Target tracking method and device under infrared active jamming based on probabilistic data association

Abstract

The application discloses a target tracking method and device under infrared active jamming based on probability data association, constructs a model set, determines target and infrared active jamming motion mode according to the model set, simulates an actual combat simulation scene; establishes a coalesced object measurement model and an infrared sensor error adaptive measurement model; estimates the actual size of the coalesced object, and pre-processes measurement information of the coalesced object measurement model and the infrared sensor error adaptive measurement model; data associates a plurality of measurement information, obtains a probability of measurement and track association; processes target motion model and the coalesced object measurement model and the infrared sensor error adaptive measurement model to predict target motion state, updates target motion state prediction; and obtains target state estimation and estimation error covariance. The application solves the problem of no accurate target measurement and measurement quality uncertainty when the target and the infrared active jamming coalesce, and improves the measurement accuracy.

Description

Technical Field

[0001] This invention relates to target tracking technology under infrared active jamming conditions, and in particular to a target tracking method and apparatus under infrared active jamming conditions based on probabilistic data association. Background Technology

[0002] With the widespread use of infrared-guided weapons, corresponding infrared jamming technologies are also constantly evolving. Infrared guidance and infrared jamming are developing under this mutually restrictive relationship.

[0003] Infrared countermeasures are basically approached from two aspects. One is to suppress the infrared radiation emitted by the target, which is known as infrared stealth. The other is to generate active infrared jamming, which involves using a strong infrared radiation source emitted by the aircraft to cause a decoy missile to miss its target.

[0004] Existing research on infrared active jamming technology focuses on signal processing and image processing, addressing issues such as radiation intensity, band radiation, and end-point imaging. Currently, mainstream infrared active jamming algorithms primarily consider the situation after the target and jammer are separated. However, in reality, various complex infrared active jamming phenomena exist. Due to the complexity of infrared active jamming forms, the duration and frequency of jamming, and the differences in target motion states, various infrared jamming scenarios arise, such as infrared active jamming overlapping with the target, forming complex "one-line" or "V-shaped" patterns. Because infrared active jamming and the target have extremely similar radiation characteristics, they cannot be directly distinguished by grayscale in infrared images. When the target and jammer are stuck together, measurement deterioration is severe, with all measurements deviating from the true target, leading to decreased measurement accuracy. To address this problem, it is necessary to consider the target tracking issue in the complex scenario of the target and jammer being stuck together in the early stages of infrared active jamming. Summary of the Invention

[0005] To address the aforementioned deficiencies in the existing technology, the present invention aims to provide a target tracking method based on probabilistic data association under infrared active interference conditions that are complex in form, have varying shapes of target and interference adhesion, and exhibit measurement quality uncertainty.

[0006] The present invention is achieved through the following technical solution.

[0007] According to one aspect of the present invention, a target tracking method based on probabilistic data association under infrared active jamming conditions is provided, comprising:

[0008] A model set including a target motion model and an infrared active jamming model is constructed. Based on the model set, the motion modes of the target and the infrared active jamming are determined to simulate actual combat scenarios.

[0009] Based on simulation scenarios, a measurement model of the adhered object and an adaptive measurement model of the infrared sensor error are established for complex scenarios of target interference and adhesion.

[0010] Based on the distance between the projectile and the target and the line of sight in the actual combat scenario, the actual size of the adhered object is estimated. The gate adaptive method is used to preprocess the measurement information of the adhesion object measurement model and the infrared sensor error adaptive measurement model.

[0011] A probabilistic data association algorithm is used to associate multiple measurement information, including targets and interference, within the gate threshold to obtain the probability of correlation between measurement and track.

[0012] A nonlinear filtering algorithm is used to process the target motion model, the measurement model of the adhered object, and the adaptive measurement model of the infrared sensor error to predict the target motion state. The target motion state prediction is then updated, and the target state estimate and estimation error covariance are obtained as the final filtering result.

[0013] Preferably, a measurement model for the adhesion objects is established for complex scenarios involving target interference and adhesion, including:

[0014] The target is adhered to or covered by interference, forming an adhered object. Based on the uncertainty of the measurement of target and interference information, the centroid of the bounding box of the adhered object, the four vertices of the bounding box and the midpoint of the four sides are used as adhesion measurement features. The number of measurements is increased, and the points in the adhesion measurement that are closer to the actual target measurement are selected to form an adhesion object measurement model.

