A multi-dimensional user data collection and analysis method based on virtual reality scenarios

Abstract

The present invention discloses a multi-dimensional user data collection and analysis method based on a virtual reality scene, which specifically relates to the field of user data collection and analysis. By analyzing the distribution of virtual crowds in the VR scene, monitoring areas are delimited, and virtual sentinels are set up to simulate an intelligent monitoring scene. The virtual sentinels can monitor specific targets according to preset tasks, and at the same time label and update the user portraits of the virtual crowds to refine the description of user characteristics, so as to integrate real-scene materials and construct dynamic virtual crowds in the VR environment. It can provide strong virtual data support for the training of relevant recognition AI models in places or scenes that require behavior monitoring, target recognition, and user characteristic analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of user data collection and analysis, and more specifically, to a method for multi-dimensional user data collection and analysis based on a virtual reality scenario. Background Art

[0002] The combination of virtual reality (VR) technology and recognition-based artificial intelligence (AI) is becoming an important direction for frontier research and practice. User recognition-based AI conducts target recognition (such as actions, movement trajectories, user portraits, etc.) by learning and processing a large amount of user data, while VR technology can create highly simulated virtual environments, providing unique training and testing scenarios to make up for the deficiencies of traditional data collection methods. Therefore, how to construct a highly realistic VR scenario and monitor the behavior of simulated dynamic crowds as virtual training data for user recognition-based AI has become the key.

[0003] To solve the above problems, a technical solution is provided. By analyzing the distribution of virtual crowds in the VR scenario, a monitoring area is delimited, and virtual sentinels are set up to simulate an intelligent monitoring scenario. The virtual sentinels can monitor specific targets according to preset tasks, and at the same time label and update the user portraits of the virtual crowds to refine the description of user characteristics. Integrating real-scene materials in the VR environment and constructing dynamic virtual crowds can provide strong technical support for behavior monitoring, target recognition, and user feature analysis. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method for multi-dimensional user data collection and analysis based on a virtual reality scenario to solve the problems put forward in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] S1: Collect real-scene materials, disassemble the real scene into elements by combining 3D modeling technology, and integrate them into the VR scenario;

[0007] S2: Extract pedestrian trajectory, action, and behavior pattern data from the real-scene materials, and simulate 3D virtual crowds in the VR scenario according to the extracted data;

[0008] S3: Deconstruct the integrated VR scenario, delimit a monitoring area based on the distribution of virtual crowds, and set virtual sentinels in the monitoring area;

[0009] S4: Issue tasks to all virtual sentinel viewpoints based on the preset target user types to be monitored, and at the same time perform a primary annotation on the user portraits of the virtual crowds;

[0010] S5: Conduct line-of-sight analysis on the virtual sentry viewpoint and perform secondary annotation on the user portraits of the individuals of concern in the virtual crowd;

[0011] S6: Extract the user characteristics of all the annotated virtual individuals and establish a virtual data sample library for target user identification.

[0012] In a preferred embodiment, in S1, collecting real-scene materials and integrating the element decomposition of the real scene into the VR scene by combining 3D modeling technology specifically includes:

[0013] Use a high-definition camera or a drone to take multi-angle shots of the designated material collection site, including horizontal view, vertical view, and local close-ups, to generate complete image data;

[0014] Use a laser scanning instrument to comprehensively scan the collection site, collect depth data and generate accurate point clouds, import the point cloud data into modeling software to convert it into a preliminary geometric model, and fuse the photogrammetric image data and point clouds to generate a high-precision three-dimensional model;

[0015] Denoise, color correct, and texture stitch the image data, and crop the image outside the designated material collection site area according to the scene requirements to reduce data redundancy;

[0016] Perform key element annotation on the main constituent elements in the three-dimensional model, separate the key elements into independent geometric modules, reduce the number of faces of the model through polygon reduction technology to improve rendering performance while retaining the key details of the elements; import the disassembled geometric modules and rearrange them according to the spatial layout of the real scene of the site to form a VR scene based on real-scene materials.

