A large-scale network multi-target tracking method based on data compression

Abstract

The application discloses a large-scale network multi-target tracking method based on data compression, and comprises the following steps: 1, constructing a mathematical model of a large-scale network under coupling attack; 2, designing a prediction-estimator; 3, for the kappa unit in the large-scale network, the prediction-estimator is used to obtain the covariance related to the prediction error; 4, using the undetermined parameter Y κ,u+1 in the prediction-estimator to determine the covariance related to the prediction error; 5, substituting Y κ,u+1 into the prediction-estimator equation to obtain the accurate estimation value at the u+1 moment; judging whether u+1<Delta is established or not, if yes, executing 6; 6, according to the and Y κ,u+1 , further obtaining the covariance related to the estimation error covariance, and setting u=u+1, continuing to execute 2 until u+1=Delta is established. The method can avoid the problems of communication network jamming and large tracking error induced by coupling attack, and provides valuable theoretical reference for communication network, data processing and target tracking.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of information technology, and relates to a target tracking method, in particular to a large-scale network multi-target tracking method based on data compression, which is used to solve problems such as multi-target tracking and positioning, motion state estimation and the like. BACKGROUND

[0002] Large-scale networks can be seen everywhere in real life and have been widely applied to fields such as traffic systems, intelligent power systems and brain networks, and the large-scale network multi-target tracking technology based on data compression is one of components of a new generation of information technology, involving communication network technology and big data processing and the like.

[0003] With the gradual increase of actual demand, the data volume will continue to increase. Since the communication network transmission data and processing data capacity is limited in practice, the data compression technology is one of feasible data processing means, and processing data based on a given data compression criterion not only reduces the data packet size but also guarantees the reliability of data transmission. On the other hand, in the large-scale network, each target is connected to each other through a wireless network, at this time, it is easy to attract the attention of malicious attackers and interfere with and tamper with the transmitted information, which will seriously reduce the implementation effect of the multi-target tracking technology. Therefore, based on the constructed coupling attack prediction-estimation equation, the influence of the coupling attack on the tracking accuracy is comprehensively analyzed.

[0004] The existing multi-target tracking technology lacks the comprehensive influence of the data compression strategy and the coupling attack under the framework of the large-scale network, thereby causing problems such as poor tracking effect and low technical level. SUMMARY

[0005] In order to solve the problem that the multi-target tracking technology of the large-scale network cannot simultaneously consider the influence of the data compression strategy and the coupling attack, resulting in problems such as invalidation of the designed prediction-estimator and poor target tracking effect, the application provides a large-scale network multi-target tracking method based on data compression. The method takes the development of a new generation of information technology as the goal, gives a development scheme of the large-scale network multi-target tracking technology based on data compression, can avoid problems such as communication network congestion and large tracking error induced by the coupling attack, and provides valuable theoretical reference for communication networks, data processing and target tracking and the like.

[0006] The object of the application is realized by the following technical scheme:

[0007] A large-scale network multi-target tracking method based on data compression comprises the following steps:

[0008] Step one, a mathematical model of the large-scale network under the coupling attack is constructed, and the dynamic model of the κth unit in the large-scale network is described as follows:

[0009]

[0010] y κ,u =C κ,u x κ,u +σ κ,u ,

[0011] where H denotes the total number of units in the large-scale network, x κ,u represents the model state mainly describing the dynamic characteristics of the κth unit at time u, x κ,u+1 represents the model state mainly describing the dynamic characteristics of the κth unit at time u+1, B κ,u and C κ,u denote the model noise driving matrix and the measurement distribution matrix of the κth unit at time u, θ κ,u and σ κ,u denote the model state noise and the sensor output noise of the κth unit at time u, g(x κ,u ) is a continuous nonlinear function satisfying the initial condition g(0) = 0, ω κκ,u characterizes the self-coupling relationship of the κth unit at time u, ω κρ,u characterizes the coupling relationship between the κth unit and the ρth unit at time u, Π describes the internal coupling relationship, η ρ,u describes the coupling attack occurring in the process of data transmission from the ρth unit to the κth unit at time u, y κ,u denotes the sensor measurement of the κth unit at time u.

[0012] Step two, design a prediction-estimator for the large-scale network model given in step one, the specific steps are as follows:

[0013] Step two one, for the large-scale network under coupling attack, construct a prediction-estimator I with the following form:

[0014]

[0015] where x denotes the dynamic estimation value of the κth unit at time u, x denotes the dynamic prediction value of the κth unit at time u for time u+1.

