XR large-space multi-person interaction safety supervision system based on real-time data processing

Abstract

The invention relates to the technical field of XR large-space multi-person interaction, in particular to an XR large-space multi-person interaction safety supervision system based on real-time data processing, which comprises a space positioning module, a behavior recognition module, a dynamic optimization module, an early warning module and a data management module. An optimal path matrix is generated through a multi-dimensional space segmentation algorithm and a path planning algorithm, and a user behavior sequence and priority setting are dynamically adjusted in combination with an intelligent optimization algorithm, so that efficient real-time supervision is realized. Meanwhile, a prediction model is constructed for behavior state prediction, an early warning signal is triggered, and corresponding measures are taken. The arrangement rationality and the operation efficiency of the interaction area can be improved, balanced distribution of user loads is ensured, resource waste is reduced, and the accuracy and the real-time performance of safety supervision are enhanced.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of virtual reality and augmented reality, specifically to an XR large space multi-person interaction safety supervision system based on real-time data processing. BACKGROUND

[0002] With the rapid development of XR (Extended Reality) technology, its application in large space multi-person interaction scenarios is increasingly widespread, such as virtual meetings, immersive training, and collaborative design. However, existing safety supervision systems have many shortcomings in dealing with XR large space multi-person interaction scenarios, making it difficult to meet the real-time, precision, and efficiency requirements in complex environments.

[0003] After searching, a project safety quality supervision system with publication number CN114091966B was disclosed on January 10, 2025. This patent realizes the safety quality supervision of engineering projects through receiving modules, creation modules, inspection modules, rectification modules, verification modules, and determination modules. However, this technical solution is mainly aimed at the field of engineering management and lacks the ability to process dynamic behavior and real-time data in XR large space multi-person interaction scenarios. Although the modular design can meet the needs of static inspection and rectification processes, it cannot achieve efficient real-time monitoring and dynamic adjustment in the face of rapidly changing XR interaction environments, which may lead to lagging supervision or misjudgment.

[0004] After searching, a construction safety supervision system with publication number CN111986461B was disclosed on May 10, 2022. This patent realizes the safety supervision of construction sites through data acquisition modules, data transmission modules, control modules, and execution modules. Although this technical solution has certain real-time data acquisition and early warning capabilities, its application scenarios are limited to construction sites and mainly rely on data input from fixed monitoring points. For XR large space multi-person interaction scenarios, this system lacks the ability to capture and analyze the dynamic behavior of multiple users, and fails to consider the spatial positioning and interaction characteristics of XR devices, making it difficult to adapt to complex three-dimensional interaction environments and affecting the comprehensiveness and accuracy of supervision.

[0005] The above problems show that existing safety supervision systems still have obvious shortcomings in real-time, dynamic behavior capture, multi-user collaborative supervision, and adaptability to complex environments when dealing with XR large space multi-person interaction scenarios. Therefore, the present application provides an XR large space multi-person interaction safety supervision system based on real-time data processing, aiming to realize comprehensive, accurate, and efficient real-time supervision of XR interaction environments by introducing advanced spatial positioning, behavior recognition, and dynamic data analysis technologies, thereby meeting the safety and reliability requirements of modern XR applications. SUMMARY

[0006] The application provides an XR large space multi-person interaction safety supervision system based on real-time data processing, which solves the problem that the existing safety supervision system cannot meet the real-time, accuracy and efficiency requirements in complex environments when dealing with XR large space multi-person interaction scenarios. In the prior art, the processing capability for dynamic behavior and real-time data is insufficient, especially in rapidly changing XR interaction environments, traditional systems cannot achieve efficient real-time monitoring and dynamic adjustment, which may lead to lagging supervision or misjudgment. In addition, the existing system has limited ability to capture multi-user dynamic behavior and analyze complex interaction characteristics, and does not fully consider the spatial positioning characteristics of XR devices, making it difficult to fully adapt to the needs of three-dimensional interaction environments.

[0007] To solve the above technical problems, the application provides the following technical solutions:

[0008] In a first aspect, the application provides an XR large space multi-person interaction safety supervision system based on real-time data processing, which includes,

[0009] A spatial positioning module collects user position information, divides the interaction area through a multi-dimensional space segmentation algorithm, and generates an optimal path matrix in combination with a path planning algorithm; the user position information includes real-time coordinates, attitude data and device status of the users participating in the interaction;

[0010] A behavior recognition module extracts priority scores based on user action features, generates user behavior sequences in combination with the optimal path matrix and performs iterative optimization to obtain an initial solution; the initial solution includes user behavior sequences and their corresponding priority setting combinations;

[0011] A dynamic optimization module generates core and auxiliary populations based on the initial solution for initialization, and performs optimization through a disturbance mechanism, a selection probability model and an information diffusion strategy in combination with a difference evaluation and a cross replacement method to obtain a final solution; the final solution includes optimized user behavior sequences and their corresponding priority setting combinations;

[0012] An early warning module constructs a prediction model to predict user behavior states, triggers final solution recalculation and issues a warning signal according to the prediction results;

[0013] A data management module displays data through a visual interface, stores data using a database, and encrypts the data using an encryption algorithm.

