A low, slow and small target tracking method based on identification information

Abstract

The application relates to a low, slow and small target tracking method based on identification information, which obtains the target type of each track by utilizing track features and target identification criteria, further establishes a height constraint condition based on the target type of each track, and updates the target state estimation value according to the target state estimation value at the current moment and the space constraint condition, so that the tracking performance of the low, slow and small target can be effectively improved. Therefore, the application realizes a target tracking method capable of improving the tracking performance of the low, slow and small target.

Description

Technical Field

[0001] This invention relates to the field of radar technology, and in particular to a method for tracking low, slow, and small targets based on identification information to improve the tracking performance of low, slow, and small targets. Background Technology

[0002] Low-altitude, slow-speed, and small-sized targets refer to aircraft that fly at low altitudes, have slow speeds, and small radar cross-sections. These aircraft are inexpensive, easy to carry and operate, have low takeoff requirements, and are highly unpredictable, posing an increasingly prominent threat to urban security and major events, and presenting greater challenges to modern security and defense systems.

[0003] As an active sensor for aerial target detection, radar has the advantages of all-weather, all-time detection and long detection range, and has gradually become an important means of UAV detection. However, during low-altitude surveillance, especially in complex urban environments, radar is easily affected by buildings, towers, and other structures, resulting in multipath propagation and impacting the detection performance of low-altitude targets. Therefore, it suffers from poor accuracy in detecting low, slow-moving, and small targets.

[0004] Existing target tracking methods do not take into account prior knowledge of the height dimension of low, slow, and small targets. They directly send the target observation points with poor accuracy into the target tracking module. As a result, the output target trajectory will have a large deviation from the actual target motion trajectory. The poor tracking accuracy will not be able to guide the low, slow, and small target handling module to intercept and handle the target. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing a method for tracking small, slow targets based on identification information.

[0006] The technical solution of the low-speed small target tracking method based on recognition information of the present invention is as follows:

[0007] S1, Based on the motion characteristic information F of the i-th track at time k-1 before the i-th time step k-1,i The target type c at time k-1 of the i-th track is obtained according to the target identification criteria. k-1,i And based on the target state estimate at time k-1 of the i-th track. Establish the associated gate G at time k k,i , where k is a positive integer greater than 1, and i is a positive integer;

[0008] S2. Obtain the associated gate G that falls within the k-th time step. k,i The set of free points Z within k According to the data association method, from the set of free point traces Z at the k-th time... k Select the associated point z of the i-th track at time k.k,i ;

[0009] S3, according to the association point trail z of the i-th track at the k-th moment k,i , according to the target tracking method, updating the target state estimation value of the i-th track at the k-th moment And the estimated covariance matrix P of the i-th track at the k-th moment k,i ;

[0010] S4, if the target type at the k-1-th moment is a low, slow and small target, a height constraint condition is established;

[0011] S5, according to the target state estimation value of the i-th track at the k-th moment The estimated covariance matrix P of the i-th track at the k-th moment k,i , and the height constraint condition, a maximum a posteriori probability optimization model is established, and the optimal target state estimation value of the i-th track at the k-th moment under the height constraint condition And the estimated covariance matrix update value of the i-th track at the k-th moment And let

[0012] S6, S1 to S5 are performed for each track at the k-1-th moment;

[0013] S7, let k=k+1, and S1 to S6 are performed.

[0014] The low, slow and small target tracking method based on identification information has the following beneficial effects:

[0015] The target type is obtained through the identification information of each track, and the height constraint condition is further established for the low, slow and small target type, a maximum a posteriori probability optimization model based on the height constraint is designed on the basis of the traditional target tracking algorithm, and the optimal target state estimation is obtained, which can effectively improve the tracking performance of the low, slow and small target, so that the low, slow and small target tracking method based on identification information for improving the tracking performance of the low, slow and small target is realized. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 It is a flowchart of a low, slow and small target tracking method based on identification information of an embodiment of the application;

[0017] Figure 2 It is an x-axis tracking accuracy comparison graph of the embodiment of the application and the traditional method;

[0018] Figure 3 It is a y-axis tracking accuracy comparison graph of the embodiment of the application and the traditional method;

[0019] Figure 4A z-axis tracking precision comparison chart between the embodiment of the present application and the conventional method. DETAILED DESCRIPTION

[0020] As shown in the figure, a low, slow and small target tracking method based on identification information according to the embodiment of the present application comprises the following steps: Figure 1

[0021] S1, obtaining a target type c of a k-1 moment of an i th track according to motion characteristic information F of the k-1 moment of the i th track k-1,i , according to a target identification criterion. k-1,i Wherein, k is a positive integer greater than 1, and i is a positive integer.

