Multi-radar maneuvering target tracking and asynchronous fusion method and device

Abstract

This invention provides a method and apparatus for multi-radar maneuvering target tracking and asynchronous fusion, comprising the following steps: dividing radar information into common field-of-view information and non-common field-of-view information; determining the target's state transition model, performing local filtering tracking, and obtaining a Gaussian mixture form of the target information's posterior probability hypothesis density; transmitting the posterior probability hypothesis density represented by the Gaussian mixture form to a nearby communicable radar station, and, combined with the radar scanning time, using the target's state transition model to recursively advance the radar information to the fusion time; constructing an information loss function, and fusing information within the common field of view based on a criterion of minimizing information difference; merging the information within the non-common field of view using a compensation strategy in the obtained common field-of-view fusion result to obtain the final fusion result. This invention achieves an effective method for tracking maneuvering targets, and features low computational complexity and high accuracy.

Description

Technical Field

[0001] This invention belongs to the field of target tracking technology, specifically relating to a distributed multi-radar maneuvering target tracking and asynchronous fusion method under different fields of view. Background Technology

[0002] With the development of target tracking technology, traditional single-radar systems are no longer able to cope with complex multi-target combat environments. Against this backdrop, thanks to the rapid development of sensor network communication technology and multi-sensor information fusion technology, distributed multi-radar systems based on multi-station radar joint tracking have attracted widespread attention. Compared to single radars, multi-radar joint tracking has advantages such as wide detection range, high tracking accuracy, and strong anti-jamming capability, and is therefore widely used in the military field. However, in real-world scenarios, due to the maneuverability of targets, traditional single-model tracking leads to a serious deterioration in tracking performance because it cannot accurately characterize the target's motion characteristics. Furthermore, due to the limited detection capabilities of radars, their field of view is generally different and not completely overlapping. Simultaneously, due to differences in radar power-on time and hardware settings, the measurement information obtained by multiple radars is not synchronized in time, preventing the system from accurately fusing information about the same target at the same time, resulting in a serious deterioration in fusion performance. Therefore, research on distributed multi-radar maneuvering target tracking and fusion methods under different field of view is crucial. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and apparatus for multi-radar maneuvering target tracking and asynchronous fusion. The solution of this invention can solve the problems existing in the prior art.

[0004] The technical solution of this invention:

[0005] According to the first aspect, a method for multi-radar maneuvering target tracking and asynchronous fusion is provided, including the following steps:

[0006] S1. Based on prior information from multiple radar fields of view, the radar's common field of view and non-common field of view are divided into two parts: common field of view information and non-common field of view information, through spatial segmentation method.

[0007] S2. After the local radar obtains the target measurement information, it determines the target's state transition model, performs local filtering and tracking, and obtains the Gaussian mixture form of the target information posterior probability hypothesis density.

[0008] S3. The posterior probability hypothesis density represented by the Gaussian mixture form is transmitted to a nearby communicable radar station. Due to the non-synchronous measurement information, the tracking results are not synchronized in time. Combined with the radar scanning time, the target state transition model is used to recursively push the radar information to the fusion time.

[0009] S4. Using the local radar information and the neighboring communicable radar information obtained in step S3, construct an information loss function and fuse information within the common field of view based on the criterion of minimizing information difference.

[0010] S5. Using the local radar information obtained in step S3, the neighboring communicable radar information, and the public field of view fusion result obtained in step S4, the information in the non-public field of view is merged through a compensation strategy to obtain the final fusion result.

[0011] Furthermore, the Gaussian mixture form of the posterior probability hypothesis density is as follows:

[0012]

[0013]

[0014]

[0015] in

[0016]

[0017]

[0018]

[0019] Let z represent the detection probability of the target state x at time k of the i-th radar, z be the measurement information of the radar node, and l be the model number;

[0020] Includes false positives. and testing items Two items; Represents a multidimensional Gaussian probability density function; For filter gain; I is the observation matrix of the measurement information; I is the identity matrix; The covariance matrix of the measurement noise; The mean is covariance is Gaussian density function; This represents the weight of the a-th predicted Gaussian component at time k in the l-model; Indicates the number of models; This represents the mean of the a-th predicted Gaussian component at time k under model l; This represents the covariance of the a-th predicted Gaussian component at time k under model l.

