Aircraft detection method and system based on adaptive interactive multi-model algorithm

Abstract

The invention discloses an aircraft detection method and system based on a self-adaptive interactive multi-model algorithm, and belongs to the technical field of strong attack and defense countermeasures. The method comprises the following steps: determining a motion model in a motion process and a Markov probability transfer matrix at a previous moment; estimating a state variable and a model probability of each motion model at a previous moment; calculating a state variable, a state vector covariance matrix and a model probability of each motion model at the next moment; state variable fusion and state vector covariance matrix fusion are carried out; when the detection is not finished, updating the Markov probability transfer matrix at the previous moment; and updating the next moment to the previous moment, and taking the updated Markov probability transition matrix as the Markov probability transition matrix of the previous moment. According to the method, errors caused by motion model matching delay and lagging can be reduced, the state estimation effect of the multi-mode maneuvering target is improved, and the detection precision of a high-value target is greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of strong attack and defense confrontation, and in particular relates to an aircraft detection method and system based on an adaptive interactive multi-model algorithm. Background Art

[0002] In a strong offensive and defensive confrontation environment, the motion law of high-value targets in a relatively certain period of time can be equivalently decomposed into multiple basic models. Using multi-modal fitting to fit the complex motion process of the target is a feasible method. When detecting and estimating the state of maneuvering targets, some traditional filtering and detection methods can achieve good results when the target is maneuvering in a single mode, but the effect is not ideal when the target is maneuvering in multiple modes.

[0003] Most of the existing research on basic IMM algorithms adjust the probability of matching models through a pre-set Markov probability transfer matrix. However, these IMM algorithms based on certain prior knowledge and statistical laws can only adjust the probability of matching models after the target modality has changed for a period of time, and there is obvious delay and lag in the model switching process. Summary of the invention

[0004] One of the purposes of the present invention is to provide an aircraft detection method based on an adaptive interactive multi-model algorithm, which can reduce the errors caused by motion model matching delay and lag, improve the state estimation effect of multi-modal maneuvering targets, and greatly improve the detection accuracy of high-value targets.

[0005] A second object of the present invention is to provide an aircraft detection system based on an adaptive interactive multi-model algorithm.

[0006] In order to achieve one of the above purposes, the present invention adopts the following technical solution: An aircraft detection method based on an adaptive interactive multi-model algorithm, the aircraft detection method comprising: Step S1, using an aircraft to detect a target to obtain a motion model of the target during motion, and determining the motion model during motion and a Markov probability transfer matrix at a previous moment; Step S2, estimating the state variables and model probabilities of each motion model at the previous moment; Step S3, using the Markov probability transfer matrix and the estimation result, calculating the state variables, state vector covariance matrix and model probability of each motion model at the next moment; Step S4, using the state variables, state vector covariance matrix and model probability of each motion model at the next moment, perform state variable fusion and state vector covariance matrix fusion in sequence; Step S5: When the detection is not finished, the Markov probability transfer matrix at the previous moment is updated using the model probability of each motion model at the next moment; Step S6, updating the next moment to the previous moment, using the updated Markov probability transfer matrix as the Markov probability transfer matrix of the previous moment, returning to step S2, and continuing to perform detection until the detection is completed.

[0007] Furthermore, the specific implementation process of step S3 includes: Step S31, using the estimation result and the Markov probability transfer matrix, calculate the initial values ​​of the state variables and the state vector covariance matrix of each motion model when filtering at the previous moment; Step S32, determining the residual, residual covariance, state variable and state vector covariance matrix of each motion model at the next moment according to the initial values ​​of the state variable and state vector covariance matrix of each motion model when filtering at the previous moment; Step S33, using the model probability estimation result of each motion model at the previous moment, the Markov probability transfer matrix, and the residual and residual covariance at the next moment, calculate the model probability of each motion model at the next moment.

[0008] Furthermore, the specific process of step S32 includes: Step S321, using the initial values ​​of the state variables and state vector covariance matrix of each motion model when filtering at the previous moment, predict the state variables and state vector covariance matrix of each motion model at the next moment; Step S322: using the state variables and state vector covariance matrix of each motion model at the next moment to make predictions, calculating the residual and residual covariance of each motion model at the corresponding moment, so as to determine the filter gain of each motion model at the corresponding moment; Step S323: using the residual, residual covariance and filter gain of each motion model at the next moment, update the prediction results of the state variables and state vector covariance matrix of each motion model at the corresponding moment.

