Multi-model filtering target positioning method, system, device, product and medium

Abstract

The invention relates to the technical field of radar detection, and provides a multi-model filtering target positioning method, system, device, product and medium, and the method comprises the steps: determining a model matching parameter, carrying out the initialization assignment, obtaining an initialization parameter, and obtaining a sampling point set and a weight coefficient through the initialization parameter; performing state one-step prediction to obtain a state one-step prediction value, and performing secondary sampling and measurement one-step prediction to obtain an autocorrelation covariance and a mixed covariance; carrying out SVD operation to obtain a decomposition value set, detecting a target by a target radar to obtain a radar detection value, obtaining a Kalman gain matrix according to the mixed covariance, and calculating state estimation and an initial covariance; updating the model probability to obtain a Markov transition probability, and normalizing the Markov transition probability to obtain a normalized probability; and state estimation and target covariance are obtained through state fusion, so that the target orientation is determined. According to the invention, accurate tracking of the high maneuvering target is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of radar detection, and particularly to a target positioning method, system, device, product and medium for multi-model filtering. Background Art

[0002] In the process of device control, the positioning and tracking technology of maneuvering targets has important application value. For example, unmanned aerial vehicles, autonomous driving vehicles, etc. require high-precision positioning and tracking technology. However, maneuvering targets usually have complex motion patterns, such as sudden acceleration, deceleration, turning, etc. These uncertainties pose great challenges to target tracking. This also makes maneuvering target tracking a typical non-linear filtering process. During the target tracking process, model mismatches, initial errors, random noise, and non-Gaussian heavy-tailed noise will all lead to a decrease in filtering accuracy or even divergence.

[0003] IMM (Interacting Multiple Model) can map the motion model of the target to a set of many known models, and each filter model works in parallel. A Markov probability transfer matrix is used to obtain the interaction and switching between different sub-models. Finally, the algorithm performs data fusion on the target state filter estimates obtained by each model filter to obtain the system parameter estimate. The selection of the filter model is another key factor of the IMM algorithm. The performance of the filter directly affects the tracking effect of the IMM algorithm. Since the motion of maneuvering targets is non-linear, the linear filtering process cannot fully match the actual situation. Existing non-linear filters are easily affected by model mismatches, random errors, and other non-ideal environments. Therefore, they are insufficient in terms of adaptability and robustness. In some engineering applications, for example, when the sensor is unreliable, tracking maneuvering targets with measurement outliers, heavy-tailed non-Gaussian process noise, and measurement noise will all lead to a decrease in filtering accuracy. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the related art. For this purpose, the present invention provides a target positioning method, system, device, product and medium for multi-model filtering, realizing low-latency and high-precision tracking of moving targets.

[0005] The present invention provides a target positioning method for multi-model filtering, including: S1: Determine the model matching parameters, initialize and assign values to the filter model according to the model matching parameters to obtain the initialization parameters, and obtain the sampling point set and weight coefficients through the initialization parameters; S2: Perform one-step state prediction using the sampling point set and the weight coefficients to obtain a one-step state prediction value. Perform secondary sampling on the one-step state prediction value and conduct one-step measurement prediction to obtain the autocorrelation covariance and the mixed covariance. S3: Perform SVD operation on the autocorrelation covariance to obtain a set of decomposition values. Obtain the Kalman gain matrix based on the mixed covariance and the set of decomposition values. The target radar detects the target to obtain a radar detection value. Obtain the initial state estimate and the initial covariance through the radar detection value and the Kalman gain matrix. S4: Obtain the updated model probability from the radar detection value. Obtain the Markov transition probability based on the updated model probability. Normalize the Markov transition probability to obtain the normalized probability. S5: Perform state fusion using the updated model probability, the initial state estimate, and the initial covariance to obtain the target state estimate and the target covariance. Determine the target azimuth through the target state estimate and the target covariance. Iterate the target azimuth using the updated model probability and the normalized probability until the target leaves the detection range of the target radar.