[0014] As a preferred scheme of the XR large space multi-person interaction safety supervision system based on real-time data processing, the collection of user position information, the division of the interaction area through a multi-dimensional space segmentation algorithm, and the generation of an optimal path matrix in combination with a path planning algorithm include the following steps:

[0015] Real-time coordinates, posture data and device status of the user participating in the interaction are collected in real time using high-precision sensors; the real-time coordinate data includes the X, Y and Z axis positions of the user in three-dimensional space; the posture data includes the head orientation angle and body tilt angle; and the device status includes the battery level and connection stability.

[0016] The interaction area is divided using a grid division method, and the weight of each data in each partition is set using a historical data analysis method, then each data is multiplied by the corresponding weight using multiplication operation, and the sum is calculated to obtain the comprehensive interaction demand intensity of each partition;

[0017] The comprehensive interaction demand intensity is sorted in ascending order, and the selection frequency is set using an empirical rule, the largest comprehensive interaction demand intensity is selected step by step according to the selection frequency, and the partition corresponding to the comprehensive interaction demand intensity is defined as the candidate user position, and the reference point is randomly selected, the distance between the reference point and all candidate user positions is calculated using the Euclidean distance formula;

[0018] Each candidate user position is segmented using a multi-dimensional space segmentation algorithm combined with distance to obtain the interaction area of each user, and if the current area and the adjacent area have a shared boundary, the area is marked as a reachable area, otherwise it is marked as an unreachable area;

[0019] The distance between the user pairs in the reachable area is calculated using the Euclidean distance formula, and the distance matrix is constructed combined with the unreachable area, and the directed graph is constructed using the Graph tool combined with the distance matrix;

[0020] The out-edge and in-edge of each user are extracted from the directed graph using the path planning algorithm to obtain the out-edge set and in-edge set of each user;

[0021] According to the out-edge set and in-edge set, the product of the number of out-edges and in-edges in the set is calculated;

[0022] After arranging the product in ascending order, the user corresponding to the product is selected as the intermediate node according to the arrangement result for iteration, and in the iteration process, the distance between the user pairs is defined as the weight, and when the user pairs can reach through the intermediate node, the path is updated through relaxation operation;

[0023] When the distance between the user pairs (i, j) in the updated path is updated from unreachable to a finite value, it means that the path of the user pair (i, j) is updated, the ith user is added to the in-edge set of the intermediate node, and the jth user is added to the out-edge set of the intermediate node;

[0024] Continue to iterate and update, and output the optimal path matrix after the number of iterations reaches the maximum number; the optimal path matrix refers to the shortest physical distance between each pair of users in the interaction area.

[0025] As a preferred scheme of the XR large space multi-person interaction safety supervision system based on real-time data processing of the application, wherein: the priority score based on user action feature extraction is combined with the optimal path matrix to generate a user behavior sequence and iteratively optimized to obtain an initial solution, including the following steps:

[0026] The weight coefficient is set using the historical data analysis method, and the priority score of each user is calculated according to the user energy consumption and the minimum distance value of the user in the optimal path matrix. According to the priority score, all users are arranged in descending order of priority score to obtain a user arrangement result, which is defined as a behavior sequence.

[0027] The adjustment coefficient value range is set using the field method, wherein represents the minimum adjustment coefficient, and represents the maximum adjustment coefficient.

[0028] An adjustment coefficient in the value range is generated using a random number generator to calculate the priority weight of each user.

[0029] The behavior sequence and priority weight are encoded and then combined using encoding technology to generate a scheduling individual. A population is randomly generated for initialization. A target function is defined, and the target function value is calculated. The target function value is used as the individual fitness value, which is arranged in ascending order. The maximum fitness value is selected for crossover and mutation operations to generate a candidate solution for iteration. When the number of iterations reaches the maximum number, the iteration is stopped, and the initial solution is output. The initial solution refers to the combination of the respective behavior sequences and priority settings of the users participating in the interaction.

[0030] As a preferred scheme of the XR large space multi-person interaction safety supervision system based on real-time data processing of the application, wherein: the initial solution is used to generate a core and auxiliary population for initialization, and through a disturbance mechanism, a selection probability model and an information diffusion strategy, combined with a difference degree evaluation and a cross-replacement method, a final solution is obtained, including the following steps:

[0031] The initial solution is used as a core individual, and Gaussian disturbance and mutation operations are used to disturb and mutate the core individual to generate a core population for initialization. θ core individuals are randomly selected from the core population and defined as auxiliary individuals. An auxiliary population is generated for initialization by combining random generation operations. The reciprocal of the initial solution target function value is defined as heuristic information, and an information diffusion factor is set using a rule.

