CN-122020496-A - Real-time warning method and system for abnormal positioning of meta universe

CN122020496ACN 122020496 ACN122020496 ACN 122020496ACN-122020496-A

Abstract

The application provides a real-time alarming method for abnormal positioning of a meta universe, which relates to the technical field of data supervision, and comprises the steps of acquiring a positioning request initiated by a client of a user in a target space, wherein the positioning request carries context information capable of positioning the position of the user; the method comprises the steps of carrying out pose resolving on a request picture set and equipment information in context information and a positioning map corresponding to a target space to obtain the number of points in the pose, judging whether positioning is abnormal according to the number of points in the pose, obtaining a historical user VPS coordinate of last successful positioning when positioning is abnormal, determining a current movement range of a user based on the historical user VPS coordinate and the equipment information, adopting an object recognition model to recognize and generate a historical object picture set of a target space point cloud model in the current movement range, carrying out anomaly analysis according to the historical object picture set and the current object picture set in the request picture set, and generating a positioning alarm of object change in the target space if the positioning is abnormal.

Inventors

YAO JUNFENG
LI BIN
WANG LE
GUO SHIHUI
LIN HAO
ZHANG XIAOLEI
FANG TONGGUANG
ZHENG SIHAI

Assignees

厦门大学
咪咕新空文化科技(厦门)有限公司

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (10)

1. A real-time alarming method for abnormal positioning of metauniverse, which is applied to a server, and is characterized in that the method comprises the following steps, Acquiring a positioning request initiated by a client of a user in a target space, wherein the positioning request carries context information capable of positioning the position of the user, and the target space refers to a real space of a virtual model with a meta universe; Carrying out pose resolving on the request picture set and the equipment information in the context information and a positioning map corresponding to the target space, and obtaining the number of points in the pose; Judging whether the position is abnormal or not according to the number of the points in the pose, acquiring the VPS coordinates of the historical user successfully positioned last time when the position is abnormal, and determining the current movement range of the user based on the VPS coordinates of the historical user and the equipment information; Recognizing and generating a historical object picture set of a target space point cloud model in a current motion range by adopting an object recognition model, wherein the object recognition model is obtained by constructing a full convolution neural network according to object point cloud data training; And carrying out anomaly analysis according to the historical object picture set and the current object picture set in the request picture set, and if the anomaly exists, generating a positioning alarm of object change in the target space.
2. The method according to claim 1, wherein determining whether the location anomaly is based on the number of points in the pose comprises: If the number of the points in the pose is greater than or equal to the number threshold, judging that the positioning request is successful and the positioning is normal, storing the context information of the positioning request and returning the context information to the current user VPS coordinate obtained by the pose solution of the client; If the number of the points in the pose is smaller than the number threshold, judging that the positioning request fails; when the positioning request fails, detecting whether the positioning request is a non-target space object change problem or not and detecting whether the VPS coordinates of the historical user can be obtained or not; If the non-target space object changes or cannot be obtained, returning to the client to fail the positioning request; If the problem of the non-target space object change is not solved and the object change can be obtained, the abnormal positioning is judged.
3. The method of claim 2, wherein the non-target spatial object change problem is a performance or system resource starvation problem for a client device.
4. The method of claim 1, wherein the pose resolving the request picture set and the device information in the context information with the positioning map corresponding to the target space comprises: Extracting a first two-dimensional characteristic point of the request picture set, and matching with a second two-dimensional characteristic in a positioning map corresponding to the target space to obtain a target two-dimensional characteristic point in the second two-dimensional characteristic point; And determining the matching pair of the first two-dimensional feature point and the three-dimensional map point and the corresponding relation of the matching pair through the binding relation of the target two-dimensional feature point and the three-dimensional map point, and performing pose resolving according to the matching pair, the corresponding relation of the matching pair and the equipment information.
5. The method of claim 1, wherein the device information includes a fastest movement speed, a slowest movement speed, and a most frequently occurring movement speed of a user of the client in the target space, and wherein determining the current movement range of the user based on the historical user VPS coordinates and the device information comprises: constructing a simpson distribution according to the fastest movement speed, the slowest movement speed and the movement speed with the largest occurrence number; in each round of Monte Carlo simulation, sampling from the Simpson distribution to obtain a plurality of random movement speeds; Equally dividing an interval formed from the fastest movement speed to the slowest movement speed into a plurality of groups; counting the occurrence times of the random movement speed in each group, calculating the probability, and determining the upper limit of the group as the target movement speed by accumulating the probability to exceed 99%; based on a plurality of target movement speeds obtained by multi-round Monte Carlo simulation, when the statistical convergence obtained by calculation based on the plurality of target movement speeds is less than 1%, confirming the target movement speed of the current round as the current speed of a user; acquiring a time interval between a positioning request successfully positioned last time and a positioning request initiated by a client; Determining a movement radius according to the time interval and the current speed of the user; and determining the current motion range of the user according to the motion radius by taking the VPS coordinates of the historical user as an origin.
6. The method of claim 1, wherein the constructing of the object recognition model comprises: Acquiring object point cloud data of each historical object in a target space, and labeling corresponding object class labels for the object point cloud data to generate training data; Training PointNet a network or a 3D U-Net network according to training data, calculating in forward propagation of a training process to obtain a predicted object type, calculating a loss value according to the predicted object type and an object type label, optimizing the weight and bias of the PointNet network or the 3D U-Net network in reverse propagation based on the loss value, and constructing to obtain the object recognition model under the condition that a preset training stop condition is met.
7. The method of claim 1, wherein the identifying and generating the historical object picture set of the target spatial point cloud model in the current motion range using the object identification model comprises: Loading a preset target space point cloud model, and cutting out local point cloud data corresponding to the current motion range from the target space point cloud model; And inputting the local point cloud data into the object identification model, outputting the point cloud data of the historical object, and drawing a historical object picture set according to the point cloud data of the historical object.
8. The method of claim 1, wherein the performing anomaly analysis based on the historical object picture set and the current object picture set in the request picture set, if there is an anomaly, generating a localization alert of an object change in the target space, comprises: performing approximation analysis on the current object picture set in the history object picture set and the request picture set, and generating a positioning alarm of new object entering in the target space when the similarity is lower than a preset value, or Extracting a first position relation between a target object and an adjacent object in a historical object picture set and a second position relation between an approximate object and the adjacent object, which are approximate to the target object, in a current object picture set; And when the first position relation and the second position relation are different, generating a positioning alarm that the target object moves in the target space.
9. The method of claim 1 or 8, after generating a localization alert of a change in an object within a target space, the method further comprising: creating a real-time acquisition space video task, and transmitting the task to a client through long chain connection; receiving a space video sent by a client, modeling the space video, and generating a block mapping model and a sparse point cloud model for updating a target space point cloud model; and supplementing a space video, a block mapping model and a sparse point cloud model to the alarm record.
10. The real-time alarm system for abnormal positioning of the metauniverse is characterized by comprising a server and a client; The server acquires a positioning request initiated by a client of a user in a target space, wherein the positioning request carries context information capable of positioning the position of the user, the target space refers to a real space of a virtual model with a meta universe, and Pose resolving is carried out on the request picture set and the equipment information in the context information and the positioning map corresponding to the target space to obtain the number of points in the pose, and Judging whether the position is abnormal according to the number of the points in the pose, acquiring the VPS coordinates of the last successfully positioned historical user when the position is abnormal, determining the current movement range of the user based on the VPS coordinates of the historical user and the equipment information, and Identifying and generating a historical object picture set of a target space point cloud model in a current motion range by adopting an object identification model, wherein the object identification model is constructed by training a full convolution neural network according to object point cloud data, and And carrying out anomaly analysis according to the historical object picture set and the current object picture set in the request picture set, and if the anomaly exists, generating a positioning alarm of object change in the target space.