[0021] Furthermore, the result of fusing the radar information at the fusion time point is as follows: in, The result of fusing radar information recursively to the fusion time; Let be the total number of posterior Gaussian components of the i-th radar at fusion time ε. Let a be the weight of the a-th posterior Gaussian component. Let represent the mean of the a-th posterior Gaussian component. Let a represent the covariance of the a-th posterior Gaussian component. The mean is covariance is Gaussian density function, ε represents the distributed radar network; ε is the fusion time; a represents the a-th Gaussian component.

[0022] Furthermore, the information loss function is as follows: Where w i D represents the weight of the i-th radar. CS (v||v i The expression for the Cauchy-Schwarz divergence between two probability hypothesis densities is: K is the unit of measurement for the hypervolume of the probability hypothesis density function space, ||vv i || 2 The difference in probability hypothesis density (vv) i The square of the L2 norm of f. v is the value of the loss function f. w The probability hypothesis density of fusion under minimum constraints, v i Let be the probability hypothesis density of the i-th radar for observation of the common field of view.

[0023] Furthermore, when the loss function f w Under the minimum conditions, the fusion result of the information within the common field of view is: in, Γ(x) is arrive The projection, For the target state space, Indicates monitoring area S i Indicator functions on,

[0024] Furthermore, the final fusion result is as follows: in, This refers to information outside the radar's public field of view.

[0025] According to the second aspect, a distributed multi-radar maneuvering target tracking and asynchronous fusion device under different fields of view is provided, comprising N distributed radars, a filter, a state synchronization calculation unit, a field of view segmentation unit, and a field of view fusion calculation unit. The N distributed radars are arranged in a distributed radar network. A local distributed radar transmits the acquired target state to the filter for filtering and tracking, obtaining a Gaussian mixture of local posterior probability hypothesis densities, and then transmits it to neighboring radars in the distributed radar network. Each radar sends the received Gaussian mixture of local posterior probability hypothesis densities along with its acquired target state to the state synchronization calculation unit. The step calculation unit pushes the target state information of all radars to the fusion time based on the obtained data, and sends the information at the fusion time to the field-of-view fusion calculation unit. The field-of-view segmentation unit segments the field of view of each distributed radar into a common field of view and a non-common field of view based on the position and prior information of the field of view of each distributed radar, and segments the radar information into common field of view information and non-common field of view information according to the segmented field of view, and sends the segmented radar field of view information to the field-of-view fusion calculation unit. The field-of-view fusion calculation unit calculates the common field of view fusion information and the non-common field of view fusion information, and merges the common field of view fusion information and the non-common field of view fusion information to obtain the final fusion result.

[0026] The beneficial effects of this invention compared to the prior art are as follows:

[0027] This invention solves the problem of asynchronous measurement information obtained from radars with different fields of view by fusing the target tracking results of multiple distributed radars under different fields of view, and realizes a method for effectively tracking moving targets. It has the characteristics of low computational complexity and high accuracy. Attached Figure Description

[0028] The accompanying drawings, which form part of this specification, are provided to further illustrate embodiments of the invention and, together with the textual description, explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0029] Figure 1 A schematic diagram illustrating the steps of a multi-radar maneuvering target tracking and asynchronous fusion method provided according to an embodiment of the present invention is shown;

[0030] Figure 2 A schematic diagram of a multi-radar maneuvering target tracking and asynchronous fusion device according to an embodiment of the present invention is shown;

[0031] Figure 3The figure shows a simulation scenario of tracking a maneuvering target in a cluttered environment under different fields of view of multiple radars, according to an embodiment of the present invention.

[0032] Figure 4 The figure shows the simulation accuracy results of maneuvering target tracking under different fields of view of multiple radars in a clutter environment according to an embodiment of the present invention;

[0033] Figure 5 The figure shows the simulation results of the basic estimation of maneuvering target tracking under different views of multiple radars in a clutter environment, according to an embodiment of the present invention. Detailed Implementation

[0034] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0036] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0037] like Figure 1As shown, according to an embodiment of the present invention, a method for distributed multi-radar maneuvering target tracking and asynchronous fusion under different fields of view is provided, including the following steps:

[0038] S1. Based on prior information from multiple radar fields of view, the radar's common field of view and non-common field of view are divided into two parts: common field of view information and non-common field of view information, through spatial segmentation method.