[0032] The weights of the heuristic information and the information diffusion factor are set by using a cross-validation method, all adjacent users of the current user are counted by using a statistical method, and an adjacent user set is constructed;

[0033] The selection probability is calculated by using a selection probability model formula;

[0034] The selection probability is selected by using a roulette, and the behavior sequence corresponding to the selected selection probability is taken as the current behavior sequence;

[0035] The weight coefficient of the user behavior is set by using an analytic hierarchy process, and a target function is defined;

[0036] The target function value is calculated, and the target function value is taken as the fitness value of the core or auxiliary individual;

[0037] Iterative updating is performed, and in the iterative process, the start state distance between the core individual and the adjacent auxiliary individual is calculated by using a difference degree evaluation method, and is defined as the start state difference degree;

[0038] The end state distance between the core individual and the adjacent auxiliary individual is further calculated by using the difference degree evaluation method, and is defined as the end state difference degree, the sum of the start state difference degree and the end state difference degree is calculated by using an addition operation formula, and is defined as the comprehensive difference, a constant is set by using an empirical rule, and the information diffusion factor is updated in combination with the comprehensive difference;

[0039] The exchange iteration number and the selection number are further set by using a field method, when the iteration number reaches the exchange iteration number, the core and auxiliary populations are arranged in ascending order according to the fitness value, and the first auxiliary individual in the auxiliary population arrangement result and the first core individual in the core population arrangement result are selected from the auxiliary population arrangement result according to the selection number to perform cross replacement, after the cross replacement, the attenuation factor is set by using a grid search method, and state path fusion is performed, the fused state path is taken as a new individual, and global information diffusion factor updating is performed;

[0040] Iterative updating is continued, and when the iteration number reaches the maximum number, the iteration is stopped, and the final solution is output; the final solution includes the respective behavior sequences of the participating interactive users and the priority setting combinations thereof at all time slices.

[0041] As a preferred scheme of the XR large space multi-person interaction safety supervision system based on real-time data processing, wherein: the step of constructing a prediction model to predict the user behavior state, and triggering the final solution recalculation and issuing a warning signal according to the prediction result comprises the following steps:

[0042] The historical running data of the participating interactive users are collected; the historical running data includes action frequency, movement trajectory and device state;

[0043] A prediction model is built using a deep learning architecture, and the mean square error is defined as the objective function, the model parameters are optimized using grid search and cross-validation, and the optimized parameters and historical running data are used as input, and the behavior state prediction value of each user is output;

[0044] The warning threshold is set using rules, and the behavior state prediction value is compared with the warning threshold, and when the current state is greater than or equal to the warning threshold, the user is excluded, and the final solution is recalculated and a warning signal is triggered, otherwise the monitoring continues.

[0045] As a preferred scheme of the XR large space multi-person interaction safety supervision system based on real-time data processing, when the warning signal is triggered and the monitoring continues, the comparison result report and optimization suggestion are generated and sent to the administrator through the transmission protocol, and the corresponding measures are implemented by the administrator.

[0046] As a preferred scheme of the XR large space multi-person interaction safety supervision system based on real-time data processing, after collecting the user position, posture and device data, preprocessing operations are performed, including denoising, interpolation and normalization.

[0047] As a preferred scheme of the XR large space multi-person interaction safety supervision system based on real-time data processing, the data is displayed through a visual interface, which means that the comparison result is converted into a chart using a tool, and the comparison result, optimization suggestion and final solution are visually displayed through the chart.

[0048] As a preferred scheme of the XR large space multi-person interaction safety supervision system based on real-time data processing, the data is stored using a database, which means that the comparison result is stored through the database, and the user ID and timestamp are added to the comparison result through the database after storage, and the comparison result is classified.

[0049] As a preferred scheme of the XR large space multi-person interaction safety supervision system based on real-time data processing, the stored data is encrypted using a symmetric encryption algorithm, and the key is changed regularly.

[0050] The XR large space multi-person interaction safety supervision system has the advantages that the multi-dimensional space segmentation algorithm, the path planning algorithm and the calculation of the shortest physical distance between each pair of users are introduced, so that the XR large space multi-person interaction safety supervision system can accurately adjust the user positions, the arrangement of the interaction area is more reasonable, the service range is maximized, and the overall interaction quality and operation efficiency are improved. BRIEF DESCRIPTION OF DRAWINGS

[0051] For the convenience of those skilled in the art, the present application will be further described below in conjunction with the drawings.