[0006] For the target positioning method based on multi-model filtering provided by the present invention, step S1 specifically includes: S11: Determine the initial Markov transition probability and the initial model matching probability as the model matching parameters. Initialize and assign values to the filter model according to the initial Markov transition probability and the initial model matching probability to obtain the mixed state estimate and the mixed covariance matrix. Use the mixed state estimate and the mixed covariance matrix as the initialization parameters. S12: Perform QR decomposition on the mixed covariance matrix to obtain the mixed decomposition matrix. Obtain the sampling point set through the mixed state estimate and the mixed decomposition matrix, and obtain the weight coefficients through the sampling point set.

[0007] For the target positioning method based on multi-model filtering provided by the present invention, step S2 specifically includes: S21: Perform non-linear processing on the sampling point set to obtain the one-step prediction value of the sampling point. Complete one-step state prediction through the one-step prediction value of the sampling point and the weight coefficients to obtain the one-step state prediction value. S22: Perform secondary sampling on the one-step state prediction value to obtain the measurement sampling point set. Conduct one-step measurement prediction through the measurement sampling point set and the weight coefficients to obtain the mixed covariance and the autocorrelation covariance.

[0008] For the target positioning method based on multi-model filtering provided by the present invention, step S3 specifically includes: S31: Perform SVD operation on the autocorrelation covariance to obtain a set of decomposition values including a first decomposition value and a second decomposition value, and construct a Kalman gain matrix through the first decomposition value, the second decomposition value, and the mixed covariance; S32: Determine the target radar, detect the target through the target radar to obtain the radar detection value, and obtain the initial state estimate and the initial covariance through the Kalman gain matrix and the radar detection value.

[0009] According to the target positioning method based on multi-model filtering provided by the present invention, step S4 specifically includes: S41: Calculate the detection value residual according to the radar detection value, calculate the residual covariance according to the autocorrelation covariance, calculate the likelihood function value through the detection value residual and the residual covariance, and obtain the updated model probability through the likelihood function value; S42: Obtain the probability change rate through the updated model probability, calculate the Markov transition probability according to the initial Markov transition probability and the probability change rate, and normalize the Markov transition probability to obtain the normalized probability.

[0010] According to the target positioning method based on multi-model filtering provided by the present invention, in step S5, extract the target position, target speed, and target acceleration estimate of the detected target from the target state estimate and the target covariance, and determine the target azimuth through the target position, the target speed, and the target acceleration estimate.

[0011] The present invention also provides a target positioning system based on multi-model filtering, including: Sampling point set module: used to determine the model matching parameters, initialize and assign values to the filter model according to the model matching parameters to obtain the initialization parameters, and obtain the sampling point set and the weight coefficient through the initialization parameters; One-step prediction module: used to perform one-step state prediction through the sampling point set and the weight coefficient to obtain the one-step state prediction value, perform secondary sampling on the one-step state prediction value, and perform one-step measurement prediction to obtain the autocorrelation covariance and the mixed covariance; Initial covariance module: used to perform SVD operation on the autocorrelation covariance to obtain a set of decomposition values, and obtain the Kalman gain matrix according to the mixed covariance and the set of decomposition values. The target radar detects the target to obtain the radar detection value, and obtains the initial state estimate and the initial covariance through the radar detection value and the Kalman gain matrix; Normalized probability module: used to obtain the updated model probability through the radar detection value, obtain the Markov transition probability according to the updated model probability, and normalize the Markov transition probability to obtain the normalized probability; Target azimuth module: It is used to perform state fusion by updating the model probability, initial state estimate, and initial covariance to obtain the target state estimate and target covariance, determine the target azimuth through the target state estimate and target covariance, and iterate the target azimuth by updating the model probability and normalizing the probability until the target leaves the detection range of the target radar.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the target positioning method of multi-model filtering as described in any one of the above.

[0013] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the target positioning method of multi-model filtering as described in any one of the above.

[0014] The present invention also provides a computer program product. The computer program product includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the steps of the target positioning method of multi-model filtering as described in any one of the above.