[0017] In a preferred embodiment, in S2, extracting pedestrian trajectory, action, and behavior pattern data from the real-scene materials and simulating the 3D virtual crowd in the VR scene specifically includes:

[0018] Select a real area where pedestrians flow densely and meet the preset specific scene requirements, and capture and record the behavior patterns and movement trajectories of pedestrians;

[0019] Extract the trajectory curves of pedestrians through a motion detection algorithm, store them as time series data, analyze the actions of pedestrians in the video, decompose the actions into joint angle changes, and classify the behavior patterns of pedestrians;

[0020] Based on the behavior pattern classification, establish behavior trigger rules to simulate the autonomous behaviors of the virtual crowd;

[0021] Convert the time series data of pedestrian trajectory curves into a mathematical model to describe the movement law of pedestrians in three-dimensional space. Utilize 3D modeling and animation techniques to generate 3D virtual crowds based on the behavior patterns and trajectory models of pedestrians, and integrate the virtual crowds into the VR scene.

[0022] In a preferred embodiment, in S3, deconstruct the integrated VR scene, delimit the monitoring area based on the distribution of virtual crowds, and set virtual sentinels in the monitoring area, specifically including:

[0023] Functionally deconstruct the VR scene, divide the VR scene into multiple custom logic layers, including a geometry layer, a dynamic layer, and an interaction layer. Among them, the geometry layer is used to define the spatial boundaries of the monitoring areas within the scene, and the dynamic layer is used to track the distribution and movement trajectories of virtual crowds;

[0024] According to the dynamic layer, real-time statistics of the distribution density heat map of virtual crowds in each area of the VR scene are carried out, areas with frequent crowd movement are screened out, marked as virtual monitoring areas, and spatial boundaries are delimited in the geometry layer;

[0025] Pre-define static viewpoints, and set static viewpoints at the spatial boundaries of the virtual monitoring areas according to the size of the virtual monitoring areas. The specific method is as follows:

[0026] Obtain the maximum observable angle and the farthest observation distance set by the static viewpoint;

[0027] Judge whether the diameter of the circumcircle of the spatial boundary of the virtual monitoring area is greater than the farthest observation distance preset by the static viewpoint;

[0028] If so, take the center of the circumcircle of the spatial boundary of the virtual monitoring area as the midpoint, and divide the virtual monitoring area into a cross shape according to the horizontal and vertical axes of the VR scene. Each small monitored area after division is used as a virtual monitoring sub-area, and the junction of the virtual monitoring sub-areas is the new spatial boundary;

[0029] Iteratively divide the virtual monitoring sub-areas until the diameters of the circumcircles of the spatial boundaries of all virtual monitoring sub-areas are greater than the farthest observation distance preset by the static viewpoint, and then end the division;

[0030] Set equidistant monitoring positions on the circumcircle of the spatial boundary of the virtual monitoring area according to the maximum observable angle set by the static viewpoint, and mark these static viewpoints as virtual sentinel viewpoints in the interaction layer.

[0031] In a preferred embodiment, in S4, based on the preset target user types to be monitored, issue tasks to all virtual sentinel viewpoints, and at the same time perform a label on the user portraits of virtual crowds, specifically including:

[0032] Set scene tasks for the virtual sentry viewpoints according to the target user types to be monitored, with each virtual sentry viewpoint corresponding to a VR device, and deliver the VR devices for test users to use;

[0033] Test users monitor the virtual crowd in the virtual monitoring area based on the scene tasks corresponding to the VR devices, and at the same time perform a primary annotation on the user portraits.