[0052] Figure 1 Fig. 1 is a schematic diagram of the overall module structure of the XR large space multi-person interaction safety supervision system, which shows the main function modules of the system and their mutual relationship.

[0053] Figure 2 Fig. 4 is a workflow diagram of the space positioning module, which describes in detail the process of user position information collection, partition division and optimal path matrix generation.

[0054] Figure 3 Fig. 5 is an optimization flowchart of the behavior recognition module, which shows the steps of extracting priority scores based on user action features and generating initial solutions.

[0055] Figure 4 Fig. 6 is an optimization flowchart of the core population and auxiliary population of the dynamic optimization module, which illustrates the application of the disturbance mechanism, the selection probability model and the information diffusion strategy.

[0056] Figure 5 Fig. 7 is a prediction and early warning triggering flowchart of the early warning module, which includes the implementation process of historical data collection, prediction model construction and early warning signal triggering. DETAILED DESCRIPTION

[0057] The technical solutions of the present application will be described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0058] The present application provides an XR large space multi-person interaction safety supervision system based on real-time data processing, and the overall module structure is as shown in Fig. 1. Figure 1As shown, it includes a spatial positioning module, a behavior recognition module, a dynamic optimization module, a warning module, and a data management module. These modules work together through specific connection relationships and collaboration mechanisms to achieve the functional goals of the system. The specific embodiments of each module will be described in detail below in conjunction with the accompanying drawings.

[0059] The spatial positioning module is the foundation of the entire system, and its workflow is as shown in Figure 2 First, this module collects user position information through high-precision sensors, which can be laser radars, infrared cameras, or ultra-wideband positioning devices. The sensors are arranged around the interaction area to ensure coverage of the entire three-dimensional space. The collected data includes the user's real-time coordinates (X, Y, Z axis positions), attitude data (head orientation angle, body tilt angle), and device status (battery level, connection stability). These data are transmitted to the central processing unit for further processing through wired or wireless means. Next, the interaction area is divided into multiple sub-zones using the grid division method, and each sub-zone is set with a weight based on historical data analysis. The weight is associated with the user's position, action frequency, and device status, and is used to calculate the comprehensive interaction demand intensity of each sub-zone. After the comprehensive interaction demand intensity is calculated through multiplication and weighted summation, it is sorted in ascending order and the sub-zones corresponding to the maximum values are selected as the candidate user positions. Then, a reference point is randomly selected and the distances from the reference point to all candidate user positions are calculated using the Euclidean distance formula. Based on this, the multi-dimensional space partitioning algorithm is used to partition the sub-zones, generating reachable and unreachable regions. Finally, by constructing a distance matrix and a directed graph, combined with a path planning algorithm, the optimal path matrix is calculated. Each step in this process depends on the results of the previous step, for example, the setting of the sub-zone weight directly affects the subsequent calculation of the comprehensive interaction demand intensity, and the comprehensive interaction demand intensity determines the selection of the candidate user position.

[0060] The workflow of the behavior recognition module is as shown in Figure 3As shown in FIG. 6, the core of the module is to extract priority scores based on user action features and generate an initial solution. After the module obtains user location information and the optimal path matrix from the spatial positioning module, it first sets the weight coefficient according to the historical data analysis method and calculates the priority score of each user by combining the user energy consumption and the minimum distance value. The calculation result of the priority score is used to generate the user arrangement result, i.e., the behavior sequence. The behavior sequence is combined with the priority weight through encoding technology to generate the scheduling individual, which enters the population generation stage after initialization. In this process, the module uses a random number generator to generate an adjustment coefficient and sets the value range of the adjustment coefficient according to the field method. The generated population calculates the fitness value through the objective function, and the individual with the highest fitness value is selected for crossover and mutation operations, thereby generating the candidate solution. After multiple iterations of optimization, the initial solution is output, which contains the respective behavior sequences of the participating interactive users and their priority setting combinations. The key of this module lies in the design of the objective function and the calculation of the fitness value. The objective function needs to consider the comprehensive influence of user energy consumption, distance relationship, and priority weight to ensure that the generated behavior sequence and priority setting can meet the actual demand.

