Probability density hypothesis track generation method and system based on minimizing cross entropy

By minimizing the probability density assumption track generation method of cross entropy, the state estimation set is divided and optimized, which solves the difficulty of track generation of traditional multi-objective tracking methods when targets are adjacent or cross-moving, and achieves more accurate goals State division and track generation improve the reliability of high-level convergence and threat assessment.

CN114491413BActive Publication Date: 2024-11-01AVIATION WARFARE SERVICE COLLEGE OF NAVAL AVIATION UNIV OF THE CHINESE PEOPLES LIBERATION ARMY
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202210088269.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-25
Publication Date
2024-11-01
Estimated Expiration
2042-01-25

AI Technical Summary

Technical Problem

The traditional multi-objective tracking method has difficulties in dealing with target adjacent or cross-movement, and cannot effectively generate a complete target track, affecting subsequent high-level fusion and threat estimation.

Method used

The probability density hypothesis track generation method based on minimizing cross entropy is adopted. By dividing the state estimation set, the posterior probability density function is determined, and the minimum cross entropy optimization method is used to gradually optimize it to generate the optimal state division to form the various goals. track.

Benefits of technology

It effectively solves the problem of track generation when targets are adjacent or cross-moving, provides more accurate target state division and track generation, and improves the reliability of subsequent high-level fusion and threat assessment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114491413B_ABST
    Figure CN114491413B_ABST
Patent Text Reader

Abstract

The present invention relates to a probability density hypothesis track generation method and system based on minimizing cross entropy, comprising: obtaining a state estimation set based on observation values ​​observed by a sensor, and dividing the state estimation set to obtain an effective partition set; determining a posterior probability density function based on the effective partition set; obtaining a parameter probability density function set based on the effective partition set; based on the parameter probability density function set, using a minimum cross entropy optimization method, gradually optimizing the posterior probability density function to obtain an optimal posterior probability density function; obtaining an optimal state partition based on the optimal posterior probability density function; and generating tracks of each target based on the optimal state set. The present invention can better solve the problem of track generation when targets are adjacent or cross-moving.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of multi-target tracking, and in particular to a probability density hypothesis track generation method and system based on minimizing cross entropy. Background Art

[0002] The traditional multi-target tracking method based on data association transforms the multi-target tracking problem into a single target tracking problem, and naturally generates the target track after completing the single target filtering. Although the multi-target tracking method based on Sequential Monte Carlo method-Probability hypothesis density (SMC-PHD) filtering can complete the state estimation of multiple targets when the number of targets changes in real time and the measurement information is uncertain, it does not give the attribution of each target state, and it is impossible to form a complete target track, which brings many difficulties to the subsequent high-level fusion (target recognition, situation assessment and threat estimation, etc.).

[0003] For the track generation problem based on SMC-PHD filtering, there are mainly two methods:

[0004] The first method is to combine SMC-PHD filtering with traditional data association methods (such as MHT and JPDA) to generate target tracks. This method can be further divided into two categories. One is to use SMC-PHD filtering as a measurement preprocessing device, whose main function is to remove clutter and false alarms from the measurement set, and then use the remaining measurements as the input of the traditional data association method to complete multi-target tracking and form target tracks; the other is to use the state estimation results of SMC-PHD as the input of the traditional data association method to generate multi-target tracks.

[0005] The second method considers incorporating track management into the filtering process. While the SMC-PHD filtering is in progress, the target attributes of the state are marked by labeling the particles, thereby outputting a labeled state estimate and subsequently generating a target track.

[0006] The above two methods still have difficulties in tracking targets that are in adjacent or intersecting motion. Summary of the invention

[0007] In view of this, the present invention provides a probability density hypothesis track generation method and system based on minimizing cross entropy, abandons the traditional data association method, and reduces the track generation problem to the state estimation set partition problem, which can better solve the track generation problem when the targets are adjacent or crossing each other.

