CN101581967B - Method for inhibiting mutual interference between magnetic force trackers in augment reality system - Google Patents

Abstract

The invention discloses a method for suppressing mutual interference of magnetic trackers in an augmented reality system. The method first fixes the magnetic tracker at a suitable position, and then interferes with the tracking of the head and hands according to the respective movement characteristics of the head and hands during interaction. Different suppression methods are used: the opponent's tracking interference adopts particle filter; the head's tracking interference first judges the state of the head, and then adopts different processing methods according to different states. When it is stationary, it uses Kalman filter; when it moves slowly, it uses intrinsic geometry. Quantitative bilateral filtering; stop filtering when moving fast. This method does not require users to change or add hardware configurations, and is easy to operate, economical and practical. In an augmented reality system where a real hand grabs a virtual object, different filtering methods are used according to different states, which can suppress the interference of the magnetic transmitter to a certain extent. It can reduce the interference of the Euler angle and make the change of each angle in the Euler angle appear smooth.

Description

Method for suppressing mutual interference of magnetic trackers in augmented reality system

technical field

The present invention relates to the field of augmented reality technology, in particular to a method for suppressing mutual interference of magnetic trackers, and more specifically to a method for suppressing the interaction mode of grabbing a virtual object by a real hand Magnetic tracker mutual interference method.

Background technique

Augmented Reality (AR-Augmented Reality) is an important new means of multimedia technology in the three-dimensional field. AR technology involves many fields of computer science, including 3D modeling, real-time tracking and 3D registration, scene fusion and many other new technologies and means.

Tracking systems are an integral part of augmented reality systems. At present, the commonly used tracking systems mainly include optical tracking, magnetic tracking, ultrasonic tracking, infrared tracking, inertial tracking and other systems. Due to the different working principles, the above tracking systems have their own advantages and disadvantages. Among them, the magnetic tracker has become the most widely used type of orientation tracker in the augmented reality system due to its advantages of fast speed, good real-time performance, simple operation, and relatively low cost. But its disadvantages are also obvious: it is susceptible to interference from the surrounding environment, such as magnetic objects, magnetic fields generated by other equipment, etc., resulting in incorrect measurements.

At present, many augmented reality systems are required to support multiple people to interact at the same time, but the effective range of existing magnetic trackers is very limited, that is, the error is controlled within 2mm, and the effective distance is generally only ±0.5m. Such an effective range must not cover the interactive area. While switching to a larger launcher can increase the effective range, the increase is limited. Therefore, the solution of only using one transmitter obviously cannot meet the requirements of large-scale interactive scenes. But when multiple transmitters are used, that is, each user corresponds to a transmitter, due to the generation of multiple magnetic fields, interference will occur between the transmitters. Interference will increase the degree of data dispersion, and the most direct manifestation is that it will cause obvious shaking of the captured virtual object.

Many existing methods for suppressing interference are processing between transmission and reception. Such as Chinese patent: its title is "Method and Signal Processing System for Suppressing Interference in a Signal Processing System", application number 200410075232.X; These magnetic trackers are well packaged and it is difficult to obtain the waveform characteristics of the transmitted signal. Even if the desired features can be obtained and the desired processing can be performed, existing hardware configurations or circuits must be added or changed. Furthermore, many methods for removing interference have restrictions on the scope of use, such as Chinese patent: its name is "Interference detection method and interference avoidance system for wireless communication lines", application number 02120426.8, Chinese patent application: its name is "a An Active Illumination Imaging Device Against Ambient Light Interference", Application No. 200810054453.7. There are also some patents that use the idea of synthesizing multiple filters, but their application fields and filtering strategies are completely different from this patent. A method and system for counteracting interference with a matching device", application number 00818873.4. This patent decomposes the original signal into multiple channels, each channel has a specific spreading sequence, and offsets the interference in the received CDMA transmission by adapting multiple despreading vectors. This method requires the transmission Signal processing; Chinese patent application: its name is "Multiple Filters Realize Broadband Frequency Selection Method", application number: 200710026456.5. This patent application is applied in the frequency selection of broadband signals, although it uses multiple filters, but each filter has the same parameters. This method is only suitable for environments with relatively simple signal types and interference.