[0016] Step two two, based on the sensor measurement, data compression data and prediction-estimator I, construct a prediction-estimator II with the following form:

[0017]

[0018] where C κ,u+1 denotes the measurement distribution matrix of the κth unit at time u+1, denotes the output compressed value of the κth unit at time u + 1, denotes the dynamic estimate of the κth unit at time u + 1, Y κ,u+1 the pending parameter in the predictor II at time u + 1;

[0019] Step three, for the κth unit in the large-scale network, obtain the covariance related to the prediction error by means of the predictor designed in step two

[0020]

[0021] where o1 is the first scaling parameter, o2 is the second scaling parameter, o3 is the third scaling parameter, o4 is the fourth scaling parameter, o5 is the fifth scaling parameter, o6 is the sixth scaling parameter, o7 is the seventh scaling parameter, o8 is the eighth scaling parameter, and denote the inverse of o1, o2, o3, o4, o5, o6, o7 and o8, respectively, Φ and denote the known real matrix and the upper bound of the attack signal, respectively, I denotes the identity matrix with appropriate dimensions, denotes the mathematical expectation of the random variable of the network attack in the process of transmitting data from the jth unit to the κth unit, ω κj,u describes the coupling relationship between the jth unit and the κth unit, denotes the covariance matrix of the κth unit with respect to the dynamic estimation error at time u, denotes the upper bound of the covariance related to the prediction error, Φ T , Π T , and denote the transpose of Φ, Π, and B κ,u , respectively, and denote the square of ω κκ,u and ω κρ,u , respectively, denotes the trace operation of , i.e., the sum of the diagonal elements of the matrix

[0022] Step four, determine the pending parameter Y κ,u+1 in the predictor by means of in step three

[0023]

[0024] where o9, o 10 , and Ξ κ,u+1respectively represent the ninth scaling parameter, the tenth scaling parameter, the data compression error upper bound and the sensor output noise covariance matrix, and respectively represent the inverse of o9 and represent the transpose of C κ,u+1 is the square of represents the inverse of

[0025] Step five, substituting the undetermined parameter Y κ,u+1 into the prediction-estimation equation to obtain the accurate estimation value at u+1 time At this time, it is judged whether the inequality u+1<Δ is established, wherein Δ represents the total prediction-estimation step length, if the inequality u+1<Δ is established, step six is executed, otherwise the algorithm is ended;

[0026] Step six, further obtaining the covariance upper bound related to the estimation error covariance according to the Y and Y κ,u+1 determined in step three and step four

[0027]

[0028] wherein, represents the covariance matrix of the κth unit about the dynamic estimation error at u+1 time, represents the square of (I-Y κ,u+1 C κ,u+1 ) T and respectively represent the transpose of (I-Y κ,u+1 C κ,u+1 ) and Y κ,u+1

[0029] Let u=u+1, and step two is continuously executed until u+1=Δ is satisfied.

[0030] Compared with the prior art, the present application has the following advantages:

[0031] ​​​​​The application focuses on a new generation of information technology, solves the problem of large-scale network multi-target tracking technology development based on data compression, and considers the overall influence of data compression strategy and coupling attack on tracking technology. Under the variance constraint index, the prediction-estimation equation depending on the coupling attack probability is constructed, and the specific form of the estimation parameter is determined under the demand performance. Compared with existing achievements, the large-scale network multi-target tracking method developed by the application considers the fact that the communication network resources are limited and network attacks occur frequently, has certain adaptability and robustness, etc., and solves the problems of poor tracking effect and even divergence of existing technologies under the large-scale network framework. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is the flow chart of the multi-target tracking technology of the large-scale network under data compression of the application;

[0033] Figure 2 is the motion trajectory x of the first target component 1 in the large-scale network 11,u and its tracking trajectory wherein represents the trajectory state of the first target, and "---" represents the tracking trajectory given by the prediction-estimation method;

[0034] Figure 3 is the motion trajectory x of the first target component 2 in the large-scale network 12,u and its tracking trajectory

[0035] Figure 4 is the motion trajectory x of the second target component 1 in the large-scale network 21,u and its tracking trajectory

[0036] Figure 5 is the motion trajectory x of the second target component 2 in the large-scale network 22,u and its tracking trajectory

[0037] Figure 6 is the motion trajectory x of the third target component 1 in the large-scale network 31,u and its tracking trajectory

[0038] Figure 7 is the motion trajectory x of the third target component 2 in the large-scale network 32,u and its tracking trajectory DETAILED DESCRIPTION

[0039] The technical solutions of the present application are further described below with reference to the drawings, but are not limited thereto, and any modification or equivalent replacement of the technical solutions of the present application without departing from the spirit and scope of the present application shall be covered in the protection scope of the present application.