Description

Real-time warning method and system for abnormal positioning of meta universe Technical Field The application relates to the technical field of data supervision, in particular to a real-time warning method and system for abnormal positioning of a meta universe. Background With the development of technology, a virtual model (such as virtual antiques, virtual tourist attractions, etc.) presenting a meta universe in real space can be realized. As the user moves, different virtual models are presented, and the stability and accuracy of the virtual model presentation depends on visual localization of the real space. Frequently, problems with visual positioning failure occur, which may be due to performance problems of equipment worn by the user, insufficient system resources, and/or changes in real space, such as movement of indoor objects, new or removed, refinished, etc. In the prior art, the situation that the real space is changed cannot be early-warned in time, after the problem of visual positioning failure occurs, the user needs to feed back the manual customer service, and then the user needs to check reasons one by one from research and development to the scene, after the fact that the real space is changed is confirmed, the changed real space data needs to be acquired again, and then the space point cloud model is repaired. Because the research and development intervention positioning and repair response time is slower, the repair time is longer, so that a customer cannot timely experience the space of the virtual model of the accurate positioning and displaying meta universe, and the overall experience of the user is influenced. Meanwhile, due to the lack of instantaneity, time cannot be won for repairing the space point cloud model for the follow-up supplementary acquisition video. Disclosure of Invention The application provides a real-time warning method and a real-time warning system for abnormal positioning of a meta-universe, which can timely sense that a target space is changed, improve convenience and real-time performance and improve service responsiveness. The first aspect of the application provides a method for real-time warning of abnormal location of a meta-universe, which is applied to a server and comprises the steps of, Acquiring a positioning request initiated by a client of a user in a target space, wherein the positioning request carries context information capable of positioning the position of the user, and the target space refers to a real space of a virtual model in a meta universe; Carrying out pose resolving on the request picture set and the equipment information in the context information and a positioning map corresponding to the target space, and obtaining the number of points in the pose; Judging whether the position is abnormal or not according to the number of the points in the pose, acquiring the VPS coordinates of the historical user successfully positioned last time when the position is abnormal, and determining the current movement range of the user based on the VPS coordinates of the historical user and the equipment information; recognizing and generating a historical object picture set of a target space point cloud model in a current motion range by adopting an object recognition model, wherein the object recognition model is obtained by constructing a full convolution neural network according to object point cloud data training; And carrying out anomaly analysis according to the historical object picture set and the current object picture set in the request picture set, and if the anomaly exists, generating a positioning alarm of object change in the target space. In some embodiments, determining whether a localization anomaly is based on the number of points in the pose comprises: if the number of the points in the pose is greater than or equal to the number threshold, the successful positioning request and normal positioning are judged, the context information of the request is stored and returned to the current user VPS coordinate obtained by the pose solution of the client. If the number of the points in the pose is smaller than the number threshold, judging that the positioning request fails; when the positioning request fails, detecting whether the positioning request is a non-target space object change problem or not and detecting whether the VPS coordinates of the historical user can be obtained or not; If the non-target space object changes or cannot be obtained, returning to the client to fail the positioning request; If the problem of the non-target space object change is not solved and the object change can be obtained, the abnormal positioning is judged. In some embodiments, the non-target spatial object change problem is a performance or system resource starvation problem for the client device. In some embodiments, pose resolving the request picture set and the device information in the context information with a positioning map corresponding to t