Target tracking method and system for virtual reality environment

Abstract

The application relates to the technical field of target tracking, in particular to a target tracking method and system for a virtual reality environment, which comprises the following steps: acquiring angular velocity, acceleration and six-degree-of-freedom pose data of a tracking target at different time points in real time, combining a Kalman filtering algorithm to acquire a posterior error covariance matrix at each time point; calculating a tracking inaccuracy at each time point; obtaining an occlusion risk degree at each time point; adjusting a process noise covariance matrix of the Kalman filtering algorithm; and estimating the pose of the tracking target by combining the adjusted process noise covariance matrix with the Kalman filtering algorithm. The application can realize robust, continuous and high-precision tracking of the tracking target under the conditions of occlusion or motion mutation, and ensure that the pose estimation of an interactive object in the virtual reality environment is always stable and reliable.

Description

Target Tracking Method and System for Virtual Reality Environments Technical Field

[0001] This application relates to the field of target tracking technology, and specifically to target tracking methods and systems for virtual reality environments. Background Technology

[0002] In virtual reality environments, target tracking technology is key to achieving immersive interactive experiences. By using real-time captured spatial positional information of the human body's limbs, eyes, and interactive devices to form natural interaction channels, it greatly enhances the immersiveness and realism of the virtual reality environment, resulting in significant technological breakthroughs in games and social applications.

[0003] In complex interactive scenarios, user body turning, interactive operations, or environmental obstacles often obstruct the field of view of the tracking device, resulting in optical occlusion. This leads to the loss or inaccuracy of the tracking optical pose data. When performing short-term pose estimation using data measured by the inertial measurement unit (IMU), the inherent noise in the measured data and the amplification of errors during the integration process cause significant drift in pose estimation. This manifests as virtual object jitter, ghosting, or even complete loss of control and disappearance, resulting in a decrease in the tracking accuracy of the target and severely damaging the user's immersion and interactive experience. Summary of the Invention

[0004] To address the aforementioned technical problems, a target tracking method and system for virtual reality environments are provided to resolve existing issues.

[0005] The solution to the technical problem addressed in this application is to provide a target tracking method and system for virtual reality environments, comprising the following steps:

[0006] In a first aspect, embodiments of this application provide a target tracking method for a virtual reality environment, the method comprising the following steps:

[0007] The system acquires the angular velocity, acceleration, and six-degree-of-freedom pose data of the tracked target at different times in real time, and combines the Kalman filter algorithm to obtain the posterior error covariance matrix at each time point.

[0008] Based on the overall bias of the posterior error covariance matrix and the bias intensity in the dominant direction, the uncertainty and error of pose estimation are assessed, and the tracking inaccuracy at each time step is calculated.

[0009] The tracking inaccuracy at different times is smoothed. The residuals are calculated by comparing the smoothing results with the tracking inaccuracy at each time. The occlusion risk at each time is obtained by combining the abnormal accumulation of the residuals.

[0010] Based on the occlusion risk level, the process noise covariance matrix of the Kalman filter algorithm is adjusted. Using the adjusted process noise covariance matrix, combined with the Kalman filter algorithm, the pose of the tracked target is estimated.

[0011] Preferably, the process of acquiring the six-degree-of-freedom pose data is as follows: an optical tracking device is deployed in the environment where the tracking target is located, the optical tracking device captures optical markers on the tracking target, and the six-degree-of-freedom pose data of the tracking target is calculated using a triangulation algorithm.

[0012] Preferably, the six-degree-of-freedom pose data includes three-dimensional position coordinates and three-degree-of-freedom attitude data.

[0013] Preferably, the method for obtaining the posterior error covariance matrix is ​​as follows: the angular velocity, acceleration and the six-degree-of-freedom pose data of the previous time step at each time step are used as inputs to the Kalman filter algorithm to obtain the posterior error covariance matrix at each time step.

[0014] Preferably, the calculation of tracking inaccuracy at each time point includes:

[0015] Perform eigenvalue decomposition on the posterior error covariance matrix to obtain all eigenvalues, and select the largest eigenvalue;

[0016] The tracking inaccuracy is the product of the trace of the posterior error covariance matrix at each time step and the largest eigenvalue.