[0009] Furthermore, in step S5, the specific process of the update includes: Step S51, using the model probability of each motion model at the next moment and the Markov probability transfer matrix at the previous moment, calculate the error compression ratio of any two motion models; Step S52: using the error compression ratio, correct the corresponding elements in the Markov probability transfer matrix at the previous moment.

[0010] Further, in step S51, the error compression ratio is:

[0011] in, For the i The motion model and j The error compression ratio of the motion model; is the first probability transfer matrix of the Markov probability transfer matrix at the previous moment. i Row and j The element value corresponding to the column; is the first probability transfer matrix of the Markov probability transfer matrix at the previous moment. j Row and i The element value corresponding to the column; and Respectively i The motion model and j The motion model k The model probability at time .

[0012] Furthermore, in step S52, correction is performed according to the following formula: ; ; in, is the first in the modified Markov probability transfer matrix of the previous moment i Row and i The element correction value corresponding to the column; is the first in the modified Markov probability transfer matrix of the previous moment i Row and j The element correction value corresponding to the column; l is the exponential factor; , , , N is the number of motion models.

[0013] In order to achieve the second objective above, the present invention adopts the following technical solution: An aircraft detection system based on an adaptive interactive multi-model algorithm, the aircraft detection system comprising: The detection module is configured to: detect the target using the aircraft to obtain the motion model of the target during the motion process, and determine the motion model during the motion process and the Markov probability transfer matrix at the previous moment; The estimation module is configured to: estimate the state variables and model probabilities of each motion model at a previous moment; The calculation module is configured to: use the Markov probability transfer matrix and the estimation result, adopt the IMM algorithm, and calculate the state variable, state vector covariance matrix and model probability of each motion model at the next moment; The fusion module is configured to: utilize the state variables, state vector covariance matrix and model probability of each motion model at the next moment to sequentially perform state variable fusion and state vector covariance matrix fusion; The judgment module is configured to: judge whether the detection is finished, if so, output the state variable fusion result and the state vector covariance matrix fusion result, and end; if not, use the model probability of each motion model at the next moment to update the Markov probability transfer matrix at the previous moment and transmit it to the update module; The updating module is configured to update the next moment to the previous moment, use the updated Markov probability transfer matrix as the Markov probability transfer matrix of the previous moment and transmit it to the estimation module to continue to perform detection until the detection is completed.

[0014] Furthermore, the calculation module includes The first calculation submodule is configured to: calculate the initial value of the state variable and the state vector covariance matrix of each motion model when filtering at the previous moment by using the estimation result and the Markov probability transfer matrix; The determination submodule is configured to: determine the residual, residual covariance, state variable and state vector covariance matrix of each motion model at the next moment according to the initial values ​​of the state variable and state vector covariance matrix of each motion model when filtering at the previous moment; The second calculation submodule is configured to calculate the model probability of each motion model at the next moment by using the model probability estimation result of each motion model at the previous moment, the Markov probability transfer matrix, and the residual and residual covariance at the next moment.

[0015] Furthermore, the determining submodule includes: The prediction subunit is configured to: predict the state variables and the state vector covariance matrix of each motion model at the next moment by using the initial values ​​of the state variables and the state vector covariance matrix of each motion model when filtering at the previous moment; The calculation subunit is configured to: use the state variables and state vector covariance matrix of each motion model at the next moment to make predictions, calculate the residual and residual covariance of each motion model at the corresponding moment, so as to determine the filter gain of each motion model at the corresponding moment; The updating subunit is configured to update the prediction results of the state variables and the state vector covariance matrix of each motion model at the corresponding moment by using the residual, residual covariance and filter gain of each motion model at the next moment.

[0016] Furthermore, the judging module includes: The third calculation submodule is configured to: calculate the error compression ratio of any two motion models using the model probability of each motion model at the next moment and the Markov probability transfer matrix at the previous moment; The correction submodule is configured to: use the error compression ratio to correct the corresponding elements in the Markov probability transfer matrix at the previous moment.