[0061] The workflow of the dynamic optimization module is shown in FIG. 7. Figure 4 As shown in FIG. 7, the main function of the module is to further optimize the initial solution and generate the final solution. The module first takes the initial solution output by the behavior recognition module as the core individual and generates the core population through Gaussian disturbance and mutation operations. The number of individuals in the core population is set by experience value, usually twice the number of initial solutions. Then, a number of core individuals are randomly selected from the core population as auxiliary individuals, and an auxiliary population is generated through random generation operations. After the initialization of the core population and the auxiliary population, the module optimizes the population through heuristic information and information diffusion factor. The heuristic information is defined by the reciprocal of the objective function value of the initial solution, and the information diffusion factor is determined by rule setting and statistical method. In the optimization process, the module calculates the selection probability using the selection probability model formula and selects the current behavior sequence through the roulette method. The analytic hierarchy process is used to set the weight coefficient of user behavior, and the calculation result of the objective function value is used as the individual fitness value. The module calculates the start state difference and end state difference between the core individual and the adjacent auxiliary individual through the difference evaluation method, and updates the information diffusion factor according to the empirical rule. When the number of iterations reaches the exchange iteration number, the module arranges the core population and the auxiliary population in ascending order of fitness value, and performs cross replacement according to the selected number. The state path after cross replacement is fused by setting the attenuation factor through grid search method, and the fused state path is taken as the new individual to enter the next iteration. Finally, the module outputs the final solution containing the respective behavior sequences of the participating interactive users and their priority setting combinations at all time slices.

[0062] The workflow of the early warning module is shown in FIG. 8. Figure 5As shown, its main task is to build a prediction model and predict the user behavior state. The module first obtains the user's historical running data, including action frequency, movement trajectory, and device status, from the spatial positioning module and behavior recognition module. These data are input into the deep learning architecture after preprocessing, and the prediction model is built. The objective function of the prediction model is defined as the mean square error, and the model parameters are optimized through grid search and cross-validation. The optimized model takes the historical running data as input and outputs the behavior state prediction value of each user. The module sets the warning threshold through rules and compares the behavior state prediction value with the warning threshold. When the prediction value is greater than or equal to the warning threshold, module 4 triggers the recalculation of the final solution and sends a warning signal. Otherwise, the module continues to monitor the user behavior state. The triggering mechanism of the warning signal depends on the accuracy of the prediction model and the reasonableness of the warning threshold, so the model needs to be calibrated and optimized regularly.

[0063] The data management module is responsible for storing, displaying, and encrypting all data generated by the system. The module first converts the comparison results into charts through visualization tools and visualizes the comparison results, optimization suggestions, and final solutions through charts. These charts can be line charts, bar charts, or heat maps, which are used to visually present the trend of user behavior state changes and optimization effects. At the same time, the module stores the comparison results through the database and adds user ID and timestamp to each record for classified query. To ensure the security of the data, the module uses a symmetric encryption algorithm to encrypt the stored data and changes the key regularly. The choice of encryption algorithm needs to balance security and computational efficiency to ensure that the encryption process does not significantly affect system performance.

[0064] The above five modules are connected through data flow and control flow to form a complete system. The spatial positioning module provides basic data for the behavior recognition module, the output of the behavior recognition module is used as the input of the dynamic optimization module, the final solution of the dynamic optimization module provides the basis for the decision of the warning module, and the data management module uniformly manages and protects the data of the entire system. This modular design not only improves the scalability and maintainability of the system, but also ensures efficient collaboration between modules.

[0065] In order to better enable relevant persons in the art to fully understand and implement the present invention, the following supplementary explanation of the specific implementation principles of the present invention is made in conjunction with a specific application scenario.

[0066] In a certain virtual meeting scenario, multiple users wear XR devices to enter a large interactive space area, and the system monitors and optimizes the user's behavior and position in real time through the safety supervision system of the present invention. The entire process is divided into the following steps:

[0067] Firstly, in the running phase of the spatial positioning module, sensors are arranged around the interaction area, including lidar, infrared camera and ultra-wideband positioning device, to collect real-time coordinates (X, Y, Z axis positions) of users, attitude data (head orientation angle, body tilt angle) and device status (battery level, connection stability). These data are transmitted to the central processing unit through wireless means, and the interaction area is divided into several subareas using grid division method. For example, when a user is located at the center point of the interaction area, the comprehensive interaction demand intensity is obtained by multiplication and weighted summation, and the weight is set by historical data analysis method, which is related to the user's position, action frequency and device status. Then, a reference point is randomly selected and the distance from the reference point to all candidate user positions is calculated by Euclidean distance formula. On this basis, the multi-dimensional space segmentation algorithm further segments the subareas to generate reachable and unreachable areas. Finally, the distance matrix and directed graph are constructed, and the optimal path matrix is calculated by combining the path planning algorithm, providing basic data support for the subsequent behavior recognition module. This step ensures the accuracy of user position information and the rationality of subarea division, thereby improving the system's adaptability to complex environments.