[0039] Assuming the monitoring area is Due to limitations in radar hardware and their geographical location within the monitoring area, the fields of view of different radars are typically different and not completely overlapping. Without loss of generality, for any two radar nodes... View range is represented as Monitoring area through spatial segmentation method It is divided into two parts: one part is the radar's common field of view (cFoV), defined as The other part is the radar's non-common field of view (eFoV), defined as... and Furthermore, the division between common and non-common fields of view for multiple radars follows the same principle as described above. It is defined as the part where the fields of view of two or more radars overlap, and the part where there is only one radar field of view is the non-common field of view.

[0040] S2. After the radar obtains the target measurement information, it uses a multi-model-based probability hypothesis density filter for local filtering and tracking to obtain the mixture Gaussian form of the posterior probability hypothesis density.

[0041] Given the target's maneuverability, a single model cannot fully characterize its motion characteristics. Therefore, each radar node utilizes a multi-model-based probability hypothesis density filter for local filtering and tracking. In one embodiment, "multi-model" refers to multiple motion models, typically including a uniform linear motion model, a cooperative left-turn motion model, and a cooperative right-turn motion model. The target state transition density model when the target switches between these motion models is represented as follows:

[0042]

[0043] in, This represents the state of target ξ at time k; This indicates the two-dimensional position information of the target in the x and y directions; This represents the two-dimensional velocity information of the target in the x and y directions; This represents the label of the model at time k; This represents a finite set containing all candidate model labels, where the candidate models can be determined based on the specific scenario. Represents the target state transition density. Let t(l) represent the target state transition density at time k under model l. k |l k-1 The transition probability matrix () represents the model transition matrix, which can be determined according to the specific scenario. By using a multi-model-based probability hypothesis density filter to track maneuvering targets, the target model automatically switches as the target's motion characteristics change. Compared with traditional single-model modeling methods, the method proposed in this invention effectively solves the model mismatch problem that occurs in maneuvering target tracking.

[0044] The iterative process of the multi-model-based probability hypothesis density filter is divided into two parts: prediction and update. Hypothesis and Let represent the posterior probability hypothesis density of the i-th radar at time k-1 and the prediction probability hypothesis density at time k, respectively. Then the prediction process can be expressed as:

[0045]

[0046] Where, p S,k Indicates the probability of the target's survival; Let ζ represent birth intensity, and ζ represent the integral variable.

[0047] The update process includes updating the posterior probability hypothesis density and updating the model probabilities:

[0048]

[0049]

[0050] in, Let represent the detection probability of the i-th radar at time k for the ξ-th target; Let represent the random finite set of the i-th radar measurements, where the radar measurement information is the distance and velocity of the target relative to the radar in a two-dimensional scene; Represents a multi-objective likelihood function; Indicates clutter intensity; p i,ξ (l k ) indicates l k The probability of the model; Indicates the number of models.

[0051] Based on the above principles, the Gaussian Mixture Method (GM) is used for implementation. The specific implementation steps are as follows:

[0052] Assume the Gaussian mixture form of the posterior probability hypothesis density at time k-1. It is known and parameterized as follows:

[0053]

[0054] in This represents the number of posterior Gaussian components at time k-1; This represents the weight of the a-th posterior Gaussian component at time k-1; This represents the mean of the a-th posterior Gaussian component at time k-1; This represents the covariance of the a-th posterior Gaussian component at time k-1. Let represent a Gaussian density function with mean m and covariance P.

[0055] Substituting equation (5) into equation (2), we can calculate the Gaussian mixture form of the prediction probability hypothesis density at time k:

[0056]

[0057] in

[0058]

[0059]

[0060]

[0061]

[0062] and Let F represent the mixed Gaussian components of surviving and spawning targets at time k, respectively; T is the radar scan period; F l k-1 (T) is the state transition matrix of model l at time k-1; Let be the state covariance matrix of model l at time k-1.

[0063] Then The mixture of Gaussian components of the survival target and the birth target can be integrated to obtain:

[0064]

[0065] in This represents the number of Gaussian components predicted at time k; This represents the weight of the a-th predicted Gaussian component at time k in the l-model; This represents the mean of the a-th predicted Gaussian component at time k under model l; This represents the covariance of the a-th predicted Gaussian component at time k under model l.

[0066] Next, by combining the measurement information to update the model probability and the prediction probability hypothesis density, and substituting equation (8) into equation (3), the Gaussian mixture form of the posterior probability hypothesis density can be obtained:

[0067]

[0068] in,

[0069]

[0070]

[0071]

[0072]

[0073]

[0074]

[0075] Includes false positives. and testing items Two items; Represents a multidimensional Gaussian probability density function; For filter gain; I is the observation matrix of the measurement information; I is the identity matrix; This is the covariance matrix of the measurement noise.

