigation decision rule system, including:

[0023] The priority of the rules corresponding to the cultural relics status analysis model is set to the highest globally. When the rules corresponding to the cultural relics status analysis model conflict with the rules corresponding to the tourist sensory analysis model and the tourist behavior analysis model, the rules corresponding to the tourist sensory analysis model and the tourist behavior analysis model are forcibly overwritten.

[0024] The tourist cognitive feedback vector output by the tourist sensory analysis model data needs to be used to enhance the tourist sensory perception within the cultural relics risk safety range corresponding to the risk parameter vector;

[0025] The tourist guide path decision vector output by the tourist behavior analysis model is verified for path accessibility based on the risk parameter vector. The tourist guide path decision vectors that fail the path accessibility verification are eliminated, and only the tourist guide path decision vectors that pass the path accessibility verification are output.

[0026] As a further solution of the present invention, model distillation is performed to obtain a lightweight cultural and museum exhibition hall guide model, including:

[0027] Based on the hardware characteristics of the cultural and museum terminals, a terminal portrait is constructed. The terminal portrait includes the visitor user terminal, the edge node terminal, and the edge cloud terminal. The gradient masking technology is used to distill the rules contained in the cultural and museum guide model into a lightweight model corresponding to the terminal portrait.

[0028] For the tourist user side, only the sensory feedback rules related to the tourist sensory analysis model in the museum guide model are distilled, and the risk response priority logic is retained;

[0029] For edge nodes, the rules related to the tourist behavior analysis model in the museum guide model are distilled to retain the tourist guidance path decision-making and multi-terminal coordination capabilities;

[0030] For the edge cloud, the rules related to the cultural relics status analysis model in the cultural and museum guide model are distilled, and the cross-system global scheduling capabilities are trained to retain the cultural relics risk assessment and model rule optimization capabilities.

[0031] As a further solution of the present invention, a lightweight museum guide model is deployed to each terminal of the museum, and tourist guide decisions are generated based on the lightweight museum guide model, including:

[0032] The tourist user end collects and analyzes the tourist's sensory data in real time, generates the tourist's sensory needs and uploads them to the edge node end simultaneously. The edge node end performs risk verification on the tourist's sensory needs and uploads the risk verification results to the edge cloud end. The edge cloud end generates corresponding tourist guide decisions for the tourist sensory needs that pass the risk verification, and reconstructs and adjusts the needs of tourists that fail the risk verification.

[0033] The edge cloud sends the generated tourist guide decision to the edge node. The edge node analyzes the tourist guide decision, generates the tourist guide decision analysis results and sends them to the tourist user end. The tourist user end performs corresponding operations based on the tourist guide decision analysis results and the tourist sensory needs.

[0034] As a further solution of the present invention, the lightweight museum guide model is optimized and adjusted according to the guide results, including:

[0035] The edge cloud periodically analyzes navigation results, identifies optimization points, and uses federated learning to incrementally iterate model parameters;

[0036] For high-frequency conflict scenarios, the core rules are re-distilled through the digital twin environment to generate lightweight model incremental packages and distribute them to the edge node and tourist user ends. The edge node and tourist user ends optimize the rules based on the lightweight model incremental packages.

[0037] As a further solution of the present invention, data extraction is performed on the cultural and museum exhibition hall to obtain composite data of the cultural and museum exhibition hall, and data preprocessing is performed on the composite data of the cultural and museum exhibition hall to generate a composite data set of the cultural and museum exhibition hall, including:

[0038] Perform spatiotemporal calibration, noise filtering, feature fusion, and privacy desensitization operations on the composite data of cultural and museum exhibition halls to generate a composite data set of cultural and museum exhibition halls;

[0039] The tourist sensory data set is represented as sensory data reflecting the tourist's visual focus and cognitive state, and at least includes tourist visual data and tourist cognitive data;

[0040] The tourist behavior dataset is represented as behavioral data describing the audience's spatial movement, interactive behavior and interest preferences, and at least includes tourist behavior pattern data, tourist interest point distribution data and tourist interaction data;

[0041] The cultural relic status data set is represented as status data reflecting the physical status of the cultural relic and the microenvironment in which it is located, and at least includes the physical status data of the cultural relic and the status data of the cultural relic microenvironment.

[0042] As a further solution of the present invention, spatiotemporal calibration, noise filtering, feature fusion, and privacy desensitization operations are performed on the composite data of cultural and museum exhibition halls to generate a composite data set of cultural and museum exhibition halls, including:

[0043] The spatiotemporal calibration is to map the extracted composite data of cultural and museum exhibition halls to the same spatial reference and synchronously align the corresponding time parameters;

[0044] The noise filtering is performed to remove redundant noise and correct data on the extracted composite data of cultural and museum exhibition halls;

[0045] The feature fusion is performed by extracting and fusing the composite data of the museum and exhibition hall that has been time-space calibrated and noise filtered, generating a tourist sensory data set, a tourist behavior data set, and a cultural relic status data set;

[0046] The privacy desensitization is performed by injecting privacy noise into the tourist sensory dataset, the tourist behavior dataset, and the cultural relic status dataset, and only uploading the composite data set of the cultural relics exhibition hall after privacy desensitization.

[0047] On the other hand, an embodiment of the present invention further provides a virtual-reality integrated tour guide system for cultural and museum exhibition halls based on digital twins, including:

[0048] A data acquisition module is used to extract data from cultural and museum exhibition halls and perform data preprocessing on the extracted composite data of cultural and museum exhibition halls to obtain a composite data set of cultural and museum exhibition halls;

[0049] A simulation module is used to perform simulation based on a composite data set of a museum and to construct a visitor sensory analysis model, a visitor behavior analysis model, and a cultural relic status analysis model;

[0050] A model fusion module is used to perform model fusion and model distillation operations on the visitor sensory analysis model, the visitor behavior analysis model, and the cultural relic status analysis model to generate a lightweight cultural and museum exhibition hall guide model;

[0051] A decision-making module, which is used to analyze the status of the cultural and museum exhibition hall in real time based on the lightweight cultural and museum exhibition hall guidance model and generate tourist guidance decisions;

[0052] The feedback optimization module is used to optimize and adjust the lightweight cultural and museum exhibition hall guidance model according to the implementation effect of the visitor guidance decision.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] Extract data from cultural and museum exhibition halls to obtain composite data of cultural and museum exhibition halls, perform data preprocessing on the composite data of cultural and museum exhibition halls, generate composite data sets of cultural and museum exhibition halls, provide high-precision input for model construction through multi-dimensional data extraction and data preprocessing, perform simulation based on the composite data sets of cultural and museum exhibition halls, construct visitor sensory analysis models, visitor behavior analysis models, and cultural relic status analysis models, realize personalized analysis and feedback of visitors by constructing visitor sensory analysis models, optimize the generation of guided tours according to the visitor behavior analysis model, and control the status of cultural relics in real time based on the cultural relic status analysis model to reduce the probability of damage to cultural relics;

[0055] A cultural and museum guide model is constructed based on the visitor sensory analysis model, the visitor behavior analysis model, and the cultural relic status analysis model. Model distillation is then performed to obtain a lightweight cultural and museum guide model. Conflicts between multi-model decisions are resolved through model fusion. During the model fusion phase, the rules corresponding to each model are integrated and compensated to generate a rule system that takes into account both visitor needs and cultural relic protection needs. This improves the accuracy of the final decision and alleviates the conflict between visitor needs and cultural relic safety needs. Furthermore, through end-to-end distillation, lightweight models that meet the terminal characteristics are deployed for different terminals, thereby reducing system power consumption and response delay.