[0008] To achieve the above object, the present invention provides the following solutions:

[0009] A probability density hypothesis track generation method based on minimizing cross entropy, comprising:

[0010] Obtaining a state estimation set based on observation values ​​observed by the sensor, and dividing the state estimation set to obtain a valid partition set;

[0011] Determine a posterior probability density function based on the effective partition set; obtain a parameter probability density function set based on the effective partition set;

[0012] Based on the parameter probability density function set, the posterior probability density function is gradually optimized using a minimum cross entropy optimization method to obtain an optimal posterior probability density function;

[0013] Based on the optimal a posteriori probability density function, an optimal state division is obtained; and based on the optimal state division, a track of each target is generated.

[0014] Preferably, the state estimation set is obtained based on the observation values ​​observed by the sensor, and the state estimation set is divided to obtain a valid partition set, including:

[0015] A state estimation set is obtained based on the observations observed by the sensor;

[0016] The state estimation set is divided into a valid partition set; each state vector in the valid partition set Where: C K is the number of targets appearing in the monitoring area in the previous K moments, τ0 represents the set of all false estimates, τ i is the state estimation vector of the i-th target;

[0017] Each state vector in the valid partition set satisfies the following constraints:

[0018] 1) Element conservation: the elements remain constant before and after division,

[0019] Where: N k is the number of state estimates at time k;

[0020] 2) Elements do not intersect: Due to τ i Each estimate in is either classified as a false estimate or corresponds to at most one target,

[0021] Where: i represents the i-th target, j represents the j-th target;

[0022] 3) τ i Orderliness of state estimates in τ i is a set of elements sorted in ascending order of time.

[0023] where n l For target i in k l The number of state estimates at the time;

[0024] 4) Uniqueness: Since each state estimate corresponds to at most one target track,

[0025] 5) Measurability: Each target track corresponds to at least two target states, namely |τ i |≥2,i=1,2,…,C K .

[0026] Preferably, determining the posterior probability density function based on the effective partition set; obtaining the parameter probability density function set based on the effective partition set comprises:

[0027] Determining the posterior probability density function based on the valid partition set;

[0028] Define a weighted directed graph;

[0029] Obtaining a probability weight matrix of the weighted directed graph;

[0030] The parameter probability density function set is constructed based on the probability weight matrix and the valid partition set.

[0031] The present invention also provides a probability density hypothesis track generation system based on minimizing cross entropy, comprising:

[0032] A data partitioning module, which obtains a state estimation set based on observation values ​​observed by the sensor, and partitions the state estimation set to obtain a valid partition set;

[0033] A probability density module, which determines a posterior probability density function based on the effective partition set; and obtains a parameter probability density function set based on the effective partition set;

[0034] An optimization module, based on the parameter probability density function set, adopts a minimum cross entropy optimization method to gradually optimize the posterior probability density function to obtain an optimal posterior probability density function;

[0035] The track generation module obtains the optimal state division based on the optimal posterior probability density function; and generates the track of each target based on the optimal state division.

[0036] Preferably, the data partitioning module comprises:

[0037] A data acquisition unit obtains a state estimation set based on the observation values ​​observed by the sensor;

[0038] A data partitioning unit is configured to partition the state estimation set to obtain a valid partition set; each state vector in the valid partition set Where: C K is the number of targets appearing in the monitoring area in the previous K moments, τ0 represents the set of all false estimates, τ i is the state estimation vector of the i-th target;

[0039] Each state vector in the valid partition set satisfies the following constraints:

[0040] 1) Element conservation: the elements remain constant before and after division,

[0041] Where: N k is the number of state estimates at time k;

[0042] 2) Elements do not intersect: Due to τ i Each estimate in is either classified as a false estimate or corresponds to at most one target,

[0043] Where: i represents the i-th target, j represents the j-th target;

[0044] 3) τ i Orderliness of state estimates in τ i is a set of elements sorted in ascending order of time.

[0045] where n l For target i in k l The number of state estimates at the time;

[0046] 4) Uniqueness: Since each state estimate corresponds to at most one target track,

[0047] 5) Measurability: Each target track corresponds to at least two target states, namely |τ i |≥2,i=1,2,…,C K .

