An ar interaction method and system for health and epidemic prevention popularization

Abstract

The application discloses an AR interaction method and system for health and epidemic prevention popularization, and relates to the technical field of AR interaction, which comprises the following steps: constructing an epidemic disease transmission mechanism equation group based on literature retrieval and parameter card, and inputting an interaction instruction in a computer interface, wherein the interaction instruction corresponds to a protection action and is mapped into a control variable of the mechanism equation. The system realizes the relationship fitting between the interaction action and the multidimensional parameters of the parameter card through a double-axis evolution knowledge graph network, realizes real-time calculation of intuitive parameters, and generates counterfactual comparison results and action contribution degrees. The results are displayed in a double-track manner on the computer interface, intuitively showing the interaction action state and the mechanism evolution curve, and providing instant feedback based on the counterfactual difference. When complex environmental conditions (such as the existence of multiple local flow fields) are detected, a mixed calculation module is triggered, a global flow field is generated through local template scaling and attention superposition, and the parameters are calibrated. The method improves the scientificity and interactivity of epidemic prevention popularization.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of AR interaction, in particular to an AR interaction method and system for health and epidemic prevention popularization. BACKGROUND

[0002] With the frequent occurrence of public health events, the demand for popularization of epidemic prevention knowledge is increasing. However, the existing popularization means of epidemic prevention still mainly rely on text propaganda, video explanation and passive transmission, and lack of interactivity and individualized feedback. Especially when facing different groups of people, it is difficult to intuitively show the causal relationship between protective actions and epidemic transmission mechanism, resulting in that the public's understanding of epidemic prevention measures stays at the surface level and lacks scientificity.

[0003] On the other hand, modern computer technology and augmented reality (AR) have been gradually applied to education and medical popularization, but most systems only stay in static visualization or simple action demonstration, and cannot realize mechanism extraction, parameterized modeling and interactive evolution analysis based on scientific literature. For example, traditional epidemiological models such as SIR and SEIR can simulate the macro transmission process, but cannot accurately couple individual actions with micro transmission factors, nor can they provide instant feedback based on counterfactual comparison, making it difficult for users to truly understand the specific contribution of a certain action to the reduction of infection risk. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an AR interaction method for health and epidemic prevention popularization to solve the problem of how to couple the mechanism knowledge in epidemic prevention literature with interactive actions to model, realize interactive transmission evolution visualization and provide scientific feedback based on counterfactual comparison.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides an AR interaction method for health and epidemic prevention popularization, which comprises: analyzing and parameterizing related literature on epidemic prevention to form a parameter card;

[0008] Based on the search of the related literature and the parameter card, the epidemic transmission mechanism equation set is generated;

[0009] An interactive instruction is input through a computer interface, the interactive instruction corresponds to the selection or switching of protective actions, and is mapped to the control quantity of the epidemic mechanism equation;

[0010] The control quantity is injected into the epidemic transmission mechanism equation set to calculate the infection probability, exposure dose, air concentration and surface pathogen load in real time, and generate a counterfactual comparison result and action contribution degree based on the interactive instruction.

[0011] The interactive action state and mechanism evolution curve are displayed in a double-track manner in the computer interface, and instant feedback is provided based on the counterfactual difference value;

[0012] In the calculation process of the epidemic transmission mechanism equation group, when a complex environmental condition is detected, a hybrid calculation module is triggered, and the parameters of the epidemic transmission mechanism equation group are calibrated using the calculation results.

[0013] As a preferred scheme of the AR interaction method for health and epidemic prevention popularization, wherein:

[0014] The relevant literature includes articles on the target epidemic and a set of credible sources constructed;

[0015] The article content of the relevant literature is disassembled to obtain the micro mechanism and macro mechanism of the epidemic action;

[0016] According to the micro mechanism and the macro mechanism, a directly corresponding parameter card is generated.

[0017] The coupling relationship between the micro mechanism, the macro mechanism, and the micro mechanism and the macro mechanism is utilized to supplement the coupling relationship of the parameter card.

[0018] As a preferred scheme of the AR interaction method for health and epidemic prevention popularization, wherein:

[0019] The micro mechanism is the individual attribute of the pathogenic microorganism, including elements: morphological characteristics of the pathogenic microorganism, sensitivity to different daily detergents, penetration of individual size to different isolation barriers, survival and decay process of individual in different environments, and pathogenic process of individual;

[0020] The macro mechanism includes elements: spatial transmission process, reaction intensity of the epidemic in different populations, and susceptible population;

[0021] The generation process of the parameter card is:

[0022] Step 1: Through the retrieval of the relevant literature, the first stage of learning is carried out; in the first stage of learning, only any mechanism element and multi-dimensional parameter corresponding to the mechanism element in the micro mechanism or the macro mechanism are retrieved, each retrieval result is taken as an evidence, and an anchor point is buried in each evidence position; the similarity of retrieval is taken as the evidence weight;

[0023] Based on the j-dimensional parameters of the mechanistic elements in each piece of evidence i, a parameter file is generated; all evidence is integrated to obtain parameters containing J dimensions of all evidence; in the J-dimensional parameter space generated from all parameter files, each parameter file is mapped and the evidence weight is marked to obtain the parameter card;

[0024] Where i represents the evidence index, j represents the parameter dimension reflected in the evidence, and J represents the dimension after taking the union of all dimensions of all evidence;

[0025] Step 2: After retrieving the relevant literature and locating the anchor points of each piece of evidence during the first stage of learning, proceed to the second stage of learning: analyze the coupling relationship between the anchor points and generate candidate edges for the rules, including: co-occurrence rules, causal triggering rules, conditionalization rules, and mutually exclusive or cooperative rules.

[0026] At the same time, the confidence level of each type of edge is used as the corresponding weight.

[0027] As a preferred embodiment of the AR interactive method for public health and epidemic prevention education described in this invention, wherein:

[0028] The disease transmission mechanism equation set includes acquiring interactive actions input on the computer interface and encoding them as interactive action operators;

[0029] In the relevant literature, after locating the anchor point, the third stage of learning is performed. Actions similar to the interactive action d are retrieved from the relevant literature, and control processes or interactive actions with similar working principles to the interactive action are retrieved through transfer learning as action nodes. Furthermore, based on the representation of each action node in the corresponding literature, it is encoded as an action node. Action operators are used to generate action nodes. The vector difference between the action operator and the interaction action operator encoded by the interaction action d is used as the vector difference for each action node. offset features ;

[0030] Where 'e' represents the index of the action node, This represents the e-th action node of the interactive action d;

[0031] The disease transmission mechanism equations are constructed using a dual-axis evolutionary knowledge graph network, generating interactive actions and the interaction relationships between parameters of each dimension in each mechanism element. The dual-axis evolutionary knowledge graph network consists of two channels, a vertical flow and a horizontal flow, as well as a cross-axis fusion head, and is connected between the two channels by a modulation cancellation layer.