[0056] The lightweight cultural and museum exhibition hall guidance model is deployed to each terminal of the cultural and museum exhibition hall, and tourist guidance decisions are generated based on the lightweight cultural and museum exhibition hall guidance model. Based on the tourist guidance decisions, a virtual and real integrated guidance of the cultural and museum exhibition hall is carried out to obtain the guidance results. The lightweight cultural and museum exhibition hall guidance model is optimized and adjusted according to the guidance results. By iteratively optimizing the model, the accuracy and stability of the model-generated decisions are maintained. Based on the above aspects, cultural and museum exhibition hall guidance decisions that take into account both the safety protection of cultural relics and the tourist visiting experience are provided to tourists. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flowchart of the steps of the virtual-reality integrated tour guide method for cultural and museum exhibition halls based on digital twins of the present invention;

[0058] Figure 2It is a schematic diagram of the virtual-reality fusion guide system for cultural and museum exhibition halls based on digital twins of the present invention. DETAILED DESCRIPTION

[0059] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a schematic diagram of the execution flow of a virtual-reality fusion navigation method for a cultural and museum exhibition hall based on digital twins provided by an embodiment of the present invention. The virtual-reality fusion navigation method for a cultural and museum exhibition hall based on digital twins is introduced in detail below.

[0060] Step S1, extract data from the cultural and museum exhibition hall, obtain composite data of the cultural and museum exhibition hall, preprocess the composite data of the cultural and museum exhibition hall, and generate a composite data set of the cultural and museum exhibition hall, which includes a tourist sensory data set, a tourist behavior data set, and a cultural relic status data set.

[0061] Specifically, the extracted composite data of cultural and museum exhibition halls are subjected to spatiotemporal calibration, noise filtering, feature fusion, and privacy desensitization operations to generate a composite data set of cultural and museum exhibition halls.

[0062] It can be understood that the spatiotemporal calibration is represented by mapping the extracted composite data of cultural and museum exhibition halls to the same spatial reference and synchronously aligning the corresponding time parameters; the noise filtering is represented by performing redundant noise elimination and data correction operations on the extracted composite data of cultural and museum exhibition halls; the feature fusion is represented by performing feature extraction and fusion on the composite data of cultural and museum exhibition halls that have undergone spatiotemporal calibration and noise filtering to generate tourist sensory data sets, tourist behavior data sets and cultural relics status data sets; the privacy desensitization is represented by injecting privacy noise into the tourist sensory data sets, tourist behavior data sets and cultural relics status data sets, and only uploading the composite data set of cultural and museum exhibition halls after privacy desensitization.

[0063] In this embodiment, step S1 includes:

[0064] Step S11: constructing a tourist sensory data set.

[0065] It can be understood that the tourist sensory data set is represented as sensory data reflecting the tourist's visual focus and cognitive state, and at least includes the tourist's visual data and the tourist's cognitive data.

[0066] Specifically, the eye tracking module integrated into AR glasses is used to obtain the visitor's pupil gaze point coordinate sequence, gaze duration distribution, and pupil diameter change curve, and the device posture and timestamp are simultaneously recorded. A Kalman filter is applied to fuse continuous gaze point trajectories to compensate for head jitter noise. A Hampel filter is also used to remove blink noise points. The pupil diameter change curve is calibrated using the average pupil diameter of the visitor in the first 30 seconds as the baseline. Events in which the duration of continuous primary viewing of a single exhibit exceeds a preset gaze duration threshold are segmented and sliced, that is, the corresponding start and end timestamps and the target artifact ID of the event are recorded.

[0067] The device pose data is mapped to the global coordinate system of the 3D model of the cultural relic through the rigid body transformation matrix. That is, the three-dimensional coordinates of each valid gaze point are mapped to the two-dimensional UV parameter space. The face number to which the gaze point belongs is determined through ray collision detection, and the corresponding UV coordinate value is recorded. For each triangular face, all gaze event data projected onto the face are aggregated according to the visitor visual weight formula: Calculate the visual hotspot weight value, where is the visual hotspot weight value of the triangle face, is the total number of primary views of the triangle, For the The duration of the second gaze, For the The pupil diameter change rate of the corresponding period, is the area of ​​the triangular patch, a visual hot zone weight matrix is ​​constructed based on the visual hot zone weight value, and the visual hot zone weight matrix is ​​output as the tourist visual data.

[0068] Furthermore, the response time difference between a visitor clicking a button and a corresponding information popping up in the museum guide system is recorded, and the interaction delay index is mapped to [0, 1] according to the exponential decay function to achieve data normalization, and the normalized value is used as the interaction delay index; the natural language question data of the audience is collected through a microphone to generate a visitor voice text, and the visitor voice text is converted into a voice text vector based on BERT, and the L2 norm of the visitor voice text is extracted, and the L2 norm is normalized to generate semantic complexity data. When there is no visitor voice text, the semantic complexity data is set to 0; the quotient of the standard deviation and the mean of the pupil diameter change curve is used as the pupil diameter variation coefficient; for continuous gaze events with a gaze duration greater than a preset continuous gaze duration threshold, a linear fit is performed on the pupil diameter change curve data segment, and the absolute value of the corresponding slope is obtained, and the slope is used as the pupil dilation slope;

[0069] Based on the pupil diameter variation coefficient and pupil dilation slope corresponding to the tourist group benchmark, standardization is performed, and the standardized pupil diameter variation coefficient and pupil dilation slope are weightedly fused with the interaction delay index and semantic complexity data to generate a fusion index. The fusion index is normalized to generate a cognitive load index, which is output as tourist cognitive data.

[0070] For example, a tourist watches a certain cultural relic for x seconds. The pupil diameter corresponding to this event is 3.92 mm on average, with a standard deviation of 0.12. The pupil diameter variation coefficient is 0.0306, the pupil dilation slope is 0.14 mm per second, the interaction delay index is 0.92, the semantic complexity data is 0.87, and the tourist group benchmarks corresponding to the pupil diameter variation coefficient include a pupil diameter variation coefficient mean of 0.028 and a pupil diameter variation coefficient standard deviation of 0.006. The tourist group benchmarks corresponding to the pupil dilation slope include a pupil dilation slope mean of 0.1 and a pupil dilation slope standard deviation of 0. 0.05, so after standardization, the pupil diameter coefficient of variation and pupil dilation slope are 0.433 and -0.2, respectively. After weighted fusion, the fusion index obtained is 0.5*0.433+0.2*(-0.2)+0.3*0.92+0.1*0.87=0.5395, where 0.5 is the weight of the pupil diameter coefficient of variation, 0.2 is the weight of the pupil dilation slope, 0.3 is the weight of the interaction delay index, and 0.1 is the weight of the semantic complexity data. The fusion index is normalized using the Sigmoid function to obtain a cognitive load index of 0.63. (This is just an example to illustrate the calculation process of the cognitive load index; the actual calculation data must be determined according to the specific situation.)