[0017] Preferably, the residual calculation process is as follows: each moment and all moments before it are recorded as a local time period; the tracking inaccuracy of all moments within the local time period is smoothed to obtain the smoothed value of each moment; the difference between the tracking inaccuracy of each moment and the smoothed value is used as the residual of each moment.

[0018] Preferably, obtaining the occlusion risk level at each time moment includes:

[0019] Anomaly detection is performed on all residuals within a local time period at each time point to obtain the cumulative offset at each time point;

[0020] The occlusion risk level is the product of the residual at each time step and the cumulative offset.

[0021] Preferably, the process of obtaining the cumulative offset is as follows: using the CUSUM algorithm to perform anomaly detection on all residuals within the local time period at each time point, and outputting the cumulative offset at each time point.

[0022] Preferred, Adjusted process noise covariance matrix at time step The calculation formula is: ,in, for The corresponding normalized occlusion risk level at any given moment. for The process noise covariance matrix before adjustment at time step 1.

[0023] Secondly, embodiments of this application also provide a target tracking system for a virtual reality environment, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the above-described target tracking methods for a virtual reality environment.

[0024] This application has at least the following beneficial effects:

[0025] This application uses the Kalman filter algorithm to obtain the posterior error covariance matrix at each time step and calculates the tracking inaccuracy at each time step. Its advantage lies in quantifying the uncertainty of pose estimation through the posterior error covariance matrix, and further considering the overall uncertainty of the posterior error covariance matrix and the error amplification in the dominant direction, thus comprehensively reflecting the quality of target tracking. The tracking inaccuracy at different time steps is smoothed, and the residual is calculated based on the difference between the smoothing result and the tracking inaccuracy at each time step. Its advantage lies in considering the deviation of the tracking inaccuracy at that time step from the recent average level. The occlusion risk level is obtained at each time step, which is beneficial because it considers the continuous accumulation of tracking quality deviations and reflects the risk level of target occlusion at that time step. The process noise covariance matrix of the Kalman filter algorithm is adjusted. By utilizing the adjusted process noise covariance matrix and combining it with the Kalman filter algorithm, the pose of the tracked target is estimated. Its beneficial effect lies in dynamically adjusting the process noise covariance matrix of the Kalman filter algorithm based on the assessment of occlusion risk, thereby enhancing the flexibility and adaptability of the Kalman filter algorithm. This reduces the Kalman filter's reliance on IMU measurement data when occlusion occurs, making it more inclined to use predicted data to estimate the pose. This suppresses integral drift, reduces the accumulation of pose estimation errors, and maximizes the continuity and stability of tracking. This achieves robust, continuous, and high-precision tracking of the target under occlusion or sudden motion changes, ensuring that the pose estimation of interactive objects in virtual reality environments remains stable and reliable. Ultimately, this provides users with a continuous and accurate virtual interactive experience in complex interactive scenarios. Attached Figure Description

[0026] The target tracking method for virtual reality environments of this application will be further described in detail below with reference to the accompanying drawings.

[0027] Figure 1 is a flowchart of the steps of the target tracking method for a virtual reality environment provided in an embodiment of this application;

[0028] Figure 2 is a flowchart of the steps of the method for obtaining the adjusted process noise covariance matrix provided in the embodiment of this application. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the target tracking method and system for virtual reality environments proposed in this application will be further described in detail below with reference to the accompanying drawings and implementation examples. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit the scope of this application.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0031] Please refer to Figure 1, which shows a flowchart of the steps of a target tracking method for a virtual reality environment provided in an embodiment of this application. The method includes the following steps:

[0032] Step 1: Real-time acquisition of the angular velocity, acceleration, and six-degree-of-freedom pose data of the tracked target at different times.

[0033] During target tracking, situations such as target occlusion are very likely to occur, and in more complex environments, it is difficult to obtain the target's posture information, affecting the accuracy and stability of tracking. This can cause virtual objects in the virtual environment to shake, drift, suddenly disappear, or lag, causing motion sickness in users and resulting in a poor human-computer interaction experience.

[0034] In this embodiment, the tracking targets are the user's head and the user's hand, therefore:

[0035] Users wear head-mounted displays, which contain an inertial measurement unit (IMU) to collect the angular velocity and acceleration of the user's head in real time.

[0036] Users wear an interactive controller, which has a built-in inertial measurement unit (IMU) to collect the angular velocity and acceleration of the user's hand in real time.