[0015] One aspect of the present invention provides a target tracking device under infrared active jamming conditions based on probability data association, comprising:

[0016] The simulation module is used to build a model set, determine the movement mode of the target and infrared active jamming based on the model set, and simulate actual combat scenarios.

[0017] The module is used to build a measurement model of the adhered object and an adaptive measurement model of the infrared sensor error in complex scenarios of target interference and adhesion.

[0018] The preprocessing module is used to preprocess the measurement information of the adhesion object measurement model and the infrared sensor error adaptive measurement model using the gate adaptive method.

[0019] The correlation module is used to correlate multiple measurement information, including target and interference information, to obtain the probability of correlation between measurement and track.

[0020] The state prediction and update module is used to process the target motion model, the measurement model of the adhered object, and the adaptive measurement model of the infrared sensor error to predict the target motion state, update the target motion state prediction, and obtain the target state estimate and the estimation error covariance.

[0021] The present invention, by adopting the above technical solution, has the following beneficial effects:

[0022] 1. This invention considers the complex scenario of target and infrared active interference sticking together, and adopts an adhesion measurement model and an error adaptive measurement model to obtain measurements that are as close as possible to the real target. It solves the problems of no accurate target measurement and measurement quality uncertainty when the target and infrared active interference stick together, and improves the measurement accuracy.

[0023] 2. This invention takes into account the problem of varying shapes and sizes of the adhered objects, real-time changes in measurement quality, and the complexity requirements of the algorithm. It determines the gate threshold based on the estimated size of the adhered objects and uses gate adaptive technology to preprocess the received multiple measurement information, eliminating measurements that are obviously impossible to come from the target, thereby reducing the computational load of the subsequent algorithm.

[0024] 3. This invention considers the situation where real target measurements and interference measurements cannot be distinguished under infrared active interference. It uses a probabilistic data association algorithm to process multiple measurement information including the target and interference, calculates the probability that each measurement may originate from the real target, solves the problem of target and measurement data association in multiple measurement scenarios, and improves target tracking accuracy. Attached Figure Description

[0025] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, do not constitute an undue limitation of the invention. In the drawings:

[0026] Figure 1 A schematic diagram of a measurement model for adhered objects;

[0027] Figure 2 The implementation process of a target tracking method under infrared active jamming conditions based on probabilistic data association;

[0028] Figures 3(a) and 4(a) are schematic diagrams of scenario 1 and scenario 2 of the present invention, respectively;

[0029] Figures 3(b) and 4(b) show the target trajectory estimation results calculated by the target tracking method based on probabilistic data association under infrared active interference conditions proposed in this invention in the corresponding scenarios.

[0030] Figures 3(c), 3(d), 4(c), and 4(d) respectively show the target line-of-sight angle estimation results calculated by the algorithm proposed in this invention in the corresponding scenarios;

[0031] Figure 5 This is a schematic diagram of a target tracking device under infrared active interference conditions based on probability data association, as shown in an embodiment of the present invention. Detailed Implementation

[0032] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention, but are not intended to limit the present invention.

[0033] This invention provides a target tracking method under infrared active interference conditions based on probabilistic data association, the implementation process of which is as follows: Figure 2 As shown, it includes the following steps:

[0034] Step 1: Construct a model set including a target motion model and an infrared active jamming model. Based on the model set, determine the motion modes of the target and the infrared active jamming, and simulate actual combat scenarios.

[0035] The specific implementation process is as follows:

[0036] 11) Target Motion Analysis: The most common motions of targets in three-dimensional space are near-uniform linear motion and turning motion. A model set is constructed using a non-maneuvering target uniform motion model (denoted as CV model) and a three-dimensional uniform turning target motion model (3D-CT model). The CV model is suitable for describing near-uniform linear motion with no or low maneuverability, and is used to match the target's conventional cruise flight. The 3D-CT model is suitable for describing near-uniform turning motion with high maneuverability, and is used to match the target's turning maneuvers in a plane and three-dimensional space. Other target motions are described by a weighted average of the two models.