[0015] One or more of the above technical solutions in the embodiments of the present invention have at least one of the following technical effects: The present invention provides a target positioning method, system, device, product, and medium for multi-model filtering. By introducing the unscented Kalman filter method through SVD (Singular Value Decomposition) operation, state one-step prediction, and measurement one-step prediction, the numerical stability and robustness of the unscented Kalman filter are improved. By introducing an updated model probability, the influence of switching the filter model on determining the target azimuth is considered. The probability change rate is also used to normalize and correct the Markov transition probability to obtain a normalized probability, optimizing the model switching strategy. By combining the unscented Kalman filter method and the Markov transition probability, accurate tracking of high-maneuver targets is achieved, and at the same time, the adaptability of the algorithm in complex environments is improved.

[0016] The additional aspects and advantages of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0018] Figure 1 It is a schematic flowchart of the target positioning method for multi-model filtering provided by the present invention.

[0019] Figure 2 It is a schematic structural diagram of the target positioning system for multi-model filtering provided by the present invention.

[0020] Figure 3 It is a schematic structural diagram of the target positioning device for multi-model filtering provided by the present invention.

[0021] Reference numerals: 100, sampling point set module; 200, one-step prediction module; 300, initial covariance module; 400, normalized probability module; 500, target azimuth module; 810, processor; 820, communication interface; 830, memory; 840, communication bus. Specific embodiments

[0022] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention. The following embodiments are used to illustrate the present invention but cannot be used to limit the scope of the present invention.

[0023] In the description of the embodiments of the present invention, it should be noted that the terms "first", "second", and "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0024] In the description of this specification, the description referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the embodiments of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0025] The following combines Figures 1 to 3 to describe the specific implementation manners of the present invention: Figure 1 is a schematic flowchart of the target positioning method for multi-model filtering provided by the present invention. As Figure 1 shown, first, determine the model matching parameters to obtain the initialization parameters, and obtain the sampling point set and the weight system through the initialization parameters; then perform a one-step state prediction to obtain the one-step state prediction value, perform secondary sampling and a one-step measurement prediction; then perform SVD calculation and obtain the radar detection value, calculate the Kalman gain matrix to obtain the initial state estimate and the initial covariance; subsequently, calculate the updated model probability to obtain the Markov transition probability, perform normalization to obtain the normalized probability; finally, perform state fusion to obtain the target state estimate and the target covariance, and determine the target azimuth.

[0026] The present invention provides a target positioning method for multi-model filtering, including: S1: Determine the model matching parameters, perform an initialization assignment to the filter model according to the model matching parameters to obtain the initialization parameters, and obtain the sampling point set and the weight coefficients through the initialization parameters; Further, the purpose of this stage is to perform an initialization assignment to the filter model to obtain the initialization parameters, and obtain the sampling point set and the weight coefficients. Specifically, step S1 specifically includes: S11: Determine the initial Markov transition probability and the initial model matching probability as the model matching parameters, perform an initialization assignment to the filter model according to the initial Markov transition probability and the initial model matching probability to obtain the mixed state estimate and the mixed covariance matrix, and use the mixed state estimate and the mixed covariance matrix as the initialization parameters; S12: Perform QR decomposition on the mixed covariance matrix to obtain the mixed decomposition matrix, obtain the sampling point set through the mixed state estimate and the mixed decomposition matrix, and obtain the weight coefficients through the sampling point set.

[0027] For the above steps, the specific implementation in this embodiment is as follows: First, obtain multiple possible filter models, and then obtain the initial Markov transition probability that the filter model selected at the previous moment, i.e., the (k - 1)-th moment, transfers from the i-th filter model to the j-th filter model and the probability that the i-th filter model at the (k - 1)-th moment is a matching model , that is, the initial model matching probability. According to the initial Markov transition probability and the initial model matching probability, initialization assignment can be performed. Calculate the mixing probability of the i-th filter model and the j-th filter model at the (k - 1)-th moment : Among them, is the normalization constant of the j-th filter model, and , N is the total number of filter models. Then, through the known state estimate of the i-th filter model at the (k - 1)-th moment and the covariance matrix of the i-th filter model at the (k - 1)-th moment the mixed state estimate and the mixed covariance matrix

[0028] at the (k - 1)-th moment can be obtained as initialization parameters: , and , where T represents the transpose of the matrix, and in each step of this embodiment, when the k-th moment is the 0-th moment, i.e., the starting moment, then each value at the (k - 1)-th moment is a preset parameter given according to experience. Among them, is the -th sampling point at the (k - 1)-th moment, n is the number of rows of the state vector determined according to the mixed decomposition matrix, is the scaling decomposition matrix, is the -th column parameter of the scaling decomposition matrix, is the scaling parameter, and , where b is the state control parameter, generally with a value range of 1 -4 to 1, and δ is a first candidate parameter determined according to experience and is non-negative.