[0017] In summary, the technical solution of the present invention has the following technical effects: This embodiment calculates the state variables, state vector covariance matrix and model probability of each motion model at the next moment through the motion model in the motion process and the Markov probability transfer matrix at the previous moment, as well as the estimation results of the state variables and model probabilities at the previous moment, so as to realize the state variable fusion and state vector covariance matrix fusion of all motion models in the motion process; and uses the model probability of each motion model at the next moment to realize the adaptive adjustment of the Markov probability transfer matrix, thereby reducing the errors caused by the delay and lag in the switching process of multiple motion models; the present invention can adaptively and quickly switch to the matching model according to the current matching situation, has a good effect in state estimation of multi-modal maneuvering targets, can greatly improve the detection accuracy of the aircraft to the maneuvering target, ensure that the aircraft effectively obtains the target detection information, and provide method support for the detection of maneuvering targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a schematic flow chart of an aircraft detection method based on an adaptive interactive multi-model algorithm of the present invention; Figure 2 Schematic diagram of aircraft detection of targets. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0021] In the actual observation environment, the maneuvering form of the target is unknown and changing, but the target's motion can be fitted by some basic motion models (such as maneuvering turning model and uniform motion model). In the detection process, the position and velocity of the target are taken as state variables, that is, , then the target state equation can be expressed as: ; (1)

[0022] in, for k System status at all times, is the state transfer matrix, is the noise transfer matrix, is the process noise variable, where the state transfer matrix All eigenvalues ​​of are in the left half of the S plane and have the characteristics of slow time variation and boundedness.

[0023] Aircraft are equipped with corresponding sensors that can independently measure distance and angle, such as Figure 2 As shown. Then the aircraft The observation equation for the maneuvering target can be expressed as: ; (2) in, .

[0024] Linearizing it gives

[0025] Then the measurement equation is: ; (3) In formulas (2) and (3), and Represents aircraft The distance and azimuth to the target, and Indicates aircraft exist k The vehicle's own position at any moment is generally calculated by its own navigation system. For the i Aircraft exist k The observed value at time, For the i Aircraft (i.e. sports model) exist k The observation matrix at time . represents the observation noise, and and are independent zero-mean Gaussian white noises, and their variances are and Equations (1) and (3) constitute the state equation and measurement equation for the aircraft to detect the target.

[0026] Based on the above principles, this embodiment provides an aircraft detection method based on an adaptive interactive multi-model algorithm. Figure 1 , the aircraft detection method comprises: Step S1: Detecting a target using an aircraft to obtain a motion model of the target during motion, and determining the motion model during motion and a Markov probability transfer matrix at a previous moment.

[0027] Assume that the IMM algorithm includes N A motion model, k At time -1, the state variables and covariance matrix of each motion model are expressed as and , and the probability of each model is .

[0028] The Markov probability transfer matrix in this embodiment is: ; (4) in, for k -1 Markov probability transfer matrix; for k -1 time in the Markov probability transfer matrix i Row and j The element value corresponding to the column (i.e. k -1 moment i The motion model is transferred to j The probability of a motion model), , , , N is the number of motion models.

[0029] Step S2: Estimate the state variables and model probabilities of each motion model at the previous moment.

[0030] Step S3, using the Markov probability transfer matrix and the estimation result, adopting the IMM algorithm, calculate the state variables, state vector covariance matrix and model probability of each motion model at the next moment.

[0031] The specific implementation process of this step includes: Step S31, using the estimation result and the Markov probability transfer matrix, calculate the initial values ​​of the state variables and the state vector covariance matrix of each motion model when filtering at the previous moment; According to the various motion models k-1 time state variables, covariance matrix and model probability value, state interaction, produce k The initial value of the moment filter.

[0032] In this embodiment, the initial value of the state variable is: ; (5) in, For the j The motion model k -1 is the initial value of the state variable during filtering; For the i The motion model k State variable estimation results at time -1; For the i The motion model k -1 time model probability estimation result. In this embodiment, the initial value of the state vector covariance matrix is:

[0033] ; (6) in, For the j The motion model k -1 is the initial value of the state vector covariance matrix when filtering; and Respectively i The motion model k The true value of the state variable at time -1 and the j The motion model k State variable estimation results at time -1; For the i The motion model k The true value of the state vector covariance matrix at time -1; is the transposed matrix.

[0034] Step S32: Determine the residual, residual covariance, state variable and state vector covariance matrix of each motion model at the next moment according to the initial values ​​of the state variable and state vector covariance matrix of each motion model when filtering at the previous moment.

[0035] The specific process of step S32 includes: Step S321: using the initial values ​​of the state variables and state vector covariance matrix of each motion model at the previous moment when filtering, predict the state variables and state vector covariance matrix of each motion model at the next moment.