Filtering technology can extract useful signals from received signals containing interference. At present, the most widely used filtering methods mainly include Wiener filtering, Kalman filtering, particle filtering and other filtering. Kalman filtering and its improved methods have been used in many fields, such as target tracking, noise suppression and so on. They have good real-time and noise resistance, but also have the defect of strong dependence on the system dynamic model. If the target is highly maneuverable, the system model cannot be accurately modeled, and the filtering effect will be greatly affected. When the user interacts, the movements of the head and hands are irregular and have great mobility. Therefore, such methods cannot be simply applied. Regarding the problem of target mobility, many solutions have been proposed, such as "current statistical model", and the most potential maneuvering target tracking method-"interactive multi-model theory", etc. However, because the data obtained by the magnetic tracker is relatively accurate, the first digit after the decimal point is relatively stable, and the last few digits will fluctuate significantly due to interference from other devices and the device itself. This requires estimates of values to be accurate to at least the second decimal place. Obviously, these methods cannot meet the requirements of augmented reality systems in terms of accuracy. Particle filter implements recursive Bayesian filter through non-parametric Monte Carlo simulation method, which is suitable for nonlinear systems. But it is not the optimal estimate. When the interference is large, although the target can be tracked and the effect is better than many other maneuvering tracking models, the suppression effect on interference is still very limited.

The spatial curve smoothing method in graphics can remove the noise in the curve very well. The spatial dispersion curve with noise is similar to the disturbed movement track mentioned in this paper, that is, the disturbance value only floats around the correct value, and the whole movement trend is more obvious. However, this method only targets static curves and cannot process dynamic data; and the original method only denoises spatial coordinates without considering Euler angles. The Euler angle determines the direction of the virtual object, and the interference of the Euler angle will also make the virtual object unstable and shake.

Contents of the invention

In view of the problems and deficiencies in the above-mentioned prior art, the object of the present invention is to provide a method for suppressing mutual interference of magnetic trackers in an augmented reality system. Different states use different filtering methods, which can suppress the interference of the magnetic transmitter to a certain extent, reduce the interference of the Euler angle, and make the change of each angle in the Euler angle appear smooth, suitable for grabbing virtual objects Large-scale interactive augmented reality system or virtual reality system as the main mode of interaction.

In order to achieve the above object, the present invention adopts the following technical idea: when interacting, the head and hands have their own movement characteristics, and different processing strategies are used for the movement tracking of the head and hands. The mobility of the hand is relatively large, but changing the position of the transmitter can make the hand very close to the transmitter, so that the interference received is very small, and the particle filter can achieve better results; the head is far away from the transmitter, and the interference There is a lot of interference. When interacting, the head is still or slowly moving most of the time. The system state model is easy to determine in the static state, and the optimal estimation Kalman filter is used; when moving slowly, the requirements for the real-time response of the system can be slightly reduced. Using the bilateral filtering method of intrinsic geometric quantity, the interference can be removed in close to real time, and the problem of removing the interference of Euler angles can be effectively solved; when the motion is fast, the filtering is stopped.

Technical solution technical scheme of the present invention is as follows:

A method for suppressing mutual interference of magnetic trackers in an augmented reality system. First, the magnetic tracker is fixed in a suitable position, and then according to the respective movement characteristics of the head and hands during interaction, different suppression methods are used for the tracking interference of the head and hands: The tracking interference of the opponent adopts particle filter; the tracking interference of the head first judges the state of the head, and then adopts different processing methods according to different states, using Kalman filtering when stationary; using intrinsic geometric bilateral filtering when moving slowly; fast moving When the filter is stopped, the specific steps are as follows:

(1) Fix the magnetic tracker, and place the transmitter directly below the active area of the hand; place the two receivers on the visor of the light-transmitting helmet and the back of the hand of the data glove respectively; the head and hands are directly above the magnetic transmitter Activities within the effective range of 0.4×0.5×0.45m; the transmitters are placed horizontally and the distance between each pair is not less than 1.8m.