[0037] 12) CV model:

[0038] The simplest maneuvering model is the white noise acceleration model, which assumes that the acceleration components are zero-mean Gaussian white noise. When the target moves at a near-uniform velocity, the state vector describing its motion only takes its position and velocity components. Assuming the target moves in three-dimensional space, it can be used... This describes the target's motion state, where x, y, and z represent the target's position components along the three axes of the navigation frame as it moves in a three-dimensional coordinate system. These represent the velocity components along the three axes of the navigation frame as the target moves in a three-dimensional coordinate system. When the target is moving at a constant velocity in a straight line in three-dimensional space, its motion state can be described by the following state equation:

[0039] x k+1 =diag[F,F,F]x k +diag[G,G,G]w k

[0040]

[0041] in, Let x represent the target state at time k. k ,y k ,z k This represents the components of the target's position on each axis of the coordinate system at time k. Let represent the components of the target's velocity along each axis in the coordinate system at time k, ... k denoted as , where G is the system process noise; G is the noise driving matrix; and T is the sampling interval.

[0042] 13) 3D-CT model

[0043] The three-dimensional uniform turning model is characterized by its angular velocity and velocity being approximately constant, and the angular velocity not necessarily being perpendicular to the velocity. Therefore, it can describe uniform motion along a non-zero torsional curve. It can be used to describe planar maneuvers such as escape maneuvers and S-maneuvers, as well as three-dimensional maneuvers such as roller maneuvers. The target state is defined as follows in the three-dimensional uniform turning model:

[0044]

[0045] Where x, y, z represent the position components of the target along the three axes of the navigation system when it moves in the three-dimensional coordinate system. ω represents the velocity components along the three axes of the navigation frame when the target moves in a three-dimensional coordinate system. x ,ω y ,ω z This represents the angular velocity components along the three axes of the navigation system when the target moves in a three-dimensional coordinate system.

[0046] Its discretization model can be obtained as follows:

[0047] x k+1 =Fx k +diag[kron[G,I3],I3]w k

[0048]

[0049] in

[0050]

[0051]

[0052]

[0053]

[0054]

[0055]

[0056] Where, x k Let I be the target's state vector at time k; let I be the identity matrix; let F be the system state transition function; let G be the noise driving matrix; let T be the sampling interval; w k ~N(0,Q) k () represents process noise.

[0057] 14) Infrared active interference model

[0058] Consider the following nonlinear system:

[0059] x k =f(x) k-1 )+G k w k-1

[0060] In the formula, x k Let f be the system state vector at time k; f be the system state function; w be the system state vector at time k. k For system process noise; G k This is the noise-driven matrix.

[0061] The target state equation above can be used in infrared active jamming scenarios such as infrared decoys and foil-type area source jamming. A corresponding model can be established based on the characteristics of the jamming, and the system state function f can be calculated.

[0062] Taking the infrared radiation source ejected by the target as an example, let's take the state vector. The state transition equation of an infrared radiation source is:

[0063] x k+1 =Fx k +Gu k +w k

[0064]

[0065] In the formula, x k+1 The target state at time k+1; x k Represents the target state at time k; F is the system state transition function; w k G represents the system process noise; G represents the input u. k The gain matrix; u k The system input at time k is T; the sampling interval is C. d ρ is the drag coefficient; g is the air density; S is the acceleration due to gravity; m is the frontal area of ​​the infrared radiation source; ρ is the air density; g is the acceleration due to gravity; S is the frontal area of ​​the infrared radiation source; m k It is the mass of the infrared radiation source at time k; Let be the components of the infrared radiation source's velocity along each axis in the coordinate system at time k-1.

[0066] Assuming air resistance is ignored, the infrared radiation source is only affected by gravity in the y-direction, and the simplified model is as follows:

[0067] x k+1 =Fx k +Gu k +w k

[0068]

[0069] Where, x k+1 The target state at time k+1; x k Represents the target state at time k; F is the system state transition function; w k G represents the system process noise; G represents the input u. k The gain matrix; u k is the system input at time k; T is the sampling interval.

[0070] Step 2: Based on the simulation scenario, establish a measurement model for the adhered object and an adaptive measurement model for the infrared sensor error in the complex scenario of target interference and adhesion.

[0071] The specific implementation process is as follows:

[0072] 21) Establish a measurement model for adhered objects in complex scenarios involving target interference and adhesion.