[0071] Step S12: constructing a tourist behavior dataset.

[0072] It can be understood that the tourist behavior dataset is represented as behavioral data describing the audience's spatial movement, interactive behavior and interest preferences, and at least includes tourist behavior pattern data, tourist interest point distribution data and tourist interaction data.

[0073] Specifically, the UWB positioning base station extracts the three-dimensional coordinate point cloud of tourists, and the millimeter-wave radar scans the crowd density heat map synchronously. Kalman filtering is applied to the three-dimensional coordinate point cloud of tourists to eliminate positioning jitter. DBSCAN clustering is used to separate individuals and remove static reflection interference from the radar point cloud contained in the crowd density heat map, and the radar point cloud is mapped to the UWB coordinate system.

[0074] The curvature radius of the tourist path segment within a fixed time is calculated, and the speed entropy within the fixed time window is calculated. The group pattern recognition is performed based on the curvature radius and the speed standard deviation. For example, when the radar scan detects that the crowd density in an area with a diameter of 2 meters is not less than 5 people and the average speed of tourists is less than 0.3 meters per second, this situation is marked as a "cluster stay" group pattern; if the tourist path segment shows a curvature greater than 3 points in a row, the group pattern is marked as "cluster stay"; , marking this situation as the “flow observation” group pattern, and finally outputting the group pattern, curvature radius, speed entropy and crowd density heat map as tourist behavior pattern data.

[0075] Specifically, data is collected from the logs of the museum guide system to obtain the duration of stay on the cultural relic interface. Events where the duration of stay on the cultural relic interface does not reach the preset duration threshold are marked as invalid interest events and eliminated. Events where the duration of stay on the cultural relic interface is greater than the preset duration threshold are marked as valid interest events. A cultural relic interest area is established with the cultural relic coordinates as the center based on the crowd density heat map.

[0076] The time dimension weight of the intensity of tourists' interest in cultural relics exhibits is calculated according to the hyperbolic tangent function, and the space dimension weight of the intensity of tourists' interest in cultural relics exhibits is calculated according to the density of people flow in the cultural relics interest area. The intensity of tourists' interest in cultural relics exhibits is obtained based on the time dimension weight and the space dimension weight, and a corresponding cultural relics interest intensity vector is generated based on the cultural relics exhibits interest intensity.

[0077] For example, let's assume the ID of a cultural relic is xxx. The mean, median, and standard deviation of the visitor dwell time on the corresponding interface in the museum's guide system are 20 seconds, 20 seconds, and 3.16, respectively. The pedestrian density within the cultural relic's area of ​​interest is 3.2 people per square meter. The corresponding temporal dimension weight is tanh[(20*20) / (10*3.16)]≈1.0, where 10 is the empirical coefficient. The spatial dimension weight is min(1,3.2 / 5)=0.64, where 5 is the saturation threshold for pedestrian density within the cultural relic's area of ​​interest. Based on the temporal and spatial dimension weights, the visitor interest intensity for the cultural relic exhibit with ID xxx is 0.7*10+0.3*0.64=0.892, and the corresponding cultural relic interest intensity vector is output as {cultural relic ID: xxx, interest intensity: 0.892}. (This is just an example to illustrate the calculation process for interest intensity; the actual calculation should be determined based on the specific situation.)

[0078] The transfer probability between exhibits is calculated based on UWB trajectories, and an interest association topology map is constructed. Gaussian smoothing and speed threshold segmentation are performed on the tourist path trajectory to identify effective stay points. For example, events with a speed of less than 0.1 meters per second for 5 consecutive frames are marked as effective stay points, and the coordinates of the effective stay points are mapped to the cultural relic interest area. When the duration of consecutive stay points in the same cultural relic interest area is not less than 3 seconds, it is marked as a valid stay event and the cultural relic ID and event start and end timestamps are recorded. A counting matrix is ​​constructed to record the transfer frequency between exhibits. Laplace smoothing is used to calculate the basic probability, and a spatiotemporal correction factor is introduced to correct and compensate the basic probability to generate the transfer probability between cultural relics. The cultural relic is used as a node, and the node weight is the average stay time of tourists at the cultural relic. When the transfer probability between cultural relics is greater than the preset transfer probability threshold, a directed edge is created, and the edge weight is the transfer probability between cultural relics. An interest association topology map is generated, and the cultural relic interest intensity vector and the interest association topology map are output as the distribution data of tourist interest points.

[0079] For example, a tourist's visiting path is to first stay in front of artifact A with artifact ID 102 for 12 seconds, then walk to artifact B with artifact ID 203 and stay for 8 seconds, and finally stay in front of artifact C with artifact ID 301 for 15 seconds. The corresponding UWB trajectory point distance is: the straight line length from artifact A to artifact B is 10 meters, and it takes 15 seconds; the straight line length from artifact B to artifact C is 8 meters, and it takes 12 seconds;

[0080] The UWB trajectory of the above-mentioned tourist visiting path pair is subjected to stop event extraction. In front of cultural relic A, the speed of 240 consecutive UWB points is less than 0.1 meters per second, and the centroid of the point set is located within the cultural relic interest range of cultural relic A. This stop event is marked as a valid stop and recorded as [ID: 102, start time: xx:x0:00, end time: xx:x0:12]. In front of cultural relic B, the speed of 150 consecutive UWB points is less than 0.1 meters per second, and the centroid of the point set is located within the cultural relic interest range of cultural relic B. This stop event is marked as a valid stop and recorded as [ID: 203, start time: xx:x0:27, end time: xx:x0:35]. In front of cultural relic C, the speed of 300 consecutive UWB points is less than 0.1 meters per second, and the centroid of the point set is located within the cultural relic interest range of cultural relic C. This stop event is marked as a valid stop and recorded as [ID=301, start time: xx:x0:47, End time: xx:x1:02];

[0081] The count matrix is ​​constructed based on the historical count plus current event technology. The historical count of the transfer direction from artifact A to artifact B is 118, 1 time this time, and 119 times cumulatively; the historical count of artifact A to artifact C is 85, 0 times this time, and 85 times cumulatively; the historical count of artifact B to artifact C is 142, 1 time this time, and 143 times cumulatively;

[0082] The calculated probability of artifact A moving to artifact B is (119+0.1) / ((119+85)+2×0.1)≈0.583. After applying temporal and spatial corrections, the calculated time attenuation factor is e^(-15 / 300)=0.951, and the calculated spatial attenuation factor is 1 / (1+0.8)≈0.556. The probability of transfer between artifacts is 0.583*0.951*0.556≈0.308. (This is just an example to illustrate the calculation process of the transfer probability between artifacts. The actual calculation data must be determined according to the specific situation.)