[0036] This embodiment performs prediction according to the following formula: ; (7) ; (8) in, For k -1 moment prediction j The motion model k The predicted value of the state variable at time t; For the j The motion model k -1 time state transfer matrix; For the j The motion model k -1 The initial value of the state variable during filtering; For k -1 moment prediction j The motion model k The predicted value of the state vector covariance matrix at time t; for k The noise transfer matrix at time -1; For the j The motion model k The variance of the process noise variable at time -1.

[0037] Step S322: Use the state variables and state vector covariance matrix of each motion model at the next moment to make predictions, calculate the residual and residual covariance of each motion model at the corresponding moment, and determine the filter gain of each motion model at the corresponding moment.

[0038] The residuals and residual covariances in this embodiment are: ; ; (9) in, and Respectively j The motion model k Residuals and residual covariances at time instants; For the j The motion model k The observed value at time; For the j The motion model k The observation matrix at time; and Respectively in k -1 moment prediction j Motion Model k The predicted values ​​of the state variables and the predicted values ​​of the state vector covariance matrix at the moment; For the j The motion model k The variance of the observation noise at time .

[0039] The filter gain in this embodiment is: ; (10) in, For the j The motion model k The filter gain at time.

[0040] Step S323: using the residual, residual covariance and filter gain of each motion model at the next moment, update the prediction results of the state variables and state vector covariance matrix of each motion model at the corresponding moment.

[0041] This embodiment is updated according to the following formula: ; (11) ; (12) in, For the j The motion model k Updated value of state variables at the moment; For k -1 moment prediction j Motion Model k The predicted value of the state variable at time t; For the j The motion model k The filter gain at the moment; For the j The motion model k The residual of the moment; For the j The motion model k The updated value of the state vector covariance matrix at time; For k -1 moment prediction j The motion model k The predicted value of the state vector covariance matrix at time t; For the j The motion model k The residual covariance at time t.

[0042] Step S33, using the model probability estimation result of each motion model at the previous moment, the Markov probability transfer matrix, and the residual and residual covariance at the next moment, calculate the model probability of each motion model at the next moment.

[0043] The model probability in this embodiment is: ; (13) ; (14) in, For the j The motion model k Model probability at time; is an intermediate variable; for k -1 time in the Markov probability transfer matrix i Row and j The element value corresponding to the column; For the i The motion model k Model probability estimation results at time -1.

[0044] Step S4: using the state variables, state vector covariance matrix and model probability of each motion model at the next moment, perform state variable fusion and state vector covariance matrix fusion in sequence.

[0045] In this embodiment, state variable fusion and state vector covariance matrix fusion are performed according to the following formula: ; (15) ; (16) in, is the state variable fusion result; is the state vector covariance matrix fusion result.

[0046] Step S5: When the detection is not completed, the model probability of each motion model at the next moment is used to update the Markov probability transfer matrix at the previous moment.

[0047] In the above IMM basic algorithm, the Markov probability transfer matrix It plays a vital role in updating the model probability. However, in the traditional IMM algorithm, Most of them are obtained through prior knowledge and remain unchanged throughout the filtering process, which means that the model probability can only be updated after the target mode changes for a period of time. The target maneuvering mode cannot be adaptively updated, and the corresponding model probability cannot be adaptively updated either, which causes obvious delays and lags in the model switching process, thereby causing certain errors. This embodiment can achieve the adaptive update of the Markov transfer matrix by giving the model error compression ratio. The specific process of adaptive adjustment and updating of the matrix includes: Step S51, using the model probability of each motion model at the next moment and the Markov probability transfer matrix at the previous moment, calculate the error compression ratio of any two motion models.

[0048] The error compression ratio in this embodiment is: ; (17) in, For the i The motion model and j The error compression ratio of the motion model; is the first probability transfer matrix of the Markov probability transfer matrix at the previous moment. i Row and j The element value corresponding to the column; is the first probability transfer matrix of the Markov probability transfer matrix at the previous moment. j Row and i The element value corresponding to the column; and Respectively i The motion model and j The motion model k The model probability at time .

[0049] Step S52: using the error compression ratio, correct the corresponding elements in the Markov probability transfer matrix at the previous moment.