[0040] The present application provides a large-scale network multi-target tracking method based on data compression, which mainly deals with the problem of multi-target tracking of large-scale network that cannot be solved by existing prediction-estimation methods, especially the problems such as poor prediction-estimation performance and even divergence when model nonlinearity, coupled attack and data compression exist simultaneously. Figure 1 As shown in the figure, the method comprises the following steps:

[0041] Step 1, constructing a mathematical model of large-scale network under coupled attack.

[0042] The dynamic model of the κth unit in the large-scale network is described as follows:

[0043]

[0044] y κ,u =C κ,u x κ,u +σ κ,u , (2)

[0045] where H represents the total number of units in the large-scale network, represents the model state mainly describing the dynamic characteristics of the κth unit at time u, represents the model state mainly describing the dynamic characteristics of the κth unit at time u+1, represents the set of all column vectors of dimension x, and respectively represent the model noise driving matrix and the measurement distribution matrix of the κth unit at time u, and respectively represent the set of all matrices of dimension x×θ and y×x. and respectively represent the model state noise and the sensor output noise of the κth unit at time u, and the corresponding positive definite covariance matrices are Θ κ,u > 0 and Ξ κ,u > 0, where and respectively represent the set of all column vectors of dimension θ and σ.g(x κ,u ) is a continuous nonlinear function satisfying the initial value condition g(0) = 0 and satisfying the inequality [g(z)-g(f)] T [g(z)-g(f)]≤(z-f) T Φ TΦ(z-f) where z and f are vectors of appropriate dimensions, Φ is a known real matrix, [g(z)-g(f)] T , (z-f) T and Φ T denote the transpose of [g(z)-g(f)], (z-f) and Φ, respectively. ω κκ,u characterizes the self-coupling relationship of the κ-th unit at time u, ω κρ,u characterizes the coupling relationship between the κ-th unit and the ρ-th unit at time u, and Π describes the internal coupling relationship. ρ,u describes the coupling attack that occurs in the process of data transmission from the ρ-th unit to the κ-th unit at time u, which is specifically expressed as η ρ,u = x ρ,u + λ ρ,u ζ ρ,u , x ρ,u denotes the dynamic characteristics of the ρ-th unit at time u, λ ρ,u denotes a random variable that occurs in the process of data transmission from the ρ-th unit to the κ-th unit, and is described by a Bernoulli random variable, whose mathematical expectation is λ ρ,u = 0 indicates that no coupling attack occurs, otherwise a coupling attack occurs. ζ ρ,u is a random attack signal sent by a hacker in the process of data transmission from the ρ-th unit to the κ-th unit at time u, and satisfies where denotes the transpose of ξ ρ,u , and is a known non-negative real number that characterizes the upper bound of the attack signal.

[0046] Shared networks are one of the main ways of transmitting data. In order to further improve the security and reliability of network data transmission, data compression technology is introduced. denotes that the sensor measurement of the κ-th unit at time u satisfies y κ,u = [y κ1,u , y κ2,u ,..., y κy,u ] T , where y κ1,u denotes the first component of y κ,u , y κ2,u denotes the second component of y κ,u , and so on y κy,u denotes the y-th component of y κ,u , denotes the set of column vectors of dimension y, [y κ1,u , y κ2,u ,..., y κy,u ] T denotes [y κ1,u , y κ2,u ,..., yκy,u ] of transpose. The compressed set is described by Ω, specifically Ω = {y κ,u ||y κε,u |≤β κ , ε = 1, 2,..., y} if y κ,u is the εth component of y κε,u , the absolute value of y κ is limited by β κ , then the data is processed by the data compression technique, β κ,u is a non-negative real number representing the data compression threshold. There exists a real number sequence belonging to the set where is a positive integer, such that the compressed value satisfies the following equation:

[0047]

[0048] The compression error satisfies the following inequality:

[0049]

[0050] where, denotes the matrix 2-norm of the data compression error, represents the sensor observation y κκ,u , and is the arithmetic square root of the dimension y.