[0073] In this embodiment, considering the complexity and high radiation energy of infrared active jamming, when a target releases infrared active jamming, the target may adhere to the jamming or even be covered by it. For example, the adhesion between the infrared active jamming and the target may present various complex shapes such as "heart-shaped," "ellipsoidal," "line-shaped," and "V-shaped." In actual combat, the adhesion measurement shape formed after the target and infrared active jamming overlap varies. The measurement information provided by the sensor is the centroid of the adhered object, and the position of the centroid is as follows: Figure 1 As shown by the blue pentagon, accurate measurement of the target cannot be obtained.

[0074] In this invention, based on the adhesion phenomenon between the target and infrared active interference, a measurement model of the adhesion object is established. Specifically, the centroid of the adhesion object's bounding box, the four vertices of the bounding box, and the midpoints of the four sides are used as adhesion measurement feature points. These feature points are then used as measurement information. The measurement after processing by the adhesion object measurement model is denoted as y. l .like Figure 1 As shown by the red triangle, the number of measurements is artificially increased, and points that are closer to the actual target measurement are selected in the adhesion measurement to increase the accuracy of the measurement.

[0075] 22) Establish an error adaptive measurement model.

[0076] In traditional measurement models, the measurement variance (σ) is typically... 2 The value is a pre-designed fixed value. However, during the test, it was found that due to the complexity and variety of the infrared active interference, as well as its high radiation intensity, the target and the interference may stick together. In this case, the obtained input measurement is the centroid of the target and the object that is stuck together. In different scenarios, the degree of mixing between the target and the infrared active interference is different, and the measurement deviation from the actual target measurement is different. Therefore, the measurement error and measurement quality vary greatly for each scenario, and the fixed parameter cannot be adapted to all scenarios.

[0077] In this embodiment of the invention, considering the above-mentioned problems, an error adaptive measurement model is established.

[0078] The estimated bounding box line-of-sight angle ε at time k. k With measurement variance (σ) 2 In conjunction with this, the measurement error changes in real time with the change of the line of sight of the bounding box, realizing the adaptive function of measurement error.

[0079] The error adaptive measurement model for infrared sensors is as follows:

[0080]

[0081]

[0082] In the formula, Z k For the angle measurement of the infrared sensor at time k; θ k The azimuth and elevation angles measured by the sensor; x k ,y k ,z k The position components of the target in the geodetic coordinate system; Noise in azimuth measurement; Noise was measured for pitch angle. The noise variance is measured for the azimuth angle. The pitch angle measurement noise variance; ε is the bounding box line-of-sight angle, and θ1 and θ3 are the pitch angle measurements at the diagonal vertices of the bounding box. For measuring the azimuth angle of the diagonal vertices of the bounding box.

[0083] Step 3: Based on the distance between the projectile and the target and the line-of-sight angle in the actual combat scenario, estimate the actual size of the adhered object, and use the gate adaptive method to preprocess the measurement information of the adhered object measurement model and the infrared sensor error adaptive measurement model.

[0084] 31) Adaptive gate algorithm

[0085] In this embodiment of the invention, the purpose of data preprocessing is to eliminate obviously impossible measurements in order to reduce the amount of subsequent computation. Thresholds are used to exclude impossible measurement track associations.

[0086] In data preprocessing, the threshold is a sub-region of the sensor detection area spanned by the predicted values ​​of the measurements, and its size is generally determined by a threshold value. Measurements falling within the threshold are considered valid and can be used for subsequent track updates.

[0087] The simplest thresholding technique is the rectangular gate, which is used when a measurement y (where the element is y) is applied. l Its measurement residual vector elements If the following formula is satisfied, the measurement is considered to be within the threshold and can be used for track updates.

[0088]

[0089] In the formula, K Gl For variance gain; y l Measurements received by the infrared sensor; For one-step predictive measurement; σ r The standard deviation of the residuals is derived from the measurement variance. and prediction variance Composition, and G is the gate threshold. In practical scenarios,

[0090] This invention proposes a gate-adaptive method for data preprocessing of measurement information in both the adhesion object measurement model and the infrared sensor error adaptive measurement model. This method considers the measurement variance in a traditional rectangular gate. It is a pre-designed fixed value, but due to the differences in the interference time, interference frequency and target motion state of infrared active jamming, the target and the interference may stick together or be covered by the interference in various forms, forming a sticky object. At this time, the input measurement obtained is the centroid of the sticky object of the target and the interference. In different scenarios, the degree of mixing between the target and the interference is different, and the measurement deviation from the real target measurement is different. Therefore, the measurement quality varies greatly for each scenario, and the fixed gate threshold cannot adapt to all scenarios.