[0050] The corrected previous moment (i.e. k -1 moment) Markov probability transfer matrix for: ; (18) This embodiment is corrected according to the following formula: ; (19) ; (20) in, is the first in the modified Markov probability transfer matrix of the previous moment i Row and i Correction values ​​of the elements corresponding to the columns (i.e. diagonal elements); is the Markov probability transfer matrix i Row and j Corrected values ​​of the elements corresponding to the columns (i.e., non-diagonal elements); l is the exponential factor; , , , N is the number of motion models.

[0051] Since the Markov probability transfer matrix It is generally a symmetric positive definite matrix with dominant main diagonal. When , it can be known from formula (17) ;when hour, This shows that the ratio of the corrected compression ratio The adaptive adjustment of the Markov matrix can be realized, and the adaptive updating method can automatically increase the transition probability of the matching model and reduce the transition probability of the unmatched model.

[0052] Step S6, updating the next moment to the previous moment, using the updated Markov probability transfer matrix as the Markov probability transfer matrix of the previous moment, returning to step S2, and continuing to perform detection until the detection is completed.

[0053] In order to verify the effectiveness of the technical solution of this embodiment, a simulation test was carried out: The simulation conditions are set as follows: the total simulation time is 180s, the sampling period is T =1s, the initial position of the aircraft is (-200m, 1000m), and the maneuvering speed is 120m / s; the target performs a 0.8° / s left turn maneuver from 30s to 80s, and a 0.8° / s right turn maneuver from 130s to 180s, and maintains a uniform speed for the rest of the time. The initial state of the target's position and speed is , the system noise and observation noise are independent of each other and are and .

[0054] In the above target maneuvering situation, the IMM algorithm uses three models, which are set as: uniform motion model, 0.8° / s left maneuvering turning model and 0.8° / s right maneuvering turning model. The initial value of the model probability is , the initial value of the Markov probability transfer matrix is:

[0055] The aircraft respectively adopted the improved adaptive IMM filtering algorithm (Improved AdaptiveIMM, IA-IMM) and the basic IMM algorithm proposed in this embodiment, and conducted 300 Monte Carlo tests respectively, and took the root mean square error (Root-mean Square Error, RMSE) of each parameter as the evaluation index.

[0056] RMSE can reflect the detection accuracy, and its calculation method is: (twenty one) In formula (21), n Indicates the number of simulations. x represents the true value, Indicated in k Estimated value at time.

[0057] Taking a simulation as an example, the motion trajectory of the target is predicted by the technical solution of this embodiment, which is basically consistent with the actual trajectory with a small error. The algorithm proposed in this embodiment (the improved adaptive IMM filtering algorithm, i.e. the technical solution of this embodiment) can quickly switch to the matching model according to the maneuvering situation of the target and significantly increase the probability of matching the model. The basic IMM algorithm switches relatively slowly and cannot effectively increase the probability of matching the model, which ultimately leads to an increase in tracking error and a decrease in filtering accuracy.

[0058] The root mean square error of position and velocity for 300 trials was calculated. In the short period when the target maneuver mode is switched, the error will fluctuate to a certain extent (such as at 1s, 30s and 130s), which is caused by the incomplete matching during the model transition process. In addition, since the proposed method can adaptively adjust the probability transfer matrix, once switched to the matching model, the adaptive algorithm can effectively reduce the tracking error. From the algorithm performance comparison results in Table 1, it can be seen that the proposed adaptive IMM algorithm effectively improves the position and velocity tracking accuracy. In the comprehensive x and y directions, the accuracy is improved by 6.21% and 8.8%, respectively.

[0059]

[0060] The above embodiment can be implemented by adopting the technical solutions given in the following embodiments: Another embodiment provides an aircraft detection system based on an adaptive interactive multiple model algorithm, the aircraft detection system comprising: The detection module is configured to: detect the target using the aircraft to obtain the motion model of the target during the motion process, and determine the motion model during the motion process and the Markov probability transfer matrix at the previous moment; The estimation module is configured to: estimate the state variables and model probabilities of each motion model at a previous moment; The calculation module is configured to: use the Markov probability transfer matrix and the estimation result, adopt the IMM algorithm, and calculate the state variable, state vector covariance matrix and model probability of each motion model at the next moment; The fusion module is configured to: utilize the state variables, state vector covariance matrix and model probability of each motion model at the next moment to sequentially perform state variable fusion and state vector covariance matrix fusion; The judgment module is configured to: judge whether the detection is finished, if so, output the state variable fusion result and the state vector covariance matrix fusion result, and end; if not, use the model probability of each motion model at the next moment to update the Markov probability transfer matrix at the previous moment and transmit it to the update module; The updating module is configured to update the next moment to the previous moment, use the updated Markov probability transfer matrix as the Markov probability transfer matrix of the previous moment and transmit it to the estimation module to continue to perform detection until the detection is completed.