(2) Read the head and hand motion trajectory data in the above two receivers, and judge whether the virtual object is captured. If it is captured, the interference data of tracking hands and the interference data of tracking head are processed separately; if not caught, only the interference data of tracking head is processed;

The method for processing the interference data of tracking hands is as follows: decompose the interference data into 6 one-dimensional vectors, then respectively apply the second-order ARP model as the system state model, and use the decomposed one-dimensional vectors as the observation model , use the SIR algorithm to perform particle filtering on the received value of the magnetic receiver, and output the spatial coordinates and Euler angles;

The described processing of tracking head interference data: judging the state of the head:

If the head is in a static state, the system state model is obtained, and the obtained 6DOF data is decomposed into two types: spatial coordinates and Euler angles, and then the spatial coordinates and Euler angles are filtered by centralized Kalman filtering. After the filtering is completed, directly Output space coordinates, Euler angles;

If the head is in a slow moving state, the interference data of the head is filtered by the intrinsic geometric quantity bilateral filtering method, and the spatial coordinates and Euler angles are directly output after the filtering is completed;

If the head is in a fast-moving state, first judge what the previous state is, if it is also a fast-moving state, stop filtering, and directly output the spatial coordinates and Euler angles without any other processing; if the previous state is static or slow In the moving state, stop filtering, and use the critical damping string to calculate the new velocity and position, and then output the space coordinates and Euler angles.

Compared with the prior art, the present invention has the following obvious outstanding substantive features and significant advantages: the method for suppressing the mutual interference of magnetic trackers in the above-mentioned augmented reality system, the method does not require the user to change or add hardware configuration, and is easy to operate, economical and practical , in an augmented reality system where a real hand grabs a virtual object, different filtering methods are used according to different states, which can have a certain inhibitory effect on the interference of the magnetic transmitter, reduce the interference of the Euler angle, and make each Euler angle The change of an angle appears smooth, and it is suitable for a large-scale interactive augmented reality system or virtual reality system with grabbing virtual objects as the main interaction method.

Description of drawings

1 is a flowchart of a method for suppressing mutual interference of magnetic trackers in an augmented reality system according to an embodiment of the present invention;

Fig. 2 is the flow chart of the intrinsic geometric quantity bilateral filtering method of the embodiment of the present invention;

Fig. 3 is a schematic diagram of placement of a magnetic tracker according to an embodiment of the present invention;

Figure 4 is a comparison diagram of the effect before and after the bilateral filtering of the intrinsic geometric quantity on the X-axis data;

Figure 5 is a comparison diagram of the effect before and after the bilateral filtering of the intrinsic geometric quantity to the Y-axis data;

Fig. 6 is a comparison diagram of the effect before and after the bilateral filtering of the intrinsic geometric quantity on the Z-axis data;

Figure 7 is a comparison diagram of the effect before and after the bilateral filtering of the intrinsic geometric quantity on the precession angle;

Fig. 8 is a comparison diagram of the effect before and after the bilateral filtering of the intrinsic geometric quantity to the interference of the nutation angle;

Fig. 9 is a comparison diagram of the effect before and after the bilateral filtering of the intrinsic geometric quantity removes the interference of the rotation angle.

Detailed ways

Embodiments of the present invention will be further described in detail below in conjunction with the accompanying drawings.

The embodiments of the present invention are described in detail below: this embodiment is implemented on the premise of the technical solution of the present invention, and detailed implementation is provided, but the protection scope of the present invention is not limited to the following embodiments.

As shown in Figure 1, a method for suppressing mutual interference of magnetic trackers in an augmented reality system, the specific steps are as follows:

(1), fix the magnetic tracker, as shown in Figure 3, the transmitter is placed directly below the active area of the hand; the two receivers are respectively placed on the brim of the light-transmitting helmet and the back of the hand of the data glove; the head and hand Activities within the effective range of 0.4×0.5×0.45m directly above the magnetic transmitter; the transmitters are placed horizontally and the distance between them is not less than 1.8m.

