Extended target detection method based on tracking information

Abstract

This invention discloses an extended target detection method based on tracking information, primarily addressing the problem of poor detection performance under low signal-to-noise ratio (SNR) conditions caused by insufficient utilization of prior target tracking information in existing extended target detection methods. The implementation involves selecting a correlation model for the extended target; based on the selected correlation model, using the extended target's tracking information, predicting the target's spatial position and the distribution of its scattering center echo in each range cell, and calculating the predicted SNR information; designing four fusion detectors for the extended target based on the predicted SNR information; and substituting the predicted SNR information into the corresponding extended target fusion detector for target detection according to the different target echo models and computational performance requirements. This invention fully utilizes the extended target's tracking information, improving target detection performance and extending the effective tracking range of radar targets under a given false alarm probability. It can be applied to radar detection.

Description

Technical Field

[0001] This invention relates to the field of radar technology, and in particular to a method for detecting extended targets, which can be applied to radar detection. Background Technology

[0002] With the advancement of electronic information technology, the signal bandwidth of radar systems, both transmitted and received, has gradually increased. This increased bandwidth leads to improved range resolution, causing target echoes to spread across multiple consecutive range cells. In this case, the target cannot be simply considered a point target, but rather an extended target with multiple scattering points. Due to differences in target models, traditional point target detection algorithms are ineffective against range-extended targets in broadband radar signals; therefore, it is necessary to design algorithms for detecting extended targets. Compared to traditional narrowband radar systems, broadband radar systems can acquire target information with greater precision, helping people to more accurately grasp the actual situation of targets, which is crucial in military and civilian fields. Therefore, research on extended target detection methods is of paramount importance.

[0003] There has been considerable research on the problem of extended target detection.

[0004] In the journal *Systems Engineering & Electronics*, 2014, 36(12):2406-2410, Zhang Xiaowei et al. proposed an extended target detection method based on compressed sensing measurements. This method first constructs a sinc basis to sparsely represent the one-dimensional range profile of the extended target; then, using an approximate message passing algorithm, it obtains the correlation coefficient of the one-dimensional range profile represented linearly by the sinc basis from the compressed sensing measurements; finally, it uses an L0-norm-based detector to achieve extended target detection. The sinc basis of this algorithm can effectively sparsely represent the one-dimensional range profile of the extended target, thus optimizing extended target detection.

[0005] In the journal *Journal of Projectiles, Rockets & Guidance*, 2021, 41(02):87-95, Li Nan proposed a range-extended target detector based on auxiliary data matrix estimation. This method first assumes that the clutter covariance matrix is ​​known and derives the target detector under this condition. Secondly, in practice, when the clutter covariance matrix is ​​unknown, the clutter covariance matrix is ​​estimated using auxiliary data. The estimated matrix is ​​then substituted into the detector, resulting in a range-extended target detector based on the auxiliary data matrix estimation, thereby achieving clutter suppression and improving target detection performance.

[0006] In the journal *Modern Radar*, 2009, 31(05):35-38, Wang Xiaohong et al. proposed an accumulation detection algorithm based on a pre-extraction and post-detection processing mode for extended target detection in broadband millimeter-wave frequency stepping radar. Specifically, the algorithm first eliminates oversampling redundancy through a target extraction algorithm to extract a one-dimensional range profile of the target; then, it utilizes the energy of multiple scattering points on the target for accumulation detection. This algorithm can fully utilize the energy of multiple strong scattering points on the target, effectively improving the detection capability of extended-range targets.

[0007] While the aforementioned extended target detection methods can achieve good detection results, their target detection and target tracking are independent. Because the detection results of each frame are sequentially used for target tracking, the target detection does not fully utilize the target tracking information, resulting in very limited detection performance. For example, in low signal-to-noise ratio conditions, they struggle to achieve superior detection performance, affecting the accurate classification and recognition of targets.