[0034] In a preferred embodiment, in S5, the line-of-sight analysis of the virtual sentry viewpoints and the secondary annotation of the user portraits of the individuals of interest in the virtual crowd specifically include:

[0035] Obtain and analyze the sequential line-of-sight data of the virtual sentry viewpoints within the set monitoring time window in the VR scene. The sequential line-of-sight data includes the line-of-sight angles of each virtual sentry viewpoint, the individual targets of the virtual crowd focused by the line of sight, and the duration of the line of sight focusing on and following the individual targets;

[0036] Filter out the data where the line of sight is not focused on the individuals in the virtual crowd, and perform sequential coincidence analysis on the remaining sequential line-of-sight data;

[0037] Statistically analyze the maximum number of viewpoints and the viewing time when each virtual individual in the virtual crowd is simultaneously gazed at by the virtual sentry viewpoints. When there are virtual individuals gazed at by more than half of the virtual sentry viewpoints and the average viewing time reaches the set duration, perform a secondary annotation on the user portraits of these virtual individuals.

[0038] In a preferred embodiment, in S6, extract the user characteristics of all the annotated virtual individuals and establish a virtual data sample library for target user recognition, including:

[0039] Obtain the virtual user portrait data, behavior patterns, and movement trajectories of the annotated individuals in the VR scene, and extract and integrate the user portrait characteristics, movement trajectory characteristics, and behavior characteristics;

[0040] Store the integrated characteristics in the training data sample library, mark the characteristic data corresponding to the virtual individuals whose user portraits have only been annotated once as the training set samples, and mark the characteristic data corresponding to the virtual individuals whose user portraits have been annotated once and twice as the validation set samples;

[0041] Use the training set samples and the validation set samples as the virtual data sample library for the target user recognition model.

[0042] The technical effects and advantages of a multi-dimensional user data collection and analysis method based on a virtual reality scene according to the present invention:

[0043] By combining real - world scene materials with 3D modeling technology, the real environment can be highly restored and integrated into the VR scene to achieve dynamic crowd simulation. Modern computer vision technology can extract pedestrian trajectories, actions, and behavior patterns from real - world scenes, providing real - data support for virtual crowds. On this basis, the monitoring area can be delimited by analyzing the crowd distribution, and virtual sentinels can be set up to simulate an intelligent monitoring scenario. The virtual sentinels can monitor specific targets according to preset tasks, and at the same time label and update the user portraits of the virtual crowds to refine the user feature descriptions. By extracting the user features of the labeled individuals, a high - quality target user recognition training dataset can be generated to support the training and verification of machine - learning models. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of a multi - dimensional user data acquisition and analysis method based on a virtual reality scene according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] Embodiment 1

[0047] Figure 1 A multi - dimensional user data acquisition and analysis method based on a virtual reality scene according to the present invention is provided, which includes the following steps:

[0048] S1: Collect real - world scene materials, disassemble the elements of the real - world scene by combining 3D modeling technology, and integrate them into the VR scene;

[0049] S2: Extract pedestrian trajectory, action, and behavior pattern data from the real - world scene materials, and simulate 3D virtual crowds in the VR scene according to the extracted data;

[0050] S3: Deconstruct the integrated VR scene, delimit the monitoring area based on the virtual crowd distribution, and set virtual sentinels in the monitoring area;

[0051] S4: Based on the preset target user types to be monitored, issue tasks to all virtual sentinel viewpoints, and at the same time perform a first - time labeling of the user portraits of the virtual crowds;

[0052] S5: Conduct line - of - sight analysis on the virtual sentinel viewpoints, and perform a second - time labeling of the user portraits of the individuals of interest in the virtual crowds;

[0053] S6: Extract the user characteristics of all marked virtual individuals and establish a virtual data sample library for target user identification.

[0054] In S1, collect real - world scene materials and, in combination with 3D modeling technology, disassemble the elements of the real - world scene and integrate them into the VR scene. The high - precision restoration of the real - world scene is an important link in the application of VR technology. By collecting high - resolution images and video materials and combining three - dimensional modeling tools (such as Blender, Unity, or Unreal Engine), complex real - world scenes can be transformed into highly realistic VR environments. Elements included in the scene (such as buildings, roads, dynamic crowds) need to be disassembled and integrated layer by layer to ensure the authenticity and interactivity of the VR scene.