[0083] Specifically, the interactive operations performed by tourists in the cultural and museum exhibition hall guidance system are collected, and the interactive operations are merged in time windows. For example, continuous clicks with an operation interval of less than 0.5 seconds are merged into a single operation. The semantic complexity data and the pupil diameter variation coefficient corresponding to the moment of the interactive operation are combined according to the preset weight ratio to generate tourist interaction data.

[0084] Step S13: constructing a cultural relic status dataset.

[0085] It can be understood that the cultural relic status data set is represented as status data reflecting the physical status of the cultural relic and the microenvironment in which it is located, and at least includes the physical status data of the cultural relic and the status data of the cultural relic microenvironment.

[0086] Specifically, the resonance frequency data of the cultural relic is obtained through the vibration sensor embedded in the base of the cultural relic, and the resonance frequency data is subjected to frequency domain filtering to separate the environmental noise interference and obtain the resonance frequency offset; the surface of the cultural relic is scanned with a hyperspectral camera to obtain the surface spectral data of the cultural relic, and the micro-damaged area on the surface of the cultural relic is identified based on the image segmentation algorithm to generate the surface damage gradient of the cultural relic; the stress accumulation of the cultural relic material is monitored in real time through the optical fiber strain sensor to generate the cultural relic cumulative stress value, and the resonance frequency offset, the surface damage gradient of the cultural relic and the cultural relic cumulative stress value are output as the physical state data of the cultural relic.

[0087] The temperature change curve and humidity fluctuations in the cultural relics display cabinet are recorded by temperature and humidity sensors to obtain the temperature and humidity stress index; the light sensor is used to monitor the visible light and infrared radiation intensity, and the band separation algorithm is used to distinguish the contributions of natural light and artificial light sources to obtain the light radiation flux; the micro-vibration unit captures the environmental mechanical disturbance and calculates the vibration risk value in combination with the resonance frequency data; the gas detection module analyzes the pollutant concentration, obtains the pollutant concentration, and outputs the temperature and humidity stress index, light radiation flux, vibration risk value and pollutant concentration as the cultural relics microenvironment status data.

[0088] Step S2: Perform simulation based on the composite data set of the museum and construct a visitor sensory analysis model, a visitor behavior analysis model, and a cultural relic status analysis model.

[0089] In this embodiment, step S2 includes:

[0090] Step S21: Perform tourist sensory feedback simulation based on the tourist sensory data set to obtain tourist sensory simulation results, analyze the tourist sensory simulation results through a neural network, generate tourist sensory feedback mapping rules, and build a tourist sensory analysis model based on the tourist sensory feedback mapping rules.

[0091] Specifically, the visual hotspot weight matrix is ​​projected onto the surface of the 3D model of the cultural relic using a ray tracing algorithm, and the saliency distribution under different viewing angles is calculated to generate a spatial visual field. The corresponding parameters are dynamically adjusted based on the cognitive load index. When the cognitive load index exceeds 0.7, the visual field radius is compressed and the noise is enhanced to simulate the visual blur effect under high cognitive load. When the cognitive load index is lower than 0.3, the field of view is expanded and the color saturation is reduced to simulate the perceptual diffusion in a relaxed state. When a specific cognitive state change is detected, the feedback intensity of the non-dominant sensory channel is automatically adjusted to maintain the overall perceptual balance and simulate the multi-sensory coupling effect. For example, when a sudden increase of 0.8 in semantic complexity data is detected, the signal-to-noise ratio of the auditory channel is reduced by 40%, and the tactile vibration feedback intensity is increased by 60%, forming a closed-loop sensory disturbance tourist perception feedback simulation, and outputting the tourist sensory simulation results.

[0092] The results of the tourist sensory simulation are input into a causal convolutional neural network that expands the temporal receptive field through void sampling. Based on this neural network, long-term dependencies are captured, such as extracting the evolution characteristics of tourist perception patterns within 72 hours, and key disturbance events are identified through a multi-head self-attention mechanism. For example, when visual field contraction and tactile enhancement occur simultaneously, the attention weight exceeds 0.85, and the event is marked as a sensory conflict event; a difference map between the simulated data and the real sensor data is constructed, and the residual vector is calculated. The residual corrector is trained through an adversarial generative network to compensate for the residual of the simulated data. Based on the above operations, a tourist perception feedback mapping rule is generated. The tourist perception feedback mapping rule is expressed as a quantitative mapping relationship between the tourist's biometric response and the multimodal feedback parameters corresponding to the cultural and museum exhibition hall, and a tourist sensory analysis model is constructed according to the tourist perception feedback mapping rule.

[0093] Step S22: simulate the behavior of a tourist group based on the tourist behavior data set, obtain the simulation results of the tourist group behavior, analyze the simulation results of the tourist group behavior through a neural network, generate the entropy change conduction rules of the tourist group behavior knowledge, and construct a tourist behavior analysis model based on the entropy change conduction rules of the tourist group behavior knowledge.

[0094] Specifically, based on the group pattern, the one-way flow state and the cluster stay state of tourist tour patterns are distinguished in real time. The path curvature radius quantifies the degree of curvature of the audience's path. For example, high curvature represents frequent turning. Speed ​​entropy evaluates the disorder of movement speed changes. For example, high entropy value indicates frequent speed change. The crowd density heat map maps the instantaneous congestion level of the cultural and museum exhibition space. The tourist behavior dataset is injected into the digital twin environment to generate a group behavior evolution sandbox. The real behavior characteristics are loaded for each virtual tourist in the group behavior evolution sandbox. The curvature radius and speed entropy are used to generate the path selection tendency. When a high path curvature radius and a low speed entropy path are detected, the probability of turning to a straight path is enhanced. When a low path curvature radius and a high speed entropy path are detected, the probability of selecting this path is enhanced. The audience movement speed is jointly adjusted based on the heat map and the group pattern. The movement speed is automatically reduced in areas with dense audiences. When a cluster stay pattern is identified, a detour avoidance strategy is triggered synchronously.

[0095] The cultural relic interest intensity vector directly controls the virtual tourist's dwell time allocation. The higher the interest intensity of an exhibit, the longer the generated dwell time, which can be extended up to 160% of the preset baseline duration. The transfer direction of tourists' visits is guided by the interest association topology map. The higher the edge weight of the cultural relic exhibit combination in the interest association topology map, the greater the possibility of tourists transferring from the current cultural relic exhibit to the related cultural relic exhibit. This transfer behavior is also inversely regulated by the crowd density of the target area to avoid overcrowding in the high-interest cultural relic exhibit area, which reduces the tourist experience.

[0096] The interaction intensity index is generated by integrating the operational complexity and semantic complexity data of visitors' interactive operations in the museum's guide system with the pupil diameter variation coefficient corresponding to the interaction operation according to the preset weight ratio. The interaction intensity index is then normalized and divided into three value ranges: low, medium, and high. The low value range simplifies the elements of the museum's guide system's interactive interface, the medium value range maintains the preset standard interface element distribution, and the high value range dynamically analyzes the depth of visitors' interactive operations and increases the corresponding interactive interface elements in a targeted manner. For example, when the pupil fluctuates violently and is accompanied by a high-complexity question, it will inevitably trigger the display of knowledge extension.