(2) Read the head and hand motion trajectory data in the above two receivers, and judge whether the virtual object is captured. If captured, then the processing of the interference data of the tracking hand and the interference data of the tracking head are processed respectively; if not caught, then only the processing of the interference data of the tracking head is transferred to step (4);

The described method for processing the interference data for tracking hands is as follows: decompose the interference data into 6 one-dimensional vectors, then respectively apply the second-order ARP model as the system state model, and use the decomposed one-dimensional vectors as the observation model, Use the SIR algorithm to perform particle filtering on the received value of the magnetic receiver, and output the spatial coordinates and Euler angles;

The method for processing the interference data of the tracking head is as follows: judging the state 8 of the head:

If the head is in a static state, the system state model is obtained, and the obtained 6DOF data is decomposed into spatial coordinates and Euler angles, and then the spatial coordinates and Euler angles are filtered by centralized Kalman filtering, and the spatial coordinates and Euler angles are directly output after the filtering is completed. Coordinates, Euler angles;

If the head is in a slow moving state, the interference data of the head is filtered by the intrinsic geometric quantity bilateral filtering method, and the spatial coordinates and Euler angles are directly output after the filtering is completed;

If the head is in a fast-moving state, first judge what the previous state is, if it is also a fast-moving state, stop filtering, and directly output the spatial coordinates and Euler angles without any other processing; if the previous state is static or slow In the moving state, stop filtering, and use the critical damping chord to calculate the new velocity and position, and then output the space coordinates and Euler angles.

The particle filter described in the above step (3) refers to the time update process in which the system state model represents the target state. The voluntary movement trend of moving targets is generally more obvious, and particle propagation can be a random movement process, which obeys the first-order ARP (autoregressive process) equation. However, since the state propagation of the target has speed or acceleration, the second-order ARP model is adopted, and the model is expressed as follows:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; B</span>}{\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}$

定义为上一时刻目标的位置；X_t、X_t-1、X_t-2为t、t-1、t-2时刻各粒子的位置；Bω_t是随机噪声；A_i为从实验估计的模型参数。in

Defined as the position of the target at the previous moment; X _t , X _t-1 , X _t-2 are the positions of each particle at time t, t-1, and t-2; Bω _t is random noise; A _i is estimated from the experiment Model parameters.

The observation model is the value taken directly from the magnetic tracking receiver. After the model is determined, the SIR algorithm is used for filtering. The whole process is: draw random samples according to the prior conditional probability p(x ₀ |y ₀ ), n is the number of random samples, n=100 in this project. From the system transfer equation, considering the random samples, the predicted samples can be obtained. Calculate new weights based on observed values and predicted samples. Based on the new weights, calculate the posterior probability. The sample is redrawn with this posterior probability, and an estimate of the state is obtained.

The centralized Kalman filter described in the above step (5) refers to: the spatial coordinates and the Euler angles are filtered by using the centralized Kalman filter fusion structure. Since the target is considered to be basically stationary, the value of the receiver at the current moment is equal to the value at the previous moment. The state equation is:

S(τ)=S(τ-1)

Among them, S(τ) is the state vector at time τ.

The measured value is the value taken by the receiver, and the measurement equation is:

T(τ)=S(τ)+v(τ)

Wherein, T(τ)=[T _x (τ), _Ty (τ), T _z (τ)] ^T is the value of the receiver detected at the time of τ in X, Y, and Z directions; v( τ) is the measurement process noise matrix, v(τ)=[v _x (τ), v _y (τ), v _z (τ)] ^T .

Then Kalman's recursive algorithm can be applied to correct the predicted value:

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; S</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; S</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; K</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; S</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>$

为系统在τ时刻的状态矢量的滤波估计， 

为利用上一状态的预测估计，K(τ)为增益矩阵。in,

is the filtered estimation of the state vector of the system at time τ,

In order to use the prediction and estimation of the previous state, K(τ) is the gain matrix.