[0055] At material collection sites such as city squares, stations, shopping malls, etc., use high - definition cameras with a resolution of not less than 4K and drones equipped with a stabilization system to shoot the site in all directions and from multiple angles. This includes horizontal - angle shooting, circling the site along the horizontal plane to obtain panoramic images; vertical - angle shooting, using drones to obtain the overall layout of the site from an aerial perspective; and taking local close - ups of key areas to capture high - precision information such as building details and textures, generating complete image data.

[0056] Use a high - precision laser scanner (such as LiDAR) to conduct a comprehensive three - dimensional scan of the collection site. Through multiple measurements, obtain the depth data of the site and generate accurate point - cloud data. During the scanning process, measure from different positions and angles to ensure that the three - dimensional structure of the site is covered without dead angles. Import the point - cloud data into modeling software to convert it into a preliminary geometric model, and fuse the photogrammetric image data and point - cloud to generate a high - precision three - dimensional model. The point - cloud data provides accurate spatial positions and constructs the basic framework of the model.

[0057] In the modeling software, perform denoising processing on the imported image data to eliminate noise and impurities generated during shooting. Conduct color correction to unify the hues and brightness under different lighting conditions and ensure the consistency of model textures. Stitch the textures of the model to eliminate stitching marks and ensure the continuity and integrity of the textures. According to the scene requirements, crop the image and model parts outside the material collection site area to reduce data redundancy and improve the efficiency of model processing and rendering.

[0058] Mark the key elements in the three - dimensional model, separate the key elements into independent geometric modules, reduce the number of faces of the model through polygon reduction technology to improve rendering performance, while retaining the key details of the elements;

[0059] Import the processed geometric modules into a VR development engine (such as Unity or Unreal Engine). These modules serve as the basic building blocks of the scene, providing a realistic physical foundation for the virtual environment. Rearrange the spatial layout of the real-world scene to form a VR scene based on real-world materials.

[0060] In S2, extract pedestrian trajectory, motion, and behavior pattern data from real-world scene materials, and simulate 3D virtual crowds within the VR scene based on the extracted data. Modern computer vision technology can automatically identify and record the motion trajectories, behavior characteristics, and interaction patterns of pedestrians, and this data can be further used for the construction and simulation of dynamic crowds in the VR scene.

[0061] Select a real-world area with dense pedestrian flow that meets the preset specific scene requirements, and capture and record the behavior patterns and movement trajectories of pedestrians.

[0062] Use a motion detection algorithm (such as the optical flow method) to extract the movement trajectories of pedestrians from the video, and represent them as time series data: r(t) = {x(t), y(t), z(t)}, where x(t), y(t), and z(t) are the coordinates in the horizontal, vertical, and height directions at time t, respectively, and r(t) is the time series movement trajectory. Smooth the trajectory curve through Kalman filtering to correct the trajectory fluctuations caused by environmental interference.

[0063] Analyze the actions of pedestrians in the video, decompose the actions into joint angle changes, and classify the behavior patterns of pedestrians. The calculation method of the joint angle is as follows: θ ab is the included angle between joints a and b, P a and P b are the position vectors of joints a and b, respectively.

[0064] Based on the behavior pattern classification, establish behavior trigger rules to simulate the autonomous behaviors of virtual crowds. The behavior pattern classification can be based on trajectory and action characteristics, and use clustering algorithms to classify the behavior patterns. The following are several typical simple behavior patterns:

[0065] Free walking: Regular joint angle changes and a relatively straight trajectory.

[0066] Stay and observe: Joint angles do not move, the movement speed is close to zero, and the position changes frequently.

[0067] Aggregation behavior: Multiple trajectories converge in the same area.