[0097] Based on the above operations, the entropy change conduction rules of tourist group behavior knowledge are generated. The entropy change conduction rules of tourist group behavior knowledge are expressed as the functional relationship between the cognitive association strength of the tourist group and the generation of the cross-space guide path. According to the entropy change conduction rules of tourist group behavior knowledge, a tourist behavior analysis model is constructed.

[0098] Step S23: simulate the risk transmission of micro-damage to cultural relics based on the cultural relics status data set, obtain the simulation results of the risk transmission of micro-damage to cultural relics, analyze the simulation results of the risk transmission of micro-damage to cultural relics through a neural network, generate the risk transmission rules of micro-damage to cultural relics, and construct a cultural relics status analysis model based on the risk transmission rules of micro-damage to cultural relics.

[0099] The physical state data of the cultural relics and the microenvironmental state data are injected into the digital twin platform, and a physical field coupling simulation is run. The simulation process simulates the dynamic transmission of micro-damage. For example, when the vibration risk value exceeds the preset vibration risk threshold, the resonance frequency offset increases synchronously; the fluctuation of the temperature and humidity stress index accelerates the change of the surface damage gradient; the pollutant concentration and the light radiation flux synergistically increase the cumulative stress value. The simulation results are generated based on the above simulation process;

[0100] The simulation results are input into the spatiotemporal graph convolutional neural network for analysis. The cultural relics exhibits are used as network nodes in the spatiotemporal graph convolutional neural network. The node attributes are the data contained in the corresponding cultural relic status dataset. The edge connection represents the conduction path between the physical state data of the cultural relic and the microenvironmental state data. The convolution layer extracts cross-grid damage conduction patterns, such as crack extension from the high-temperature and humidity stress zone to the adjacent low-stiffness zone. The timing layer captures cumulative stress mutation events, such as the resonance frequency deviation step caused by the continuous exceeding of the vibration risk value.

[0101] The network output layer distillation extracts the risk transmission rules for micro-damage to cultural relics. These rules represent the risk state transmission path for micro-damage caused by changes in the cultural relic microenvironment to the cultural relic structure. For example, when the temperature and humidity stress index is greater than 0.7 and the pollutant concentration exceeds 3 ppb, the surface damage gradient growth rate increases to 1.2 times the baseline value. When the vibration risk value is higher than 0.1g and the light radiation flux is greater than 4kJ / ㎡, the resonant frequency offset increases linearly with an increase of 0.5Hz per unit of vibration risk value and 0.3Hz per unit of light radiation. (This is just an example to illustrate the form of the cultural relic micro-damage risk transmission rules; the actual rules must be determined according to specific circumstances.)

[0102] Step S3: construct a cultural and museum exhibition hall guide model based on the visitor sensory analysis model, the visitor behavior analysis model, and the cultural relic status analysis model, and perform model distillation to obtain a lightweight cultural and museum exhibition hall guide model.

[0103] In this embodiment, step S3 includes:

[0104] Step S31 : constructing a museum guide model based on the visitor sensory analysis model, the visitor behavior analysis model, and the cultural relic status analysis model.

[0105] Specifically, the tourist cognitive feedback vector generated by the tourist sensory analysis model, the tourist guide path decision vector generated by the tourist behavior analysis model, and the risk parameter vector generated by the cultural relic status analysis model are temporally aligned and normalized. The cognitive load index is input into the tourist behavior analysis model to correct the virtual tourist's stay time distribution. For example, when the cognitive load index is greater than 0.6, the tourist stay time is extended by 30%. The crowd density heat map is input into the cultural relic status analysis model to calibrate the vibration risk value. For example, when the crowd density within the cultural relic interest range of the cultural relic exhibit is greater than 5 people per square meter, the risk coefficient is given a compensation weight of 1.5 times.

[0106] Among them, the tourist cognitive feedback vector is used to express the tourist cognitive status and interaction needs, the tourist guide path decision vector is used to express the tourist guide decision path and the tourist spatiotemporal distribution status, and the risk parameter vector is used to express the cultural relics damage risk and protection needs.

[0107] The rules corresponding to each model are integrated, and the priority of the rules corresponding to the cultural relics status analysis model is set to the highest globally. When the rules corresponding to the cultural relics status analysis model conflict with the rules corresponding to the tourist sensory analysis model and the tourist behavior analysis model, the rules corresponding to the tourist sensory analysis model and the tourist behavior analysis model are forcibly overwritten;

[0108] The tourist cognitive feedback vector output by the tourist sensory analysis model data needs to be used to enhance the tourist sensory perception within the cultural relics risk safety range corresponding to the risk parameter vector;

[0109] The tourist guide path decision vectors output by the tourist behavior analysis model are verified for path accessibility based on the risk parameter vector. The tourist guide path decision vectors that fail the path accessibility verification are eliminated, and only the tourist guide path decision vectors that pass the path accessibility verification are output, thereby constructing a tourist guide decision rule system.

[0110] Furthermore, based on the real-time scene status of the cultural and museum exhibition hall, the weight distribution coefficients of the tourist sensory analysis model, the tourist behavior analysis model and the cultural relic status analysis model are obtained, and the weight ratio of the corresponding feature vectors of each model is dynamically adjusted according to the weight distribution coefficient to generate a joint feature vector of the cultural and museum exhibition hall.

[0111] For example, during peak hours at a certain museum, the real-time output data from the three models were as follows: the visitor sensory analysis model detected an average cognitive load index of 0.68 and a visual thermal weight of 0.85 for the xxx cultural relic exhibition area; the visitor behavior analysis model outputted a transfer probability of 0.79 for exhibits from cultural relic A to cultural relic B, and monitored a crowd density of 5.2 people per square meter; the cultural relic status analysis model detected a resonant frequency offset of +3.1Hz, exceeding the preset safety threshold of 3Hz, and a surface damage gradient of 1.2 microns per hour;

[0112] The dynamic weight coefficients are calculated based on the preset weight distribution mechanism. After weighted calculation and normalization to ensure the total is 100%, the weight of the cultural relics status analysis model is 47.8%, the weight of the tourist behavior analysis model is 29.8%, and the weight of the tourist sensory analysis model is 22.4%. After weighting the corresponding data of the tourist sensory analysis model, tourist behavior analysis model, and cultural relics status analysis model, the output vectors are {0.68*0.224=0.152,0.85*0.224=0.190}; {0.358, 0.268}; {0.296, 0.574}, respectively. The first and second dimensions of the above three vectors are fused to generate {0.806, 1.032}. Since the first dimension data exceeds the preset safety threshold, the cultural relics protection strategy is triggered and the corresponding operation is executed. The second dimension data exceeds the preset guide threshold, triggering the forced diversion mechanism and the corresponding operation. (This is just an example of dynamically adjusting the weight ratio of the corresponding feature vectors of each model according to the weight distribution coefficient to generate the form of the joint feature vector of cultural and museum exhibition halls. The actual data needs to be determined by the specific situation.)

[0113] Step S32: Perform model distillation on the museum guide model to obtain a lightweight museum guide model.