[0091] This invention employs gate adaptive technology. The degree of adhesion between the target and infrared active interference varies across different scenarios, and the differences in the shape and size of the adhered measurement object significantly affect measurement accuracy. When the adhered object is heart-shaped, ellipsoidal, or similar in shape with a small length and width, the system's measurement deviation from the actual target is small, and a smaller gate threshold should be used. When the adhered object is straight, V-shaped, or similar in shape with a larger length and width, the system's measurement deviation from the actual target is larger, and a larger gate threshold should be used.

[0092] Specifically, based on the missile-target distance r estimated at time k-1 k-1 and the bounding box view angle ε k Calculate the bounding box size L k :

[0093] L k =r k-1 *ε k

[0094] The estimated bounding box view angle ε k With measurement variance In connection with this, as described in step 2, section 22) regarding the adaptive measurement model for infrared sensor errors, the measurement error changes in real time with the line-of-sight angle of the tracking frame, and the gate threshold also changes in real time accordingly, thus realizing the adaptive function of the gate threshold. If the measurement y received by the infrared sensor... l If the following equation is satisfied, the measurement is considered a valid measurement falling within the track wave threshold, denoted as z. l It can be used for subsequent track updates.

[0095]

[0096] In the formula, K Gl y is the variance gain; l is the measurement number; y l Measurements received by the infrared sensor; For one-step predictive measurement; σ r The standard deviation of the residuals is derived from the measurement variance. and prediction variance Composition, and In real-world scenarios,

[0097] Step 4: Use a probabilistic data association algorithm to process multiple measurement information, including targets and interference, within the gate threshold to obtain the probability of correlation between measurements and tracks.

[0098] The specific implementation process is as follows:

[0099] 41) Probabilistic Data Association Algorithm

[0100] The Probabilistic Data Association (PDA) algorithm assumes that any valid measurement within the gate may originate from the target, but the probability of each measurement originating from the target differs. The goal of the PDA algorithm is to calculate the likelihood of each measurement being associated with the target trajectory at the current moment. This probability (Bayesian information) is used in the PDA tracking algorithm, which considers the uncertainty of measurement origin. If the system model is linear, the resulting PDA algorithm is based on Kalman filtering. If the system model is nonlinear, probabilistic data association is considered within the framework of extended Kalman filtering, particle filtering, etc. Probabilistic data association algorithms have low computational cost, simple structure, and are easy to implement in programming; they remain widely used today for solving single-target tracking and sparsely distributed multi-target tracking problems.

[0101] Define P D P represents the detection probability of the target, i.e., the probability that the target will be detected by the sensor, which is determined by the sensor's performance. G The probability of a target appearing within the gate depends on the gate's design. It is assumed that clutter unrelated to the target's trajectory follows a Poisson distribution with density β (β includes new targets and erroneous measurements). Given N measurements within the gate of track i, the distance from measurement j to track i is... For trajectory i, there are N+1 possible hypotheses. The first hypothesis (defined as H0) is that there are no valid measurements. The probability of H0 occurring is p′. i0 Proportional, where p′ i0 The calculation is shown below.

[0102] p′ i0 =β N (1-P G P D )

[0103] Similarly, measurement j is the hypothesis H of an effective measurement. j The probability of (j=1,2,……N) occurring and p′ ij Proportional.

[0104]

[0105] Finally, the probability p of associating measurement j with track i is calculated by normalizing the above formula. ij .

[0106]

[0107] Element β N-1 It was eliminated during the normalization process, in p ij This is not required in the calculation. Therefore, a simplified form is shown below, where j = 0 indicates that there is no valid measurement.

[0108]

[0109] In the formula:

[0110]

[0111]

[0112] Where M is the dimension of the measurement, S i This is the measurement prediction covariance matrix.