[0068] Fast movement: The speed increases abnormally, and the joint angle changes frequently, usually corresponding to avoidance or emergency behaviors.

[0069] Convert the time series data of the pedestrian trajectory curve into a continuous function R(t), where R(t) = r0 + ∫v(τ)dτ, to describe the movement law of pedestrians in three-dimensional space. r0 and v(τ) are the initial position of the pedestrian and the instantaneous speed and direction at the time integral variable τ respectively. Using 3D modeling and animation technology, generate a 3D virtual crowd based on the behavior patterns and trajectory models of pedestrians. Import the virtual crowd into a VR development engine (such as Unreal Engine), allocate the initial positions of the characters according to the spatial layout of the real venue, and set the dynamic paths based on the trajectory mathematical model to simulate the flow trend of the crowd.

[0070] In S3, deconstruct the integrated VR scene, delimit the monitoring area based on the distribution of the virtual crowd, and set virtual sentinels in the monitoring area. As the "observers" of intelligent monitoring, the virtual sentinels can monitor specific crowds according to the set tasks and viewpoints, simulating the role of the monitoring system in the real scene.

[0071] Conduct a functional deconstruction of the VR scene, divide the VR scene into multiple custom logic layers, including a geometry layer, a dynamic layer, and an interaction layer. Among them, the geometry layer is used to define the spatial boundaries of the monitoring areas within the scene, and the dynamic layer is used to track the distribution and movement trajectories of the virtual crowd;

[0072] According to the dynamic layer, statistically generate a heat map of the distribution density of the virtual crowd in each area of the VR scene in real time. Generate the heat map according to the following formula: ρ is the population density of the area, and N and A represent the number of virtual people in the area and the area defined by the geometry layer respectively.

[0073] Screen the areas with frequent crowd movement, mark them as virtual monitoring areas, and delimit the spatial boundaries in the geometry layer;

[0074] Pre-define static viewpoints, and set static viewpoints at the spatial boundaries of the virtual monitoring areas according to the size of the virtual monitoring areas. The specific method is as follows:

[0075] Obtain the maximum observable angle and the farthest observation distance set by the static viewpoints;

[0076] Judge whether the diameter of the circumscribed circle of the spatial boundary of the virtual monitoring area is greater than the farthest observation distance preset by the static viewpoint;

[0077] If so, take the center of the circumscribed circle of the spatial boundary of the virtual monitoring area as the midpoint, and divide the virtual monitoring area into four parts in the horizontal and vertical axis directions of the VR scene. Each small monitored area after division is used as a virtual monitoring sub-area, and the junction of the virtual monitoring sub-areas is the new spatial boundary.

[0078] Iteratively divide the virtual monitoring sub-regions until the diameter of the circumscribed circle of the spatial boundaries of all virtual monitoring sub-regions is greater than the preset maximum observation distance of the static viewpoints, and then end the division.

[0079] Set equidistant monitoring positions on the circumscribed circle of the spatial boundary of the virtual monitoring area according to the maximum observable angle set by the static viewpoints, and mark these static viewpoints as virtual sentry viewpoints on the interaction layer.

[0080] Based on the equidistant angular distribution of the static viewpoints on the circumscribed circle of the spatial boundary, assuming the normal viewing range of a person, if the maximum observable angle set by the static viewpoints is 120 degrees, then at least 360 degrees / 120 degrees = 3 static viewpoints are required to fully cover the virtual monitoring area. At the same time, it means that the static viewpoints should be set at the three equal division points on the circumscribed circle of the spatial boundary.

[0081] In S4, based on the preset target user type to be monitored, issue tasks to all virtual sentry viewpoints, and at the same time perform a labeling of the user portraits of the virtual crowd.

[0082] According to the set target user type to be monitored, set scenario tasks for the virtual sentry viewpoints. Each virtual sentry viewpoint corresponds to a VR device, and the VR device is provided for the test users to use. Assuming the target user type to be monitored is a suspicious old person, then the test users continuously monitor and identify the virtual individuals that meet the characteristics in the virtual crowd according to the monitoring area within their own perspectives, and then perform dynamic labeling.