Intrinsic geometric quantity bilateral filtering method described in above-mentioned steps (6), as shown in Figure 2, its steps are as follows:

(6-1), extracting spatial coordinate information;

(6-2), transform the spatial coordinate information into an intrinsic geometric quantity representation: the side length of the side formed by connecting two points, and the angle between the side and the positive direction of the X-axis and the positive direction of the Z-axis are represented;

(6-3), then filter the angle between the side and the positive direction of the X-axis and the positive direction of the Z-axis;

(6-4), use the angle after filtering as a constraint condition, construct the objective function to invert the vertex of the curve;

(6-5), extracting Euler angles;

(6-6), for the incoming data, use the quasi-real-time model for dynamic processing;

(6-7). Filter both sides of the Euler angles.

The quasi-real-time model described in the above steps (6-6) is: the value of a certain moment of the incoming data is only related to the value of the previous 1 moment and the value of the next m moment. Supplement new observation data, discard the front-end data at the same time, keep the length of the sliding window as n, and re-filter the n data that is constantly updated.

Its advantage is that it does not require a large number of historical observations, and the estimated value can be adjusted according to the successive arrival of new data. If the frequency of sampling points is high, it can visually approach the effect of real-time measurement. Another advantage is that the iteration function is retained, and the number of iterations is increased by increasing the length of the sliding window. Because n=2+m, when n becomes larger, m also becomes larger. In this way, the number of data points to be filtered increases each time, and only one filter value is output each time, which means that the data points need to be filtered more than once, that is, iterations are realized, and m>2 at this time.

The Euler angle bilateral filtering described in the above steps (6-7) is: the Euler angle weight factor of a certain point is not only related to the geometric distance between its adjacent points, but also related to the difference between the Euler angle values . Whenever the system obtains a new coordinate estimate, Euler angle bilateral filtering is performed in combination with Euler angle.

Since the processed spatial coordinate data is relatively accurate data, it should be used for Euler angle filtering as much as possible. The image denoising method proposed by Tomas and Manduchi - bilateral filtering, is to replace the gray value of the current point with the weighted average of the gray value of the surrounding points. The weight factor is not only related to the geometric distance between two points, but also to them The gray value difference is related, so it is called bilateral filtering. We refer to this idea, that is, the Euler angle weight factor of a point is not only related to the geometric distance between its adjacent points, but also related to the difference between the Euler angle values.

According to the discrete expression of bilateral filtering, the Euler angle bilateral filtering expression can be changed to:

${\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}$

${\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}$

${\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>{{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}$

σ_c、σ_s、σ_c′、σ_s′、σ_c″、σ_s″分别是它们的自由参数，用户可以根据自己的需要适当选取。上面式子是按一阶领域展开的。where ψ _i , θ _i , and φ _i are the precession angle, nutation foot, and rotation angle in the Euler angles; e _i represents the geometric distance between two points; W _c , W _s , W _c ′, W _s ′, W _c ″, W _s ″ are Gaussian functions,

σ _c , σ _s , σ _c ′, σ _s ′, σ _c ″, and σ _s ″ are their free parameters respectively, and users can select them according to their needs. The above formula is expanded according to the first-order field.

σ _c (σ _c ′, σ _c ″), σ _s (σ _s ′, σ _s ″) have a greater impact on the anti-interference effect. The larger σ _c is, the greater the influence of distance in the weight factor is, and the larger σ _s is, the influence of Euler angle difference of adjacent vertices on the weight factor will be strengthened, and vice versa. Since the processed spatial coordinate information is relatively correct and trustworthy, we need to increase the ratio of σ _c to σ _s . The filtering of Euler angles is closely related to the filtering of spatial coordinates. Whenever the system obtains new coordinate estimates, it combines Euler angles with Euler angles for bilateral filtering. So the spatial coordinate filtering iterates a few times, it also iterates a few times. Since the first-order domain is used, the Euler angle estimate will be output one beat later than the coordinate estimate.