[0061] Furthermore, the calculation module includes The first calculation submodule is configured to: calculate the initial value of the state variable and the state vector covariance matrix of each motion model when filtering at the previous moment by using the estimation result and the Markov probability transfer matrix; The determination submodule is configured to: determine the residual, residual covariance, state variable and state vector covariance matrix of each motion model at the next moment according to the initial values ​​of the state variable and state vector covariance matrix of each motion model when filtering at the previous moment; The second calculation submodule is configured to calculate the model probability of each motion model at the next moment by using the model probability estimation result of each motion model at the previous moment, the Markov probability transfer matrix, and the residual and residual covariance at the next moment.

[0062] Furthermore, the determining submodule includes: The prediction subunit is configured to: predict the state variables and the state vector covariance matrix of each motion model at the next moment by using the initial values ​​of the state variables and the state vector covariance matrix of each motion model when filtering at the previous moment; The calculation subunit is configured to: use the state variables and state vector covariance matrix of each motion model at the next moment to make predictions, calculate the residual and residual covariance of each motion model at the corresponding moment, so as to determine the filter gain of each motion model at the corresponding moment; The updating subunit is configured to update the prediction results of the state variables and the state vector covariance matrix of each motion model at the corresponding moment by using the residual, residual covariance and filter gain of each motion model at the next moment.

[0063] Furthermore, the judging module includes: The third calculation submodule is configured to: calculate the error compression ratio of any two motion models using the model probability of each motion model at the next moment and the Markov probability transfer matrix at the previous moment; The correction submodule is configured to: use the error compression ratio to correct the corresponding elements in the Markov probability transfer matrix at the previous moment.

[0064] The principles, formulas and parameter definitions involved in the above embodiments are all applicable and will not be described in detail here.

[0065] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. The above embodiments only express several implementation methods of the present application, and their descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the present application, several deformations and improvements can be made, which all belong to the protection scope of the present application. Therefore, the protection scope of the patent of this application shall be based on the attached claims.

Claims

1. An aircraft detection method based on an adaptive interactive multi-model algorithm, characterized in that: The aircraft detection method comprises: Step S1, using an aircraft to detect a target to obtain a motion model of the target during motion, and determining the motion model during motion and a Markov probability transfer matrix at a previous moment; Step S2, estimating the state variables and model probabilities of each motion model at the previous moment; Step S3, using the Markov probability transfer matrix and the estimation result, calculating the state variables, state vector covariance matrix and model probability of each motion model at the next moment; Step S4, using the state variables, state vector covariance matrix and model probability of each motion model at the next moment, perform state variable fusion and state vector covariance matrix fusion in sequence; Step S5: When the detection is not finished, the Markov probability transfer matrix of the previous moment is updated using the model probability of each motion model at the next moment; Step S6, updating the next moment to the previous moment, using the updated Markov probability transfer matrix as the Markov probability transfer matrix of the previous moment, returning to step S2, and continuing to perform detection until the detection is completed.

2. The aircraft detection method according to claim 1, characterized in that: The specific implementation process of step S3 includes: Step S31, using the estimation result and the Markov probability transfer matrix, calculate the initial values ​​of the state variables and the state vector covariance matrix of each motion model when filtering at the previous moment; Step S32, determining the residual, residual covariance, state variable and state vector covariance matrix of each motion model at the next moment according to the initial values ​​of the state variable and state vector covariance matrix of each motion model when filtering at the previous moment; Step S33, using the model probability estimation result of each motion model at the previous moment, the Markov probability transfer matrix, and the residual and residual covariance at the next moment, calculate the model probability of each motion model at the next moment.

3. The aircraft detection method according to claim 2, characterized in that: The specific process of step S32 includes: Step S321, using the initial values ​​of the state variables and state vector covariance matrix of each motion model when filtering at the previous moment, predict the state variables and state vector covariance matrix of each motion model at the next moment; Step S322: using the state variables and state vector covariance matrix of each motion model at the next moment to make predictions, calculating the residual and residual covariance of each motion model at the corresponding moment, so as to determine the filter gain of each motion model at the corresponding moment; Step S323: using the residual, residual covariance and filter gain of each motion model at the next moment, update the prediction results of the state variables and state vector covariance matrix of each motion model at the corresponding moment.