[0032] As a preferred embodiment of the AR interactive method for public health and epidemic prevention education described in this invention, wherein:

[0033] The lateral flow consists of two layers of graph convolutional networks with attention mechanisms. It takes a bipartite graph of action nodes and multidimensional parameter nodes of parameter cards as input. During information transmission, attention modulation is performed through the direction and magnitude of the offset vector, edge type, and evidence confidence. It is used to learn the structural differences between interactive actions and multidimensional parameters of parameter cards and output lateral adjustment cues dimension by dimension.

[0034] The longitudinal flow is composed of a fully connected neural network with monotonic constraints. Taking the action depth as input, it can generate a monotonic influence curve between the action depth and the multidimensional parameters of the parameter card. During aggregation, evidence confidence and condition labels are introduced to automatically reduce the weight or block evidence that does not meet the applicable conditions: if the evidence only partially matches the conditions, the automatic weight reduction operation is performed; if the evidence does not match the conditions at all, the automatic blocking operation is performed.

[0035] The modulation cancellation layer is placed between the transverse and longitudinal flows. It achieves mutual cancellation of transverse and longitudinal features through residual orthogonalization and decorrelation constraints, thereby avoiding coupling interference between transverse and longitudinal channels in information learning.

[0036] The cross-axis fusion head consists of two fully connected networks and features a physical feasible domain projection and a bidirectional smoothing mechanism in the output stage.

[0037] As a preferred embodiment of the AR interactive method for public health and epidemic prevention education described in this invention, wherein:

[0038] The control quantity is injected into the disease transmission mechanism equation set to generate the corresponding value of each dimension parameter on the parameter card; and according to the corresponding value, it is mapped to: infection probability, exposure dose, air concentration and surface pathogen load through a preset mapping function.

[0039] Simultaneously, the actual state of the interactive action and the corresponding counterfactual baseline state are calculated; wherein, the counterfactual baseline state is generated by setting the interactive action corresponding to the interactive instruction to a virtual state of not being executed or having the minimum execution strength, while keeping other actions unchanged;

[0040] By comparing the calculation results of the disease transmission mechanism equations under actual conditions and counterfactual baseline conditions, the differences in infection probability, exposure dose, air concentration, and surface pathogen load are obtained as counterfactual comparison results.

[0041] Based on the counterfactual comparison results, and combined with the evidence weights of each dimension parameter in the parameter card, the action contribution of each interactive action is calculated.

[0042] As a preferred embodiment of the AR interactive method for public health and epidemic prevention education described in this invention, wherein:

[0043] The complex environment condition comprises, in the process of simulating the epidemic disease transmission, multiple local different air flow fields exist;

[0044] The mixed computing module comprises, according to the scaling and superposition of the pre-constructed local flow field template, the global flow field is made to conform to the current environment, according to the scaling and superposition process of the local flow field template, the global attention parameter is distributed, and the global attention weight is obtained through global normalization;

[0045] The output result of injecting the control quantity into the epidemic disease transmission mechanism equation group is globally distributed according to attention;

[0046] The local flow field template comprises, assuming that the global is composed of multiple fixed unit minimum spaces, the local typical flow field space is obtained through simulation of a typical space; each space unit is provided with an attention parameter, which is proportional to the overall strength of the local flow field template.

[0047] In a second aspect, the present application provides an AR interaction system for health epidemic science popularization, comprising: a collection unit, which analyzes and parameterizes relevant literature on epidemic prevention to form a parameter card;

[0048] An analysis unit generates the epidemic disease transmission mechanism equation group based on the search of the relevant literature and the parameter card;

[0049] A control unit inputs an interactive instruction through a computer interface, the interactive instruction corresponds to the selection or switching of a protective action, and is mapped into a control quantity of the epidemic disease mechanism equation;

[0050] A feedback unit injects the control quantity into the epidemic disease transmission mechanism equation group, and calculates the infection probability, exposure dose, air concentration and surface pathogen load in real time, and generates a counterfactual comparison result and action contribution degree based on the interactive instruction; the interactive action state and mechanism evolution curve are displayed in a double-track manner in the computer interface, and instant feedback is provided based on the counterfactual difference.

[0051] In a third aspect, the present application provides a computer device comprising a memory and a processor, the memory stores a computer program, wherein: the computer program is executed by the processor to realize any step of the AR interaction method for health epidemic science popularization according to the first aspect of the present application.

[0052] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, wherein: the computer program is executed by the processor to realize any step of the AR interaction method for health epidemic science popularization according to the first aspect of the present application.

[0053] The application has the beneficial effects that: through literature analysis and parameter card construction, the epidemic prevention knowledge is converted from abstract text into quantifiable mechanism parameters; through the epidemic disease transmission mechanism equation set and the double-axis evolution knowledge graph network, the interactive action and the multi-dimensional parameter are accurately fitted, and the scientific effect of the protection behavior can be presented in real time. Compared with the traditional one-way popular science mode, the application can generate counterfactual comparison results and action contribution, directly reveal "if a certain action is not executed, how will the infection risk change", and greatly enhance the interactivity and persuasiveness of popular science. At the same time, the mixed calculation module is introduced in the complex environment, the local template flow field and the attention scaling superposition are generated to generate the global flow field matrix, the model is calibrated, and the calculation precision and environmental adaptability are effectively improved. In summary, the application not only improves the scientific cognition of the public to the epidemic prevention measures, but also helps the user to understand the effect of the combination of multiple actions, so as to realize the personalized and interactive health and epidemic prevention popular science, and has high application and popularization value. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creating laborious work.

[0055] Figure 1 The flow chart of the AR interactive method for health and epidemic prevention popularization. DETAILED DESCRIPTION

[0056] In order to make the above-mentioned purposes, features and advantages of the application more apparent and easy to understand, the specific embodiments of the application will be described in detail below with reference to the drawings of the specification.

[0057] In the following description, many specific details are set forth in order to provide a thorough understanding of the application, but the application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the application, therefore the application is not limited to the specific embodiments disclosed below.

[0058] Secondly, "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0059] REFERENCE Figure 1 For one embodiment of the application, the embodiment provides an AR interactive method for health and epidemic prevention popularization, including the following steps:

[0060] S1: Analyze and extract parameters from relevant literature on epidemic prevention to form a parameter card.