[0048] Preferably, the probability density module comprises:

[0049] A posterior probability density unit, which determines the posterior probability density function based on the valid partition set;

[0050] Directed graph unit, defines a weighted directed graph;

[0051] A weight matrix unit, which obtains a probability weight matrix of the weight directed graph;

[0052] A parameter probability density unit constructs the parameter probability density function set based on the probability weight matrix and the valid partition set.

[0053] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0054] The present invention relates to a probability density hypothesis track generation method and system based on minimizing cross entropy, including: obtaining a state estimation set based on observation values ​​observed by sensors, and dividing the state estimation set to obtain a valid partition set; determining a posterior probability density function based on the valid partition set; obtaining a parameter probability density function set based on the valid partition set; based on the parameter probability density function set, using a minimum cross entropy optimization method, gradually optimizing the posterior probability density function to obtain an optimal posterior probability density function; obtaining an optimal state partition based on the optimal posterior probability density function; generating tracks for each target based on the optimal state set. The present invention can better solve the problem of track generation when targets are in adjacent or intersecting motion. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0056] Figure 1 The flowchart of the method for generating a probability density hypothesis track based on minimizing cross entropy of the present invention is as follows;

[0057] Figure 2 This is a structural diagram of the track generation system based on the probability density hypothesis of minimizing the cross entropy of the present invention.

[0058] Explanation of symbols: 1-data partition module, 2-probability density module, 3-optimization module, 4-track generation module. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The purpose of the present invention is to provide a probability density hypothesis track generation method and system based on minimizing cross entropy, abandon the traditional data association method, reduce the track generation problem to the state estimation set partition problem, and better solve the track generation problem when the targets are adjacent or cross-moving.

[0061] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] Figure 1 The flowchart of the probability density hypothesis track generation method based on minimizing cross entropy of the present invention is shown in the figure. As shown in the figure, the present invention provides a probability density hypothesis track generation method based on minimizing cross entropy, comprising:

[0063] Step S1, obtaining a state estimation set based on the observation values ​​observed by the sensor, and dividing the state estimation set to obtain a valid partition set. Specifically:

[0064] A set of state estimates is obtained based on the observations observed by the sensors.

[0065] The state estimation set is divided into a valid partition set; each state vector in the valid partition set Where: C K is the number of targets appearing in the monitoring area in the previous K moments, τ0 represents the set of all false estimates, τ i is the state estimation vector of the i-th target.

[0066] Each state vector in the valid partition set satisfies the following constraints:

[0067] 1) Element conservation: the elements remain constant before and after division,

[0068] Where: N k is the number of state estimates at time k.

[0069] 2) Elements do not intersect: Due to τ i Each estimate in is either classified as a false estimate or corresponds to at most one target,

[0070] Where: i represents the i-th target, j represents the j-th target.

[0071] 3) τ i Orderliness of state estimates in τ i is a set of elements sorted in ascending order of time.

[0072] Where n lFor target i in k l The estimated number of states at time.

[0073] 4) Uniqueness: Since each state estimate corresponds to at most one target track,

[0074] 5) Measurability: Each target track corresponds to at least two target states, namely |τ i |≥2,i=1,2,…,C K .

[0075] Step S2, determining a posterior probability density function based on the effective partition set; and obtaining a parameter probability density function set based on the effective partition set.

[0076] Furthermore, the step S2 is specifically as follows:

[0077] We will find the optimal state partition ω * The problem of is transformed into the optimal estimation problem of ω in the sense of maximizing the posterior probability density function (Maximuma-Posteriori probability, MAP), that is,

[0078]

[0079] Where: is the posterior probability density function.

[0080] For any valid partition ω, assuming the number of targets at time k is n k , where the number of new targets is b k , the number of extinction targets is m k , the number of targets detected by the sensor is d k , the number of undetected targets is u k , the number of false estimates is f k , if the state estimate at this time is N k , then the above parameters have the following algebraic relationship:

[0081]

[0082] The posterior probability density function expression is as follows:

[0083]

[0084] Where: C0 is a constant that does not depend on the effective partition ω; N(x; μ, Δ) represents a Gaussian distribution with mean μ and variance Δ; τ i is the state estimation vector of target i, C is the number of all estimates in the valid partition ω excluding the false estimate τ0, is the step prediction of the state; is the innovation covariance; p m is the target extinction probability; p d is the target detection probability; p b is the target newborn probability; p f is the target false alarm probability.