[0114] Specifically, a terminal portrait is constructed based on the hardware characteristics of the cultural and museum exhibition hall terminal. The terminal portrait includes the visitor user end, the edge node end and the edge cloud end. The gradient mask technology is used to directionally distill the rules contained in the cultural and museum exhibition hall guide model into a lightweight model corresponding to the terminal portrait.

[0115] Understandably, for the visitor user side, only the sensory feedback rules related to the visitor sensory analysis model in the museum guide model are distilled, and the risk response priority logic is retained. For example, a distilled sensory feedback rule is that when the cognitive load index is greater than 0.7 and the visitor's visual hotspot weight value is greater than 0.8, the AR projection brightness is reduced to 60% of the baseline value and the voice commentary is compressed to 50% of its original length. When a risk alert occurs, the sensory rule is forcibly overridden. For example, if the resonance frequency offset exceeds the threshold, the AR vibration effect is preferentially disabled to ensure that the risk of cultural relics is controlled. (This is just an example of the form of the visitor user side distillation rule; the actual method needs to be determined by the specific situation.)

[0116] As you can understand, edge nodes distill rules related to the visitor behavior analysis model from the museum's navigation model, preserving visitor routing decisions and multi-terminal coordination capabilities. For example, a distilled rule related to the visitor behavior analysis model specifies that when the crowd density in the target cultural relics exhibition area exceeds 4 people per square meter, an alternative route is generated to avoid areas with a crowd density greater than 5 people per square meter. (This is just an example of the form of edge node distillation rules; actual distillation needs to be determined based on specific circumstances.)

[0117] It's understandable that edge cloud distillation of rules related to the cultural relic status analysis model within the museum navigation model, along with training for cross-system global scheduling capabilities, preserves cultural relic risk assessment and model rule optimization capabilities. For example, a rule related to a distilled cultural relic status analysis model states that if the resonant frequency offset of an exhibit exceeds a preset threshold for five consecutive minutes, the artifact is marked as high-risk and the maximum number of visitors to the artifact exhibition area is restricted. (This is just an example of the form of edge cloud distillation rules; actual rules will be determined based on specific circumstances.)

[0118] Step S4: deploy the lightweight museum guide model to each terminal of the museum, including the visitor user terminal, edge node terminal and edge cloud terminal, and generate visitor guide decisions based on the lightweight museum guide model.

[0119] Specifically, the tourist user end collects and analyzes the tourist's sensory data in real time, generates the tourist's sensory needs and uploads them to the edge node end simultaneously. The edge node end performs risk verification on the tourist's sensory needs and uploads the results of the risk verification to the edge cloud. The edge cloud generates corresponding tourist guide decisions for the tourist's sensory needs that pass the risk verification, and reconstructs and adjusts the needs for the tourist's sensory needs that fail the risk verification.

[0120] Furthermore, the edge cloud sends the generated tourist guide decision to the edge node end, the edge node end analyzes the tourist guide decision, generates the tourist guide decision analysis result and sends it to the tourist user end, and the tourist user end performs corresponding operations based on the tourist guide decision analysis result and the tourist sensory needs.

[0121] For example, when a tourist looks at cultural relic A, AR glasses capture the visual focus hot zone weight of 0.85 through eye tracking, and simultaneously detect the cognitive load index of 0.75, and generate the tourist demand "add 3D animation demonstration of the craft details of cultural relic A with a duration of 60 seconds and a brightness of 80%". The tourist demand is converted into vector form and uploaded to the edge node end with encryption. The edge node end retrieves the real-time data corresponding to the status of cultural relic A for risk verification. The verification result is "demand triggers high risk of cultural relic status". The edge node end encapsulates the verification result and the tourist demand into vector form and uploads it to the edge cloud. The edge cloud end reconstructs the unsuccessful tourist demand and generates a reconstruction plan that meets the status of cultural relic A. The solution is to "add a high-precision static model of the craftsmanship details of Artifact A, limit the brightness to 60%, and compress the display duration to 30 seconds." This generates a guide decision vector {operation: static model display, parameters: brightness = 60%, duration = 30 seconds} and sends it to the edge node. The edge node decodes the guide decision vector and adapts it to the local device, generating a decoding result: "Project a static 3D model of Artifact A on the (X:120, Y:90, Z:50) coordinate mat for 30 seconds, and simultaneously reduce the brightness of the display cabinet LED to 60%." The corresponding instruction is sent to the corresponding terminal. Ultimately, the visitor's user terminal will be presented with a static model of Artifact A at 60% brightness for 30 seconds in the visitor's field of view. (This is just an example to illustrate the process of generating visitor guide decisions using a lightweight cultural and museum exhibition guide model. The actual process needs to be determined according to specific circumstances.)

[0122] Step S5: Based on the visitor guide decision, a virtual-reality integrated guide is conducted for the cultural and museum exhibition hall, a guide result is obtained, and the lightweight cultural and museum exhibition hall guide model is optimized and adjusted according to the guide result.

[0123] Specifically, the edge cloud periodically aggregates desensitized tour guide data from external museums. For example, if the average cognitive load index in front of artifact A's display case is 0.62, cross-museum collaborative iteration is achieved through a federated learning framework. Each museum locally encrypts sensitive information and then uploads feature parameters. For example, if Museum X provides the incremental vibration response weight for the type of artifact corresponding to artifact A, the edge cloud aggregates and generates new model weights. For example, the update rule states: When the semantic complexity of the visitor interaction text for the type of artifact corresponding to artifact A is greater than 0.8, the AR information density is reduced from 50% to 60%, and tactile guidance is added. (This is just an example of the process of periodically analyzing the tour guide results on the edge cloud to identify optimization points and using federated learning to incrementally iterate the model parameters. The actual process will need to be determined according to the specific situation.)

[0124] For high-frequency conflict scenarios, the core rules are re-distilled through the digital twin environment to generate lightweight model incremental packages and distribute them to the edge node and tourist user ends. The edge node and tourist user ends optimize the rules based on the lightweight model incremental packages.

[0125] Figure 2 A schematic diagram of a virtual-reality integrated tour guide system for cultural and museum exhibition halls based on digital twins, which is provided in some embodiments of the present application and can realize the ideas of the present application, is shown.

[0126] Specifically, the digital twin-based virtual-reality integrated tour guide system for cultural and museum exhibition halls includes:

[0127] The data acquisition module is used to extract data from the cultural and museum exhibition hall, and perform data preprocessing on the extracted cultural and museum exhibition hall composite data to obtain a cultural and museum exhibition hall composite data set.

[0128] The simulation module is used to perform simulation based on the composite data set of the museum and to construct a visitor sensory analysis model, a visitor behavior analysis model and a cultural relic status analysis model.

[0129] The model fusion module is used to perform model fusion and model distillation operations on the visitor sensory analysis model, the visitor behavior analysis model and the cultural relic status analysis model to generate a lightweight cultural and museum exhibition hall guide model.

[0130] The decision-making generation module is used to analyze the status of the cultural and museum exhibition hall in real time according to the lightweight cultural and museum exhibition hall guidance model and generate tourist guidance decisions.