[0037] Multiple optical tracking devices are deployed in the user's spatial environment to capture optical markers attached to the handle and the head-mounted display device. Using triangulation algorithms, the three-dimensional position coordinates of the tracking target and the three-degree-of-freedom attitude data, including pitch angle, roll angle and yaw angle, are calculated in real time to form six-degree-of-freedom pose data of the user's head and six-degree-of-freedom pose data of the handle.

[0038] It should be noted that the format of the six-DOF pose data is as follows: ,in, These are the coordinates on the X-axis. In order to be in The coordinate values ​​of the axis. In order to be in The coordinate values ​​of the axis. Indicates pitch angle, Indicates the roll angle. Indicates the yaw angle.

[0039] In this embodiment, the data acquisition frequency is 10Hz. As for other implementation methods, the implementer can set it according to the actual situation.

[0040] The collected data is normalized to obtain the angular velocity, acceleration, and six-degree-of-freedom pose data of the user's head and hands at each moment during the user's movement.

[0041] In this embodiment, the minimum-maximum normalization method is used for normalization processing. The minimum-maximum normalization method is a well-known technique and will not be described in detail here.

[0042] At this point, the angular velocity, acceleration, and six-degree-of-freedom pose data of the user's head and hands at each moment are obtained.

[0043] Step 2: Using the Kalman filter algorithm, obtain the posterior error covariance matrix at each time step; based on the overall deviation of the posterior error covariance matrix and the deviation intensity in the dominant direction, assess the uncertainty and error of pose estimation, and calculate the tracking inaccuracy at each time step.

[0044] In a virtual reality environment, if the target is temporarily out of the tracking range of the optical tracking device due to the user rotating their body, using the controller, or having their head obscured, the optical tracking device will be unable to provide visual positioning data and will only rely on the inertial measurement unit for attitude estimation. This will result in the accumulation of tracking errors, unsmooth movement of objects in the virtual environment, and even the object disappearing from the user's field of vision, thus damaging the user's immersive experience and the completion of interactive tasks.

[0045] The state equations and measurement equations of the Kalman filter system are as follows:

[0046] ;

[0047] ;

[0048] in, for The state variable at time t, for The state variable at time t, State variables arrive The state transition matrix, for The control input matrix at each time step, for Time-based control input, for The measurement variable at time, for The measurement matrix at time, For process noise, To measure the noise, we assume that the noise particles are independent and follow a normal distribution. Then:

[0049] , ;

[0050] in, Let k be the process noise covariance matrix at time k. Let be the measurement noise covariance matrix at time k.

[0051] Therefore, the specific implementation steps of the Kalman filter algorithm are as follows:

[0052] Based on the posterior state estimate from the previous time step First, predict the prior state estimate. and the prior error covariance matrix Specifically:

[0053] State prediction:

[0054] ;

[0055] in, for Prior state estimation at time 10:00 for The state transition matrix at time t, for Posterior state estimation at time 10:00. for The control input matrix at each time step, for Time-based control input;

[0056] Covariance prediction:

[0057] ;

[0058] in, for The prior error covariance matrix at time t. for The posterior error covariance matrix at time t. for The transpose of the matrix;

[0059] State update: Calculate the residual, which is the error between the actual measured value and the prior state estimate, and correct the calculated prior state estimate using the minimum variance principle to obtain the posterior state estimate. Simultaneously, based on the prior error covariance matrix and the Kalman filter gain, the posterior error covariance matrix is ​​solved. .

[0060] ;

[0061] ;

[0062] in, for Posterior state estimation at time 10:00. for The measurement variable at time, Kalman gain; for The posterior error covariance matrix at time t. Represents the identity matrix;

[0063] Kalman gain Calculation formula:

[0064] ;

[0065] in, for The transpose of the matrix;

[0066] Therefore, the user's head is placed in angular velocity, acceleration and The six-DOF pose data at time 1 is used as input to the Kalman filter algorithm to obtain the user's head position. The posterior error covariance matrix at time t;

[0067] Place the user's hand on angular velocity, acceleration and The six-DOF pose data at time 1 is used as input to the Kalman filter algorithm to obtain the user's hand position. The posterior error covariance matrix at time t;

[0068] In this embodiment, the initial measurement noise covariance matrix is ​​set as follows: ,in, To represent a 6×6 identity matrix, The six-DOF pose data at time t is used as The posterior state estimation at time 1 will Angular velocity and acceleration at time t are used as The control input at any time is used to obtain... The posterior error covariance matrix at time t.