[0076] Furthermore, to address the issue of the increasing number of Gaussian components with each iteration, a pruning and merging method is considered to handle useless Gaussian components. The parameterized representation of the Gaussian component sequence after pruning is as follows:

[0077]

[0078] in Let be the number of posterior Gaussian components at the i-th radar time k. Let be the weight of the a-th Gaussian component, and α be the set pruning threshold. The merging method uses a modified Mahalanobis distance-based approach to integrate multiple similar Gaussian components into a single Gaussian component. The processed Gaussian component sequence is represented as I. m It satisfies:

[0079]

[0080] Where β is the set merging threshold value, and For I m The mean of any two Gaussian components in the equation.

[0081] Finally, regarding the issues of target state extraction and target number estimation, we consider assigning weights greater than the extraction threshold γ. w Treating the Gaussian components as the target to be preserved, we can obtain the final filtered Gaussian component sequence:

[0082]

[0083] Therefore, based on the properties of the probability hypothesis density function, I f The mean of the Gaussian components is the target state, I. f The sum of the weights of all Gaussian components in the equation is the estimate of the number of targets.

[0084] S3. The posterior probability hypothesis density represented by the Gaussian mixture form is transmitted to a nearby communicable radar station. Due to the non-synchronous measurement information, the tracking results are not synchronized in time. Based on the radar scanning time, the target state transition model is used to recursively push the radar information to the fusion time.

[0085] In one embodiment, a distributed network is assumed. There are N radar nodes, each capable of sensing environmental targets, generating measurement data, and performing local data processing. The radar's time series model is an equally spaced time series, and the fusion time interval is τ. Traditional fusion based on minimizing information differences does not consider the influence of different viewpoints; the fusion expression is:

[0086]

[0087] in It describes the weights of each radar information in traditional fusion methods. However, due to the non-time synchronization of radar measurement information, the local radar may not have a filtered output at the fusion time. If the i-th local radar has a filtered output at the ε-fusion time, then proceed directly to step S4; otherwise, the radar information from the previous fusion time is recursively pushed to the ε-fusion time, and the recursive result is expressed in Gaussian mixture form as follows:

[0088]

[0089] in

[0090]

[0091]

[0092] Where τ is the fusion time interval. Here is the state transition matrix of model l at time ε-1; Let l be the state covariance matrix of model l at time ε-1. For the l-model at the fusion time ε The weight of the a-th posterior Gaussian component under the given condition. Let be the total number of posterior Gaussian components of the i-th radar at fusion time ε-1; and Let be the mean and covariance of the a-th Gaussian component after fusion at time ε-1, respectively. Therefore, the fused result at time ε is:

[0093]

[0094] in, The result of fusing radar information recursively to the fusion time; Let be the total number of posterior Gaussian components of the i-th radar at fusion time ε. Let a be the weight of the a-th posterior Gaussian component. Let represent the mean of the a-th posterior Gaussian component. Let a represent the covariance of the a-th posterior Gaussian component. The mean is covariance is Gaussian density function, ε represents the distributed radar network; ε is the fusion time; a represents the a-th Gaussian component.

[0095] By merging the state information of different radars at different times at the fusion time, the problem of non-time synchronization of information from multiple radars can be solved.

[0096] S4. Using the fusion time radar information obtained in step S3, construct an information loss function and fuse radar information in the common field of view based on the criterion of minimizing information difference.

[0097] In one embodiment, step S4 is specifically implemented as follows: based on the local radar information and the neighboring communicable radar information obtained in step S2, an information loss function f is constructed. w :

[0098]

[0099] Among them, w i D represents the weight of the i-th radar. CS (v||v i The expression ) represents the Cauchy-Schwarz divergence between two probability hypothesis densities, and is calculated as follows:

[0100]

[0101] Where v is the value in the loss function f w The probability hypothesis density of fusion under minimum constraints, v i Let |v| represent the probability hypothesis density of the i-th radar observing the common field of view; K is the hypervolume unit of measurement for the probability hypothesis density function space, ||v| i || 2 The difference in probability hypothesis density (vv) i The square of the L2 norm of ).