[0061] Further, the relevant literature includes articles on the target epidemic, and a collection of credible sources is constructed. The article content of the relevant literature is disassembled to obtain the micro and macro mechanisms of the epidemic. First, a collection of credible sources is established, which includes articles related to the target epidemic selected from medical databases, public health agencies, academic journals, and authoritative guidelines. By searching and organizing the relevant literature, the system disassembles the article content and extracts the micro and macro mechanisms involved.

[0062] According to the micro and macro mechanisms, the corresponding parameter card is generated. At the same time, the coupling relationship between the micro and macro mechanisms is used to supplement the parameter card.

[0063] The micro mechanism is the individual attribute of the pathogenic microorganism, including elements such as the morphological characteristics of the pathogenic microorganism, the sensitivity to different daily detergents, the penetration of individual size to different isolation barriers, the survival and decay process of individuals in different environments, and the pathogenic process of individuals. By extracting the micro mechanism, the parameter card can provide parameters at the "pathogen itself characteristics" level. It is used to demonstrate the role of the crowd and achieve the explanation of epidemic prevention.

[0064] The macro mechanism includes elements such as the transmission process in space, the reaction intensity of the epidemic in different populations, and the susceptible population. The extraction of macro mechanism provides parameter support at the "group transmission and environmental conditions" level for the parameter card.

[0065] It is worth mentioning that the purpose of the micro mechanism is to directly convert "pathogen itself characteristics" into bottom-layer parameters that can drive interactive and AR demonstrations, so that every operation of the user can produce a visual response with a reasonable and evidence-based response. After the morphological characteristics, sensitivity to detergents, particle size and barrier penetration, survival / decay in different temperature and humidity conditions, and pathogenic process are unified and deposited as controllable quantities and boundary conditions in the parameter card, the system can map these quantities one by one to the source term, inactivation term, barrier penetration term, and receptor susceptible site of the mechanism equation. When the user selects or adjusts "cleaning / disinfection, wearing barriers, changing environmental conditions" and other actions on the computer interface, the AR layer will simultaneously demonstrate the "microscopic action" process of microorganisms being inactivated, blocked, diluted, or still penetrating, and the numerical evolution is constrained by the evidence range and monotonic direction given by the parameter card, thereby achieving an explainable explanation of "why this action can reduce the risk", rather than relying solely on static animations.

[0066] The macroscopic mechanism aims to aggregate the influence of "group transmission and environmental conditions" into scene-level control quantities, so that the system can give risk evolution results coupled with interactive actions and counterfactual comparisons on the spatial and crowd scales. After modeling the transmission path and intensity, the reaction intensity and susceptibility of different crowds as the ventilation / mixing efficiency, near-field exchange, contact network edge weight, and crowd susceptibility coefficient in the parameter card, the mechanism equation can be solved on the room, floor, or crowd contact graph to obtain the dynamic curves of "air concentration, surface load, exposure dose, and infection probability" in real time. When the user superimposes or switches actions, the system can immediately generate a counterfactual baseline for "if the action is not taken" and calculate the action contribution. The macroscopic mechanism also provides target quantities for the calibration of complex environments (such as local concentration peaks, ventilation short circuits, and partition differences), which facilitate the triggering of the mixing calculation module to generate a global flow field by template flow field + attention scaling superposition, and online correction of the model. In this way, the macroscopic layer not only supports situational and personalized science popularization interactions but also ensures that numerical changes at the microscopic level can reflect actual effects on the spatial and crowd levels, forming a "action-mechanism-visualization-counterfactual" closed loop.

[0067] The generation process of the parameter card is as follows:

[0068] Step one: learn in the first stage by searching the relevant literature; in the first stage of learning, only the micro-mechanism or any mechanism element and the corresponding multi-dimensional parameters in the macro-mechanism are searched, for example, in the literature describing the survival of microorganisms, there may be three dimensions of temperature, humidity, and survival time. Each search result is taken as a piece of evidence, and an anchor point is embedded in each evidence position; the similarity of the search is taken as the evidence weight.

[0069] According to the j-dimensional parameters of the mechanism elements in each evidence i, a parameter profile is generated; all evidence is integrated to obtain parameters containing J dimensions in all evidence; in the J-dimensional parameter space generated by all parameter profiles, each parameter profile is mapped and labeled with evidence weight to obtain the parameter card.

[0070] Where i represents the evidence index, j represents the parameter dimension reflected in the evidence, and J represents the dimension after taking the union of all dimensions of all evidence.

[0071] Step two: After the positioning of the anchor points for each piece of evidence in the first stage of learning, the second stage of learning is entered by searching the relevant literature: analyze the coupling relationship between anchor points to generate candidate edges about rules, including: co-occurrence rules, causal trigger rules, conditional rules, and mutually exclusive or synergistic rules; Co-occurrence rules: mechanism parameters that frequently co-occur in multiple documents, such as the co-occurrence between "ventilation rate improvement" and "aerosol concentration decrease". Causal trigger rules: the change of one mechanism parameter directly leads to the significant change of another parameter, for example, "temperature rise" leads to "pathogen survival time shortening". Conditional rules: a relationship only holds under certain conditions, for example, "the effect of humidity on transmission intensity is significant only in low temperature environments". Mutually exclusive or synergistic rules: there are mutually exclusive or synergistic effects between different parameters, for example, there is an additive effect rather than a mutually exclusive relationship.

[0072] Each rule generates a candidate edge, and different weights are assigned according to the confidence and frequency of the literature evidence. Finally, the system not only retains the parameter values corresponding to individual mechanism elements in the parameter card, but also supplements the coupling relationship between different parameters through these rules, making the parameter card a comprehensive structure containing both numerical ranges and relationship networks.

[0073] S2: Based on the search of the relevant literature and the parameter card, generate the epidemic disease transmission mechanism equation set.

[0074] The epidemic disease transmission mechanism equation set includes obtaining interactive actions inputted in the computer interface and encoding them as interactive action operators.

[0075] After locating the anchor points in the relevant literature, the third stage of learning is entered, in which similar actions to the interactive action d are searched in the relevant literature, and control processes or interactive actions similar in principle to the interactive action are searched through transfer learning as action nodes ; and according to the expression of each action node in the corresponding literature, encode it as the action operator of the action node , thereby generating the vector difference between the action operator of the action node and the interactive action operator encoded by the interactive action d as the offset feature of each action node .