[0069] It should be noted that the Kalman filtering algorithm is a well-known technique and will not be elaborated upon here.

[0070] Furthermore, the posterior error covariance matrix quantifies the uncertainty of state estimation. Therefore, based on the output posterior error covariance matrix, the tracking inaccuracy is calculated as follows:

[0071] Perform eigenvalue decomposition on the posterior error covariance matrix to obtain all eigenvalues, and select the largest eigenvalue;

[0072] In this embodiment, the Eigenvalue decomposition (EVD) algorithm is used to perform feature decomposition and obtain feature values. The EVD algorithm is a well-known technology and will not be described in detail here. As other implementation methods, implementers may use other methods of existing technology, such as the SVD algorithm, etc. This embodiment does not impose any special restrictions on this.

[0073] Calculate the trace of the posterior error covariance matrix;

[0074] It should be noted that the calculation process of the trace of a matrix is ​​a well-known technique and will not be elaborated here.

[0075] The product of the trace of the posterior error covariance matrix of the user's head at each time step and the largest eigenvalue is used as the tracking inaccuracy of the user's head at each time step.

[0076] The product of the trace of the posterior error covariance matrix of the user's hand at each time step and the largest eigenvalue is used as the tracking inaccuracy of the user's hand at each time step.

[0077] It should be noted that the trace of the posterior error covariance matrix reflects the overall uncertainty of the pose estimation at this time. The larger the value of the trace, the greater the overall uncertainty of the state estimation, the more divergent the pose error distribution of the Kalman filter estimation, and the lower the reliability of the target tracking state. The maximum eigenvalue reflects the error magnitude of the matrix in the direction of the eigenvector corresponding to the maximum eigenvalue. It increases with directional occlusion or sudden motion. The larger the maximum eigenvalue, the lower the reliability of the state estimation, which is mainly concentrated in the direction of the eigenvector corresponding to the maximum eigenvalue. It reflects the higher possibility of directional occlusion or sudden motion in this direction. The greater the obtained tracking inaccuracy, the worse the target tracking effect is at this time, and the greater the uncertainty and error of the pose estimation.

[0078] At this point, the tracking inaccuracies of the user's head and hands at various times are obtained.

[0079] Step 3: Smooth the tracking inaccuracy at different times. Calculate the residuals by comparing the smoothing results with the tracking inaccuracy at each time. Combine the abnormal accumulation of the residuals to obtain the occlusion risk at each time.

[0080] Furthermore, since optical occlusion in virtual reality environments is mostly sudden and brief, the tracking inaccuracy assessment evaluates the stability and accuracy of the Kalman filter at this time, but cannot determine the changes in tracking stability, nor can it predict whether the occlusion is just beginning or about to end, which will lead to response delays. As a result, it is impossible to grasp the best compensation time, and it is difficult to avoid pose estimation drift, resulting in a poor user interaction experience.

[0081] Based on the above analysis, the occlusion risk is calculated by analyzing the changes in tracking inaccuracy at different times to assess the deviation state of the Kalman filter. Specifically:

[0082] Each moment and all moments preceding it are recorded as a local time interval;

[0083] Smooth the tracking inaccuracy of the user's head at all times within a local time period to obtain the smoothed value of the user's head at each time point;

[0084] In this embodiment, an exponentially weighted moving average algorithm is used for smoothing, wherein the smoothing factor is set to 0.1, so that the smoothing result pays more attention to recent changes in tracking quality. The exponentially weighted moving average algorithm is a well-known technology and will not be described in detail here.

[0085] It should be noted that the smoothing value reflects the short-term trend of target tracking quality and assesses the recent average level of target tracking quality.

[0086] The difference between the tracking inaccuracy of the user's head at each time step and the smoothing value is used as the residual;

[0087] It should be noted that the residual reflects the degree of deviation of the current tracking inaccuracy from the recent average level. The larger the residual, the higher the deviation, indicating that the tracking quality has changed suddenly, possibly due to optical obstruction.