[0083] The "test users" here can be real humans or other trained AI models for special personnel identification. The test users monitor the virtual crowd in the virtual monitoring area based on the scenario tasks corresponding to the VR devices, and at the same time perform a labeling of the user portraits. The user portraits mainly include basic feature data such as age group, appearance ratio, body type and gait reflected by the virtual individuals in the VR scenario.

[0084] In S5, perform line-of-sight analysis on the virtual sentry viewpoints and perform secondary labeling on the user portraits of the individuals of concern in the virtual crowd.

[0085] Obtain and analyze the sequential line-of-sight data of the virtual sentry viewpoints within the set monitoring time window in the VR scenario. The sequential line-of-sight data includes the line-of-sight angles of each virtual sentry viewpoint, the individual targets of the virtual crowd focused by the line of sight, and the duration of the line-of-sight focus following the individual targets.

[0086] Filter out the data where the line of sight is not focused on the virtual crowd individuals, and perform time-series coincidence analysis on the remaining time-series line-of-sight data (combine the data of all virtual sentry viewpoints within the set monitoring time window with the monitored virtual individuals for calculation), and count the maximum number of viewpoints and the fixation time when each virtual individual in the virtual crowd is simultaneously gazed at by the virtual sentry viewpoints. When there is a virtual individual who is simultaneously gazed at by more than half of the virtual sentry viewpoints and the average fixation time reaches the set duration, perform secondary annotation on the user profile of this virtual individual.

[0087] Count the number of virtual sentry viewpoints N that simultaneously gaze at the virtual target individual ID numbered i at each time point within the set monitoring time window i of the virtual sentry sentinel(k,IDi) , and its calculation expression is: where M is the total number of all virtual sentry viewpoints, k is the number of the time point within the set monitoring time window, and δ ij ( k ) = 1 indicates that the j-th sentry is gazing at the target individual ID at the k-th time point i , otherwise δ ij ( k ) = 0.

[0088] When the maximum number of virtual sentry viewpoints N max reaches half of the total number of all virtual sentry viewpoints (i.e., N max ≥ [M / 2]), it meets the fixation requirement for secondary annotation.

[0089] In S6, extract the user characteristics of all the annotated virtual individuals, establish a virtual data sample library for target user recognition. By extracting features from the virtual crowd being concerned in the VR scene, structured data samples can be generated. Combining with the annotated user profile information, these data samples can be used for the training and verification of machine learning models to further improve the recognition accuracy of target users.

[0090] Obtain the virtual user profile data, behavior patterns, and movement trajectories of the annotated individuals in the VR scene, and extract and integrate the user profile features, movement trajectory features, and behavior features;

[0091] Obtain all the data of the annotated virtual individuals from the user profile system in the VR scene, which are divided into primary annotation and secondary annotation. Among them, the virtual individuals with obvious features are primarily annotated and used for targeted training of the user recognition model. The individuals with less obvious features are secondarily annotated. If they are not annotated after being examined by multiple sentries, it can be determined that this individual is imperfect or suspicious data. Such data can test the robustness and generalization ability of the model and avoid the model overfitting to obvious features.

[0092] Store the integrated features in the training data sample library. Mark the feature data corresponding to the virtual individuals with only one annotation in the user profile as the training set samples, and mark the feature data corresponding to the virtual individuals with one annotation and a second annotation in the user profile as the validation set samples.

[0093] Use the training set samples and the validation set samples as the virtual data sample library of the target user recognition model. Through feature integration and optimization, divide the data into a training set and a validation set to ensure the diversity and representativeness of the sample library covering the target users. Finally, use the sample library to train and validate the target user recognition model, providing strong technical support for user management and behavior analysis in the virtual environment.