[0008] In the journal *Journal of Electronics and Information Technology*, 2016, 38(05):1072-1078, Liu Hongliang et al. proposed an integrated target detection and tracking algorithm based on the concept of cognitive detection. This algorithm targets narrowband radar point targets, utilizing tracking information to assist target detection and improve detection performance. While this method solves the problem of independent target detection and tracking, it cannot be directly applied to extended targets. Therefore, how to utilize tracking information to assist in the detection of extended targets remains an urgent research topic. Summary of the Invention

[0009] The purpose of this invention is to address the shortcomings of the prior art by proposing an extended target detection method based on tracking information, so as to fully utilize the tracking information of the extended target and effectively improve the detection performance of the extended target.

[0010] The objective of this invention is achieved as follows:

[0011] I. Technical Principles

[0012] Cognitive detection refers to a dynamic closed-loop system with high environmental awareness. It perceives its surroundings through interactive autonomous learning and improves the overall performance of the radar system by adjusting signal transmission, reception, and processing methods based on environmental feedback. This cognitive detection, based on prior target information, can fully utilize historical observation information to enhance target detection performance. Existing methods based on cognitive detection primarily detect point targets. This invention transfers this concept to the field of extended target detection, designing a tracking-assisted detection algorithm for extended targets to improve their detection performance.

[0013] II. Implementation Plan

[0014] Based on the above technical principles, this invention first predicts the spatial position of the target and the distribution of its scattering center echo in each range cell based on the tracking information of the extended target; then, it designs a fusion detector for the extended target based on the predicted information, thus achieving a closed loop from tracking to detection. The implementation steps include the following:

[0015] (1) Suppose that the number of range cells occupied by the extended target is N, and select the target motion model, the target scattering center change model, the radar observation model, and the echo signal model as the relevant models of the extended target;

[0016] (2) Predict the signal-to-noise ratio of the target:

[0017] (2a) Based on the observation of the target motion model, Kalman filtering is used to track and predict the target position, and the radial distance prediction value at time k to time k+1 is obtained.

[0018] (2b) Based on the observation of the target scattering center variation model, Kalman filtering is used to track and predict the target RCS, and the predicted RCS value of the i-th range cell at time k+1 is obtained at time k.

[0019] (2c) Combining the radar equations, and based on the prediction information from (2a) and (2b), calculate the predicted signal-to-noise ratio (SNR) at time k for the i-th range cell at time k+1.

[0020] (3) Based on the echo signal model, design an extended target fusion detector based on predicted signal-to-noise ratio:

[0021] (3a) Design the optimal fusion detector T for the Swerling 1 type fluctuating target echo sub-model in the echo signal model. TDET-O1 and suboptimal fusion detector T TDET-S1 :

[0022]

[0023]

[0024] Where n represents the reference window length for estimating the noise, This represents the range cell echo data of the Swerling 1 type undulating target echo sub-model.

[0025] (3b) Design the optimal fusion detector T for the Swerling 3 type fluctuating target echo sub-model in the echo signal model. TDET-O3 and suboptimal fusion detector T TDET-S3 :

[0026]

[0027]

[0028] This represents the range cell echo data of the Swerling 3 type undulating target echo sub-model.

[0029] (4) The signal-to-noise ratio prediction value of the i-th distance cell at time k+1 obtained based on the tracking information at time k is... Substitute the results into the fusion detector designed in step (3) to obtain the detection results of the extended target.

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

[0031] This invention fully utilizes the tracking information of extended targets by calculating the predicted signal-to-noise ratio (SNR) based on the tracking information. Furthermore, by combining the predicted SNR with the design of a fusion detector for extended targets and incorporating the predicted SNR into the fusion detector for extended target detection, it achieves extended target detection assisted by tracking information. Thus, given a false alarm probability, it improves detection performance and extends the effective tracking range of the radar for targets. Attached Figure Description

[0032] Figure 1 This is a flowchart illustrating the implementation of the present invention;

[0033] Figure 2 This is a schematic diagram of the elliptic gate in the existing Kalman filter;

[0034] Figure 3 This is a schematic diagram illustrating the detection performance of the Swerling 1 type fluctuating target echo sub-model in the case of uniform signal-to-noise ratio distribution.

[0035] Figure 4 This is a schematic diagram illustrating the detection performance of the Swerling 1 type fluctuating target echo sub-model in the case of uneven signal-to-noise ratio distribution.