[0085] Assuming the total number of estimates is N, then

[0086]

[0087] Define a weighted directed graph G = (V, E, P), where V is the set of nodes in the graph, E is the edge of the graph, and P is the weight matrix. For V and E, the expression is:

[0088]

[0089]

[0090] Where V max is the upper limit of the target speed; k2-k1≤2 means that only the correlation between the estimated values ​​of two consecutive frames is examined.

[0091] For a node without edge connection, we call it an isolated point, which is a false alarm by default. In actual processing, it is moved to the false track τ0. The problem of generating the target track is to divide the nodes in the graph so that the posterior probability density function maximum.

[0092] If N i , i = 1, ..., N represents the nodes of the graph. Since the state estimation points are time-series, that is, the graph G is directed, the node N i and N j The connected edge is represented as (i,j). To construct the parameter vector, the following definition is made:

[0093] 1) Node N i The probability of connecting the starting point of a certain edge in the graph G is p b (i).

[0094] 2) Node N i The probability of a terminal point on a certain edge in the connected graph G is p f (i).

[0095] 3) A path in the connection graph passes through node N i , and then through the node N j The probability is p ij .

[0096] From the above definition, we can see that for any N i ∈V, then:

[0097] p f (i)+∑p ij =1;

[0098] The above formula represents the node N in the connection graph i Appears as a terminal point or an intermediate node. Therefore, the probability weight matrix is

[0099]

[0100] In fact, P is a parameterized vector, corresponding to v in the parameterized probability density function. In order to solve the track association problem when the target crosses, consider integrating the target state information and heading information into the probability weight matrix. Assume that node N i and N j The corresponding moments are k1 and k2, and k2>k1, and the weight probability p of the edge in the connection graph ij Define as follows:

[0101]

[0102] In the formula, ||·||2 represents the 2-norm; Represents state estimation k2-k1-step prediction of θ ij Represents vector With vector The angle between ij The smaller the node N i and N j The greater the possibility of association; β1 and β2 are constants related to process noise and measurement noise; C i is a normalization parameter, and its value can be given by the following formula:

[0103]

[0104] After the definition of the probability weight matrix P is completed, a path τ=(i1,…,i m ) is:

[0105]

[0106] For a valid partition Assuming that the initial node set is Θ, the parameter probability density function is:

[0107]

[0108] Step S3, based on the parameter probability density function set, adopting the minimum cross entropy optimization method, gradually optimizes the posterior probability density function to obtain the optimal posterior probability density function.

[0109] Specifically, let Y be the target state space, S(·) be the performance function defined on Y, then the optimal value of the performance function γ * for:

[0110]

[0111] where y * is the target state when the performance function is optimal. To obtain the optimal value of the performance function γ * , assuming that f and h are m-dimensional parameter probability density functions associated with the performance function S(·), the cross entropy (also known as Kullback-Leibler distance) between them is defined as:

[0112]

[0113] In the formula, E f represents expectation, D(f|h) reflects the closeness of two probability density functions, and the process of finding the closest two probability density functions under certain constraints is the process of minimizing cross entropy. Assuming h(·) is known, both f and h are parameterized probability density functions, which can be represented by f(y,v) and h(y,u), v and u are parameter vectors, and the minimum cross entropy is represented by min D(f|h), and its performance constraints are:

[0114] E f S(Y) = γ;

[0115] Under the above constraints, the solution of min D(f|h) is:

[0116]

[0117] In the formula, v * is the parameter vector corresponding to the solution of the minimum cross entropy, λ is a constant, which can be obtained by the following formula

[0118]

