Radar-based expressway target tracking method and related equipment

Abstract

The embodiment of the invention provides a radar-based expressway target tracking method and related equipment. The method comprises the following steps: constructing a target model according to a target feature-free Kalman filtering framework and a target gating recurrent neural network; generating a target motion state compensation result according to the target model; and executing target tracking operation according to the target motion state compensation result. In this way, the applicability of target tracking in a complex vehicle driving scene can be improved, and the target tracking precision is improved.

Description

Technical Field

[0001] Embodiments of the present application relate to the field of radar technology, and in particular, to a method for tracking highway targets based on radar and related devices. Background Art

[0002] Highway target tracking is mainly used to monitor and manage the dynamics of vehicles on the road to improve traffic safety and optimize traffic flow.

[0003] In related technologies, most rely on prior models to achieve highway target tracking. The above methods have problems such as poor applicability to complex vehicle driving scenarios and poor tracking accuracy. Summary of the Invention

[0004] According to the embodiments of the present application, a method for tracking highway targets based on radar and related devices is provided, which can improve the applicability of target tracking in complex vehicle driving scenarios and improve the target tracking accuracy.

[0005] In the first aspect of the present application, a method for tracking highway targets based on radar is provided, including: Constructing a target model according to the target unscented Kalman filter framework and the target gated recurrent neural network; Generating a target motion state compensation result according to the target model; Performing a target tracking operation according to the target motion state compensation result.

[0006] In some feasible embodiments, the generating a target motion state compensation result according to the target model includes: Identifying target hidden state information according to the target data set; Determining the target model error according to the target hidden state information; Generating a target motion state compensation result according to the target model error and the target UKF motion state prediction result.

[0007] In some feasible embodiments, the target model error includes: a target state transition model error, and / or, a target measurement model error.

[0008] In some feasible embodiments, the target UKF motion state prediction result includes: A target UKF motion state vector prediction result, a target UKF motion measurement vector prediction result, a covariance matrix prediction result corresponding to the target UKF motion state vector prediction result, and / or, a covariance matrix prediction result corresponding to the target UKF motion measurement vector prediction result.

[0009] In some feasible embodiments, the method further includes: Determine the predicted result of the target UKF motion state vector based on the following formula:

[0010] where, is used to represent the predicted value of the state vector at the k th moment for the k +1th moment; is used to represent the number of points at the target sigma ; is used to represent the weight of the i th target sigma point; is used to represent the predicted state value of the i th target sigma point at the k +1th moment; is used to represent the target measurement model error at the k th moment; Determine the predicted result of the covariance matrix corresponding to the predicted result of the target UKF motion state vector based on the following formula:

[0011] where, is used to represent the predicted value of the state covariance matrix at the kth moment for the k +1th moment; is used to represent the number of points at the target sigma ; is used to represent the weight of the i th target sigma point; is used to represent the state prediction deviation vector of the i th target sigma point at the k +1th moment; is used to represent the transpose of the state prediction deviation vector of the i th target sigma point at the k +1th moment; is used to represent the process noise covariance matrix; is used to represent the measurement model error covariance matrix; Determine the predicted result of the target UKF motion measurement vector based on the following formula:

[0012] where, is used to represent the predicted value of the measurement vector at the k th moment for the k +1th moment; is used to represent the number of points at the target sigma ; is used to represent the ia target sigma the weight of the point; used to represent the i a target sigma predicted value of the measurement vector of the point at the k k+1 moment; used to represent the k target state transition model error at the k+1 moment; determined based on the following formula to determine the predicted result of the covariance matrix corresponding to the target UKF motion measurement vector prediction result:

[0013] wherein, used to represent the predicted value of the measurement covariance matrix at the kth moment for the k k+1 moment; used to represent the number of points in the target sigma ; used to represent the i a target sigma the weight of the point; used to represent the i a target sigma predicted deviation vector of the measurement of the point at the k k+1 moment; used to represent the i a target sigma transpose of the predicted deviation vector of the measurement of the point at the k k+1 moment; used to represent the measurement noise covariance matrix; used to represent the k measurement model error covariance matrix at the k+1 moment.