[0131] The feedback optimization module is used to optimize and adjust the lightweight cultural and museum exhibition hall guidance model according to the implementation effect of the visitor guidance decision.

[0132] The specific usage and function of this embodiment are described below:

[0133] First, data is extracted from the cultural and museum exhibition hall to obtain composite data of the cultural and museum exhibition hall, and data preprocessing is performed on the composite data of the cultural and museum exhibition hall to generate a composite data set of the cultural and museum exhibition hall. By performing multi-dimensional data extraction and data preprocessing on the cultural and museum exhibition hall, high-precision input is provided for subsequent model construction, thereby improving the accuracy of subsequent model construction. Simulation is performed based on the composite data set of the cultural and museum exhibition hall to construct a tourist sensory analysis model, a tourist behavior analysis model, and a cultural relic status analysis model. By constructing the tourist sensory analysis model, personalized analysis and sensory feedback of tourist needs are achieved. The generation of tourist guide paths is optimized according to the tourist behavior analysis model, and the status of cultural relics is monitored in real time based on the cultural relic status analysis model to reduce the probability of damage to cultural relics.

[0134] Next, a cultural and museum exhibition hall navigation model was constructed based on the visitor sensory analysis model, the visitor behavior analysis model, and the cultural relic status analysis model. Model distillation was then performed to obtain a lightweight cultural and museum exhibition hall navigation model. Model fusion was used to resolve conflicts between multi-model decisions. At the same time, during the model fusion stage, the rules corresponding to each model were integrated and compensated to generate a rule system that takes into account both visitor needs and cultural relic protection needs. This ensures that both visitor experience and cultural relic risk prevention are taken into account while ensuring accurate decision-making. Furthermore, through end-to-end distillation, lightweight models that meet the terminal characteristics are deployed for different terminals, thereby reducing the power consumption and response delay of the navigation system.

[0135] Finally, the lightweight cultural and museum exhibition hall guidance model is deployed to each terminal of the cultural and museum exhibition hall, and tourist guidance decisions are generated based on the lightweight cultural and museum exhibition hall guidance model. Based on the tourist guidance decisions, the cultural and museum exhibition hall is guided by a virtual-reality fusion guidance method to obtain guidance results. According to the guidance results, the lightweight cultural and museum exhibition hall guidance model is optimized and adjusted. By iteratively optimizing the model, the accuracy and stability of the model-generated decisions are maintained. Based on the above aspects, tourists are provided with cultural and museum exhibition hall guidance decisions that take into account both the safety protection of cultural relics and the tourist visiting experience.

[0136] An electronic device, comprising:

[0137] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method proposed in the first embodiment of the present invention.

[0138] The following is a detailed introduction to the various components of electronic equipment:

[0139] The term "processor" is the control center of an electronic device and can be a single processor or a collective term for multiple processing elements. For example, the processor can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the first embodiment of the present invention, such as one or more digital signal processors (DSPs) or one or more field programmable gate arrays (FPGAs).

[0140] The processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0141] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0142] The memory may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory may be integrated with the processor or exist independently and be coupled to the processor via an interface circuit of the electronic device, and this is not specifically limited in the embodiments of the present invention.

[0143] The above embodiments can be implemented in whole or in part via software, hardware (e.g., circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wireless communication (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer, or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0144] It should be understood that the term "and / or" as used herein simply describes an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent the existence of A alone, the existence of both A and B, or the existence of B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the related objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0145] It should be understood that in the embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0146] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A virtual-reality integrated tour guide method for cultural and museum exhibition halls based on digital twins is characterized by: The following steps are included: Extracting data from the cultural and museum exhibition hall to obtain composite data of the cultural and museum exhibition hall, performing data preprocessing on the composite data of the cultural and museum exhibition hall to generate a composite data set of the cultural and museum exhibition hall, wherein the composite data set of the cultural and museum exhibition hall includes a visitor sensory data set, a visitor behavior data set, and a cultural relic status data set; Conduct simulations based on the complex data set of cultural and museum exhibition halls to build visitor sensory analysis models, visitor behavior analysis models, and cultural relic status analysis models; A cultural and museum exhibition guide model is constructed based on the visitor sensory analysis model, visitor behavior analysis model, and cultural relic status analysis model, and a lightweight cultural and museum exhibition guide model is obtained through model distillation. Deploy the lightweight museum guide model to each terminal in the museum, including the visitor user terminal, edge node terminal, and edge cloud terminal, and generate visitor guide decisions based on the lightweight museum guide model. Based on the tourist guide decision, a virtual-reality integrated guide is conducted for the cultural and museum exhibition hall, a guide result is obtained, and a lightweight cultural and museum exhibition hall guide model is optimized and adjusted according to the guide result.

2. The virtual-reality integrated navigation method for cultural and museum exhibition halls based on digital twins according to claim 1 is characterized in that: Based on the complex data set of cultural and museum exhibition halls, simulation is carried out to build visitor sensory analysis models, visitor behavior analysis models, and cultural relic status analysis models, including: Perform tourist sensory feedback simulation based on the tourist sensory data set, obtain the tourist sensory simulation results, analyze the tourist sensory simulation results through a neural network, generate tourist sensory feedback mapping rules, and build a tourist sensory analysis model based on the tourist sensory feedback mapping rules; Among them, the tourist perception feedback mapping rule is expressed as the quantitative mapping relationship between the tourist's biometric response and the multimodal feedback parameters corresponding to the cultural and museum exhibition hall; Based on the tourist behavior data set, tourist group behavior simulation is carried out to obtain the tourist group behavior simulation results, and the tourist group behavior simulation results are analyzed through a neural network to generate the tourist group behavior knowledge entropy change conduction rules, and a tourist behavior analysis model is constructed based on the tourist group behavior knowledge entropy change conduction rules; Among them, the conduction rule of tourist group behavior knowledge entropy change is expressed as the functional relationship between the tourist group cognitive association strength and the generation of cross-space guided path; Conduct a micro-damage risk transmission simulation on cultural relics based on the cultural relics status data set, obtain the simulation results of micro-damage risk transmission on cultural relics, analyze the simulation results of micro-damage risk transmission on cultural relics through neural networks, generate micro-damage risk transmission rules for cultural relics, and build a cultural relics status analysis model based on the micro-damage risk transmission rules for cultural relics; Among them, the risk transmission rule of micro-damage to cultural relics is expressed as the risk state transmission path of micro-damage caused by changes in the micro-environment of cultural relics to the structure of cultural relics.