[0036] Figure 5 This is a schematic diagram illustrating the detection performance of the Swerling 3 type undulating target echo sub-model under a uniform signal-to-noise ratio distribution.

[0037] Figure 6 This is a schematic diagram illustrating the detection performance of the Swerling 3 type fluctuating target echo sub-model in the case of uneven signal-to-noise ratio distribution.

[0038] Figure 7 This is a schematic diagram of the detection range of the Swerling 1 type undulating target echo sub-model under the condition of uniform signal-to-noise ratio distribution in this invention;

[0039] Figure 8 This is a schematic diagram of the detection range of the Swerling 1 type undulating target echo sub-model in the case of uneven signal-to-noise ratio distribution. Detailed Implementation

[0040] The embodiments and effects of the present invention will be further described in detail below with reference to the accompanying drawings.

[0041] Reference Figure 1 The implementation steps for this example are as follows:

[0042] Step 1: Select the relevant model for the extended objective;

[0043] Let N be the number of range cells occupied by the extended target. In this example, the target motion model, the target scattering center change model, the radar observation model, and the echo signal model are selected as the relevant models for the extended target. The specific structures of each relevant model are as follows:

[0044] The target motion model includes a uniform velocity model, a constant acceleration model, and a cooperative turning model. This example uses, but is not limited to, the uniform velocity model. In a two-dimensional plane, the target motion state equation of this model is as follows:

[0045] X(k+1)=F(k)X(k)+Γ(k)v(k)

[0046] In the formula, x(k) and y(k) represent the displacements along the x and y axes, respectively. and Let v(k) represent the velocity components along the x and y axes, respectively; v(k) = [v x (k),v y (k)] T v x (k) and v y (k) is a sequence of mutually independent zero-mean Gaussian white noise, representing the process noise along the x and y axes, respectively; F(k) and Γ(k) are the state transition matrix and the process noise distribution matrix, respectively, as follows:

[0047] Where T represents the sampling interval.

[0048] The target scattering center variation model is commonly established using a statistical model. With the radar system parameters remaining constant, the distribution of the target scattering center echo intensity in each range cell can be considered related to the radar cross section (RCS) within each range cell. In practical applications, based on historical observations, the extended target's RCS can be considered to change slowly and independently during target movement. Therefore, this example uses an existing linear model as the target scattering center variation model, and its RCS state equation is as follows:

[0049]

[0050] In the formula, This represents the RCS value within the i-th distance cell at time k. express rate of change, each They are independent of each other; Zero-mean Gaussian white noise, representing the process noise of the RCS within the i-th distance cell, each Independent of each other, F RCS (k), Γ RCS (k) represents the state transition matrix and the process noise distribution matrix, respectively, as follows:

[0051] Where T represents the sampling interval.

[0052] The radar observation model includes a one-dimensional observation model, a two-dimensional observation model, and a three-dimensional observation model. In this example, a two-dimensional observation model is selected to observe the target motion model based on the two-dimensional state equation of the target motion model; a one-dimensional observation model is selected to observe the target scattering center change model based on the one-dimensional state equation of the scattering center change model, wherein:

[0053] The selected radar observation model measures the target position in the target motion model as the radial distance r and azimuth angle θ in polar coordinates. A transformation measurement is applied to r and θ to obtain the target measurement value Z(k) in rectangular coordinates.

[0054] Z(k)=H(k)X(k)+W(k)

[0055] In the formula, W(k) is zero-mean Gaussian white noise, representing measurement noise; H(k) is the measurement matrix, expressed as:

[0056]

[0057] The selected radar observation model uses the RCS measurement value in the target scattering center variation model. for:

[0058]

[0059] In the formula The measured noise of the RCS within the i-th distance cell is represented by each... All are zero-mean Gaussian white noise and are independent of each other.