[0088] The CUSUM algorithm is used to detect anomalies in all residuals of the user's head within local time periods at each time point, and the cumulative offset of the user's head at each time point is obtained.

[0089] It should be noted that the CUSUM algorithm is a well-known technology and will not be described in detail here. The threshold in the CUSUM algorithm is set to h=0.05. This value is set according to the tolerance for occlusion false alarm rate. As other implementation methods, implementers can set it according to the actual situation.

[0090] Calculate the product of the residual of the user's head at each time step and the cumulative offset, and use it as the occlusion risk of the user's head at each time step.

[0091] To assess the tracking inaccuracy of the user's hand at various time points, the same calculation method used for the cumulative offset of the user's head at various time points is employed to calculate the occlusion risk of the user's hand at each time point. Specifically:

[0092] Smooth the tracking inaccuracy of the user's hand at all times within a local time period to obtain the smoothing value of the user's hand at each time point;

[0093] The difference between the tracking inaccuracy of the user's hand at each moment and the smoothing value is used as the residual;

[0094] Anomaly detection is performed on all residuals of the user's hand within local time periods at each time point to obtain the cumulative offset of the user's hand at each time point;

[0095] Calculate the product of the residual of the user's hand at each time step and the cumulative offset, and use it as the occlusion risk of the user's hand at each time step;

[0096] It should be noted that the cumulative offset reflects the degree of cumulative abrupt change in tracking quality, indicating whether the Kalman filter has been continuously moving away from the normal state or whether there is a hidden drift or gradual degradation process. The larger the cumulative offset, the more the deviation in tracking quality is accumulating, and the Kalman filter is gradually moving away from the stable tracking state. The greater the occlusion risk, the higher the risk of the tracked target being occluded. In this case, the lower the reliability of the pose estimation, and there is a possibility that tracking failure is about to occur or has already occurred.

[0097] This gives us the risk of head and hand occlusion at various times.

[0098] Step 4: Based on the occlusion risk level, adjust the process noise covariance matrix of the Kalman filter algorithm. Using the adjusted process noise covariance matrix, combined with the Kalman filter algorithm, estimate the pose of the tracked target.

[0099] Furthermore, in virtual reality environments, the tracked target may temporarily disappear from the field of view of the optical tracking device due to the user turning around or being blocked by occlusions in the environment. At this time, the pose estimation is performed using only the angular velocity and acceleration collected by the inertial measurement unit, which has inherent noise and integral drift problems. This causes the pose estimation error to accumulate over time, resulting in distortion of the user's tracking trajectory and the virtual object shaking or disappearing.

[0100] Based on the above analysis, when the occlusion risk is high, the Kalman filter needs to be more flexible in adapting to sudden motion changes. By increasing the process noise covariance matrix of the Kalman filter algorithm, the degree of trust of the Kalman filter in IMU measurement data is reduced. In the case of missing visual positioning data, the Kalman filter is more inclined to use the prediction of the motion model to estimate the pose, thereby suppressing integral drift and reducing the accumulation of pose estimation error.

[0101] Therefore, based on the occlusion risk level, the process noise covariance matrix of the Kalman filter algorithm is adjusted as follows:

[0102] ;

[0103] ;

[0104] in, For the user's head in The adjusted process noise covariance matrix at time t. For the user's head in The corresponding normalized occlusion risk level at any given moment. For the user's head in The process noise covariance matrix before adjustment at time step; For the user's hand in The adjusted process noise covariance matrix at time t. For the user's hand in The corresponding normalized occlusion risk level at any given moment. For the user's hand in The process noise covariance matrix before adjustment at time step;

[0105] In this embodiment, the maximum-minimum normalization method is used for normalization processing. The maximum-minimum normalization method is a well-known technique and will not be described in detail here. As other implementation methods, implementers may use other methods of the prior art, such as the tanh function, etc. This embodiment does not impose any special restrictions on this.

[0106] It should be noted that the greater the occlusion risk, the more likely occlusion or instability will occur. In this case, the process noise covariance matrix is ​​increased to reduce the Kalman filter's trust in the motion model, enabling the Kalman filter to use predicted IMU data instead of IMU measurement data containing significant noise to estimate the pose. After the occlusion ends, the occlusion risk decreases, and the process noise covariance matrix gradually recovers to its original size to smoothly return to the normal tracking state. The flowchart of the method for obtaining the adjusted process noise covariance matrix provided in this application embodiment is shown in Figure 2.