[0076] where e represents the index of the action node, represents the e-th action node of the interactive action d. The construction of the epidemic disease transmission mechanism equation set is performed through a double-axis evolving knowledge graph network to generate the interactive relationship between the interactive action and each dimension parameter in each mechanism element; the double-axis evolving knowledge graph network consists of two channels of vertical flow and horizontal flow and a cross-axis fusion head, connected between the two channels through a modulation elimination layer.​

[0077] To be said, first, the user in the computer interface input of the interactive action is coded as an interactive action operator, the purpose is to convert the originally fuzzy interaction behavior (for example, "open window ventilation" or "wear a mask") into a numerical input that can be operated by the mechanism equation, so that the action can directly act on the parameters of the propagation equation. Secondly, locate the anchor point in the relevant literature and retrieve similar actions, the purpose is to find a control group for the current interaction. Through transfer learning, further expansion to the control process with similar action principles, can make the system not only rely on a single description, but also integrate evidence from different sources to form a set of "action nodes". The action node is coded as an action operator, and the difference between the action operator and the interactive action operator forms the offset feature. The purpose of this step is to quantify the difference between the "current user action" and the "action in the existing scientific evidence". This difference includes both direction (the direction of the action on the mechanism) and size (the influence of the action strength on the parameter), providing data support for subsequent relationship fitting. Finally, introduce a double-axis evolution knowledge graph network, the purpose of which is to systematically learn complex relationships. The horizontal flow focuses on learning the horizontal correlation and offset direction and size of different action nodes in the parameter dimension; the vertical flow focuses on learning the monotonic influence curve between action depth and parameter dimension. The purpose of the modulation elimination layer is to eliminate the mutual interference of horizontal and vertical features, ensuring that the learning results are more stable and more interpretable. The purpose of the cross-axis fusion head is to fuse the two results into a consistent adjustment amount, which is directly used for numerical updating of the epidemic transmission mechanism equation set.

[0078] Specifically, the horizontal flow input is a bipartite graph composed of action nodes and parameter dimension nodes (in the previous text, the parameter card is the analytical result of micro-mechanism and macro-mechanism, and each mechanism element is decomposed into multiple dimension parameters; for example: temperature survival rate, humidity survival rate, aerosol particle size distribution, population density, ventilation rate, etc. The parameter dimension node is to explicitly express these dimension parameters as nodes, each node representing a numerical dimension that can be independently adjusted and analyzed). First, do embedding preprocessing for both types of nodes, the method is to get stable initial embeddings by linear layer and layer normalization respectively for the action semantic vector obtained by text analysis and the multi-dimensional vector of the parameter card; then calculate the anisotropic attention weight on each edge from action to parameter dimension, so that information transmission is modulated by offset direction and size, edge type and evidence confidence. Two-layer graph convolution performs message aggregation and non-linear transformation in turn, and adds residual and layer normalization between layers to ensure numerical stability and robustness.

[0079] Attention weight calculation:

[0080] ;

[0081] denotes the weight of the e-th action node passing message to the k-th parameter dimension node; denotes the softmax operation normalizing over all incoming edges to the same receiving node k; denotes the learnable coefficient of the direction consistency term; denotes the inner product of the unit offset direction and the expected action direction; denotes the expected action direction unit vector encoded by the edge type and conditional label; denotes the learnable coefficient of the offset magnitude term; denotes a monotonic mapping function of the offset magnitude, e.g. logarithmic or linear compression; denotes the learnable coefficient of the evidence confidence term; denotes the evidence confidence weight of the edge; denotes the action node index; denotes the parameter dimension node index; denotes the unit direction of the offset vector; denotes the magnitude of the offset vector.

[0082] The aggregated output of the first layer graph convolution:

[0083] ;

[0084] denotes the lateral representation of the parameter dimension node k; denotes a nonlinear activation function, e.g. GELU or ReLU; denotes the self-loop update matrix (which is an extension of the adjacency matrix with self-loops, ensuring that the node preserves its own features while absorbing neighbor information during graph convolution propagation, for realizing stable evolution of parameter dimension and action nodes); denotes the initial embedding of node k; denotes the set of action nodes connected to node k; denotes the message transformation matrix; denotes the concatenation of the action node embedding and its offset vector; denotes the e-th action node embedding; denotes the offset vector of the e-th action node.

[0085] The second layer graph convolution reuses the same weight mechanism but runs on the output of the previous layer to enhance the nonlinear representation. The final output of the lateral flow is the direction and magnitude cues of each parameter dimension, indicating which direction the dimension should adjust and the approximate magnitude range. The whole process of the lateral flow is constrained by the numerical upper and lower bounds and monotonic direction in the parameter card, and clipping and sign preservation are implemented for out-of-bound components during training and inference to ensure physical feasibility and interpretability.

[0086] The implementation of longitudinal flow, the longitudinal flow input is the depth vector of the interactive action, that is, the intensity or compliance of the user at each control site. To ensure that the physical effect of grasping should be monotonic, for example, air exchange enhancement should not cause air concentration to rise, a multilayer perceptron with monotonic constraints is used; the constraints are achieved by imposing non-negativity on the weights and using non-negative activation. The longitudinal flow first normalizes and clips the input depth to the feasible region, then establishes a monotonic influence curve in each parameter dimension, and introduces evidence confidence and conditional labels in the aggregation stage for amplitude modulation and shielding. In the aggregation stage of longitudinal flow, a dynamic modulation mechanism based on evidence confidence and conditional labels is introduced to solve the problem of the applicability difference of different evidences in a specific interactive scenario. This mechanism includes two types of processing methods: "automatic weight reduction" and "automatic shielding". The two are in a sequential logical relationship, not a parallel scheme. When a certain evidence only partially matches the current interactive environment, the system performs an automatic weight reduction operation. For example, if a literature evidence is based on an experimental condition of 30% humidity, and the current interactive environment humidity is 50%, the system generates a correction factor less than 1 through the difference degree function, so that the amplitude contribution of this evidence in the aggregation of the longitudinal curve is reduced, but the reference effect is still retained. When a certain evidence is completely incompatible with the current interactive environment, or there is a logical conflict with other high-confidence evidence, the system performs an automatic shielding operation. Shielding is achieved by directly setting the evidence weight to zero, so that the evidence no longer has an impact on the output of the longitudinal curve. For example, if a certain evidence is only applicable to "bacterial transmission in anaerobic environment", and the current interactive scenario is aerobic environment, then the evidence is completely shielded. For example, if the conditional label shows that the evidence is only applicable to children, and the current interactive object is the elderly, then the evidence is also shielded.