[0022] The target identification criterion includes but is not limited to support vector machine criterion, decision tree criterion, neural network criterion, etc. In the embodiment, the decision tree criterion is selected.

[0023] The target type output by the target identification criterion at least contains low, slow and small targets, and other target types can be formulated according to radar task requirements. In the embodiment, the target type contains low, slow and small targets and high-altitude aircraft targets.

[0024] S2, establishing an association gate G of a k moment according to a target state estimation value of a k-1 moment of an i th track k,i The specific steps are as follows:

[0025] S20, obtaining a target state prediction value of a k moment of an i th track through the following formula according to a target state estimation value of a k-1 moment of an i th track

[0026]

[0027] Wherein, represents a state transition matrix, and in the embodiment, the expression of is,

[0028]

[0029] Wherein, T s represents a radar search period, and in the embodiment, it is taken as 1 second.

[0030] In the embodiment, the target state dimension is 6, the target state estimation value of a k-1 moment of an i th track In the embodiment, the form of is, and respectively represent an x-axis target position estimation value and an x-axis target velocity estimation value of a k-1 moment of an i th track, and​​​ Let represent the estimated y-axis target position and y-axis target velocity at time k-1 of the i-th trajectory, respectively. and Let Z and Z represent the estimated position and velocity of the target along the z-axis at time k-1 of the i-th track, respectively.

[0031] S21. Based on the target state prediction value at time k of the i-th track. The predicted position value of the i-th track at time k is obtained according to the following formula. in, This represents the distance prediction value for the i-th track at time k. Let represent the azimuth prediction value of the i-th track at time k. This represents the pitch prediction value at time k for the i-th trajectory.

[0032]

[0033] Among them, h k (·) represents the radar observation function at time k, arctan(·) represents the arctangent function, and the function value is in degrees; arcsin(·) represents the arcsine function, and the function value is in degrees.

[0034] S22, Using the position prediction value of the i-th track at time k as the associated gate G at time k. k,i At the center, the associated wavegate G at time k is established according to the following formula. k,i ,

[0035]

[0036] Among them, y k R represents the range of free points associated with the i-th track at time k. k Indicates y k The distance measurement value in A k Indicates y k The azimuth measurement value, E k Indicates y k In the pitch measurement values, ΔR represents the range-correlated gate size, ΔA represents the azimuth-correlated gate size, ΔE represents the pitch-correlated gate size, and min[a,b] represents finding the smaller value of input variables a and b.

[0037] In this embodiment, the distance-related gate size ΔR is set to 300 meters, the azimuth-related gate size ΔA is set to 4 degrees, and the pitch-related gate size ΔE is set to 4 degrees.

[0038] S3. Obtain the associated gate G that falls within the k-th time step. k,i The set of free points Z withink According to the data association method, from the set of free point traces Z at the k-th time... k Select the associated point z of the i-th track at time k. k,i ;

[0039] The data association method includes, but is not limited to, the nearest neighbor method and the global nearest neighbor method. In this embodiment, the nearest neighbor method is selected, and the method includes the following steps:

[0040] S31. Calculate the set of free points Z k The j-th free point z k,j Position prediction value of the i-th track at time k The correlation measure d i,j , where z k,j =[R k,j A k,j E k,j ] T In this embodiment, the formula for calculating the correlation measure is:

[0041]

[0042] Where, m k,i Z represents the set of free point traces k The number of elements, δ R,i,j δ represents the distance difference. A,i,j δ represents the azimuth difference. E,i,j The pitch difference is expressed as follows:

[0043]

[0044]

[0045]

[0046] S32. Select the free point with the smallest correlation measure as the correlation point z of the i-th track at time k. k,i .