[0029] In addition, the weight coefficients can also be obtained through the set of sampling points: Among them, is the weight coefficient of the first-order statistical characteristics of the 0th sampling point, is the weight coefficient of the second-order statistical characteristics of the 0th sampling point, is the weight coefficient of the first-order statistical characteristics of the th sampling point except the 0th sampling point, is the weight coefficient of the second-order statistical characteristics of the th sampling point except the 0th sampling point, is the second candidate parameter determined according to experience and is non-negative.

[0030] S2: Perform a one-step state prediction through the set of sampling points and the weight coefficients to obtain a one-step state prediction value, perform secondary sampling on the one-step state prediction value, and perform a one-step measurement prediction to obtain the autocorrelation covariance and the cross-covariance; Furthermore, the purpose of this stage is to perform a one-step state prediction and a one-step measurement prediction, so as to obtain the autocorrelation covariance and the cross-covariance. Specifically, step S2 specifically includes: S21: Perform a non-linear processing on the set of sampling points to obtain a one-step prediction value of the sampling points, and complete a one-step state prediction through the one-step prediction value of the sampling points and the weight coefficients to obtain a one-step state prediction value; S22: Perform secondary sampling on the one-step state prediction value to obtain a set of measurement sampling points, and perform a one-step measurement prediction through the set of measurement sampling points and the weight coefficients to obtain the cross-covariance and the autocorrelation covariance.

[0031] For the above steps, the specific implementation manners in this embodiment are as follows: First, perform a non-linear processing on the set of sampling points to obtain the one-step prediction value of the sampling points between the kth moment and the (k - 1)th moment of the th sampling point Among them, is the non-linear function in the system state equation. Then, complete a one-step state prediction through the one-step prediction value of the sampling points and the weight coefficients to obtain the one-step state prediction value of the th sampling point between the kth moment and the (k - 1)th moment Subsequently, obtain the square root of the existing process noise covariance matrix at the (k - 1)th moment, and construct the first weighted center matrix between the kth moment and the k-1th moment : Then, the square root of the process noise covariance matrix can be obtained by the square root of the first weighted center matrix and the process noise covariance matrix: : Among them, qr() is the QR decomposition of the content in the brackets. Then the state one-step prediction value can be subsampled to obtain the measurement sampling point set: in, is the number of steps between the kth moment and the k-1th moment. Measurement sampling points, is the quadratic scaling decomposition matrix, is the second scaling decomposition matrix Column parameters.

[0032] Then, the first step prediction is measured by measuring the sampling point set and the weight coefficient. First, the measurement sampling point set is processed nonlinearly to obtain the first One-step prediction value of the measurement sampling point between the kth moment and the k-1th moment : Among them, h() is a nonlinear function in the system measurement equation, and then the measurement one-step prediction value between the kth moment and the k-1th moment can be calculated. : This completes the measurement one-step prediction. The mixed covariance of the state one-step prediction value and the measurement one-step prediction value between the kth moment and the k-1th moment can be calculated through the measurement one-step prediction value. : Then, obtain the square root of the existing k-1th moment measurement noise covariance matrix , and construct the second weighted center matrix between the kth moment and the k-1th moment by measuring the one-step prediction value and the one-step prediction value of the measurement sampling point : The autocorrelation covariance between the kth moment and the k-1th moment is obtained by taking the square root of the second weighted center matrix and the measurement noise covariance matrix: : S3: Perform SVD operation on the autocorrelation covariance to obtain a decomposition value set, and obtain a Kalman gain matrix based on the mixed covariance and the decomposition value set. The target radar detects the target to obtain a radar detection value, and obtains an initial state estimate and an initial covariance through the radar detection value and the Kalman gain matrix. Furthermore, the purpose of this stage is to perform SVD (Singular Value Decomposition) operations, construct a Kalman gain matrix, and detect the target radar to obtain radar detection values, thereby obtaining initial state estimation and initial covariance. Specifically, step S3 specifically includes: S31: Performing SVD operation on the autocorrelation covariance to obtain a decomposition value set including a first decomposition value and a second decomposition value, and constructing a Kalman gain matrix through the first decomposition value, the second decomposition value and the mixed covariance; S32: Determine a target radar, detect the target through the target radar to obtain the radar detection value, and obtain the initial state estimation and the initial covariance through the Kalman gain matrix and the radar detection value.