The practical use of this method has obvious effects. In order to further illustrate its effect, in the above embodiment, the distance between the transmitters is shortened to 1.5m, and the movement range of the receiver is increased

, the following tests are performed respectively:

We intercepted a trajectory (250 points) of hand movement, and extracted the disturbed 6DOF data and the processed 6DOF data respectively, and compared the effects of each dimension of the data before and after de-interference by means of charts.

1. Two filter curves before and after the X-axis data is de-interferenced by the intrinsic geometric quantity bilateral filter, a thin solid line is the original data curve, and a dotted line is the processed curve. The vertical axis is the data axis, indicating the X-axis coordinate value when the hand moves, in inches, and the horizontal axis is the time axis, in order of the appearance of these 250 points, the unit is the serial number, see Figure 4;

2. Two filter curves before and after the Y-axis data is de-interferenced by the bilateral filter of the intrinsic geometric quantity. A thin solid line is the original data curve, and a dotted line is the processed curve. The vertical axis is the data axis, indicating the Y-axis coordinate value when the hand moves, in inches, and the horizontal axis is the time axis, in order of the appearance of these 250 points, and the unit is the serial number, as shown in Figure 5;

3. Two filter curves before and after the Z-axis data is de-interferenced by the intrinsic geometric quantity bilateral filter, a thin solid line is the original data curve, and a dotted line is the processed curve. The vertical axis is the data axis, indicating the Z-axis coordinate value when the hand moves, in inches, and the horizontal axis is the time axis, in order of the appearance of these 250 points, and the unit is the serial number, as shown in Figure 6;

4. The two filtering curves before and after the precession angle is removed by bilateral filtering of intrinsic geometric quantities, one thin solid line is the original data curve, and the other is the dashed line after processing. Among them, the vertical axis is the data axis, indicating the precession angle value when the hand moves, the unit is angle, and the horizontal axis is the time axis, in order of the appearance of these 250 points, the unit is the serial number, see Figure 7;

5. Two filtering curves before and after the nutation angle is removed by bilateral filtering of intrinsic geometric quantities. One thin solid line is the original data curve, and the other dashed line is the processed curve. Among them, the vertical axis is the data axis, indicating the nutation angle value when the hand moves, the unit is angle, and the horizontal axis is the time axis, in order of the appearance of these 250 points, the unit is the serial number, see Figure 8;

6. Two filtering curves before and after removing the interference of the rotation angle by the bilateral filtering of the intrinsic geometric quantity, a thin solid line is the original data curve, and a dotted line is the processed curve. Among them, the vertical axis is the data axis, indicating the rotation angle value when the hand moves, the unit is angle, and the horizontal axis is the time axis. The order of appearance of these 250 points is the order, and the unit is the serial number, as shown in Figure 9.

From the above-mentioned Figures 4, 5, 6, 7, 8, and 9, it can be clearly seen that the movement on each axis of the space coordinate system and the change of each angle in the Euler angle appear smooth and stable.

Claims (4)

1. A method for restraining mutual interference of magnetic trackers in an augmented reality system is characterized in that the magnetic trackers are fixed at proper positions, and then different restraining methods are adopted for the tracking interference of a head and a hand according to respective motion characteristics of the head and the hand during interaction: particle filtering is adopted for tracking interference of the opponent; for tracking interference of the head, judging the state of the head, adopting different processing methods according to different states, and adopting Kalman filtering when the head is static; adopting internal geometric quantity bilateral filtering during slow movement; stopping filtering when the mobile terminal moves fast, and specifically comprising the following steps:

(1) fixing the magnetic tracker, and placing the emitter right below the moving area of the hand; the two receivers are respectively arranged on the brim of the light-transmitting helmet and the back of the hand of the data glove; the head and the hand move in an effective range of 0.4 multiplied by 0.5 multiplied by 0.45m right above the emitter; the emitters are horizontally arranged and the distance between every two emitters is not less than 1.8 m;