[0119] Since for the continuous probability distribution f(y,v * ) sampling method, so if we want to * ) Sampling gradually approaches the global optimal y * It is difficult to achieve. Assume h(y,u) and f(y,v * ) is a multidimensional discrete distribution, Y=(Y1,…,Y n ) is a random vector and Y i From the set {a1,…,a n}, then the estimated

[0120]

[0121] Its minimum cross entropy solution is

[0122]

[0123] In the formula, I {·} represents an event, P represents probability, a k For the set {a1,…,a n}. Under the above assumptions, the following formula can be used instead:

[0124]

[0125] Construct parameter vector sequence v1,v2,…,v t , gradually approaching the global optimal y through iterative sampling * .

[0126] In this embodiment, the parameter probability density function set obtained in step S2 is gradually approached to the global optimum through iterative sampling to obtain the optimal posterior probability density function.

[0127] Step S4, obtaining an optimal state partition based on the optimal a posteriori probability density function; and generating a track of each target based on the optimal state set.

[0128] Figure 2 FIG. 1 is a structural diagram of a track generation system based on the probability density hypothesis of minimizing cross entropy according to the present invention. Figure 2 As shown, the present invention provides a probability density hypothesis track generation system based on minimizing cross entropy, including: a data division module 1, a probability density module 2, an optimization module 3 and a track generation module 4.

[0129] The data partitioning module 1 obtains a state estimation set based on the observation values ​​observed by the sensor, and partitions the state estimation set to obtain a valid partition set.

[0130] The probability density module 2 determines a posterior probability density function based on the effective partition set; and obtains a parameter probability density function set based on the effective partition set.

[0131] The optimization module 3 uses a minimum cross entropy optimization method based on the parameter probability density function set to gradually optimize the posterior probability density function to obtain an optimal posterior probability density function.

[0132] The track generation module 4 obtains the optimal state division based on the optimal posterior probability density function; and generates the track of each target based on the optimal state set.

[0133] As an optional implementation, the data partitioning module 1 of the present invention includes: a data acquisition unit and a data partitioning unit.

[0134] The data acquisition unit obtains a state estimation set based on observation values ​​observed by the sensor.

[0135] The data partitioning unit partitions the state estimation set to obtain a valid partition set; each state vector in the valid partition set Where: C K is the number of targets appearing in the monitoring area in the previous K moments, τ0 represents the set of all false estimates, τ i is the state estimation vector of the i-th target;

[0136] Each state vector in the valid partition set satisfies the following constraints:

[0137] 1) Element conservation: the elements remain constant before and after division,

[0138] Where: N k is the number of state estimates at time k;

[0139] 2) Elements do not intersect: Due to τ i Each estimate in is either classified as a false estimate or corresponds to at most one target,

[0140] Where: i represents the i-th target, j represents the j-th target;

[0141] 3) τ i Orderliness of state estimates in τ i is a set of elements sorted in ascending order of time.

[0142] where n l For target i in k l The number of state estimates at the time;

[0143] 4) Uniqueness: Since each state estimate corresponds to at most one target track,

[0144] 5) Measurability: Each target track corresponds to at least two target states, namely |τ i |≥2,i=1,2,…,C K .

[0145] As an optional implementation, the probability density module 2 of the present invention includes: a posterior probability density unit, a directed graph unit, a weight matrix unit and a parameter probability density unit.

[0146] The a posteriori probability density unit determines the a posteriori probability density function based on the valid partition set.

[0147] The directed graph unit defines a weighted directed graph.

[0148] The weight matrix unit obtains the probability weight matrix of the weight directed graph.

[0149] The parameter probability density unit constructs the parameter probability density function set based on the probability weight matrix and the valid partition set.