[0060] The echo signal model includes a non-undulating target echo model, a Swelling 1 type undulating target echo model, a Swelling 2 type undulating target echo model, a Swelling 3 type undulating target echo model, a Swelling 4 type undulating target echo model, and a Swelling 5 type undulating target echo model. This example selects, but is not limited to, the Swelling 1 type undulating target echo model and the Swelling 3 type undulating target echo model as sub-models of the echo signal model, and names them Swelling 1 type undulating target echo sub-model and Swelling 3 type undulating target echo sub-model, respectively.

[0061] The Swerling 1 type undulating target echo sub-model performs matched filtering and coherent accumulation processing on the echo signals of each range cell, and then takes the square of the signal magnitude to obtain the echo data of its reference cell. and the echo data of the unit under test Where i = 1, 2, ..., N, j = 1, 2, ..., n, and n represents the reference window length for estimating noise; in practice, the echo data received by the radar... and It is a random variable, and the characteristics of a random variable need to be described using a probability density function;

[0062] Swerling Type 1 target echo reference cell echo data probability density function The probability density function of the echo data of the unit under test They are represented as follows:

[0063]

[0064]

[0065]

[0066] In the formula, Let be the probability density function of the echo data of the target echo detection unit of Swerling 1 type under the condition that the target does not exist. Let be the probability density function of the echo data of the target echo detection unit of Swerling 1 type under the condition of target presence. express The probability density function independent variable, u (1) express The probability density function independent variable, S i To expand the true signal-to-noise ratio of the echo from the i-th range cell in the target, Let ε(·) represent the noise power of complex Gaussian white noise, and let ε(·) be the unit step function.

[0067] The Swerling 3 type undulating target echo sub-model performs matched filtering and coherent accumulation processing on the echo signals of each range cell, and then squares the signal magnitude to obtain the echo data of its reference cell. and the echo data of the unit under test

[0068] Swerling Type 3 target echo reference cell echo data probability density function The probability density function of the echo data of the unit under test They are represented as follows:

[0069]

[0070]

[0071]

[0072] In the formula, Let be the probability density function of the echo data of the Swerling 3 type target echo detection unit under the condition that the target does not exist. Let y be the probability density function of the echo data of the Swerling 3 type target echo detection unit under the condition of target presence. (3) express The probability density function independent variable, u (3) express The probability density function independent variable.

[0073] Step 2: Predict the signal-to-noise ratio of the target.

[0074] (2.1) Based on the observation of the target motion model, the target is tracked using a Kalman filter, and the predicted x-axis displacement at time k+1 is obtained respectively. and y-axis displacement prediction value Then, based on the predicted values ​​in these two directions, the predicted radial distance from time k to time k+1 is calculated.

[0075]

[0076] (2.2) Based on the observation of the target scattering center variation model, Kalman filtering is used to track and predict the target RCS, and the predicted RCS value of the i-th range cell at time k+1 is obtained.

[0077] (2.3) In the Kalman filtering process, its elliptic correlation gate is established, such as... Figure 2As shown, the aim is to improve computational performance while maintaining a constant false alarm rate. The Kalman filter tracks the in-plane measurement Z(k+1) that meets the following conditions:

[0078]

[0079] In the formula, M(k+1) represents the new information covariance matrix. This represents the prediction of the measurement value at time k+1 from time k, where γ is a parameter representing the gate size, and P... G This represents the gate probability. Since this example uses a two-dimensional observation model, P... G The expression is as follows:

[0080] P G = 1 - exp(-γ / 2)

[0081] (2.4) Calculate the predicted signal-to-noise ratio (SNR) at time k for the i-th distance cell at time k+1. The formula is as follows:

[0082]

[0083] In the formula, P t Indicates radar transmit power; G t Indicates the radar transmitting antenna gain; G r λ represents the radar receiving antenna gain; λ represents the radar transmitting signal wavelength; F n Indicates the noise figure; B represents the Boltzmann constant; T0 = 290K; B represents the radar bandwidth; C B ξ represents the bandwidth correction factor; ξ represents the loss coefficient introduced by the loss of each component.

[0084] Step 3: Based on the echo signal model, design an extended target fusion detector based on the predicted signal-to-noise ratio.