[0113] When the number of actual target tracks is 1, the track number i can be omitted.

[0114] 42) Filter filtering

[0115] An extended Kalman filter (EPF) is selected to handle the nonlinearity of both the estimation and measurement models. Based on the CV and 3D-CT models, the EPF is used for prediction and updating to obtain the target's state estimate at the current time step. covariance P k|k .

[0116] The Extended Kalman Filter (EKF) is developed based on the Kalman Filter. It approximates the nonlinear system by performing a Taylor series expansion of the nonlinear function and taking the lower-order terms, then uses Kalman filtering to handle the filtering problem of the nonlinear system. Therefore, the Extended Kalman Filter is essentially a state estimator that achieves near-optimal estimation by approximating the linearization of the nonlinear system.

[0117] Consider the following nonlinear system:

[0118] x k =f(x) k-1 )+G k w k-1

[0119] z k =h(x k )+v k

[0120] In the formula, x k Let f be the system state vector at time k; f be the system state function; w be the system state vector at time k. k For system process noise; G k The noise driving matrix; z k Let v be the measurement vector at time k; h be the measurement function; v k For measuring noise.

[0121] General assumption w k and v k The mean is 0 and the variances are Q. k and Rk The Gaussian white noise is independent of each other, that is, w k ~N(O,Q k ), v k ~N(O,R k The nonlinear functions f(·) and h(·) are linearized to the first order, that is, after Taylor expansion, only the first two terms are retained, and then the Kalman filter algorithm is used for calculation.

[0122] ①Prediction phase

[0123] Calculate the Jacobian matrix F of the state equations of a dynamic system. k :

[0124]

[0125] Calculate one-step predictive state estimation And the prediction estimation error covariance P k|k-1 :

[0126]

[0127]

[0128] Calculate the Jacobian matrix H of the measurement equation k :

[0129]

[0130] One-step prediction of the measurement value Predicted new vector Measurement prediction covariance matrix S k ;

[0131]

[0132]

[0133]

[0134] In the formula, Let be the j-th measurement value at time k within the gate threshold.

[0135] ②Update Phase

[0136] Calculate the Kalman gain K k and equivalent new information vector

[0137]

[0138]

[0139] In the formula: N represents the number of measurements within the gate; p j The probability of matching j with the track is calculated as described in step 4.1) of the probability data association algorithm.

[0140] Calculate state estimation And estimation error covariance P k|k .

[0141] Extended Kalman filter equation:

[0142]

[0143] Estimated error covariance:

[0144]

[0145] dP represents the Kalman covariance when there is a measurement return value and that measurement value is correctly correlated with the track. k It is an increment that reflects the impact of uncertain correlations on covariance.

[0146] and dP k The calculation formula is:

[0147]

[0148]

[0149]

[0150] In the formula, For target state estimation; P k|k To estimate the error covariance; p0 is the probability that no valid measurement is associated with the track; I is the identity matrix; This represents the covariance of the estimation error for the original Kalman filter.

[0151] The estimation effect of this invention can be illustrated by the following simulation examples:

[0152] To verify the effectiveness of the target tracking method proposed in this invention, which uses a gate adaptive algorithm to handle measurement quality uncertainty and is based on probability data correlation under infrared active interference conditions, the estimation results of the target trajectory and the target line-of-sight angle are shown through two simulation scenarios.

[0153] Scenario 1 and Scenario 2 are shown in Figure 3(a) and Figure 4(a) respectively. In the scenario of Figure 3(a), the target makes a uniform turning motion in a plane. The target releases two batches of infrared active interference at 6 seconds and 12 seconds. Each batch releases 5 interference radiation sources, with a release interval of 0.1 seconds. The sensor platform guides the target to approach it proportionally.

[0154] In the scenario shown in Figure 4(a), the target is making a uniform turning motion in a plane. The target starts releasing infrared active interference at the 4th second. The interference release method is to release interference radiation sources to both sides of the target's flight direction in sequence, with a release interval of 0.1 seconds. The sensor platform guides the target in a proportional manner.

[0155] The initial states of the target in the inertial frame are set as [12000,8000,6000,200,90,0] and [12000,8000,6000,300,50,0t]. The target motion is described in detail in the attached figure. In the simulation scenario, the missiles approach the target with proportional guidance.