[0014] In some feasible embodiments, the above target gated recurrent neural network includes: a target memory update gate, a target state update gate, and a target state prediction gate; wherein, the target state update gate, and / or, the target state prediction gate, is provided with a target activation function; The target activation function includes:

[0015] wherein, used to represent the target activation function; used to represent the input of the gated network; used to represent the weights of each layer of the neural network; used to represent the biases of each layer of the neural network; used to represent the weights of the first layer of the neural network; used to represent the weights of the second layer of the neural network; used to represent the biases of the first layer of the neural network; Used to represent the bias of the second-layer neural network; ; Represents the hyperbolic tangent activation function.

[0016] In some feasible embodiments, the above-mentioned target motion state compensation result includes: The target motion state compensation result corresponding to the target divergence section; And / or, the target motion state compensation result corresponding to the target convergence section.

[0017] In a second aspect of the present application, a radar-based highway target tracking device is provided, including: A construction unit, configured to construct a target model according to the target featureless Kalman filter framework and the target gated recurrent neural network; A generation unit, configured to generate a target motion state compensation result according to the target model; An execution unit, configured to perform a target tracking operation according to the target motion state compensation result.

[0018] In a third aspect of the present application, an electronic device is provided, including a memory and a processor. A computer program is stored on the memory, and when the processor executes the computer program, the method described in any one of the above is implemented.

[0019] In a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method described in any one of the above is implemented.

[0020] The embodiments of the present application provide a radar-based highway target tracking method and related devices. The method includes: constructing a target model according to the target featureless Kalman filter framework and the target gated recurrent neural network; generating a target motion state compensation result according to the target model; performing a target tracking operation according to the target motion state compensation result. The present application can improve the applicability of target tracking in complex vehicle driving scenarios and improve the target tracking accuracy.

[0021] It should be understood that the content described in the summary of the invention is not intended to limit the key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Combined with the drawings and referring to the following detailed description, the above and other features, advantages and aspects of the embodiments of the present application will become more obvious. In the drawings, the same or similar reference numerals represent the same or similar elements, where: Figure 1It is a flowchart of a radar-based highway target tracking method according to an embodiment of the present application; Figure 2 It is a structural diagram of a radar-based highway target tracking device according to an embodiment of the present application; Figure 3 It is a structural diagram of an electronic device suitable for implementing the embodiments of the present application. Detailed implementation manners

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present disclosure. Apparently, the described embodiments are some but not all of the embodiments of the present disclosure. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present disclosure without creative efforts shall fall within the scope of protection of the present disclosure.

[0024] In addition, the term "and / or" in this article is only a relational description of associated objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, both A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after.

[0025] In the first aspect of the present application, a radar-based highway target tracking method is provided. Figure 1 It is a flowchart of a radar-based highway target tracking method 100 according to an embodiment of the present application, as Figure 1 shown. The method 100 includes: Step S1; constructing a target model according to the target featureless Kalman filter framework and the target gated recurrent neural network.

[0026] Exemplarily, the target featureless Kalman filter framework can be constructed based on the following manner: Specifically, the first-order Markov nonlinear discrete-time stochastic system can be determined based on the following formulas (1)-(2): (1) where , is used to represent the state vector at the k th moment; , is used to represent the nonlinear state transition function; is used to represent the Gaussian process noise at the k th moment; is used to represent the corresponding covariance matrix.

[0027] (2) Among them, , which is used to represent the measurement vector at the k + 1 moment; , which is used to represent the non-linear measurement function; is used to represent the k Gaussian process noise at the + 1 moment; is used to represent corresponding covariance matrix.

[0028] Specifically, based on, the Unscented Kalman Filter (UKF) can be derived to update the mean and covariance matrix.

[0029] Among them, the principle of the Unscented Transform (UT) is as follows: According to the mean and covariance matrix of the original random variable distribution, a set of target sigma points and weights are selected to make the mean and covariance matrix of the weighted points match those of the original random variable distribution. Then, the UT is applied to the samples, and the mean and covariance of the original samples are approximately predicted using the weighted mean and covariance matrix of the transformed samples.

[0030] Specifically, the above-mentioned target sigma points can be determined based on the following formula: (3)

[0031] Among them, is used to represent index; is used to represent the mean of the original random variable; is used to represent the covariance matrix of the original random variable; is used to represent the preset scaling constant; is used to represent the matrix the i column of the square root.