[0102] Therefore, based on the criterion of minimizing information differences, the fusion result of information within the common field of vision can be obtained as follows:

[0103]

[0104] in

[0105]

[0106] w i By minimizing the information loss function f w We obtain Γ(x) as arrive The projection, For the target state space, Indicates monitoring area S i The indicator function on is specifically represented as:

[0107]

[0108] S5. Using the local radar information obtained in step S3, the neighboring communicable radar information, and the public field of view fusion result obtained in step S4, the information in the non-public field of view is merged to obtain the final fusion result.

[0109] In one embodiment, step S5 is specifically implemented as follows: information within the non-public field of view is merged using a compensation strategy to obtain the final fusion result.

[0110]

[0111] in, This refers to information outside the radar's public field of view.

[0112] According to the embodiment of the second aspect, a distributed multi-radar maneuvering target tracking and asynchronous fusion device under different fields of view is provided, comprising N distributed radars, a filter, a state synchronization calculation unit, a field of view segmentation unit, and a field of view fusion calculation unit. The N distributed radars are arranged in a distributed radar network. A distributed radar transmits the acquired target state to the filter for filtering and tracking, obtaining a Gaussian mixture of posterior probability hypothesis densities, and transmits it to the radars of neighboring communicable nodes in the distributed radar network. The radar sends the received local Gaussian mixture of posterior probability hypothesis densities and its acquired target state together to the state synchronization calculation unit. The step calculation unit pushes the target state information of all radars to the fusion time based on the obtained data, and sends the information at the fusion time to the field of view fusion calculation unit. The field of view segmentation unit segments the field of view of each distributed radar into a common field of view and a non-common field of view based on the position and prior information of the field of view of each distributed radar, and segments the radar information into common field of view information and non-common field of view information according to the segmented field of view, and sends the segmented radar field of view information to the field of view fusion calculation unit. The field of view fusion calculation unit calculates the common field of view fusion information and the non-common field of view fusion information, and merges the common field of view fusion information and the non-common field of view fusion information to obtain the final fusion result.

[0113] In a further embodiment, the filter employs a multi-model-based probability hypothesis density filter, which facilitates tracking multiple targets in various operational modes. The filter model is selected based on a comprehensive consideration of actual needs, including but not limited to accuracy, size, energy consumption, and weight.

[0114] In a further embodiment, the Gaussian mixture form of the target posterior probability hypothesis density is:

[0115] Gaussian mixture form of the posterior probability hypothesis density:

[0116]

[0117] in:

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124] Includes false positives. and testing items Two items; Represents a multidimensional Gaussian probability density function; For filter gain; I is the observation matrix of the measurement information; I is the identity matrix; This is the covariance matrix of the measurement noise.

[0125] In one specific embodiment, to address the issue of the increasing number of Gaussian components with increasing iteration count, a pruning and merging method is considered to process useless Gaussian components. The parameterized representation of the Gaussian component sequence after pruning is as follows:

[0126]

[0127] in Let be the number of posterior Gaussian components at the i-th radar time k. Let α be the weight of the a-th Gaussian component, and α be the set pruning threshold value.

[0128] The merging method uses a modified Mahalanobis distance-based approach to integrate multiple similar Gaussian components into a single Gaussian component. The processed Gaussian component sequence is represented as I. m It satisfies:

[0129]

[0130] Where β is the set merging threshold value, and For I m The mean of any two Gaussian components in the equation.

[0131] Finally, regarding the issues of target state extraction and target number estimation, we consider assigning weights greater than the extraction threshold γ. w Treating the Gaussian components as the target to be preserved, we can obtain the final filtered Gaussian component sequence:

[0132]

[0133] Therefore, based on the properties of the probability hypothesis density function, I f The mean of the Gaussian components is the target state, I. f The sum of the weights of all Gaussian components in the equation is the estimate of the number of targets.

[0134] In a further embodiment, the state synchronization calculation unit, the view segmentation unit, and the view fusion calculation unit are integrated into the onboard computer.

[0135] In another embodiment, the fusion result of information within the common field of view is:

[0136]

[0137] in

[0138]

[0139] Γ(x) is arrive The projection, For the target state space, Indicates monitoring area S i The indicator function on is specifically represented as:

[0140]

[0141] Information within the non-public field of view is obtained directly through filters.

[0142] In another embodiment, the fusion result of information within the common field of view is: in, Γ(x) is arrive The projection, For the target state space, Indicates monitoring area S i Indicator functions on,

[0143] In one further embodiment, the final fusion result is:

[0144]

[0145] in, This refers to information outside the radar's public field of view.