(2) reading head and hand motion track data in the two receivers, judging whether a virtual object is grabbed, and if the virtual object is grabbed, respectively processing interference data of a tracking hand and interference data of a tracking head; if the head part is not caught, only the interference data of the tracking head part is processed;

the processing method of the interference data of the tracking hand comprises the following steps: decomposing interference data into 6 one-dimensional vectors, respectively applying a second-order ARP model as a system state model, taking the decomposed one-dimensional vectors as an observation model, performing particle filtering on the received values of a receiver by adopting an SIR algorithm, and outputting a spatial coordinate and an Euler angle;

the processing method for the interference data of the tracking head comprises the following steps: judging the state of the head:

if the head is in a static state, a system state model is obtained, the obtained 6DOF data are decomposed into two types of space coordinates and Euler angles, then Kalman filtering is respectively carried out on the space coordinates and the Euler angles by adopting centralized Kalman filtering, and the space coordinates and the Euler angles are directly output after the filtering is finished;

if the head is in a slow moving state, filtering interference data of the head by adopting an internal geometric quantity bilateral filtering method, and directly outputting a space coordinate and an Euler angle after filtering is finished; the method for bilateral filtering of the intrinsic geometric quantity comprises the following steps:

a. the spatial coordinate information is extracted out of the space,

b. converting the spatial coordinate information into an intrinsic geometric representation: the side length of the side formed by connecting the two points is represented by the included angle between the side and the positive direction of the X axis and the positive direction of the Z axis;

c. then filtering the included angles between the edge and the positive directions of the X axis and the Z axis;

d. constructing an objective function to reversely solve the vertex of the curve by taking the filtered angle as a constraint condition;

e. extracting an Euler angle;

f. for the data coming successively, using a quasi-real-time model to perform dynamic processing;

g. bilateral filtering is carried out on the Euler angle;

the quasi-real-time model is as follows: for the value of the data coming successively at a certain moment, the value is only related to the values of the first 1 moment and the last m moments, new observed data are continuously supplemented as the sampling is carried out, and the most new observed data are discarded

The data at the front end keeps the length of the sliding window as n, and filtering is carried out again aiming at n data which are continuously updated;

if the head is in a fast moving state, judging what the previous state is, if the head is also in the fast moving state, stopping filtering, and directly outputting a space coordinate and an Euler angle without other processing; and if the former state is a static state or a slow moving state, stopping filtering, calculating new speed and position by using the critical damping string, and outputting the space coordinate and the Euler angle.

2. The method for suppressing the mutual interference of the magnetic trackers in the augmented reality system according to claim 1, wherein the particle filtering is a time updating process of a system state model representing a target state, a second-order ARP model is adopted, and a model expression is as follows:

wherein,

defining the position of the target at the last moment; x_t、X_t-1、X_t-2The positions of the particles at the moments t, t-1 and t-2 are shown; b omega_tIs random noise; a. the_iAre model parameters estimated from experiments.

3. The method of claim 1, wherein the centralized kalman filtering is performed by using a centralized kalman filtering fusion structure to filter the spatial coordinates and the euler angles, and the equation of state is as follows:

S(τ)＝S(τ-1)

wherein S (tau) is a state vector at time tau,

the measured value is the value obtained by the receiver, and the measurement equation is as follows:

T(τ)＝S(τ)+v(τ)

wherein T (τ) ═ T_x(τ)，T_y(τ)，T_z(τ)]^TThe values in X, Y, Z for the receiver detected at time τ; v (τ) is the measurement process noise matrix, v (τ) ═ v_x(τ)，v_y(τ)，v_z(τ)]^T，

Then, the prediction value can be corrected by applying a Kalman recursion algorithm:

wherein,for a filtered estimation of the state vector of the system at time instant tau,

to utilize the predicted estimate of the last state, K (τ) is the gain matrix.

4. The method of claim 1, wherein the euler angle bilateral filtering is: the Euler angle weight factor of a certain point is not only related to the geometrical distance between adjacent points but also related to the difference between Euler angle values, and the Euler angle bilateral filtering is carried out by combining the Euler angles every time the system obtains a new coordinate estimation value.

Images