[0033] With respect to the above steps, the specific implementation methods in this embodiment are as follows: First, perform SVD operation on the autocorrelation covariance to obtain a set of decomposition values: Among them, SVD() means to perform SVD operation on the content in the brackets. is the first decomposition value between the kth moment and the k-1th moment, is the second decomposition value between the kth moment and the k-1th moment, is the third decomposition value between the kth moment and the k-1th moment, and the first decomposition value, the second decomposition value and the third decomposition value constitute a decomposition value set.

[0034] Then, the Kalman gain matrix at the current kth moment is constructed by the first decomposition value, the second decomposition value and the mixed covariance in the decomposition value set: : Then determine the target radar and use the target radar to detect the detected target, read the data obtained by the radar detection, and obtain the radar detection value. The radar detection value is a matrix. The initial state estimate of the jth filter model at the kth moment is obtained through the Kalman gain matrix and the radar detection value. and the initial covariance of the jth filter model at the kth moment : in, is the radar detection value at the kth moment. Here, the initial state estimate and initial covariance of the jth filter model at the kth moment can be used as the state estimate and covariance matrix of the filter model at the k-1th moment in the next future moment.

[0035] S4: obtaining the updated model probability through the radar detection value, obtaining the Markov transition probability according to the updated model probability, and normalizing the Markov transition probability to obtain the normalized probability; Furthermore, the purpose of this stage is to obtain the updated model probability, thereby obtaining the Markov transition probability, and finally normalizing to obtain the normalized probability. Specifically, step S4 specifically includes: S41: calculating the detection value residual according to the radar detection value, and calculating the residual covariance according to the autocorrelation covariance, calculating the likelihood function value through the detection value residual and the residual covariance, and obtaining the updated model probability through the likelihood function value; S42: obtaining a probability change rate through the updated model probability, calculating a Markov transition probability according to the initial Markov transition probability and the probability change rate, and normalizing the Markov transition probability to obtain the normalized probability.

[0036] With respect to the above steps, the specific implementation methods in this embodiment are as follows: First, the residual of the detection value of the jth filter model at the kth moment is calculated based on the radar detection value : In addition, it is necessary to calculate the residual covariance of the jth filter model at the kth moment based on the autocorrelation covariance : Then, the likelihood function value of the jth filter model at the kth moment can be calculated by the detection value residual and residual covariance : Among them, exp[] is an exponential operation on the content in the brackets, and then the updated model probability of the jth filter model at the kth moment is obtained by the likelihood function value and the normalization constant of the jth filter model. : Where c is the normalization coefficient, and , the updated model probability is the probability that the jth filter model is the matching model at the kth moment. Then, the probability change rate of the jth filter model at the kth moment is obtained by updating the model probability : in, is the updated model probability of the j-th filter model at the k-1th moment, that is, the updated model probability at the kth moment calculated at the previous moment, and when k is 0, is an initial value preset based on experience, is the change rate reference value set based on experience. Then, the Markov transition probability of the filter model selected at the kth moment transferring from the i-th filter model to the j-th filter model is calculated based on the initial Markov transition probability and the probability change rate. : Normalize the Markov transition probability to obtain the normalized probability that the filter model selected at the kth moment is transferred from the i-th filter model to the j-th filter model : S5: Perform state fusion by updating the model probability, initial state estimation and initial covariance to obtain the target state estimation and target covariance, determine the target direction by the target state estimation and target covariance, and iterate the target direction by updating the model probability and normalized probability until the target leaves the detection range of the target radar.