[0094] All the above formulas are dimensionless and take their numerical values for calculation. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters and threshold selection in the formula are set by those skilled in the art according to the actual situation.

[0095] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any arbitrary combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When loading or executing the computer instructions or computer programs on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center containing one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0096] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0097] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0098] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms.

[0099] The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules. They can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0100] In addition, in each embodiment of the present application, the various functional modules can be integrated in a processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0101] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0102] As described above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims described above.

[0103] Finally: The above description is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A multi-dimensional user data collection and analysis method based on a virtual reality scenario, characterized in that, It includes the following steps: S1: Collect real - world scene materials, disassemble the elements of the real - world scene by combining 3D modeling technology and integrate them into the VR scene; S2: Extract pedestrian trajectories, actions, and behavior pattern data from the real - world scene materials, and simulate 3D virtual crowds in the VR scene according to the extracted data; S3: Deconstruct the integrated VR scene, delimit monitoring areas based on the distribution of virtual crowds, and set virtual sentinels in the monitoring areas; S4: Based on the preset target user types to be monitored, issue tasks to all virtual sentinel viewpoints, and at the same time perform a first annotation on the user portraits of the virtual crowds; S5: Conduct line - of - sight analysis on the virtual sentinel viewpoints, and perform a second annotation on the user portraits of the individuals of interest in the virtual crowds; S6: Extract the user characteristics of all the annotated virtual individuals, and establish a virtual data sample library for target user recognition; In S3, deconstructing the integrated VR scene, delimiting monitoring areas based on the distribution of virtual crowds, and setting virtual sentinels in the monitoring areas specifically includes: Functionally deconstruct the VR scene, divide the VR scene into multiple custom logic layers, including a geometry layer, a dynamic layer, and an interaction layer. Among them, the geometry layer is used to define the spatial boundaries of the monitoring areas in the scene, and the dynamic layer is used to track the distribution and movement trajectories of virtual crowds; According to the dynamic layer, real - time statistics on the distribution density heat map of virtual crowds in each area of the VR scene, screen out the areas with frequent crowd movement, mark them as virtual monitoring areas, and delimit the spatial boundaries in the geometry layer; Pre - define static viewpoints, and set static viewpoints at the spatial boundaries of the virtual monitoring areas according to the size of the virtual monitoring areas. The specific method is: Obtain the maximum observable angle and the farthest observation distance set by the static viewpoints; Judge whether the diameter of the circumscribed circle of the spatial boundary of the virtual monitoring area is greater than the farthest observation distance preset by the static viewpoints; If so, take the center of the circumscribed circle of the spatial boundary of the virtual monitoring area as the mid - point, and divide the virtual monitoring area cross - wise according to the horizontal and vertical axes of the VR scene. Each small divided monitoring area is used as a virtual monitoring sub - area, and the junction of the virtual monitoring sub - areas is the new spatial boundary; Iteratively divide the virtual monitoring sub - areas until the diameters of the circumscribed circles of the spatial boundaries of all virtual monitoring sub - areas are greater than the farthest observation distance preset by the static viewpoints, and then end the division; Set equidistant monitoring positions on the circumscribed circle of the spatial boundary of the virtual monitoring area according to the maximum observable angle set by the static viewpoints, and mark these static viewpoints as virtual sentinel viewpoints on the interaction layer; In S5, conducting line - of - sight analysis on the virtual sentinel viewpoints, and performing a second annotation on the user portraits of the individuals of interest in the virtual crowds specifically includes: Obtain and analyze the time - series line - of - sight data of the virtual sentinel viewpoints in the set monitoring time window in the VR scene. The time - series line - of - sight data includes the line - of - sight angle of each virtual sentinel viewpoint, the individual target of the virtual crowd focused by the line of sight, and the duration of the line - of - sight focus following the individual target; Filter out the data where the line of sight is not focused on the individuals of the virtual crowd, and perform time - series coincidence analysis on the remaining time - series line - of - sight data; Statistically calculate the maximum number of viewpoints and the fixation time when each virtual individual in the virtual crowd is fixated by the virtual sentry viewpoints simultaneously. When there is a virtual individual fixated by more than half of the virtual sentry viewpoints simultaneously and the average fixation time reaches the set duration, perform secondary annotation on the user profile of this virtual individual.