[0085] (3.1) Design the optimal fusion detector T for the Swerling 1 type fluctuating target echo sub-model in the echo signal model. TDET-O1 and suboptimal fusion detector T TDET-S1 The implementation is as follows:

[0086] (3.1.1) For the unknown noise power in step (1.4) Calculate its estimated value, that is, the estimated value of the echo noise power of the i-th distance cell: in Represents the reference cell echo data of the Swerling 1 type fluctuating target echo sub-model;

[0087] (3.1.2) Let the random variable This represents the echo data of the unit under test in the Swerling 1 type fluctuating target echo sub-model.

[0088] (3.1.3) Launch Probability density function under the condition that the target does not exist and the probability density function given the existence of the target. They are represented as follows:

[0089]

[0090]

[0091] In the formula, S i To extend the true signal-to-noise ratio of the echo from the i-th range cell in the target, ε(·) is the unit step function, and z represents The independent variable of the probability density function;

[0092] (3.1.4) Based on the two probability density functions obtained in step (3.1.3), calculate the likelihood ratio detector Λ(Z) for the presence and absence of targets within N distance cells:

[0093]

[0094] (3.1.5) Set a threshold μ based on the false alarm probability of target detection, and compare the likelihood ratio detector Λ(Z) with the threshold μ:

[0095] When Λ(Z)>μ, condition H1 holds, meaning the target exists;

[0096] When Λ(Z) < μ, the H0 condition holds, meaning the target does not exist;

[0097] (3.1.6) Design the optimal fusion detector T for the Swerling 1 type fluctuating target echo sub-model. TDET-O1 and suboptimal fusion detector T TDET-S1 :

[0098] Because in practice, S in equation (3.1.4) i Since the signal-to-noise ratio (SNR) of the i-th distance cell at time k+1 is unknown, when performing target detection at time k+1, it is necessary to use the predicted SNR value of the ith distance cell at time k+1. Replace S i To simplify the formula, let Find the log-likelihood ratio of Λ(Z) and move the constant term to the right side of the equation, and let... Obtain the optimal fusion detector T TDET-O1 :

[0099]

[0100] It can be observed that T TDET-O1 The form is relatively complex. In practical engineering applications, since the estimation of noise power using historical information is more accurate, it can be simplified by taking the limit as n→∞, thus obtaining the suboptimal fusion detector T. TDET-S1 :

[0101]

[0102] (3.2) Using the same method as in step (3.1), design the optimal fusion detector T for the Swerling 3 type fluctuating target echo sub-model. TDET-O3 and suboptimal fusion detector T TDET-S3 :

[0103]

[0104]

[0105] In the formula, This represents the range cell echo data of the Swerling 3 type undulating target echo sub-model.

[0106] Step 4: Predict the signal-to-noise ratio of the i-th distance cell at time k+1 based on the tracking information obtained at time k. Substitute this into the fusion detector designed in step 3 to perform extended target detection at time k+1.

[0107] (4.1) When the target echo originates from a Swelling 1 type undulating target, and higher computational complexity is permissible, when performing extended target detection at time k+1, the following will be used: Substitute the optimal fusion detector T into step (3.1.6). TDET-O1 Perform target detection;

[0108] (4.2) When the target echo originates from a Swerling 1 type undulating target, and high computational complexity is not allowed, when performing extended target detection at time k+1, the following will be used: Substitute the suboptimal fusion detector T into step (3.1.6) TDET-S1 Perform target detection;

[0109] (4.3) When the target echo originates from a Swerling 3 type undulating target, and higher computational complexity is permissible, when performing extended target detection at time k+1, the following will be used: Substitute the optimal fusion detector T into step (3.2) TDET-O3 Perform target detection;

[0110] (4.4) When the target echo originates from a Swerling 3 type undulating target, and high computational complexity is not allowed, when performing extended target detection at time k+1, the following will be used: Substituting the suboptimal fusion detector T into step (3.2) TDET-S3 Perform target detection.

[0111] The effects of this invention are further illustrated by the following simulation experiments:

[0112] I. Simulation Conditions

[0113] The extended target parameter settings for this simulation experiment are shown in Table 1.

[0114] Table 1 Extended Target Parameter Settings

[0115]

[0116] The parameter settings for this simulated radar detection system are shown in Table 2.