[0087] Through this sequential processing mechanism, the longitudinal flow can detect whether there is a complete conflict or inapplicable situation in the aggregation process, and if so, perform shielding; if not, perform weight reduction according to the difference degree. This way not only ensures the scientific rigor and result stability of the longitudinal curve, but also improves the adaptability of the parameter to different interactive scenarios. The final output of the longitudinal regulation curve not only meets the monotonicity and physical feasibility constraints, but also has higher interpretability and robustness, thereby providing reliable input for the dynamic control of the mechanism equation.

[0088] Single-dimensional monotonic mapping:

[0089] ;

[0090] represents the adjustment amount component of the longitudinal flow to the parameter dimension k; represents a family of monotonic mapping functions for dimension k; represents the depth scalar of the interactive action at control site k; represents the number of hidden units; denotes the non-negative coefficient of the r-th hidden unit to the output to ensure overall monotonicity; denotes a non-negative activation function such as ReLU; denotes a non-negative weight to the input; denotes a bias.

[0091] Evidence modulation and conditional aggregation:

[0092] ;

[0093] denotes the normalized weight of the e-th action evidence to the dimension k; denotes the evidence confidence weight; denotes an indicator of whether the conditional label is satisfied, taking value zero or one; denotes a small positive number that is numerically stable; denotes the aggregated longitudinal adjustment. is not a new independent variable, but a summation index symbol, meaning: in the denominator, all action nodes (indexed by e') connected to the same parameter dimension k are traversed and their weights and conditional labels are all summed up.

[0094] The longitudinal flow output is a dimension-wise adjustment curve or adjustment amount at a given depth; the output is also constrained by the upper and lower bounds of the parameter card and the monotonic direction and will be projected back to the feasible region at the boundary.

[0095] The modulation elimination layer is located between the horizontal and longitudinal, and its responsibility is to eliminate the predictable components of the two pieces of information from each other before entering the output, avoiding the statistical confusion of structural differences and depth effects. The implementation adopts bidirectional residual orthogonalization supplemented by decorrelation loss.

[0096] Bidirectional residual orthogonalization:

[0097] ;

[0098] denotes the horizontal representation after mutual elimination; denotes the horizontal representation before mutual elimination; denotes the learnable linear mapping matrix from the longitudinal subspace to the horizontal subspace; denotes the longitudinal representation after mutual elimination; denotes the longitudinal representation before mutual elimination; denotes the learnable linear mapping matrix from the horizontal subspace to the longitudinal subspace.

[0099] Decorrelation loss:

[0100] ;

[0101] denotes the decorrelation penalty term; denotes the empirical covariance matrix between the transverse and longitudinal representations after mutual cancellation; denotes the Frobenius norm.

[0102] After processing by the layer, two mutually interfering residual representations are obtained for subsequent fusion.

[0103] Implementation of the cross-axis fusion head: the fusion head receives two mutually cancelled representations, first performs feature splicing and passes through two fully connected networks to regress the adjustment amount of each parameter dimension; then the output is subjected to physical feasible region projection and bidirectional smoothing, so that the result not only complies with the mechanism constraint but also has the stability of interactive presentation.

[0104] Fusion and projection:

[0105] ;

[0106] denotes the final adjustment amount of dimension k; Proj denotes the projection operation according to the upper and lower bounds and monotonicity requirements given by the parameter c, including clipping and sign constraints; denotes the fusion layer weight matrix; denotes the vector splicing of the two representations; denotes the fusion layer bias.

[0107] Longitudinal smoothing:

[0108] ;

[0109] denotes the adjustment amount after time or interaction step smoothing; denotes that the smoothing coefficient value is between zero and one; denotes the smoothed value of the previous time or the previous interaction step; denotes the current fusion output.

[0110] Transverse smoothing:

[0111] ;

[0112] denotes the final adjustment amount vector after consistency constraint on the parameter dimension subgraph; denotes the identity matrix consistent with the number of parameter dimensions; denotes the graph smoothing strength coefficient, which is non-negative and not greater than one; denotes the undirected Laplacian matrix of the parameter dimension subgraph; denotes the stacked vector after smoothing of all dimensions.

[0113] The final adjustment amount is written into the mechanism equation corresponding to the coefficient or boundary in the control site mapping table, which drives the real-time update of indicators such as infection probability, exposure dose, air concentration and surface load, and provides input for counterfactual baseline and action contribution calculation.

[0114] The advantage of this structure is that it can capture the "correspondence between action and parameter" and the "continuous influence of action intensity on parameter" at the same time, without the fragmentation caused by the simple mapping of action and parameter in traditional methods. The transverse flow through the two-part graph convolution can model the difference information between different action and parameter dimensions, reflect the comprehensive effect of offset direction, offset amplitude, edge relationship and evidence confidence, avoid the rigid problem of relying on experience rules to set weights, and make the results closer to the actual mechanism.

[0115] The longitudinal flow uses a monotonic constraint neural network to keep the depth change of interactive action consistent in numerical mapping at all times. In this way, when the user adjusts the action intensity on the interface, the infection probability and air concentration indicators output by the model will not appear "counterintuitive fluctuations", greatly enhancing the reliability and credibility of the system.

[0116] The introduction of the modulation elimination layer solves the coupling interference problem of transverse and longitudinal. In the conventional method, the structural difference and depth effect are often confused with each other, which can easily lead to repeated calculation or bias amplification. Through residual orthogonalization and decorrelation constraints, this layer makes the transverse responsible for difference learning and the longitudinal responsible for intensity learning, ensuring that the model has clear division of labor and non-redundant information expression, thereby improving the model convergence speed and explanation clarity.

[0117] The advantage of the cross-axis fusion head is that it can not only integrate the features of the transverse and longitudinal, but also ensure that the results are reasonable and visually friendly through physical feasible region projection and bidirectional smoothing mechanism. This prevents parameter out-of-bounds and numerical abnormalities, and makes the output curve smooth and coherent in AR display, improving the user's interactive experience.

[0118] In summary, the advantages of this structure compared to existing technologies are: it realizes high-precision matching of action and mechanism parameters, maintains the consistency of physical laws, eliminates interference between different dimensions of learning, and finally outputs an interpretable result that conforms to scientific mechanism and is suitable for interactive display. This makes epidemic prevention popular science not only stay at the demonstration level, but also provide users with intuitive and reliable feedback in a data-driven manner.

[0119] S3: input interactive instructions through the computer interface, the interactive instructions correspond to the selection or switching of protective actions, and are mapped to the control quantity of the epidemic mechanism equation.