[0051] Step two, design a predictor-estimator for the large-scale network model given in step one.

[0052] First, for the large-scale network under the coupling attack, a predictor-estimator I with the following form is constructed:

[0053]

[0054] where, ω κκ,u , Π, H, and ω κρ,u are defined in equation (1) respectively represent a nonlinear function, a self-coupling relationship, an internal coupling relationship, the total number of units in the large-scale network, the mathematical expectation of the coupling attack, and the coupling relationship of different units. represents the dynamic estimation value of the κth unit at time u, represents the dynamic prediction value of the κth unit at time u for time u+1.

[0055] Second, based on the sensor measurement, the data compression data, and the predictor-estimator I, a predictor-estimator II model with the following form is constructed:

[0056]

[0057] where, denotes the measurement distribution matrix of the k-th unit at time u+1, denotes the output compressed value of the k-th unit at time u+1, denotes the dynamic prediction value of the k-th unit at time u for time u+1, denotes the dynamic estimation value of the k-th unit at time u+1, κ,u+1 the pending parameters in the predictor II at time u+1.

[0058] Step three, for the k-th unit in the large-scale network, obtain the covariance related to the prediction error by means of the predictor designed in step two

[0059] The specific expression of the covariance related to the prediction error is:

[0060]

[0061] where, o1 is the first scaling parameter, o2 is the second scaling parameter, o3 is the third scaling parameter, o4 is the fourth scaling parameter, o5 is the fifth scaling parameter, o6 is the sixth scaling parameter, o7 is the seventh scaling parameter, o8 is the eighth scaling parameter, and denote the reciprocal of o1, o2, o3, o4, o5, o6, o7 and o8, respectively. κκ,u 、 κρ,u 、 B κ,u and Θ κ,u denote the self-coupling relationship of the p-th unit at time u, the mathematical expectation of the coupling attack, the mathematical expectation of the coupling attack and the coupling relationship of different units, the dynamic estimation value of the k-th unit, the model noise driven matrix of the k-th unit and the covariance matrix of the model state noise, respectively, which are consistent with the previously defined parameter meanings. H, Φ, Π and denote the total number of units in the large-scale network, the known real matrix, the internal coupling relationship and the attack signal upper bound, respectively. I denotes a unit matrix with appropriate dimensions, denote the mathematical expectation of the random variable of the network attack occurring in the process of the j-th unit transmitting data to the k-th unit, ω κj,u describes the coupling relationship between the j-th unit and the k-th unit. denotes the covariance matrix of the k-th unit with respect to the dynamic estimation error at time u, denotes the covariance upper bound related to the prediction error. Φ T , Π T 、 and ​They represent Φ, Π, and and B κ,u transpose, and Representing ω κκ,u and ω κρ,u The square of, express The trace operation, i.e., matrix The sum of the diagonal elements.

[0062] Step 4: Utilize the information from Step 3 Determine the undetermined parameter Y in the forecast-estimater κ,u+1 .

[0063] Undetermined parameter Υ κ,u+1 The structural form is as follows:

[0064]

[0065] Among them o9, o 10 , C κ,u+1 , and Ξ κ,u+1 Consistent with previous definitions, these represent the ninth scaling parameter, the tenth scaling parameter, the upper bound of the covariance related to the prediction error, the measurement distribution matrix, the upper bound of the data compression error, and the sensor output noise covariance matrix, respectively. and They represent o9 and The reciprocal, C represents κ,u+1 transpose, yes The square of , where I is the identity matrix. express The reverse.

[0066] Step 5: The undetermined parameter Y determined in Step 4... κ,u+1 Substitute the equations into the predictor-estimator equations to obtain the accurate estimate at time u+1. At this point, determine whether the inequality u+1 < Δ is true, where Δ represents the total prediction-estimation step size. If u+1 < Δ is true, proceed to step six; otherwise, the algorithm execution ends.

[0067] Step Six: Based on the determinations made in Steps Three and Four respectively and Υ κ,u+1 Further obtain the upper bound of the covariance related to the dynamic estimation error covariance. Let u = u+1, and continue to execute step two until u+1 = Δ is satisfied.