[0150] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0151] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A probability density hypothesis track generation method based on minimizing cross entropy, It is characterized in that include: A state estimation set is obtained based on the observation values ​​observed by the sensor, and the state estimation set is divided to obtain a valid partition set, which specifically includes: A state estimation set is obtained based on the observations observed by the sensor; The state estimation set is divided into a valid partition set; each state vector in the valid partition set Where: C K is the number of targets that appeared in the monitoring area in the previous K moments, τ 0 represents the set of all false estimates, τ i is the state estimation vector of the i-th target; Each state vector in the valid partition set satisfies the following constraints: 1) Element conservation: the elements remain constant before and after division, Where: N k is the number of state estimates at time k; 2) Elements do not intersect: Due to τ i Each estimate in is either classified as a false estimate or corresponds to at most one target, Where: i represents the i-th target, j represents the j-th target; 3) τ i Orderliness of state estimates in τ i is a set of elements sorted in ascending order of time. Where n l For target i in k l The number of state estimates at the time; 4) Uniqueness: Since each state estimate corresponds to at most one target track, 5) Measurability: Each target track corresponds to at least two target states, namely τ i ≥2,i=1,2,…,C K ; Determine a posterior probability density function based on the effective partition set; obtain a parameter probability density function set based on the effective partition set; Based on the parameter probability density function set, the posterior probability density function is gradually optimized using a minimum cross entropy optimization method to obtain an optimal posterior probability density function; Based on the optimal a posteriori probability density function, an optimal state division is obtained; and based on the optimal state division, a track of each target is generated.

2. The method for generating a probability density hypothesis track based on minimizing cross entropy according to claim 1, It is characterized in that Determining a posterior probability density function based on the effective partition set; A parameter probability density function set is obtained based on the effective partition set, including: Determining the posterior probability density function based on the valid partition set; Define a weighted directed graph; Obtaining a probability weight matrix of the weighted directed graph; The parameter probability density function set is constructed based on the probability weight matrix and the valid partition set.

3. A probability density hypothesis track generation system based on minimizing cross entropy, It is characterized in that include: The data partitioning module obtains a state estimation set based on the observation values ​​observed by the sensor, and partitions the state estimation set to obtain a valid partition set, specifically including: A data acquisition unit obtains a state estimation set based on the observation values ​​observed by the sensor; A data partitioning unit is configured to partition the state estimation set to obtain a valid partition set; each state vector in the valid partition set Where: C K is the number of targets that appeared in the monitoring area in the previous K moments, τ 0 represents the set of all false estimates, τ i is the state estimation vector of the i-th target; Each state vector in the valid partition set satisfies the following constraints: 1) Element conservation: the elements remain constant before and after division, Where: N k is the number of state estimates at time k; 2) Elements do not intersect: Due to τ i Each estimate in is either classified as a false estimate or corresponds to at most one target, Where: i represents the i-th target, j represents the j-th target; 3) τ i Orderliness of state estimates in τ i is a set of elements sorted in ascending order of time. Where n l For target i in k l The number of state estimates at the time; 4) Uniqueness: Since each state estimate corresponds to at most one target track, 5) Measurability: Each target track corresponds to at least two target states, namely τ i ≥2,i=1,2,…,C K ; A probability density module, which determines a posterior probability density function based on the effective partition set; and obtains a parameter probability density function set based on the effective partition set; An optimization module, based on the parameter probability density function set, adopts a minimum cross entropy optimization method to gradually optimize the posterior probability density function to obtain an optimal posterior probability density function; The track generation module obtains the optimal state division based on the optimal posterior probability density function; and generates the track of each target based on the optimal state division.

4. The probability density hypothesis track generation system based on minimizing cross entropy according to claim 3, It is characterized in that The probability density module includes: A posterior probability density unit, which determines the posterior probability density function based on the valid partition set; Directed graph unit, defines a weighted directed graph; A weight matrix unit, which obtains a probability weight matrix of the weight directed graph; A parameter probability density unit constructs the parameter probability density function set based on the probability weight matrix and the valid partition set.

Citation Information

Patent Citations

  • Motion target detection method based on improved minimum cross entropy

    CN102800107A

  • Multi-target tracking method based on theory of random sets

    CN106707272A

  • Adaptive distributed track data fusion method based on covariance indicator function

    CN107192998A

  • Classification and identification method for low, slow small targets

    CN112434643A

  • Deep learning-based track vector and target type joint estimation system and estimation method

    CN113569465A