[0037] Furthermore, the purpose of this stage is to perform state fusion to obtain target state estimation and target covariance, and finally determine the target orientation. Specifically, in step S5, the target position, target velocity and target acceleration estimation of the detected target are extracted from the target state estimation and the target covariance, and the target orientation is determined by the target position, the target velocity and the target acceleration estimation.

[0038] With respect to the above steps, the specific implementation methods in this embodiment are as follows: By updating the model probability, initial state estimate and initial covariance for state fusion, the target state estimate at the kth moment can be obtained. and the target covariance at the kth moment : After obtaining the target state estimate and the target covariance, the target position, target velocity and target acceleration estimate of the detected target can be extracted therefrom, and the target position, target velocity and target acceleration estimate are used to determine the target orientation, and the target movement direction and movement trend can also be determined. Subsequently, since the updated model probability of all filter models at the kth moment is obtained in the process of state fusion, the normalized probability of the filter model selected at the kth moment being transferred from the i-th filter model to the j-th filter model is also obtained in S4, so at the next moment, that is, the k+1th moment, the normalized probability at the kth moment can be used as the initial Markov transfer probability when calculating the target orientation at the k+1th moment, and the initial model matching probability when calculating the target orientation at the k+1th moment is obtained from the updated model probability of all filter models at the kth moment, and steps S1 to S5 are repeatedly executed to continuously iterate the target orientation and realize continuous updating of the target orientation until the target leaves the detection range of the target radar.

[0039] The present invention effectively realizes the precise tracking of highly maneuverable targets by using SVD operation, one-step state prediction and one-step measurement prediction, and calculating the probability of updating the model, while improving the adaptability of the algorithm in complex environments.

[0040] The target positioning device of the multi-model filtering provided by the present invention is described below. The target positioning device of the multi-model filtering described below and the target positioning method of the multi-model filtering described above can be referred to each other.

[0041] Figure 2 The schematic diagram of the structure of the target positioning system with multi-model filtering is shown as follows: Figure 2 As shown, the target positioning method for performing the multi-model filtering as described above includes: Sampling point set module 100: used to determine model matching parameters, initialize and assign values ​​to the filter model according to the model matching parameters to obtain initialization parameters, and obtain the sampling point set and weight coefficients through the initialization parameters; One-step prediction module 200: used to perform one-step prediction of the state through the sampling point set and the weight coefficient to obtain the one-step prediction value of the state, perform secondary sampling on the one-step prediction value of the state, and perform one-step prediction of measurement to obtain the autocorrelation covariance and the mixed covariance; Initial covariance module 300: used to perform SVD operation on the autocorrelation covariance to obtain a decomposition value set, and obtain a Kalman gain matrix according to the mixed covariance and the decomposition value set. The target radar detects the target to obtain a radar detection value, and obtains an initial state estimate and an initial covariance through the radar detection value and the Kalman gain matrix; Normalized probability module 400: used to obtain the updated model probability through the radar detection value, obtain the Markov transition probability according to the updated model probability, and normalize the Markov transition probability to obtain the normalized probability; Target orientation module 500: used to perform state fusion by updating model probability, initial state estimation and initial covariance to obtain target state estimation and target covariance, determine target orientation by target state estimation and target covariance, and iterate target orientation by updating model probability and normalized probability until the target leaves the detection range of the target radar.

[0042] on the other hand, Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the target positioning method of multi-model filtering, and the method includes: S1: Determine model matching parameters, initialize and assign values ​​to the filter model according to the model matching parameters to obtain initialization parameters, and obtain a sampling point set and a weight coefficient through the initialization parameters; S2: performing one-step state prediction through the sampling point set and the weight coefficient to obtain a one-step state prediction value, performing secondary sampling on the one-step state prediction value, and performing one-step measurement prediction to obtain an autocorrelation covariance and a mixed covariance; S3: Perform SVD operation on the autocorrelation covariance to obtain a decomposition value set, and obtain a Kalman gain matrix based on the mixed covariance and the decomposition value set. The target radar detects the target to obtain a radar detection value, and obtains an initial state estimate and an initial covariance through the radar detection value and the Kalman gain matrix. S4: obtaining the updated model probability through the radar detection value, obtaining the Markov transition probability according to the updated model probability, and normalizing the Markov transition probability to obtain the normalized probability; S5: Perform state fusion by updating the model probability, initial state estimation and initial covariance to obtain the target state estimation and target covariance, determine the target direction by the target state estimation and target covariance, and iterate the target direction by updating the model probability and normalized probability until the target leaves the detection range of the target radar.