[0120] The user's operation on the computer interface is first collected as a selection signal, which carries the control identifier, the original value, and the event type. The system de-bounces and shapes the selection signal to ensure that only valid operations are retained, and converts it to the corresponding action semantics, such as "mask selection", "window ventilation", or "surface cleaning", through a control-action mapping table. Subsequently, the original value is normalized or gear-reduced to unify different forms of input into intensity or level, such as normalizing the displacement of the slide bar to the range of 0-1, or reducing the mask model to different protection levels. The system further binds the action area or object for the action in combination with the scene context, and uses the parameter card for range checking and conflict resolution to ensure that the generated signal is numerically and logically consistent with the actual situation. Finally, the system configures the confidence for the operation, and encapsulates the action type, intensity, scope, and confidence into an interaction signal. In this way, the selection signal on the interface is converted into an interaction signal that can be directly called by the epidemic transmission mechanism equation, realizing the automatic mapping of user operation to mechanism parameter control.

[0121] S4: Inject the control quantity into the epidemic transmission mechanism equation set, and real-time calculate the infection probability, exposure dose, air concentration, and surface pathogen load, and generate the counterfactual comparison result and action contribution degree based on the interaction instruction.

[0122] Inject the control quantity into the epidemic transmission mechanism equation set to generate the corresponding numerical value of each dimension parameter on the parameter card; and according to the corresponding numerical value, map it to the infection probability, exposure dose, air concentration, and surface pathogen load through a preset mapping function.

[0123] The micro layer takes "pathogen itself characteristics" as the core, drives the processes of inactivation, particle size related penetration, and attenuation on different surfaces / environments, and is used to display the "close-range mechanism animation" of pathogenesis and protection in AR. This layer does not involve complex spatial environment, so it does not trigger the hybrid computing module. The macro layer takes spatial and crowd transmission as the core, continuously calculates the near-field and far-field air concentration, surface pathogen load caused by deposition, contact-resuspension path, and removal process generated by changes in ventilation / filtration. The control quantity acts on the corresponding coefficient or boundary condition in real time, forming an evolving trajectory that updates with user operation. At each time (or each interaction step), the mechanism state quantity is mapped to four types of user understandable indicators: air concentration, surface pathogen load, exposure dose, and infection probability. The air concentration directly comes from the near / far field state of the macro layer; the surface pathogen load comes from the dynamic balance of deposition and inactivation; the exposure dose comes from the time domain accumulation of air concentration and contact path in the scope where the user is located; and the infection probability comes from the risk value obtained by substituting the dose into the preset dose-response relationship. The system unifies the units, thresholds, and color scale encodings of these outputs to provide ready-to-use data for AR superposition and double-track curve display.

[0124] In the calculation process of the epidemic transmission mechanism equation group, when a complex environmental condition is detected, a hybrid calculation module is triggered, and the calculation result is used to calibrate the parameters of the epidemic transmission mechanism equation group. The complex environmental condition includes that, in the process of simulating the transmission of the epidemic (the AR process implemented by the present scheme includes a microscopic pathogenic process and an epidemic prevention process, which does not involve a complex environment; and a macroscopic transmission process and an epidemic prevention process, which may face transmission in a complex environment), there are multiple local different air flow fields.

[0125] The system continuously monitors whether to enter a complex environment: when the scene contains characteristics such as heterogeneous ventilation, local air flow channels, obstacle partitions, and multiple sources of superposition, or when the deviation between the macroscopic layer prediction and the observation / experience template exceeds the threshold, it is determined that there are multiple local different air flow fields. At this time, the hybrid calculation module is triggered; if it is in the pure microscopic explanation (close-range pathogenesis and protection) stage, it is not triggered to ensure performance and smooth interaction.

[0126] The hybrid calculation module includes scaling and superimposing according to a pre-constructed local flow field template to make the global flow field conform to the current environment, distributing global attention parameters according to the scaling and superimposing process of the local flow field template, and obtaining global attention weights through global normalization. The output result of injecting the control quantity into the epidemic transmission mechanism equation group is globally distributed according to attention. The local flow field template includes that the global is composed of multiple fixed unit minimum spaces, the local typical flow field space is obtained by simulating a typical space, and each space unit is equipped with an attention parameter which is proportional to the overall strength of the local flow field template.

[0127] Specific implementation process of the hybrid calculation module:

[0128] 1. Local template extraction and scaling. The system divides the global space into multiple fixed minimum space units (grid or room partitions), and pre-establishes a local flow field template library (including typical patterns such as air inlet / outlet, window door opening degree, fan jet, and thermal plume) for each type of typical space. During runtime, the template is scaled in size and intensity according to the current scene parameters and action intensity, so that the unit-level flow field matches the current environment.

[0129] 2. Template superposition to generate a global flow field. Align and superimpose the local flow fields of each unit in the global coordinates, and perform consistency correction and boundary continuity check during the superimposition process to ensure that the overall field domain has no obvious discontinuity.

[0130] 3. Global attention distribution. Each unit is assigned an attention parameter, which is related to the overall intensity of the unit template, the position relationship with the source, the ventilation path, the obstacle shielding, and the matching degree of the action scope; and then global normalization is performed to obtain the global attention weight of each unit at the current time.

[0131] 4. Global allocation of mechanism outputs by attention. The outputs of the system of equations or components of them that are affected by controlled quantities are spatially allocated and re-weighted according to global attention weights to obtain concentration and deposition distributions that better fit the real multi-flow field conditions.

[0132] To say, first of all, to ensure the accuracy under the condition of multi-flow field. The conventional macroscopic propagation model often assumes that the space air is uniformly mixed, but in reality there are often heterogeneous ventilation, obstacles, multiple sources superposition and other situations, resulting in significant local concentration difference. If a single global parameter is used to estimate, the output of air concentration, surface deposition and other results will deviate from the actual. And through the local template scaling and superposition mechanism, the real distribution of multi-flow field superposition can be flexibly restored, making the simulation results more consistent with the actual scene. Secondly, the adaptability and robustness of the model are improved. Through the attention parameter mechanism, the system can automatically identify which local space has a greater impact on the overall propagation result, such as air flow channel and main personnel gathering area, so as to dynamically allocate calculation weights. This global attention allocation not only avoids the waste of resources caused by uniform processing, but also ensures that the propagation and protection effect of key areas is highlighted. Thirdly, it provides the ability of parameter calibration and feedback. Finally, the coherence of AR interaction experience is optimized. The actions made by the user in the interface, such as opening the window, turning on the fan or moving the position, often change the local flow field. The mixed calculation module can immediately reflect the effect of these operations in the global flow field, and through attention weighting, the real propagation distribution is smoothly presented on the visualization, avoiding users seeing abrupt jumps in values or images, thereby enhancing the immersion and credibility of the interaction.