[0117] Table 2 Radar Detection System Parameter Settings

[0118] Radar detection system parameters numerical values Distance resolution Δρ 10m <![CDATA[3dB beamwidth θ 3dB > 5° <![CDATA[False alarm probability P FA > 0.00001 <![CDATA[Door probability P G > 0.7

[0119] For both the Swerling Type 1 and Swerling Type 3 undulating target echo sub-models, when the signal-to-noise ratio distribution is uniform, the initial signal-to-noise ratio is set to S. i (0) = 23dB, i = 1, 2, ..., 10;

[0120] The initial signal-to-noise ratio settings for both the Swerling 1 type undulating target echo sub-model and the Swerling 3 type undulating target echo sub-model when the signal-to-noise ratio distribution is uneven are shown in Table 3.

[0121] Table 3 Initial value settings for uneven signal-to-noise ratio distribution

[0122]

[0123]

[0124] II. Simulation Content

[0125] Simulation 1: For the Swerling 1 type fluctuating target echo sub-model, under the condition of uniform signal-to-noise ratio distribution, the optimal fusion detector T designed in this invention is used. TDET-O1 Suboptimal fusion detector T TDET-S1 Compared with existing traditional incoherent accumulation detectors T NCI Simulated target detection, with 1000 Monte Carlo experiments, results are as follows. Figure 3 As shown.

[0126] Depend on Figure 3 It can be seen that in the Swerling 1 type undulating target echo sub-model, the signal-to-noise ratio gradually decreases and the detection probability gradually decreases as the target moves away from the radar; the suboptimal fusion detector T of this invention TDET-S1 Compared with existing traditional incoherent accumulation detectors T NCI The detection performance is roughly the same, because when the signal-to-noise ratio distribution of the extended target is uniform, T TDET-S1 In Since they are approximately equal, their detection performance is roughly the same, but this invention is the optimal fusion detector T. TDET-O1 Optimal detection performance.

[0127] Simulation 2, for the Swerling 1 type fluctuating target echo sub-model, under the condition of non-uniform signal-to-noise ratio distribution, uses the optimal fusion detector T designed in this invention. TDET-O1 Suboptimal fusion detector T TDET-S1 Compared with existing traditional incoherent accumulation detectors T NCI Simulated target detection, with 1000 Monte Carlo experiments, results are as follows. Figure 4 As shown.

[0128] Depend on Figure 4 It can be seen that when the signal-to-noise ratio distribution is uneven, the optimal fusion detector T of this invention... TDET-O1 The second-best fusion detector T of this invention has the best detection performance. TDET-S1 The detection performance is suboptimal, while the existing traditional incoherent accumulation detector T NCI Its detection performance is the worst.

[0129] Simulation 3, for the Swerling 3 type fluctuating target echo sub-model, under the condition of uniform signal-to-noise ratio distribution, uses the optimal fusion detector T designed in this invention. TDET-O3 Suboptimal fusion detector T TDET-S3 Compared with existing traditional incoherent accumulation detectors T NCI Simulated target detection, with 1000 Monte Carlo experiments, results are as follows. Figure 5 As shown.

[0130] Simulation 4, for the Swerling 3 type fluctuating target echo sub-model, under the condition of non-uniform signal-to-noise ratio distribution, uses the optimal fusion detector T designed in this invention. TDET-O3 Suboptimal fusion detector T TDET-S3 Compared with existing traditional incoherent accumulation detectors T NCI Simulated target detection, with 1000 Monte Carlo experiments, results are as follows. Figure 6 As shown.

[0131] Depend on Figure 5 , Figure 6 It can be seen that the optimal fusion detector T of the present invention TDET-O3 The second-best fusion detector T of this invention has the best detection performance. TDET-S3 The detection performance is suboptimal, while the existing traditional incoherent accumulation detector T NCI Its detection performance is the worst.