[0156] Figures 3(b) and 4(b) show the target trajectory estimation results in the corresponding scenarios. The comparison results in the figures demonstrate that the proposed algorithm can accurately estimate the position information of the real target.

[0157] Figures 3(c), 3(d), 4(c), and 4(d) show the results of target line-of-sight angle estimation in the corresponding scenarios.

[0158] In the figure, the blue dashed line represents the target angle information provided by the infrared camera after image processing, i.e., the measurement information of the centroid of the adhered object. The green solid line represents the target line-of-sight angle estimation result calculated by the algorithm proposed in this invention. The comparison results in the figure show that, under infrared active interference, the method proposed in this invention improves the accuracy of the actual target angle estimation compared to the image processing algorithm.

[0159] like Figure 5 As shown, according to an embodiment of the present invention, a target tracking device 100 is provided for a target tracking method under infrared active interference conditions based on probability data association, for implementing the method, comprising:

[0160] Simulation module 110 is used to establish a model set, determine the target and infrared active jamming movement modes based on the model set, and simulate actual combat scenarios.

[0161] Module 120 is used to establish a measurement model of the adhered object and an adaptive measurement model of the infrared sensor error in complex scenarios of target interference and adhesion.

[0162] Preprocessing module 130 is used to preprocess the measurement information of the adhesion object measurement model and the infrared sensor error adaptive measurement model using the gate adaptive method;

[0163] The association module 140 is used to perform data association on multiple measurement information, including targets and interference, within the gate threshold to obtain the probability of correlation between measurement and track.

[0164] The state prediction update module 150 is used to process the target motion model, the adhesive object measurement model, and the infrared sensor error adaptive measurement model to predict the target motion state, update the target motion state prediction, and obtain the target state estimate and estimation error covariance.

[0165] This device corresponds one-to-one with the above method, and the technical solutions and effects involved are the same as those of the above method.

[0166] In summary, this invention can correctly correlate target measurements, distinguish between infrared active interference and the target, estimate the target's angle information, and accurately track the target's trajectory.

[0167] This invention is not limited to the above embodiments. Based on the technical solutions disclosed in this invention, those skilled in the art can make some substitutions and modifications to some of the technical features without creative effort, and all such substitutions and modifications are within the protection scope of this invention.

Claims

1. A target tracking method under infrared active jamming conditions based on probabilistic data association, characterized in that, include: A model set including a target motion model and an infrared active jamming model is constructed. Based on the model set, the motion modes of the target and the infrared active jamming are determined to simulate actual combat scenarios. Based on simulation scenarios, a measurement model of the adhered object and an adaptive measurement model of the infrared sensor error are established for complex scenarios of target interference and adhesion. Based on the distance between the projectile and the target and the line of sight in the actual combat scenario, the actual size of the adhered object is estimated. The gate adaptive method is used to preprocess the measurement information of the adhesion object measurement model and the infrared sensor error adaptive measurement model. A probabilistic data association algorithm is used to process multiple measurement information, including targets and interference, within the gate threshold to obtain the probability of correlation between measurement and track. A nonlinear filtering algorithm is used to process the target motion model, the adhesion object measurement model, and the infrared sensor error adaptive measurement model to predict the target motion state, and the target motion state prediction is updated. The target's state estimate and estimation error covariance are obtained as the final filtering result; For complex scenarios involving target interference and adhesion, a measurement model for the adhesion objects is established, including: The target is adhered to or covered by interference, forming an adhered object. Based on the uncertainty of the measurement of target and interference information, the centroid of the bounding box of the adhered object, the four vertices of the bounding box and the midpoint of the four sides are used as adhesion measurement features. The number of measurements is increased, and the points in the adhesion measurement that are closer to the actual target measurement are selected to form an adhesion object measurement model. The adaptive measurement model for infrared sensor error is established as follows: In the formula, Infrared sensor Angle measurement at a given moment; The azimuth and elevation angles measured by the sensor; The position components of the target in the geodetic coordinate system; Noise in azimuth measurement; Noise was measured for pitch angle; The noise variance is measured for the azimuth angle. To measure the noise variance for the pitch angle; For the bounding box view angle, , Measure the pitch angle of the diagonal vertices of the bounding box. , For measuring the azimuth angle of the diagonal vertices of the bounding box.