[0133] S5: Display the interactive action state and mechanism evolution curve in a double-track way in the computer interface, and provide immediate feedback based on counterfactual difference.

[0134] In this step, the system first calculates the actual state of the user's input interactive action, including infection probability, exposure dose, air concentration and surface pathogen load. At the same time, the system will automatically build a counterfactual baseline state corresponding to the interactive action. The generation method of counterfactual baseline is to set the action corresponding to the interactive instruction as "not executed" or "minimum intensity execution", while keeping the rest of the actions and environmental parameters unchanged, thus forming a control situation that only lacks the influence of the action.

[0135] After the baseline is completed, the system performs parallel calculations on the epidemic transmission mechanism equation set under the actual state and the counterfactual baseline state respectively, obtaining two sets of results. By comparing the differences between the two sets of results item by item, the system can obtain the difference values of four types of indicators: the reduction amplitude of infection probability, the reduction degree of exposure dose, the dilution effect of pathogen concentration in the air, and the attenuation level of surface pathogen load. The above difference values are the counterfactual comparison results, which can directly reveal the protection effect brought by the interactive action under the current conditions.

[0136] In order to further quantify the importance of the interactive action, the system combines the counterfactual comparison results with the multi-dimensional parameters in the parameter card and performs weighted analysis using the evidence weight therein. The evidence weight is derived from the credibility and applicability evaluation of each parameter in the literature analysis stage, so it can ensure the scientificity of the contribution calculation. In this way, the system calculates the action contribution degree of each interactive action in the overall protection process, which is expressed as a numerical value or a percentage, reflecting the relative value of the action in reducing the transmission risk.

[0137] In the computer interface, the system uses a double-track method for visualization display. The upper track is used to display the actual state and change process of the interactive action, clearly marking which actions the user has performed and the intensity or duration of the action. The lower track dynamically displays the mechanism evolution process in the form of a curve, and simultaneously plots the results of the actual state and the counterfactual baseline in the curve, highlighting the immediate protection effect of the action through shadow area or difference value marking. The user can intuitively perceive how the transmission risk will change if the action is not performed.

[0138] Through this "actual state - counterfactual baseline - difference feedback" mechanism, the invention not only realizes the quantitative display of the effect of the interactive action, but also ensures the immediacy and scientificity of the feedback. The user can clearly see the value of their own operation in the interaction, thereby enhancing the intuitiveness and persuasiveness of the health and epidemic prevention popularization.

[0139] The embodiment also provides an AR interactive system for health and epidemic prevention popularization, which comprises:

[0140] A collection unit analyzes and extracts parameters from relevant literature on epidemic prevention to form a parameter card.

[0141] An analysis unit generates the epidemic transmission mechanism equation set based on the retrieval of the relevant literature and the parameter card.

[0142] A control unit inputs interactive instructions through a computer interface, wherein the interactive instructions correspond to the selection or switching of protection actions and are mapped to the control quantity of the epidemic mechanism equation.

[0143] The feedback unit injects the control quantity into the epidemic transmission mechanism equation set, calculates the infection probability, exposure dose, air concentration and surface pathogen load in real time, and generates a counterfactual comparison result and action contribution degree based on the interaction instruction; and displays the interaction action state and mechanism evolution curve in a double-track manner in the computer interface, and provides instant feedback based on the counterfactual difference.

[0144] The embodiment also provides a computer device suitable for the AR interaction method for health and epidemic prevention popularization, which comprises a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the AR interaction method for health and epidemic prevention popularization proposed in the above embodiment.

[0145] The computer device can be a terminal, which comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0146] The embodiment also provides a storage medium having a computer program stored thereon, which is executed by a processor to realize the AR interaction method for health and epidemic prevention popularization proposed in the above embodiment. The storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0147] In summary, the present application converts unstructured knowledge into computable data base by analyzing micro and macro mechanisms from epidemic prevention documents, constructing parameter cards containing numerical ranges, evidence levels and coupling relationships. Based on parameter cards and action evidence anchors, the mechanism equation set of epidemic transmission is established, and the relationship between action and multi-dimensional parameters is learned through "interactive action operator-action node offset characteristics-biaxial evolution knowledge graph network", ensuring the explainable mapping from action to mechanism. The selected signals are standardized as interactive signals in the computer interface, which are directly injected into the mechanism equation as control variables to realize real-time adjustment of control sites such as source, ventilation, sedimentation and contact. The mechanism output is standardized as four types of indexes: air concentration, surface pathogen load, exposure dose and infection probability, and the "actual state- counterfactual baseline-difference" mechanism is used to generate instant feedback and action contribution, which intuitively reveals the protection value of single or combined action. When detecting complex environmental conditions such as heterogeneous ventilation, airflow channel and partition barrier, the mixed calculation module is triggered to generate global flow field by scaling and superimposing local flow field templates and global attention allocation, ensuring the accuracy and stability in complex scenes. The "action state trajectory" and "mechanism evolution curve" are displayed synchronously in double-track visualization, and the counterfactual difference is highlighted by color scale and labeled, forming a public-oriented, interactive, quantifiable and traceable epidemic prevention science popularization experience.

[0148] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, which should be covered by the scope of the claims of the present application.