[0068] Upper bound of covariance related to dynamic estimation error It must satisfy the following form:

[0069]

[0070] wherein, denotes the covariance matrix of the dynamic estimation error of the kth unit at time u+1, o9 and o 10 denote the ninth and tenth scaling parameters, respectively, and denote the transpose of o9 and o 10 , respectively. κ,u+1 , C κ,u+1 and Ξ κ,u+1 are consistent with the previous definitions, and denote the prediction-estimator II pending parameters, measurement distribution matrix and sensor output noise covariance matrix for the kth unit at time u+1, respectively. is the covariance upper bound related to the prediction error defined in (6), denotes the data compression error upper bound, and I denotes the unit matrix with appropriate dimensions. denotes the square of (I-Υ κ,u+1 C κ,u+1 ) T and denote the transpose of (I-Υ κ,u+1 C κ,u+1 ) and Υ κ,u+1 , respectively.

[0071] The main theories of steps one to six include:

[0072] The present application develops a large-scale network multi-target tracking method based on data compression, and further gives a novel prediction-estimation method under the model nonlinearity, coupling attack and data compression strategy. Due to the full consideration of factors such as model nonlinearity, coupling attack and data compression strategy, it is difficult to obtain the covariance matrix form related to the prediction error and the estimation error. Therefore, with the help of random analysis, matrix differential theory and information processing theory, the corresponding covariance upper bound expression is obtained, and the pending parameters in the prediction-estimation equation are optimized and designed.

[0073] The present application takes the minimum mean square error as the constraint target, that is, the selected prediction-estimation equation parameters can ensure the minimization of the estimation error covariance upper bound in real time, which meets the high-precision requirement of multi-target tracking in actual engineering.

[0074] According to the proposed prediction-estimation method, the multi-target tracking effect of the present application is demonstrated by using case one:

[0075] Case one:

[0076] The application can solve the multi-target tracking problem of large-scale network, the case considers three mobile targets, the prediction-estimation algorithm executes step length of 100, and the model parameters are set as follows:

[0077]

[0078] C 1,u =[0.04 1.02], C 2,u =[0.04-0.1cos(0.2u) -0.28],

[0079] C 3,u =[0.02 -0.04], ω κκ,ι =-0.024, ω κρ,ι =0.012 (κ≠ρ).

[0080] The initial values of the algorithm execution are selected as follows:

[0081] Wherein and are the initial values of the first target, the second target and the third target in the large-scale network respectively, [2 2] T , [4 4] T and [8 8] T respectively represent the transposition of

[22] , [4 4] and [8 8], and the initial value of the upper bound of the estimation error covariance satisfies:

[0082]

[0083] Wherein and respectively represent the initial value of the upper bound of the estimation error covariance of the first target, the second target and the third target.

[0084] The first scaling parameter to the tenth scaling parameter is selected as: o1=0.8, o2=0.2, o3=0.1, o4=0.1, o5=0.3, o6=0.1, o7=0.2, o8=0.2, o9=0.1 and o 10 =0.1, the coupling attack probability The noise covariance is selected as Θ 1,u =Θ 2,u =Θ 3,u =0.1 and Ξ 1,u =Ξ 2,u =Ξ 3,u =2. The nonlinear function satisfies:

[0085]

[0086] Prediction and estimation effect:

[0087] Multi-target tracking effect demonstration as follows Figures 2-7 As shown, it is not difficult to see that for large-scale networks, the multi-target tracking method based on data compression of this invention has significant effects and achieves state estimation within an acceptable accuracy range.