[0068] Secondly, in the running phase of the behavior recognition module, based on the user position information and optimal path matrix obtained from the spatial positioning module, the system sets the weight coefficient according to the historical data analysis method, and calculates the priority score of each user by combining the user's energy consumption and minimum distance value. For example, when a user has high energy consumption due to frequent movement and is far from the target position, his priority score is low. The priority score is used to generate the user arrangement result, i.e. the behavior sequence. The behavior sequence is combined with the priority weight by encoding technology to generate the scheduling individual, which enters the population generation phase after initialization. In this process, the random number generator generates the adjustment coefficient, and the value range of the adjustment coefficient is set by the field method. The generated population calculates the fitness value through the objective function, and the individual with the highest fitness value is selected for crossover and mutation operations, thereby generating candidate solutions. After multiple iterations of optimization, the initial solution is output, which contains the behavior sequence of each user participating in the interaction and the combination of their priority settings. This step dynamically adjusts the user behavior sequence and priority settings, achieving accurate recognition of user behavior and providing high-quality input data for the subsequent dynamic optimization module.

[0069] Next, in the running phase of the dynamic optimization module, the initial solution is used as the core individual to generate the core population through Gaussian perturbation and mutation operations. The number of individuals in the core population is set by an empirical value, typically twice the number of initial solutions. Then, a number of core individuals are randomly selected from the core population as auxiliary individuals, and an auxiliary population is generated by combining random generation operations. After the initialization of the core population and the auxiliary population, the module optimizes the population through heuristic information and information diffusion factors. The heuristic information is defined by the inverse of the objective function value of the initial solution, and the information diffusion factor is determined by rule setting and statistical methods. During the optimization process, the module calculates the selection probability using the selection probability model formula and selects the current behavior sequence through the roulette method. The analytic hierarchy process is used to set the weight coefficient of user behavior, and the calculation result of the objective function value is used as the individual fitness value. The module calculates the starting state difference and the ending state difference between the core individual and the adjacent auxiliary individual through the difference evaluation method, and updates the information diffusion factor according to the empirical rule. When the number of iterations reaches the exchange iteration number, the module arranges the core population and the auxiliary population in ascending order of fitness value, and performs cross replacement according to the selected number. The state path after cross replacement is fused by setting the decay factor through the grid search method, and the fused state path is used as the new individual to enter the next iteration. Finally, the module outputs the final solution containing the behavior sequence of each user and the priority setting combination of the participating users in each time slice. This step realizes the dynamic adjustment of user behavior sequence and priority setting through intelligent optimization algorithm, improving the efficiency and adaptability of the system.

[0070] Then, in the running phase of the early warning module, the system obtains the historical running data of users from the spatial positioning module and the behavior recognition module, including action frequency, movement trajectory, and device status. These data are preprocessed and input into a deep learning architecture to build a prediction model. The objective function of the prediction model is defined as the mean square error, and the model parameters are optimized through grid search and cross-validation. The optimized model takes the historical running data as input and outputs the behavior state prediction value of each user. The module sets the warning threshold by rule and compares the behavior state prediction value with the warning threshold. For example, when the behavior state prediction value of a user is greater than or equal to the warning threshold, the module triggers the recalculation of the final solution and sends a warning signal. Otherwise, the module continues to monitor the user's behavior state. This step ensures that the system can timely discover potential risks and take appropriate measures through the accuracy of the prediction model and the rationality of the warning threshold.

[0071] Finally, in the running phase of the data management module, the system converts the comparison results into charts through visualization tools and visualizes the comparison results, optimization suggestions, and final solutions through the charts. For example, when the interaction behavior frequency of a user is abnormal, the system generates a line chart or a heat map to intuitively present the behavior change trend. At the same time, the module stores the comparison results through the database and adds user ID and timestamp to each record for classified query. In order to ensure the security of the data, the module uses a symmetric encryption algorithm to encrypt the stored data and changes the key regularly. This step not only facilitates the management personnel to quickly understand the running state of the system through the visualization and encryption storage of the data, but also effectively guarantees the security of the data.

[0072] The above modules are connected with each other through data flow and control flow to form a complete system. The spatial positioning module provides basic data for the behavior recognition module, the output of the behavior recognition module is used as the input of the dynamic optimization module, the final solution of the dynamic optimization module provides decision basis for the early warning module, and the data management module uniformly manages and protects the data of the entire system. Such modular design not only improves the scalability and maintainability of the system, but also ensures the efficient cooperation between the modules. Through the steps of the above specific application scenarios combined with the principles, the technical solution of the present application is fully disclosed, so that the person skilled in the art can implement the technology according to the content of the specification.

[0073] The preferred embodiments of the application disclosed above are only used to help explain the application. The preferred embodiments do not describe all the details and do not limit the application to the specific embodiments described. Obviously, many modifications and changes can be made according to the content of the specification. The present application selects and describes these embodiments in order to better explain the principles and practical applications of the present application, so that the person skilled in the art can well understand and utilize the present application. The present application is limited only by the claims and their entire scope and equivalents.