[0043] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0044] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, when the program instructions are executed by a computer, the computer can execute the target positioning method of multi-model filtering provided by the above methods, the method comprising: S1: Determine model matching parameters, initialize and assign values ​​to the filter model according to the model matching parameters to obtain initialization parameters, and obtain a sampling point set and a weight coefficient through the initialization parameters; S2: performing one-step state prediction through the sampling point set and the weight coefficient to obtain a one-step state prediction value, performing secondary sampling on the one-step state prediction value, and performing one-step measurement prediction to obtain an autocorrelation covariance and a mixed covariance; S3: Perform SVD operation on the autocorrelation covariance to obtain a decomposition value set, and obtain a Kalman gain matrix based on the mixed covariance and the decomposition value set. The target radar detects the target to obtain a radar detection value, and obtains an initial state estimate and an initial covariance through the radar detection value and the Kalman gain matrix. S4: obtaining the updated model probability through the radar detection value, obtaining the Markov transition probability according to the updated model probability, and normalizing the Markov transition probability to obtain the normalized probability; S5: Perform state fusion by updating the model probability, initial state estimation and initial covariance to obtain the target state estimation and target covariance, determine the target direction by the target state estimation and target covariance, and iterate the target direction by updating the model probability and normalized probability until the target leaves the detection range of the target radar.

[0045] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to perform the target positioning method of the multi-model filtering provided by the above methods, the method comprising: S1: Determine model matching parameters, initialize and assign values ​​to the filter model according to the model matching parameters to obtain initialization parameters, and obtain a sampling point set and a weight coefficient through the initialization parameters; S2: performing one-step state prediction through the sampling point set and the weight coefficient to obtain a one-step state prediction value, performing secondary sampling on the one-step state prediction value, and performing one-step measurement prediction to obtain an autocorrelation covariance and a mixed covariance; S3: Perform SVD operation on the autocorrelation covariance to obtain a decomposition value set, and obtain a Kalman gain matrix based on the mixed covariance and the decomposition value set. The target radar detects the target to obtain a radar detection value, and obtains an initial state estimate and an initial covariance through the radar detection value and the Kalman gain matrix. S4: obtaining the updated model probability through the radar detection value, obtaining the Markov transition probability according to the updated model probability, and normalizing the Markov transition probability to obtain the normalized probability; S5: Perform state fusion by updating the model probability, initial state estimation and initial covariance to obtain the target state estimation and target covariance, determine the target direction by the target state estimation and target covariance, and iterate the target direction by updating the model probability and normalized probability until the target leaves the detection range of the target radar.

[0046] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0047] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. The target positioning method of multi-model filtering is characterized by: include: S1: Determine model matching parameters, initialize and assign values ​​to the filter model according to the model matching parameters to obtain initialization parameters, and obtain a sampling point set and a weight coefficient through the initialization parameters; S2: performing one-step state prediction through the sampling point set and the weight coefficient to obtain a one-step state prediction value, performing secondary sampling on the one-step state prediction value, and performing one-step measurement prediction to obtain an autocorrelation covariance and a mixed covariance; S3: Perform SVD operation on the autocorrelation covariance to obtain a decomposition value set, and obtain a Kalman gain matrix based on the mixed covariance and the decomposition value set. The target radar detects the target to obtain a radar detection value, and obtains an initial state estimate and an initial covariance through the radar detection value and the Kalman gain matrix. S4: obtaining the updated model probability through the radar detection value, obtaining the Markov transition probability according to the updated model probability, and normalizing the Markov transition probability to obtain the normalized probability; S5: Perform state fusion by updating the model probability, initial state estimation and initial covariance to obtain the target state estimation and target covariance, determine the target direction by the target state estimation and target covariance, and iterate the target direction by updating the model probability and normalized probability until the target leaves the detection range of the target radar.