2. The multi-dimensional user data collection and analysis method based on a virtual reality scenario according to claim 1, characterized in that, In S1, collect real - world scene materials, and combine 3D modeling technology to disassemble and integrate the elements of the real - world scene into the VR scene, specifically including: Use a high - definition camera or drone to take multi - angle photos of the designated material collection site, including horizontal perspective, vertical perspective, and local close - ups, to generate complete image data; Use a laser scanning instrument to comprehensively scan the collection site, collect depth data and generate accurate point clouds. Import the point cloud data into modeling software to convert it into a preliminary geometric model, and fuse the photogrammetric image data and point clouds to generate a high - precision 3D model; Denoise, color - correct, and texture - stitch the image data, and crop the images outside the designated material collection site area according to the scene requirements to reduce data redundancy; Perform key - element annotation on the main constituent elements in the 3D model, separate the key elements into independent geometric modules, reduce the number of faces of the model through polygon reduction technology to improve rendering performance, while retaining the key details of the elements; Import the disassembled geometric modules and rearrange them according to the spatial layout of the real - world scene of the site to form a VR scene based on real - world materials.

3. A multi-dimensional user data collection and analysis method based on a virtual reality scenario according to claim 2, characterized in that, In S2, extract pedestrian trajectory, action, and behavior pattern data from the real - world scene materials, and simulate the 3D virtual crowd in the VR scene according to the extracted data, specifically including: Select a real - world area with dense pedestrian flow and meeting the preset specific scene requirements, and capture and record the behavior patterns and movement trajectories of pedestrians; Extract the trajectory curves of pedestrians through motion detection algorithms, store them as time - series data, analyze the actions of pedestrians in the video, decompose the actions into joint - angle changes, and classify the behavior patterns of pedestrians; Based on the behavior pattern classification, establish behavior - triggering rules to simulate the autonomous behaviors of the virtual crowd; Convert the time - series data of the pedestrian trajectory curves into a mathematical model to describe the movement laws of pedestrians in three - dimensional space. Use 3D modeling and animation technologies to generate a 3D virtual crowd based on the behavior patterns and trajectory models of pedestrians, and integrate the virtual crowd into the VR scene.

4. A multi-dimensional user data collection and analysis method based on a virtual reality scenario according to claim 3, characterized in that, In S4, based on the preset target user types to be monitored, issue tasks to all virtual sentry viewpoints, and at the same time perform primary annotation on the user profiles of the virtual crowd, specifically including: According to the set target user types to be monitored, set scene tasks for the virtual sentry viewpoints. Each virtual sentry viewpoint corresponds to a VR device, and distribute the VR devices to test users for use; The test users monitor the virtual crowd in the virtual monitoring area based on the scene tasks corresponding to the VR devices, and at the same time perform primary annotation on the user profiles.

5. A multi-dimensional user data collection and analysis method based on a virtual reality scenario according to claim 4, characterized in that, In S6, extract the user characteristics of all the annotated virtual individuals, and establish a virtual data sample library for target user recognition, including: Obtain the virtual user portrait data, behavior patterns, and movement trajectories of the labeled individuals within the VR scene, and extract user portrait features, movement trajectory features, and behavior features for feature integration; Store the integrated features in the training data sample library. Mark the corresponding feature data of the virtual individuals with only one annotation of the user portrait as the training set samples, and mark the corresponding feature data of the virtual individuals with one annotation and two annotations of the user portrait as the validation set samples; Use the training set samples and the validation set samples as the virtual data sample library of the target user recognition model.

Images