[0132] Simulation 5, for the Swerling 1 type fluctuating target echo sub-model, under the condition of uniform signal-to-noise ratio distribution, uses the optimal fusion detector T designed in this invention. TDET-O1 Suboptimal fusion detector T TDET-S1 Compared with existing traditional incoherent accumulation detectors T NCI The simulation of target detection range involved 10 Monte Carlo experiments. If a target was detected, it was marked at the corresponding measurement location. The results are as follows: Figure 7 As shown.

[0133] Simulation 6, for the Swerling 1 type fluctuating target echo sub-model, under the condition of non-uniform signal-to-noise ratio distribution, uses the optimal fusion detector T designed in this invention. TDET-O1 Suboptimal fusion detector T TDET-S1 Compared with existing traditional incoherent accumulation detectors T NCI The simulation of target detection range involved 10 Monte Carlo experiments. If a target was detected, it was marked at the corresponding measurement location. The results are as follows: Figure 8 As shown.

[0134] Depend on Figure 7 , Figure 8 As can be seen, when the target is close to the radar, the signal-to-noise ratio is high, and all three detectors can detect the target well, resulting in a very dense measurement point; as the target moves away from the radar, the signal-to-noise ratio decreases, the detection probability decreases, and the measurement points gradually become sparse.

[0135] When the signal-to-noise ratio is uniformly distributed, the suboptimal fusion detector T of the present invention TDET-S1 Compared with existing traditional incoherent accumulation detectors T NCI Both have roughly the same detection performance, and both struggle to detect targets at long distances. However, the optimal fusion detector T of this invention... TDET-O1 It clearly has the longest detection range, indicating that it has the longest effective tracking range for targets;

[0136] When the signal-to-noise ratio is not uniformly distributed, T TDET-O1 The longest detection range, T TDET-S1 Secondly, T NCI Recently, it was explained that T TDET-O1 It has the longest effective tracking distance for targets.

[0137] Based on the simulation results above, the extended target detection method assisted by tracking information of the present invention, compared with the traditional extended target detection method, can not only improve the extended target detection performance under the premise of a given false alarm probability, but also extend the effective tracking distance of the radar on the target.

[0138] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method of extended object detection assisted by tracking information, characterized in that, The method comprises the following steps: (1) setting the number of distance units occupied by the extended target as N, selecting a target motion model, a target scattering center variation model, a radar observation model and an echo signal model as the related models of the extended target; (2) predicting the signal-to-noise ratio of the target: (2a) using Kalman filtering to track and predict the target position according to the observation of the target motion model, to obtain a radial distance prediction value from k time to k+1 time (2b) using Kalman filter to track and predict the RCS of the target according to the observation of the target scattering center variation model, to obtain the RCS prediction value of the i-th range cell at k+1 time (2c) Combining the radar equation, calculate the predicted value of the signal-to-noise ratio in the ith range cell from time k to time k+1 according to the predicted information of (2a) and (2b) (3) designing an extended target fusion detector based on the predicted signal-to-noise ratio according to the echo signal model: (3a) Optimal fusion detector T for Swerling 1 fluctuating target echo sub-model in echo signal model design TDET-O1 and suboptimal fusion detector T TDET-S1 : where n denotes a reference window length of estimating noise, range cell echo data representing a Swerling 1 type fluctuating target echo sub-model, (3b) Optimal fusion detector T for Swerling 3 fluctuating target echo sub-model in echo signal model design TDET-O3 and suboptimal fusion detector T TDET-S3 : range cell echo data representing a Swerling 3 type fluctuating target echo sub-model; (4) The k+1 time SNR prediction value of the ith distance unit based on the tracking information The detection result of the extended target is obtained by substituting into the fusion detector designed in step (3).

2. The method of claim 1, wherein, The target motion model selected in step (1) has the following target motion state equation: X(k+1) = F(k)X(k) + Γ(k)v(k) wherein x(k) and y(k) represent the displacement of the x and y axes, respectively, and v(k) = [v x (k), v y (k)] T , v x (k) and v y (k) are mutually independent zero-mean Gaussian white noise sequences representing the process noise of the x and y axes, respectively; F(k), Γ(k) are the state transition matrix and the process noise distribution matrix, respectively, and are given by: Wherein T represents a sampling interval.