3. The virtual-reality integrated navigation method for cultural and museum exhibition halls based on digital twins according to claim 1 is characterized in that: A cultural and museum guide model is constructed based on the visitor sensory analysis model, visitor behavior analysis model, and cultural relic status analysis model, including: The tourist cognitive feedback vector generated by the tourist sensory analysis model, the tourist guide path decision vector generated by the tourist behavior analysis model, and the risk parameter vector generated by the cultural relic status analysis model are temporally aligned and normalized, and the rules corresponding to each model are integrated to construct a guide decision rule system. Among them, the tourist cognitive feedback vector is used to express the tourist cognitive state and interaction needs, the tourist guide path decision vector is used to express the tourist guide decision path and the tourist spatiotemporal distribution state, and the risk parameter vector is used to express the cultural relic damage risk and protection needs; Based on the real-time scene status of the cultural and museum exhibition hall, the weight distribution coefficients of the tourist sensory analysis model, tourist behavior analysis model and cultural relic status analysis model are obtained. The weight ratio of the corresponding feature vectors of each model is dynamically adjusted according to the weight distribution coefficient to generate a joint feature vector of the cultural and museum exhibition hall.

4. The virtual-reality integrated navigation method for cultural and museum exhibition halls based on digital twins according to claim 3 is characterized in that: The rules corresponding to each model are integrated to build a navigation decision rule system, including: The priority of the rules corresponding to the cultural relics status analysis model is set to the highest globally. When the rules corresponding to the cultural relics status analysis model conflict with the rules corresponding to the tourist sensory analysis model and the tourist behavior analysis model, the rules corresponding to the tourist sensory analysis model and the tourist behavior analysis model are forcibly overwritten. The tourist cognitive feedback vector output by the tourist sensory analysis model data needs to be used to enhance the tourist sensory perception within the cultural relics risk safety range corresponding to the risk parameter vector; The tourist guide path decision vector output by the tourist behavior analysis model is verified for path accessibility based on the risk parameter vector. The tourist guide path decision vectors that fail the path accessibility verification are eliminated, and only the tourist guide path decision vectors that pass the path accessibility verification are output.

5. The virtual-reality integrated navigation method for cultural and museum exhibition halls based on digital twins according to claim 1 is characterized in that: Perform model distillation to obtain a lightweight museum guide model, including: Based on the hardware characteristics of the cultural and museum terminals, a terminal portrait is constructed. The terminal portrait includes the visitor user terminal, the edge node terminal, and the edge cloud terminal. The gradient masking technology is used to distill the rules contained in the cultural and museum guide model into a lightweight model corresponding to the terminal portrait. For the tourist user side, only the sensory feedback rules related to the tourist sensory analysis model in the museum guide model are distilled, and the risk response priority logic is retained; For edge nodes, the rules related to the tourist behavior analysis model in the museum guide model are distilled to retain the tourist guidance path decision-making and multi-terminal coordination capabilities; For the edge cloud, the rules related to the cultural relics status analysis model in the cultural and museum guide model are distilled, and the cross-system global scheduling capabilities are trained to retain the cultural relics risk assessment and model rule optimization capabilities.

6. The virtual-reality integrated navigation method for cultural and museum exhibition halls based on digital twins according to claim 1 is characterized in that: The lightweight museum guide model is deployed to each terminal in the museum, and tourist guide decisions are generated based on the lightweight museum guide model, including: The tourist user end collects and analyzes the tourist's sensory data in real time, generates the tourist's sensory needs and uploads them to the edge node end simultaneously. The edge node end performs risk verification on the tourist's sensory needs and uploads the risk verification results to the edge cloud end. The edge cloud end generates corresponding tourist guide decisions for the tourist sensory needs that pass the risk verification, and reconstructs and adjusts the needs of tourists that fail the risk verification. The edge cloud sends the generated tourist guide decision to the edge node. The edge node analyzes the tourist guide decision, generates the tourist guide decision analysis results and sends them to the tourist user end. The tourist user end performs corresponding operations based on the tourist guide decision analysis results and the tourist sensory needs.

7. The virtual-reality integrated navigation method for cultural and museum exhibition halls based on digital twins according to claim 1 is characterized in that: Based on the guidance results, the lightweight cultural and museum exhibition hall guidance model is optimized and adjusted, including: The edge cloud periodically analyzes navigation results, identifies optimization points, and uses federated learning to incrementally iterate model parameters; For high-frequency conflict scenarios, the core rules are re-distilled through the digital twin environment to generate lightweight model incremental packages and distribute them to the edge node and tourist user ends. The edge node and tourist user ends optimize the rules based on the lightweight model incremental packages.

8. The virtual-reality integrated navigation method for cultural and museum exhibition halls based on digital twins according to claim 1 is characterized in that: Extract data from the cultural and museum exhibition hall to obtain composite data of the cultural and museum exhibition hall, perform data preprocessing on the composite data of the cultural and museum exhibition hall, and generate a composite data set of the cultural and museum exhibition hall, including: Perform spatiotemporal calibration, noise filtering, feature fusion, and privacy desensitization operations on the composite data of cultural and museum exhibition halls to generate a composite data set of cultural and museum exhibition halls; The tourist sensory data set is represented as sensory data reflecting the tourist's visual focus and cognitive state, and at least includes tourist visual data and tourist cognitive data; The tourist behavior dataset is represented as behavioral data describing the audience's spatial movement, interactive behavior and interest preferences, and at least includes tourist behavior pattern data, tourist interest point distribution data and tourist interaction data; The cultural relic status data set is represented as status data reflecting the physical status of the cultural relic and the microenvironment in which it is located, and at least includes the physical status data of the cultural relic and the status data of the cultural relic microenvironment.

9. The virtual-reality integrated navigation method for cultural and museum exhibition halls based on digital twins according to claim 8 is characterized in that: Perform spatiotemporal calibration, noise filtering, feature fusion, and privacy desensitization on the composite data of cultural and museum exhibition halls to generate a composite data set of cultural and museum exhibition halls, including: The spatiotemporal calibration is to map the extracted composite data of cultural and museum exhibition halls to the same spatial reference and synchronously align the corresponding time parameters; The noise filtering is performed to remove redundant noise and correct data on the extracted composite data of cultural and museum exhibition halls; The feature fusion is performed by extracting and fusing the composite data of the museum and exhibition hall that has been time-space calibrated and noise filtered, generating a tourist sensory data set, a tourist behavior data set, and a cultural relic status data set; The privacy desensitization is performed by injecting privacy noise into the tourist sensory dataset, the tourist behavior dataset, and the cultural relic status dataset, and only uploading the composite data set of the cultural relics exhibition hall after privacy desensitization.

10. The virtual-reality integrated tour guide system for cultural and museum exhibition halls based on digital twins is characterized by: include: A data acquisition module is used to extract data from cultural and museum exhibition halls and perform data preprocessing on the extracted composite data of cultural and museum exhibition halls to obtain a composite data set of cultural and museum exhibition halls; A simulation module is used to perform simulation based on a composite data set of a museum and to construct a visitor sensory analysis model, a visitor behavior analysis model, and a cultural relic status analysis model; A model fusion module is used to perform model fusion and model distillation operations on the visitor sensory analysis model, the visitor behavior analysis model, and the cultural relic status analysis model to generate a lightweight cultural and museum exhibition hall guide model; A decision-making module, which is used to analyze the status of the cultural and museum exhibition hall in real time based on the lightweight cultural and museum exhibition hall guidance model and generate tourist guidance decisions; The feedback optimization module is used to optimize and adjust the lightweight cultural and museum exhibition hall guidance model according to the implementation effect of the visitor guidance decision.

Images