[0047] S4. Based on the associated point z of the i-th track at time k. k,i According to the target tracking method, update the target state estimate of the i-th track at time k. And the estimated covariance matrix P of the i-th track at time k. k,i ;

[0048] The target tracking method includes, but is not limited to, the extended Kalman filter method, the insensitive Kalman filter method, and the conversion measurement Kalman filter method. In this embodiment, the extended Kalman filter method is selected, which includes the following steps:

[0049] S41. Based on the estimated covariance matrix P of the i-th track at time k-1. k-1,i The prediction covariance matrix P of the i-th track at time k is calculated according to the following formula. k|k-1,i ,

[0050]

[0051] Among them, Q k|k-1 The process noise covariance matrix is ​​represented by the following formula in this embodiment.

[0052]

[0053] in, This represents the process noise variance, which is set to 0.1 in this embodiment.

[0054] S42. Calculate the new information covariance matrix S of the i-th track at time k according to the following formula. k,i ,

[0055]

[0056] Among them, H k The radar observation function h at time k is represented by k The Jacobian matrix of (·), R k The radar observation noise covariance matrix is ​​represented in the following form:

[0057]

[0058] Where, σ R This indicates the distance accuracy, which is set to 40 meters in this embodiment; σ A This indicates the azimuth accuracy, which is taken as 0.4 degrees in this embodiment; σ E This indicates the pitch accuracy, which is set to 0.4 degrees in this embodiment.

[0059] S43. Based on the new information covariance matrix S of the i-th trajectory at time k. k,i Calculate the Kalman filter gain K according to the following formula. k,i ,

[0060]

[0061] S44, Based on the Kalman filter gain K k,i Update the target state estimate of the i-th track at time k according to the following formula. And the estimated covariance matrix P of the i-th track at time k. k,i ,

[0062]

[0063] S5, if the target type at the k-1 moment is a low, slow and small target, a height constraint condition is established, including the following steps,

[0064] S50, determining the maximum height h of the low, slow and small target type max In the embodiment, h max is 200 m;

[0065] S51, establishing the height constraint condition according to the following formula,

[0066] Hx k ≤ h max ,

[0067] wherein H represents a height observation vector, x k represents the real state of the target corresponding to the i th track at the k moment.

[0068] In the embodiment, the expression of the height observation vector H is H = [0, 0, 0, 0, 1, 0].

[0069] In the embodiment, the real state x k of the target corresponding to the i th track at the k moment is in the form of wherein x k and respectively represent the x-axis target position and the x-axis target speed at the k moment, y k and respectively represent the y-axis target position and the y-axis target speed of the i th track at the k moment, z k and respectively represent the z-axis target position and the z-axis target speed of the i th track at the k moment.

[0070] S6, according to the target state estimation value x at the k moment of the i th track, the estimation covariance matrix P k,i at the k moment of the i th track, and the height constraint condition, a maximum a posteriori probability optimization model is established according to the following formula,

[0071]

[0072] constraint: Hx k ≤ h max

[0073] S7, according to the maximum a posteriori probability optimization model, the optimal target state estimation value x at the k moment of the i th track under the height constraint condition and the estimation covariance matrix update value P at the k moment of the i th track are obtained, and x including the following steps:

[0074] S71, the optimal solution of the maximum a posteriori probability optimization model is an optimal target state estimation value of the i-th track at the k-th moment under a high constraint condition The specific expression is,

[0075]

[0076] Wherein, Indicates the Lagrange multiplier.

[0077] S72, the estimated covariance matrix update value of the i-th track at the k-th moment The expression of,

[0078]

[0079] Wherein, G=P k,i H T (HP k,i H T ) -1 Indicates an intermediate variable, and alpha indicates the uncertainty caused by the change of target height.

[0080] S73, let

[0081] S8, performing S1 to S7 for each track at the k-1 moment;

[0082] S9, let k=k+1, and performing S1 to S8.

[0083] The beneficial effects of the low, slow and small target tracking method based on identification information are described below through simulation comparison test:

[0084] Experimental scene: the radar working frequency band is C band, the distance precision σ R of the radar measurement is 40 meters, the azimuth precision σ A is 0.4 degrees, the elevation precision σ E is 0.4 degrees, the initial position of the target is [40km, 15km, 0km], and the uniform speed is [10m / s, 10m / s, 0m / s].