4. The aircraft detection method according to any one of claims 1 to 3, characterized in that: In step S5, the specific process of the update includes: Step S51, using the model probability of each motion model at the next moment and the Markov probability transfer matrix at the previous moment, calculate the error compression ratio of any two motion models; Step S52: using the error compression ratio, correct the corresponding elements in the Markov probability transfer matrix at the previous moment.

5. The aircraft detection method according to claim 4, characterized in that: In step S51, the error compression ratio is: ， in, For the i The motion model and j The error compression ratio of the motion model; is the first probability transfer matrix of the Markov probability transfer matrix at the previous moment. i Row and j The element value corresponding to the column; is the Markov probability transfer matrix of the previous moment. j Row and i The element value corresponding to the column; and Respectively i The motion model and j The motion model k The model probability at time .

6. The aircraft detection method according to claim 5, characterized in that: In step S52, correction is performed according to the following formula: ； ； in, is the first in the modified Markov probability transfer matrix of the previous moment i Row and i The element correction value corresponding to the column; is the first in the modified Markov probability transfer matrix of the previous moment i Row and j The element correction value corresponding to the column; l is the exponential factor; , , , N is the number of motion models.

7. An aircraft detection system based on an adaptive interactive multi-model algorithm, characterized in that: The aircraft detection system comprises: The detection module is configured to: detect the target using the aircraft to obtain the motion model of the target during the motion process, and determine the motion model during the motion process and the Markov probability transfer matrix at the previous moment; The estimation module is configured to: estimate the state variables and model probabilities of each motion model at a previous moment; The calculation module is configured to: use the Markov probability transfer matrix and the estimation result, adopt the IMM algorithm, and calculate the state variable, state vector covariance matrix and model probability of each motion model at the next moment; The fusion module is configured to: utilize the state variables, state vector covariance matrix and model probability of each motion model at the next moment to sequentially perform state variable fusion and state vector covariance matrix fusion; The judgment module is configured to: judge whether the detection is finished, if so, output the state variable fusion result and the state vector covariance matrix fusion result, and end; if not, use the model probability of each motion model at the next moment to update the Markov probability transfer matrix at the previous moment and transmit it to the update module; The updating module is configured to update the next moment to the previous moment, use the updated Markov probability transfer matrix as the Markov probability transfer matrix of the previous moment and transmit it to the estimation module to continue to perform detection until the detection is completed.

8. The aircraft detection system according to claim 7, characterized in that: The computing module includes The first calculation submodule is configured to: calculate the initial value of the state variable and the state vector covariance matrix of each motion model when filtering at the previous moment by using the estimation result and the Markov probability transfer matrix; The determination submodule is configured to: determine the residual, residual covariance, state variable and state vector covariance matrix of each motion model at the next moment according to the initial values ​​of the state variable and state vector covariance matrix of each motion model when filtering at the previous moment; The second calculation submodule is configured to calculate the model probability of each motion model at the next moment by using the model probability estimation result of each motion model at the previous moment, the Markov probability transfer matrix, and the residual and residual covariance at the next moment.

9. The aircraft detection system according to claim 8, characterized in that: The determination submodule comprises: The prediction subunit is configured to: predict the state variables and the state vector covariance matrix of each motion model at the next moment by using the initial values ​​of the state variables and the state vector covariance matrix of each motion model when filtering at the previous moment; The calculation subunit is configured to: use the state variables and state vector covariance matrix of each motion model at the next moment to make predictions, calculate the residual and residual covariance of each motion model at the corresponding moment, so as to determine the filter gain of each motion model at the corresponding moment; The updating subunit is configured to update the prediction results of the state variables and the state vector covariance matrix of each motion model at the corresponding moment by using the residual, residual covariance and filter gain of each motion model at the next moment.

10. The aircraft detection system according to any one of claims 7 to 9, characterized in that: The judging module comprises: The third calculation submodule is configured to: calculate the error compression ratio of any two motion models using the model probability of each motion model at the next moment and the Markov probability transfer matrix at the previous moment; The correction submodule is configured to: use the error compression ratio to correct the corresponding elements in the Markov probability transfer matrix at the previous moment.

Images