[0146] To better illustrate the present invention, the following detailed description is provided in conjunction with specific embodiments.

[0147] Two radars with fan-shaped fields of view were simulated to track four maneuvering targets in a cluttered scene. The scene size was 45km × 35km, with the two radars located at (22km, 0) and (22km, 35km) respectively, a detection range of 35km, and a detection angle range of 80°. Each radar independently sensed targets in its respective field of view, generating measurement information, including measurements from real targets and interference measurements from clutter and noise. The number of interference measurements followed a Poisson distribution with λ = 10, and their locations were uniformly distributed. Furthermore, the scanning periods of the two radars were 1s and 2s respectively, and the fusion timing was once every 1s, thus exhibiting asynchronous timing.

[0148] There are four targets in the scene that exhibit maneuvering motion. The starting positions of the targets are respectively... and The target trajectory is characterized by multiple motion models, including a uniform linear motion model, a cooperative left-turn motion model with a turning rate of 5 rad / s, and a right-turn motion model. Arbitrarily selecting four different model combinations constitutes the trajectory shown below. Figure 3 The trajectory shown.

[0149] Regarding motion model matching, this invention employs a multi-model-based probability hypothesis density filter to achieve effective multi-target tracking. It considers using three different models to match the motion states of multiple targets in real time, and simultaneously achieves intelligent model switching based on probability. The model transition probability matrix is ​​as follows:

[0150]

[0151] The state transition matrix and state covariance matrix of the uniform linear motion (CV) model and the cooperative turning (CT) model are as follows:

[0152]

[0153]

[0154] Where σ x =σ y =10m, T s =1s, I2 represents the second-order identity matrix. This represents the Kronecker product. Additionally, for measurement noise, assume the covariance of the x and y axes is σ. x =σ y =10m, then the covariance matrix of the measurement noise is:

[0155]

[0156] Regarding the fusion of radar measurement information that is not time-synchronized, since the scanning periods of the two radars are 1s and 2s respectively, and the fusion time series is to fuse the radar information once every 1s. For fusion tracking under multiple radar fields of view that are different and not completely overlapping, this invention designs radar detection probability functions for both common and non-common fields of view. In the simulation, the detection probability of the filter is set as follows:

[0157]

[0158] This means that the detection probability of targets within the radar's field of view is 0.95, while targets outside the radar's field of view cannot be detected. The survival probability of the target is set to p. S,k =0.99.

[0159] Figure 4 and Figure 5The figures show the simulation accuracy and cardinality distribution results of maneuvering target tracking under different views of multiple radars in a cluttered environment, respectively. The evaluation metric is Optimal Submode Allocation (OSPA), with parameters set to c=100 and p=1. To prevent random errors in the experiment, the simulation results are averaged after 300 Monte Carlo experiments. As can be seen from the figures, under the same scenario, compared with traditional single-model tracking and standard difference minimization fusion methods, the distributed multi-radar maneuvering target tracking and fusion method under different views of this invention can achieve higher tracking accuracy and more accurate cardinality estimation.

[0160] In summary, the present invention offers at least the following advantages compared to existing technologies:

[0161] This invention solves the problem of asynchronous measurement information obtained from radars with different fields of view by fusing the target tracking results of multiple distributed radars under different fields of view, and realizes a method for effectively tracking moving targets. It has the characteristics of low computational complexity and high accuracy.

[0162] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for tracking and asynchronously fusing multiple radar maneuvering targets, characterized in that, Includes the following steps: S1. Based on prior information from multiple radar fields of view, the radar's common field of view and non-common field of view are divided into two parts: common field of view information and non-common field of view information, through spatial segmentation method. S2. After the local radar obtains the target measurement information, it determines the target's state transition model, performs local filtering and tracking, and obtains the Gaussian mixture form of the target information posterior probability hypothesis density. S3. The target information posterior probability hypothesis density represented by the Gaussian mixture form is transmitted to a nearby communicable radar station. Due to the non-synchronous measurement information, the tracking results are not synchronized in time. Combining the radar scanning time, the target state transition model is used to push all radar information to the fusion time. This step specifically includes: the nearby radar sends the received target information posterior probability hypothesis density in the Gaussian mixture form and the target state obtained by the nearby radar itself to the state synchronization calculation unit. The state synchronization calculation unit pushes the target state information of all radars to the fusion time according to the obtained data and sends the information of the fusion time to the field of view fusion calculation unit. The field of view segmentation unit divides the radar's field of view into a common field of view and a non-common field of view according to the position and field of view prior information of each radar. The common field of view and the non-common field of view are radar information. S4. Using the local radar information and the neighboring communicable radar information obtained in step S3, construct an information loss function and fuse information within the common field of view based on the criterion of minimizing information difference. S5. Using the local radar information obtained in step S3, the neighboring communicable radar information, and the public field of view fusion result obtained in step S4, the information in the non-public field of view is merged through a compensation strategy to obtain the final fusion result.