[0085] The traditional target tracking method and the method of the application are respectively used to track the low, slow and small target, the total number of tracking moments is 100 times, the distance tracking precision is as shown in Figure 2 , the azimuth tracking precision is as shown in Figure 3 , and the elevation tracking precision is as shown in Figure 4 .

[0086] It can be seen by comparison that the method has the same accuracy as the traditional target tracking method in the x-axis and y-axis, and has higher accuracy than the traditional target tracking method in the z-axis representing the height dimension, proving the effectiveness of the method. The reason for the improvement in tracking accuracy in the z-axis representing the height dimension is that the identification information is fully utilized and the height constraint is introduced into the target tracking process, and the target state estimate value is updated by the maximum posterior probability optimization model, thereby improving the tracking accuracy in the height dimension.

Claims

1. A low, slow and small target tracking method based on identification information, characterized by, include: S1, obtaining motion feature information F at the k-1th moment according to the ith track k-1,i obtaining a target type c at the k-1th moment of the ith track according to a target recognition criterion k-1,i establishing an association gate G at the kth moment according to a target state estimation value at the k-1th moment of the ith track k,i wherein k is a positive integer greater than 1, and i is a positive integer. S2, obtaining a free plot set Z falling into the associated gate G of the kth moment k,i of the ith track k , selecting an associated plot z of the kth moment of the ith track from the free plot set Z of the kth moment according to a data association method k of the ith track k,i ; S3, the associated point trail z of the kth moment of the ith track k,i , according to the target tracking method, updating the target state estimation value of the kth moment of the ith track and the estimated covariance matrix P of the kth moment of the ith track k,i ; S4. If the target type at time k-1 is a low, slow, and small target, then establish a height constraint condition. S5, the target state estimation value of the i th track at the k th moment , the estimation covariance matrix P of the i th track at the k th moment k,i , and the height constraint condition, a maximum a posteriori probability optimization model is established to obtain the optimal target state estimation value of the i th track at the k th moment under the height constraint condition , and the estimation covariance matrix update value of the i th track at the k th moment , and let , ; S6. Perform S1 to S5 for each track at time k-1; S7. Let k = k + 1, and execute S1 to S6; The establishment of height constraints includes: S40, determining the maximum height h of the low and slow small target type max ; S41. The height constraint condition is obtained through the second formula, which is: Hx k ≤h max , where H represents the height observation vector, x k represents the true state of the target corresponding to the i th track at the k th moment. The establishment of the maximum a posteriori probability optimization model includes: S50, using the third formula, establishes a maximum a posteriori probability optimization model, wherein the third formula is: ； Constraint: Hx k ≤ h max ; where H represents a height observation vector, h max represents a target height maximum value.

2. The low, slow and small target tracking method based on identification information according to claim 1, characterized in that, The target state estimation value at the k-1 time point according to the i th track Establish the k time point correlation gate G k,i , comprising: S10, the target state estimation value of the i th track at the k-1 th moment obtaining the target state prediction value of the i th track at the k th moment ; S11, obtaining a target state prediction value of the i th track at the k th moment obtaining a position prediction value of the i th track at the k th moment, the position prediction value at the k th moment comprising: a distance prediction value at the k th moment , an azimuth prediction value at the k th moment , and a pitch prediction value at the k th moment ; S12, taking the position prediction value of the kth moment as the center of the kth moment correlation gate G k,i , the first formula being: k,i , the first formula being: where y k represents the range of the free point track associated with the kth time and ith track, R k represents the range measurement in y k , A k represents the azimuth measurement in y k , E k represents the elevation measurement in y k , ΔR represents the range association gate size, ΔA represents the azimuth association gate size, ΔE represents the elevation association gate size, and min[a,b] represents the minimum of input variables a and b.

3. The method for tracking small, slow targets based on identification information according to claim 1, characterized in that, The optimal target state estimation value of the i-th track at the k-th moment under the acquisition height constraint condition And the estimated covariance matrix update value of the i-th track at the k-th moment , comprising: S51, the optimal solution of the maximum a posteriori probability optimization model is the optimal target state estimation value of the i th track at the k th moment under the high constraint condition The specific expression is ； wherein denotes the Lagrange multiplier; S52, the i-th track k moment of estimation covariance matrix update value The expression is ； wherein denotes an intermediate variable, and a denotes the uncertainty due to the target height variation.

Images