[0107] Based on the adjusted process noise covariance matrix and combined with the Kalman filter algorithm, the six degrees of freedom poses of the user's head and hands are estimated respectively, so as to achieve robust, continuous and high-precision tracking of the tracked target under occlusion or sudden motion, ensuring that the pose estimation of interactive objects in the virtual reality environment is always stable and reliable, thus providing users with a continuous and accurate virtual interactive experience in relatively complex interactive scenarios.

[0108] Based on the same inventive concept as the above methods, embodiments of this application also provide a target tracking system for a virtual reality environment, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any one of the above-described target tracking methods for a virtual reality environment.

[0109] It should be understood that although the steps in the flowchart of Figure 1 are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in Figure 1 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0110] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0111] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application, without departing from the content of the technical solution of this application, shall fall within the protection scope of the technical solution of this application.

Claims

1. A target tracking method for virtual reality environments, characterized in that, The method includes the following steps: acquiring the angular velocity, acceleration, and six-degree-of-freedom pose data of the tracked target at different times in real time, and obtaining the posterior error covariance matrix at each time time by combining the Kalman filter algorithm; evaluating the uncertainty and error of pose estimation based on the overall deviation of the posterior error covariance matrix and the deviation intensity in the dominant direction, and calculating the tracking inaccuracy at each time time; smoothing the tracking inaccuracy at different times, calculating the residuals by the difference between the smoothing results and the tracking inaccuracy at each time time, and obtaining the occlusion risk at each time time by combining the abnormal accumulation of the residuals; adjusting the process noise covariance matrix of the Kalman filter algorithm based on the occlusion risk, and estimating the pose of the tracked target using the adjusted process noise covariance matrix combined with the Kalman filter algorithm; Adjusted process noise covariance matrix at time step The calculation formula is: ,in, for The corresponding normalized occlusion risk level at any given moment. for The process noise covariance matrix before adjustment at time step 1.

2. The target tracking method for a virtual reality environment as described in claim 1, characterized in that, The process of acquiring the six-degree-of-freedom pose data is as follows: an optical tracking device is deployed in the environment where the target is located. The optical tracking device captures optical markers on the target and uses a triangulation algorithm to calculate the target's pose data in six degrees of freedom.

3. The target tracking method for a virtual reality environment as described in claim 1, characterized in that, The six-degree-of-freedom pose data includes three-dimensional position coordinates and three-degree-of-freedom pose data.

4. The target tracking method for a virtual reality environment as described in claim 1, characterized in that, The method for obtaining the posterior error covariance matrix is ​​as follows: the angular velocity, acceleration, and the six-degree-of-freedom pose data of the previous time step at each time step are used as inputs to the Kalman filter algorithm to obtain the posterior error covariance matrix at each time step.

5. The target tracking method for a virtual reality environment as described in claim 1, characterized in that, The calculation of tracking inaccuracy at each time step includes: performing eigenvalue decomposition on the posterior error covariance matrix to obtain all eigenvalues, and selecting the largest eigenvalue; the tracking inaccuracy is the product of the trace of the posterior error covariance matrix at each time step and the largest eigenvalue.

6. The target tracking method for a virtual reality environment as described in claim 1, characterized in that, The residual calculation process is as follows: each moment and all moments before it are recorded as a local time period; the tracking inaccuracy of all moments within the local time period is smoothed to obtain the smoothed value of each moment; the difference between the tracking inaccuracy of each moment and the smoothed value is taken as the residual of each moment.

7. The target tracking method for a virtual reality environment as described in claim 6, characterized in that, The process of obtaining the occlusion risk level at each time point includes: performing anomaly detection on all residuals within a local time period at each time point to obtain the cumulative offset at each time point; the occlusion risk level is the product of the residual at each time point and the cumulative offset.

8. The target tracking method for a virtual reality environment as described in claim 7, characterized in that, The process of obtaining the cumulative offset is as follows: the CUSUM algorithm is used to detect anomalies in all residuals within the local time period at each time point, and the cumulative offset at each time point is output.

9. A target tracking system for a virtual reality environment, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the target tracking method for a virtual reality environment as described in any one of claims 1-8.

Images