2. The target positioning method of multi-model filtering according to claim 1 is characterized in that: Step S1 specifically includes: S11: Determine an initial Markov transition probability and an initial model matching probability as the model matching parameters, initialize and assign values ​​to the filter model according to the initial Markov transition probability and the initial model matching probability, obtain a mixed state estimate and a mixed covariance matrix, and use the mixed state estimate and the mixed covariance matrix as the initialization parameters; S12: performing QR decomposition on the mixed covariance matrix to obtain a mixed decomposition matrix, obtaining the sampling point set through the mixed state estimation and the mixed decomposition matrix, and obtaining the weight coefficient through the sampling point set.

3. The target positioning method of multi-model filtering according to claim 1 is characterized in that: Step S2 specifically includes: S21: performing nonlinear processing on the sampling point set to obtain a one-step prediction value of the sampling point, completing a one-step prediction of the state through the one-step prediction value of the sampling point and the weight coefficient to obtain a one-step prediction value of the state; S22: performing secondary sampling on the state one-step prediction value to obtain a measurement sampling point set, performing one-step measurement prediction using the measurement sampling point set and the weight coefficient, and obtaining a mixed covariance and an autocorrelation covariance.

4. The target positioning method of multi-model filtering according to claim 1 is characterized in that: Step S3 specifically includes: S31: Performing SVD operation on the autocorrelation covariance to obtain a decomposition value set including a first decomposition value and a second decomposition value, and constructing a Kalman gain matrix through the first decomposition value, the second decomposition value and the mixed covariance; S32: Determine a target radar, detect the target through the target radar to obtain the radar detection value, and obtain the initial state estimation and the initial covariance through the Kalman gain matrix and the radar detection value.

5. The target positioning method of multi-model filtering according to claim 1 is characterized in that: Step S4 specifically includes: S41: calculating the detection value residual according to the radar detection value, and calculating the residual covariance according to the autocorrelation covariance, calculating the likelihood function value through the detection value residual and the residual covariance, and obtaining the updated model probability through the likelihood function value; S42: obtaining a probability change rate through the updated model probability, calculating a Markov transition probability according to the initial Markov transition probability and the probability change rate, and normalizing the Markov transition probability to obtain the normalized probability.

6. The target positioning method of multi-model filtering according to claim 1, characterized in that: In step S5, the target position, target velocity and target acceleration estimate of the detected target are extracted from the target state estimate and the target covariance, and the target orientation is determined by the target position, the target velocity and the target acceleration estimate.

7. A target positioning system for multi-model filtering, used to execute the target positioning method for multi-model filtering according to any one of claims 1 to 6, characterized in that: include: Sampling point set module: used to determine model matching parameters, initialize and assign values ​​to the filter model according to the model matching parameters, obtain initialization parameters, and obtain the sampling point set and weight coefficients through the initialization parameters; One-step prediction module: used to perform one-step prediction of the state through the sampling point set and the weight coefficient to obtain the one-step prediction value of the state, perform secondary sampling on the one-step prediction value of the state, and perform one-step prediction of measurement to obtain the autocorrelation covariance and the mixed covariance; Initial covariance module: used to perform SVD operation on the autocorrelation covariance to obtain a decomposition value set, and obtain the Kalman gain matrix based on the mixed covariance and the decomposition value set. The target radar detects the target to obtain the radar detection value, and the initial state estimation and initial covariance are obtained through the radar detection value and the Kalman gain matrix; Normalized probability module: used to obtain the updated model probability through radar detection value, obtain the Markov transition probability according to the updated model probability, normalize the Markov transition probability, and obtain the normalized probability; Target direction module: It is used to perform state fusion by updating the model probability, initial state estimation and initial covariance to obtain the target state estimation and target covariance, determine the target direction by the target state estimation and target covariance, and iterate the target direction by updating the model probability and normalized probability until the target leaves the detection range of the target radar.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the target positioning method of multi-model filtering as described in any one of claims 1 to 6 are implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the target positioning method of multi-model filtering as described in any one of claims 1 to 6 are implemented.

10. A computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, characterized in that: When the program instructions are executed by a computer, the computer can execute the steps of the target positioning method of multi-model filtering as described in any one of claims 1 to 6.

Images