3. The method of claim 1, wherein, The target scattering center variation model selected in step (1) has the following RCS state equation: In the formula, This represents the RCS value within the i-th distance cell at time k. express rate of change, each They are independent of each other; Zero-mean Gaussian white noise, representing the process noise of the RCS within the i-th distance cell, each Independent of each other, F RCS (k), Γ RCS (k) represents the state transition matrix and the process noise distribution matrix, respectively, as follows: Wherein T represents a sampling interval.

4. The method of claim 1, wherein, The radar observation model selected in step (1) is used to observe the target motion model and the target scattering center variation model, wherein: The measurement value of the target position in the target motion model is the target radial distance r and the azimuth angle θ in the polar coordinate, and the conversion measurement is used for r and θ to obtain the measurement value Z(k) of the target in the rectangular coordinate: Z(k) = H(k)X(k) + W(k) In the formula, W(k) is a zero-mean Gaussian white noise, representing measurement noise; H(k) is a measurement matrix, represented as: Measurement of target RCS in target scatter center variation model is: wherein represents the measurement noise of the RCS in the ith range cell, each are zero-mean Gaussian white noises and are mutually independent.

5. The method of claim 1, wherein, The echo signal model selected in step (1) comprises a Swerling 1 type fluctuation target echo sub-model and a Swerling 3 type fluctuation target echo sub-model; The echo signal of each distance unit is matched filtered and coherent accumulated by the Swerling 1 type fluctuation target echo sub-model, and the square of the signal modulus is taken to obtain reference unit echo data of the target and the echo data of the unit to be detected Wherein, i = 1, 2,..., N, j = 1, 2,..., n; The Swerling 3 type fluctuation target echo sub-model matches and filters the echo signals of each distance unit, carries out coherent accumulation processing, and obtains the reference unit echo data after taking the square of the signal modulus and the to-be-detected unit echo data 6. The method of claim 1, wherein, In step (2a), the target position is tracked and predicted using Kalman filtering according to the observation of the target motion model. First, the Kalman filtering is used to obtain the predicted values of the x-axis displacement and the y-axis displacement of the target at time k+1 with respect to time k Then, the predicted value of the radial distance of the target at time k+1 with respect to time k is calculated according to the predicted values in the two directions ​ 7. The method of claim 1, wherein, The signal-to-noise ratio prediction value at time k+1 in the i-th distance unit is calculated in step (2c) The formula is as follows: where P t represents the radar transmit power; G t represents the radar transmit antenna gain; G r represents the radar receive antenna gain; λ represents the radar transmit signal wavelength; F n represents the noise figure; K represents the Boltzmann constant; T0= 290 K; B represents the radar bandwidth; C B represents the bandwidth correction factor; ξ represents the loss coefficient introduced by each part loss.

8. The method of claim 1, wherein, The optimal fusion detector T of the Swerling 1 fluctuation target echo sub-model in the echo signal model designed in step (3a) TDET-O1 and the suboptimal fusion detector T TDET-S1 are implemented as follows: (3a1) Calculate an estimate of the ith range cell echo noise power: Reference cell echo data representing a Swerling 1 fluctuating target echo sub-model; (3a2) Let the random variable The detected cell echo data representing a Swerling 1 type fluctuation target echo sub-model, deduce The probability density function under the condition that the target does not exist And the probability density function under the condition that the target exists Respectively represent as follows: In the formula, S i is the real signal-to-noise ratio of the echo of the i-th range cell in the target, ε(·) is a unit step function, and z represents the argument of the probability density function; (3a3) calculating the likelihood ratio detector Λ(Z) of the existence of the target and the non-existence of the target in the N distance units according to the two probability density functions obtained in step (3a2): In the formula, μ represents the threshold of the likelihood ratio detector; (3a4) Take the log-likelihood ratio of Λ(Z) and move the constant term to the right side of the equation, and let The optimal fusion detector T for the Swerling 1 fluctuating target echo sub-model is obtained TDET-O1 : Taking the limit as n→∞ to simplify the above equation, the suboptimal fusion detector T of Swerling 1 fluctuation target echo submodel is obtained TDET-S1 :

Images