Claims

1. A large-space, multi-person interactive safety monitoring system based on real-time data processing, characterized in that: include The spatial positioning module collects user location information, divides the interaction area using a multi-dimensional spatial segmentation algorithm, and generates an optimal path matrix by combining it with a path planning algorithm; the user location information includes the real-time coordinates, posture data, and device status of the user participating in the interaction. The behavior recognition module extracts priority scores based on user action features, combines them with the optimal path matrix to generate user behavior sequences, and performs iterative optimization to obtain an initial solution. The initial solution includes a sequence of user behaviors and their corresponding priority settings; The dynamic optimization module initializes the core and auxiliary population based on the initial solution generation, and optimizes it through a perturbation mechanism, a selection probability model, and an information diffusion strategy, combined with difference assessment and cross-substitution methods, to obtain the final solution; the final solution includes the optimized user behavior sequence and its corresponding priority setting combination. The early warning module constructs a prediction model to predict user behavior status, triggers the final solution to be recalculated based on the prediction results, and issues an early warning signal. The data management module displays data through a visual interface, stores data in a database, and then encrypts it using an encryption algorithm.

2. The XR large-space multi-person interactive security monitoring system based on real-time data processing as described in claim 1, characterized in that: The process of collecting user location information, dividing the interaction area using a multi-dimensional spatial segmentation algorithm, and generating an optimal path matrix using a path planning algorithm includes the following steps: High-precision sensors are used to collect real-time coordinates, posture data, and device status of users participating in the interaction. The real-time coordinate data includes the user's X, Y, and Z axis positions in three-dimensional space. The posture data includes head orientation angle and body tilt angle. The device status includes battery level and connection stability. The interaction area is divided into partitions using a grid partitioning method. After setting the weight of each data in each partition using historical data analysis, multiplication is used to multiply each data by its corresponding weight and sum them up to obtain the comprehensive interaction demand intensity of each partition. Sort the intensity of comprehensive interaction needs in ascending order, use a rule of thumb to set the number of selections, iterate through the sorted results, and gradually select the intensity of comprehensive interaction needs based on the number of selections. Define the partition corresponding to the intensity of comprehensive interaction needs as the candidate user position, iterate through each candidate user position, and randomly select a reference point. Use the Euclidean distance formula to calculate the distance from the reference point to all candidate user positions. A multidimensional spatial segmentation algorithm combined with distance is used to segment the location of each candidate user to obtain the interaction area of ​​each user. In the interaction area, if the current area shares a boundary with the adjacent area, the area is marked as a reachable area; otherwise, it is marked as an unreachable area. The distance between user pairs in reachable regions is calculated using the Euclidean distance formula, and a distance matrix is ​​constructed by combining it with the distance matrix in unreachable regions. A directed graph is then constructed using the Graph tool in conjunction with the distance matrix. The path planning algorithm is used to extract the outgoing and incoming edges of each user from the directed graph, resulting in the outgoing and incoming edge sets for each user. Given the set of outgoing edges and the set of incoming edges, calculate the product of the number of outgoing edges and the number of incoming edges in the set; After sorting the products in ascending order, the users corresponding to the products are selected as intermediate nodes for iteration according to the sorting results. During the iteration process, the distance between user pairs is defined as the weight. When user pairs can be reached through intermediate nodes, the path is updated through relaxation operation. When the distance between user pairs (,) in the updated path changes from unreachable to finite, it means that the path of user pair (,) has been updated. The i-th user is added to the set of incoming edges of the intermediate node, and the j-th user is added to the set of outgoing edges of the intermediate node. Continue iterative updates, and after the maximum number of iterations is reached, output the optimal path matrix; the optimal path matrix refers to the shortest physical distance between each pair of users within the interaction area.

3. The XR large-space multi-person interactive security monitoring system based on real-time data processing as described in claim 2, characterized in that: The process of extracting priority scores based on user action features, generating user behavior sequences by combining them with the optimal path matrix, and iteratively optimizing them to obtain an initial solution includes the following steps: The weighting coefficients are set using historical data analysis. Based on the user's energy consumption and the user's minimum distance in the optimal path matrix, the priority score of each user is calculated. Based on the priority score, all users are sorted in descending order to obtain the user sorting result, which is defined as a behavior sequence. The range of adjustment coefficient values ​​is set using the domain method, where represents the minimum adjustment coefficient and represents the maximum adjustment coefficient. An adjustment coefficient within a range of values ​​is generated using a random number generator, and the priority weight for each user is calculated. The behavior sequence and priority weight are encoded and combined using encoding techniques to generate scheduling individuals. A random population is generated for initialization, an objective function is defined, and the objective function value is calculated. These objective function values ​​are then used as individual fitness values ​​and sorted in ascending order. The largest fitness value is selected for crossover and mutation operations to generate candidate solutions for iteration. When the maximum number of iterations is reached, iteration stops, and the initial solution is output. The initial solution refers to the individual behavior sequences and priority settings of the participating users.