Claims

1. A large-scale network multi-target tracking method based on data compression, characterized in that The method comprises the following steps: Step one, construct the mathematical model of large-scale network under coupling attack, the dynamic model of the first unit in large-scale network is described as follows: Step one, construct the mathematical model of large-scale network under coupling attack, the dynamic model of the first unit in large-scale network is described as follows: in, This represents the total number of units in a large-scale network. The representative model state description is in Time of the first The dynamic characteristics of each unit The representative model state description is in Time of the first The dynamic characteristics of each unit and They represent in Time of the first The model noise driving matrix and measurement distribution matrix of each unit. and They represent in Time of the first The model state noise and sensor output noise of each unit, It satisfies the initial value conditions. Continuous nonlinear function, Depicted in time The self-coupling relationship of the unit. portrayal Time of the first The unit and the first The coupling relationship of each unit, Describe the internal coupling relationship. Description in Time of the first Unit to the first Coupling attacks that occur during unit data transmission Indicates in Time for the first Sensor measurements per unit; Step two, design a prediction-estimator for the mathematical model of the large-scale network given in step one, and the specific steps are as follows: Step two (1), for the large-scale network under the coupling attack, a prediction-estimator I with the following form is constructed: wherein, represents at the moment a dynamic estimate value of the unit, represents a dynamic prediction value of the unit at the moment to the moment, represents the first unit, a random variable of network attack occurring in the data transmission process of the first unit to the second unit, described by a Bernoulli random variable, and its mathematical expectation is ; Step two (2), based on the sensor measurement, the data compression data and the prediction-estimator I, a prediction-estimator II with the following form is constructed: wherein, represents the measurement distribution matrix of the k-th unit at the time instant t, represents the output compression value of the k-th unit at the time instant t, represents the dynamic estimate value of the k-th unit at the time instant t, represents the measurement distribution matrix of the k-th unit at the time instant t, represents the output compression value of the k-th unit at the time instant t, represents the dynamic estimate value of the k-th unit at the time instant t, represents the measurement distribution matrix of the k-th unit at the time instant t, represents the output compression value of the k-th unit at the time instant t, represents the dynamic estimate value of the k-th unit at the time instant t, represents the measurement distribution matrix of the k-th unit at the time instant t, represents the output compression value of the k-th unit at the time instant t, Step three, for each cell in the large-scale network, obtain the covariance associated with the forecast error using the predictor-corrector designed in step two : wherein is a first scaling parameter, is a second scaling parameter, is a third scaling parameter, is a fourth scaling parameter, is a fifth scaling parameter, is a sixth scaling parameter, is a seventh scaling parameter, is an eighth scaling parameter, , , , , , , and denote the inverse of , , , , , , and respectively, and denote known real matrices and an attack signal upper bound respectively, denotes an identity matrix of appropriate dimension, denotes the mathematical expectation of the network attack random variable in the process of data transmission from the th unit to the th unit, describes the coupling relationship between the th unit and the th unit, denotes the covariance matrix of the dynamic estimation error of the th unit at time , , , and denote the transpose of , , and respectively, and denote the square of and respectively, denotes the trace operation of , i.e. the sum of the diagonal elements of the matrix ; and denote the sum sensor output noise of the th unit at time , is the model state noise covariance matrix; Step four, using the result of step three in Determining the pending parameters in the predictor II : wherein , , and denote a ninth scaling parameter, a tenth scaling parameter, an upper bound on the data compression error, and a sensor output noise covariance matrix, respectively, and denote the inverse of and , denotes the transpose of , is the square of , denotes the inverse of ; Step five, determine the unknown parameters to the equations of the predictor II, obtain the accurate estimate of the time instant At this time, determine whether the inequality is valid, where denotes the total step of the prediction and estimation, if is valid, then execute step six, otherwise the algorithm ends. Step six, determining the covariance upper bound related to the estimation error covariance from step three and step four respectively and further obtaining a covariance upper bound related to an estimation error covariance : wherein represents at the time instant the unit about the covariance matrix of the dynamic estimation error, represents the square of and represent the transpose of and respectively; Let , continue to perform step two until .

2. The large-scale network multi-target tracking method based on data compression according to claim 1, characterized in that The Inequality where, and denote vectors of appropriate dimension, is a known real matrix, , and denote the transpose of , and respectively.

3. The large-scale network multi-target tracking method based on data compression according to claim 1, characterized in that The , indicates the dynamic characteristics of the unit at the moment, indicates the random variable of the network attack occurring in the data transmission process from the unit to the unit, and the mathematical expectation is described by a Bernoulli random variable , wherein indicates that no coupling attack occurs, otherwise a coupling attack phenomenon occurs, at the moment, the random attack signal sent by a hacker in the data transmission process from the unit to the unit satisfies , wherein indicates the transpose of , is a known non-negative real number characterizing the upper bound of the attack signal.

4. The large-scale network multi-target tracking method based on data compression according to claim 1, characterized in that The ,in express The first component express The second component, express The One portion, express The transpose of; the compressed set is composed of Description, in specific form ,if The Each component The absolute value is limited by The data is then processed using compression techniques. A non-negative real number represents the data compression threshold; There exists a real sequence belonging to the set where is a positive integer, such that the compressed value satisfies the following equation: compression error satisfies the following inequality: wherein the matrix 2-norm representing the data compression error, representing the sensor observation dimension the arithmetic square root.

Images