2. The multi-radar maneuvering target tracking and asynchronous fusion method according to claim 1, characterized in that, The Gaussian mixture form of the posterior probability hypothesis density: in Let z represent the detection probability of the target state x at time k of the i-th radar, z be the measurement information of the radar node, and l be the model number; Includes false positives. and testing items Two items; Represents a multidimensional Gaussian probability density function; For filter gain; The observation matrix for measurement information; It is the identity matrix; The covariance matrix of the measurement noise; The mean is covariance is Gaussian density function; This indicates that at time k, the a-th predicted Gaussian component is... Weights under the model; Indicates the number of models; This indicates that at time k, the a-th predicted Gaussian component is... The mean under the model; This indicates that at time k, the a-th predicted Gaussian component is... Covariance under the model; for The mixed Gaussian components of the surviving target and the spawning target; This represents the number of Gaussian components predicted at time k; This indicates the intensity of clutter.

3. The multi-radar maneuvering target tracking and asynchronous fusion method according to claim 1, characterized in that, The result of fusion after the radar information is recursively pushed to the fusion time is: ,in, The result of fusion after radar information is recursively pushed to the fusion time; For the i-th radar at the fusion time The total number of posterior Gaussian components, Let a be the weight of the a-th posterior Gaussian component. Let represent the mean of the a-th posterior Gaussian component. Let a represent the covariance of the a-th posterior Gaussian component. The mean is covariance is Gaussian density function, This represents a network of distributed radars; The fusion time is denoted by 'a'; 'a' represents the a-th Gaussian component.

4. The multi-radar maneuvering target tracking and asynchronous fusion method according to claim 1, characterized in that, The information loss function is: ,in This represents the weight of the i-th radar. The Cauchy-Schwarz divergence between two probability hypothesis densities is expressed as: , Let the hypervolume be the unit of measurement for the probability hypothesis density function space. Difference in probability hypothesis density The square of the L2 norm, For the loss function The probability hypothesis density of fusion under minimum constraints. Let be the probability hypothesis density of the i-th radar for observation of the common field of view.

5. The multi-radar maneuvering target tracking and asynchronous fusion method according to claim 1, characterized in that, In order to make the loss function Under the minimum conditions, the fusion result of the information within the common field of view is: ,in, , for arrive The projection, For the target state space, Indicates the monitoring area Indicator functions on, .

6. The multi-radar maneuvering target tracking and asynchronous fusion method according to claim 5, wherein the final fusion result is: ,in, This refers to information outside the radar's public field of view.

7. A multi-radar maneuvering target tracking and asynchronous fusion device using the method described in any one of claims 1 to 6, characterized in that: The system comprises N distributed radars, filters, a state synchronization calculation unit, a view domain segmentation unit, and a view domain fusion calculation unit. The N distributed radars are arranged in a distributed radar network. Each local distributed radar transmits its acquired target state to the filter for filtering and tracking, obtaining a Gaussian mixture of local posterior probability hypothesis densities, which is then transmitted to neighboring radars in the distributed radar network. Each radar sends the received Gaussian mixture of local posterior probability hypothesis densities along with its acquired target state to the state synchronization calculation unit. The state synchronization calculation unit pushes the target state information of all radars to the fusion time based on the acquired data and sends the fusion time information to the view domain fusion calculation unit. The view domain segmentation unit, based on the position and prior view information of each distributed radar, segments the radar's view domain into a common view domain and a non-common view domain. It then segments the radar information into common view domain information and non-common view domain information according to the segmented view domains and sends the segmented radar view domain information to the view domain fusion calculation unit. The view domain fusion calculation unit calculates the common view domain fusion information and the non-common view domain fusion information, and merges them to obtain the final fusion result.

Images