2. The target tracking method under infrared active interference conditions based on probabilistic data association according to claim 1, characterized in that, A set of target motion models is constructed using a non-maneuvering target uniform motion model and a three-dimensional uniform turning target motion model; The model for the uniform motion of a non-maneuvering target is as follows: In the formula, express Constant target status, express The components of the target's position on each axis of the coordinate system at any given time. express The components of the target's velocity along each axis of the coordinate system at any given moment; Represents a diagonal matrix; This is the system state transition function; This refers to system process noise. This is the noise driving matrix; The sampling interval; The three-dimensional uniform turning target motion model is as follows: ， In the formula, This represents the position components of the target along the three axes of the navigation system as it moves in a three-dimensional coordinate system. This represents the velocity components along the three axes of the navigation frame when the target moves in a three-dimensional coordinate system. This represents the angular velocity components along the three axes of the navigation system when the target moves in a three-dimensional coordinate system.

3. The target tracking method under infrared active jamming conditions based on probabilistic data association according to claim 1, characterized in that, The infrared active interference model is constructed as follows: In the formula, for The target status at any given time; express The target status at any given time; This is the system state transition function; This refers to system process noise. For input The gain matrix; for Time system input; The sampling interval; This refers to the drag coefficient; air density; It is the acceleration due to gravity; The windward area of ​​the infrared radiation source; yes The quality of the infrared radiation source at any given moment; for The components of the velocity of the infrared radiation source at any given moment along each axis of the coordinate system.

4. The target tracking method under infrared active interference conditions based on probabilistic data association according to claim 1, characterized in that, The gate adaptive method is used to preprocess the measurement information of the adhesion object measurement model and the infrared sensor error adaptive measurement model, using the following formula: In the formula, Variance gain; Measurements received by the infrared sensor; For one-step predictive measurement; The standard deviation of the residuals. This is the threshold of the wave gate.

5. The target tracking method under infrared active jamming conditions based on probabilistic data association according to claim 1, characterized in that, Probability of measurement trajectory association as follows: In the formula, For measurement It is a flight path The probability of an effective measurement hypothesis; For measurement It is a flight path The probability of an effective measurement hypothesis; l For the summation variable; N For the flight path The number of measurements given within the gate.

6. The target tracking method under infrared active jamming conditions based on probabilistic data association according to claim 1, characterized in that, Nonlinear filtering algorithms are used to process the target motion model, the adhesion object measurement model, and the error adaptive measurement model to predict the target motion state, including: a. Perform predictive state estimation: In the formula, for Time-prediction state estimation for Estimation of the target's motion state at any given time; This is the system state function; b. Calculate the one-step predicted value of the measurement: In the formula, This is a one-step prediction of the measured value; For measurement functions.

7. The target tracking method under infrared active jamming conditions based on probabilistic data association according to claim 1, characterized in that, Update the target motion state prediction, including: Equivalent innovation vector : In the formula, For measurement The probability associated with the flight path; for Time measurement The predicted information vector; N The number of measurements given within the track gate; For measurement number; Target state estimation : In the formula, For predicting state estimation, for Time-based Kalman gain; for The time-equivalent innovation vector; Estimation error covariance : In the formula, This represents the Kalman covariance when there is a measurement return value and that measurement value is correctly correlated with the track. This represents the increment of the effect of uncertain correlation on covariance.

8. A target tracking device under infrared active jamming conditions based on probabilistic data association according to any one of claims 1-7, characterized in that, include: The simulation module is used to establish a model set, determine the movement patterns of targets and infrared active jamming based on the model set, and simulate actual combat scenarios. The module is used to build a measurement model of the adhered object and an adaptive measurement model of the infrared sensor error in complex scenarios of target interference and adhesion. The preprocessing module is used to preprocess the measurement information of the adhesion object measurement model and the infrared sensor error adaptive measurement model using the gate adaptive method. The correlation module is used to correlate multiple measurement information, including targets and interference, within the gate threshold to obtain the probability of correlation between measurement and track. The state prediction and update module is used to process the target motion model, the adhesive object measurement model, and the infrared sensor error adaptive measurement model to predict the target motion state and update the target motion state prediction. The state estimate and estimation error covariance of the target are obtained.

Images