Claims

1. An AR interactive method for popularizing health and epidemic prevention science, characterized in that: This includes analyzing and extracting parameters from relevant literature on epidemic prevention to create parameter cards; Based on the retrieval of the relevant literature and the parameter card, a set of equations for the disease transmission mechanism is generated; Interactive commands are input through a computer interface. These commands correspond to the selection or switching of protective actions and are mapped to control quantities in the disease mechanism equation. The control quantity is injected into the disease transmission mechanism equation set to calculate the infection probability, exposure dose, air concentration and surface pathogen load in real time, and generate counterfactual comparison results and action contribution based on the interactive instructions; The interactive action status and mechanism evolution curve are displayed in a dual-track manner on the computer interface, and real-time feedback is provided based on counterfactual difference. During the calculation of the disease transmission mechanism equation set, when complex environmental conditions are detected, the hybrid calculation module is triggered, and the calculation results are used to calibrate the parameters of the disease transmission mechanism equation set. The parameter card generation process is as follows: Step 1: Conduct the first stage of learning by searching the relevant literature; in the first stage of learning, only search for any mechanistic element and the corresponding multidimensional parameters in the micro or macro mechanism, and each search result is used as evidence, and anchor points are embedded at each evidence position; The similarity of the retrieved items is used as the weight of the evidence. Based on the j-dimensional parameters of the mechanistic elements in each piece of evidence i, a parameter file is generated; all evidence is integrated to obtain parameters containing J dimensions of all evidence; in the J-dimensional parameter space generated from all parameter files, each parameter file is mapped and the evidence weight is marked to obtain the parameter card; Where i represents the evidence index, j represents the parameter dimension reflected in the evidence, and J represents the dimension after taking the union of all dimensions of all evidence; Step 2: After retrieving the relevant literature and locating the anchor points of each piece of evidence during the first stage of learning, proceed to the second stage of learning: analyze the coupling relationship between the anchor points and generate candidate edges for the rules, including: co-occurrence rules, causal triggering rules, conditionalization rules, and mutually exclusive or cooperative rules. At the same time, the confidence level of each type of edge is used as the corresponding weight; The disease transmission mechanism equation set includes acquiring interactive actions input on the computer interface and encoding them as interactive action operators; In the relevant literature, after locating the anchor point, the third stage of learning is performed. This involves retrieving actions similar to the interactive action d from the relevant literature, and then, through transfer learning, retrieving control processes or interactive actions with similar working principles as action nodes d. e Furthermore, based on the representation of each action node in the corresponding literature, it is encoded as action node d. e The action operator is used to generate the action node d. e The vector difference between the action operator and the interaction action operator encoded by the interaction action d is used as the vector difference for each action node d. e offset feature p e ; Where e represents the index of the action node, d e This represents the e-th action node of the interactive action d; The disease transmission mechanism equation set is constructed using a dual-axis evolutionary knowledge graph network, generating interactive actions and the interaction relationship between parameters of each dimension in each mechanism element; the dual-axis evolutionary knowledge graph network consists of two channels, a vertical flow and a horizontal flow, and a cross-axis fusion head, which are connected between the two channels by a modulation cancellation layer; The lateral flow consists of two layers of graph convolutional networks with attention mechanisms. It takes a bipartite graph of action nodes and multidimensional parameter nodes of parameter cards as input. During information transmission, attention modulation is performed through the direction and magnitude of the offset vector, edge type, and evidence confidence. It is used to learn the structural differences between interactive actions and multidimensional parameters of parameter cards and output lateral adjustment cues dimension by dimension. The longitudinal flow is composed of a fully connected neural network with monotonic constraints. Taking the action depth as input, it can generate a monotonic influence curve between the action depth and the multidimensional parameters of the parameter card. During aggregation, evidence confidence and condition labels are introduced to automatically reduce the weight or block evidence that does not meet the applicable conditions: if the evidence only partially matches the conditions, the automatic weight reduction operation is performed; if the evidence does not match the conditions at all, the automatic blocking operation is performed. The modulation cancellation layer is placed between the transverse and longitudinal flows. It achieves mutual cancellation of transverse and longitudinal features through residual orthogonalization and decorrelation constraints, thereby avoiding coupling interference between transverse and longitudinal channels in information learning. The cross-axis fusion head consists of two fully connected networks and features a physical feasible domain projection and a bidirectional smoothing mechanism in the output stage. The control quantity is injected into the disease transmission mechanism equation set to generate the corresponding value of each dimension parameter on the parameter card; and according to the corresponding value, it is mapped to: infection probability, exposure dose, air concentration and surface pathogen load through a preset mapping function. Simultaneously, the actual state of the interaction action and the corresponding counterfactual baseline state are calculated; wherein, the counterfactual baseline state is generated by setting the interaction action corresponding to the interaction instruction to a virtual state of not being executed or having the minimum execution strength, while keeping other actions unchanged; By comparing the calculation results of the disease transmission mechanism equations under actual conditions and counterfactual baseline conditions, the differences in infection probability, exposure dose, air concentration, and surface pathogen load are obtained as counterfactual comparison results. Based on the counterfactual comparison results, and combined with the evidence weights of each dimension parameter in the parameter card, the action contribution of each interactive action is calculated.

2. The AR interactive method for popularizing health and epidemic prevention science as described in claim 1, characterized in that: The relevant literature includes articles on the target disease and a set of reliable sources. By analyzing the content of the relevant literature, the microscopic and macroscopic mechanisms of disease action can be obtained. Based on the microscopic and macroscopic mechanisms, directly corresponding parameter cards are generated; The coupling relationships between microscopic mechanisms, macroscopic mechanisms, and microscopic and macroscopic mechanisms are used to supplement the coupling relationships of the parameter card.

3. The AR interactive method for popularizing health and epidemic prevention science as described in claim 2, characterized in that: The microscopic mechanism refers to the individual attributes of pathogenic microorganisms, including the following elements: the morphological characteristics of the pathogenic microorganisms, their sensitivity to different daily detergents, the penetration of different isolation barriers by their size, their survival and attenuation processes in different environments, and their pathogenic processes. The macroscopic mechanism includes the following elements: the spatial transmission process, the intensity of the disease response in different populations, and susceptible populations.

4. The AR interactive method for popularizing health and epidemic prevention science as described in claim 3, characterized in that: The complex environmental conditions include the existence of multiple locally different airflow fields during the simulated disease transmission process; The hybrid computing module includes scaling and superimposing a pre-constructed local flow field template to make the global flow field conform to the current environment, allocating global attention parameters according to the scaling and superimposition process of the local flow field template, and obtaining global attention weights through global normalization. The output of the control variable injected into the disease transmission mechanism equation set is globally distributed according to attention. The local flow field template includes a global flow field consisting of multiple fixed unit minimum spaces. By simulating typical spaces, a typical local flow field space is obtained. Each spatial unit is equipped with an attention parameter that is proportional to the overall intensity of the local flow field template.

5. An AR interactive system for popularizing public health and epidemic prevention knowledge, based on the AR interactive method for popularizing public health and epidemic prevention knowledge according to any one of claims 1 to 4, characterized in that: This includes a data collection unit that analyzes and extracts parameters from relevant literature on epidemic prevention to create a parameter card; The analysis unit generates the disease transmission mechanism equation set based on the retrieval of relevant literature and the parameter card; The control unit receives interactive commands via a computer interface. These commands correspond to the selection or switching of protective actions and are mapped to control quantities in the disease mechanism equation. The feedback unit injects the control quantity into the disease transmission mechanism equation set, calculates the infection probability, exposure dose, air concentration and surface pathogen load in real time, and generates counterfactual comparison results and action contribution based on the interactive instructions; the interactive action status and mechanism evolution curve are displayed in a dual-track manner on the computer interface, and provide immediate feedback based on the counterfactual difference.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the AR interactive method for popularizing health and epidemic prevention science as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the AR interactive method for popularizing health and epidemic prevention science as described in any one of claims 1 to 4.

Images