4. The XR large-space multi-person interactive security monitoring system based on real-time data processing as described in claim 3, characterized in that: The initialization process, which generates a core and an auxiliary population based on the initial solution, and optimization through a perturbation mechanism, a selection probability model, and an information diffusion strategy, combined with difference assessment and crossover methods, to obtain the final solution includes the following steps: The initial solution is taken as the core individual, and Gaussian perturbation and mutation operations are used to perturb and mutate the core individual to generate a core population for initialization. θ core individuals are randomly selected from the core population and defined as auxiliary individuals. Combined with random generation operations, an auxiliary population is generated for initialization. The reciprocal of the objective function value of the initial solution is defined as heuristic information, and the information diffusion factor is set using rules. Cross-validation was used to set the weights of heuristic information and information diffusion factor, and statistical methods were used to count all neighboring users of the current user and construct a set of neighboring users. Calculate the selection probability using the selection probability model formula; Use roulette wheel to select probabilities, and use the action sequence corresponding to the selected probabilities as the current action sequence. Use the analytic hierarchy process (AHP) to set the weight coefficients for user behaviors and define the objective function; Calculate the objective function value and use it as the fitness value of the core or auxiliary individuals; Iterative updates are performed. During the iteration process, the difference evaluation method is used to calculate the starting state distance between the core individual and the adjacent auxiliary individuals, which is defined as the starting state difference. Further, the difference assessment method is used to calculate the endpoint state distance between the core individual and the adjacent auxiliary individuals, which is defined as the endpoint state difference. The sum of the differences between the starting point and the endpoint is calculated using an addition formula, which is defined as the comprehensive difference. A constant is set using empirical rules, and the information diffusion factor is updated in combination with the comprehensive difference. Further, the domain method is used to set the number of exchange iterations and the number of selections. When the number of iterations reaches the number of exchange iterations, the core and auxiliary populations are sorted in ascending order according to their fitness values. Based on the number of selections, the first auxiliary individual in the auxiliary population arrangement is selected and cross-replaced with the first core individual in the core population arrangement. After the cross-replacement, the grid search method is used to set the decay factor and perform state path fusion. The fused state path is used as the new individual and the global information diffusion factor is updated. Continue iterative updates. When the number of iterations reaches the maximum, stop iterating and output the final solution. The final solution contains the individual behavior sequences and priority settings of the participating users in all time slices.

5. The XR large-space multi-person interactive security monitoring system based on real-time data processing as described in claim 4, characterized in that: The process of constructing a prediction model to predict user behavior states, and triggering recalculation of the final solution and issuing an early warning signal based on the prediction results includes the following steps: Collect historical operational data of users participating in the interaction; the historical operational data includes action frequency, movement trajectory, and device status; A prediction model is built using a deep learning architecture, with mean squared error defined as the objective function. The model parameters are optimized using grid search and cross-validation, and the optimized parameters and historical running data are used as inputs to output the predicted behavioral state value for each user. The system uses rules to set warning thresholds. It compares the predicted value of the behavior status with the warning threshold. If the current status is greater than or equal to the warning threshold, the user is removed, and the final solution is recalculated and a warning signal is triggered. Otherwise, monitoring continues.

6. The XR large-space multi-person interactive security monitoring system based on real-time data processing as described in claim 5, characterized in that: The implementation of corresponding measures refers to generating a report of comparison results and optimization suggestions when an early warning signal is triggered and monitoring continues, sending it to the management personnel through a transmission protocol, and then having the management personnel implement the corresponding measures.

7. The XR large-space multi-person interactive security monitoring system based on real-time data processing as described in claim 1, characterized in that: After collecting the user location, posture, and device data, preprocessing operations are performed, including noise reduction, interpolation, and normalization.

8. The XR large-space multi-person interactive security monitoring system based on real-time data processing as described in claim 6, characterized in that: The process of displaying data through a visual interface refers to using tools to convert the comparison results into charts, and then visually displaying the comparison results, optimization suggestions, and final solutions through these charts.

9. The XR large-space multi-person interactive security monitoring system based on real-time data processing as described in claim 7, characterized in that: The use of a database for data storage refers to storing the comparison results in a database, adding user IDs and timestamps to the comparison results in the database, and then classifying them.

10. The XR large-space multi-person interactive security monitoring system based on real-time data processing as described in claim 8, characterized in that: Encryption via encryption algorithm refers to using a symmetric encryption algorithm to encrypt the stored data and changing the key periodically.

Images