[0032] Specifically, the above-mentioned weights can be determined based on the following formula: (4) Among them, is used to represent the weight of the target mean point; and are used to represent the weights of the symmetric target sigma points; is used to represent the dimension of the original random variable, that is, the number of random variables; is used to represent the preset scaling constant.

[0033] It should be noted that in the following text, the above is used to represent the Unscented Transform.

[0034] Specifically, the state can be converted to the state corresponding to the target sigma point based on the following formula: (5) where represents the predicted value of the state of the i-th target sigma point at the k-th moment; represents the weight of the i-th target sigma point; represents the predicted value of the state vector at the k-th moment; represents the predicted value of the state covariance matrix at the k-th moment.

[0035] Specifically, the model transfer function can be used to evaluate the state target sigma point based on the following formula: (6) where represents the predicted value of the state of the i-th target sigma point at the (k + 1)-th moment.

[0036] Specifically, the weighted average value and covariance matrix of the predicted state can be estimated based on the following formula: (7) (8) (9) where represents the predicted state deviation vector of the i-th target sigma point at the (k + 1)-th moment; represents the predicted value of the state covariance matrix at the k-th moment for the (k + 1)-th moment; represents the process noise covariance matrix.

[0037] Specifically, the propagated target sigma point can be evaluated using the measurement function based on the following formula: (10) where represents the predicted value of the measurement vector of the i-th target sigma point at the (k + 1)-th moment.

[0038] Specifically, the weighted average value and covariance matrix of the predicted observation can be estimated based on the following formula: (11) where represents the predicted value of the measurement vector at the k-th moment for the (k + 1)-th moment. (12) (13) where represents the predicted value of the measurement covariance matrix at the k-th moment for the (k + 1)-th moment; for representing the measurement noise covariance matrix; for representing the measurement prediction deviation vector of the i-th target sigma point at the (k + 1)-th moment.

[0039] Specifically, the covariance matrix between the state and the measurement can be estimated based on the following formula: (14) where is used to represent the covariance value between the estimated state and the measurement state at the k-th moment for the (k + 1)-th moment.

[0040] Specifically, the Kalman gain can be calculated based on the following formula: (15) where is used to represent the Kalman gain at the (k + 1)-th moment.

[0041] Specifically, the weighted mean and covariance matrix of the predicted observation can be estimated based on the following formula: (16) (17) where , is used to represent the measurement vector at the (k + 1)-th moment; is used to represent the predicted value of the state at the (k + 1)-th moment; is used to represent the predicted value of the state covariance matrix at the (k + 1)-th moment.

[0042] It should be noted that a target model can be constructed based on the above target featureless Kalman filter framework and the above target gated recurrent neural network.

[0043] Step S2; Generate a target motion state compensation result according to the target model.

[0044] In some feasible embodiments, the above step S2; Generate a target motion state compensation result according to the target model, including: Step S21; Identify the target hidden state information according to the target data set.

[0045] Step S22; Determine the target model error according to the target hidden state information.

[0046] It should be noted that the above target model error corresponds to the error between the system prior model and the actual model.

[0047] Exemplarily, the target hidden state information in the target data set can be learned based on the target model, and the above target hidden state can be converted into the above target model error.

[0048] Specifically, the target hidden state information in the target dataset can be memorized based on the target recurrent network structure, and the target gated network structure converts the above target hidden state information into the target model error.

[0049] Thus, the above method can accurately identify the target hidden state information in the target dataset; based on the target hidden state information, the target model error can be accurately determined to improve the determination accuracy of the target motion state compensation result.

[0050] In some feasible embodiments, the above target model error includes: a target state transition model error, and / or, a target measurement model error.

[0051] It should be noted that, based on the following formula, according to the actual state transition function and the actual measurement function, the state vector at the k +1st moment and the measurement vector at the k +1st moment can be determined: (18) (19) Wherein, is used to represent the state vector at the (k + 1)th moment; is used to represent the actual measurement value at the kth moment; is used to represent the Gaussian process noise at the kth moment; is used to represent the Gaussian process noise at the (k + 1)th moment; , is used to represent the measurement vector at the (k + 1)th moment; is used to represent the actual state value at the (k + 1)th moment; is used to represent the Gaussian process noise at the (k + 1)th moment.

[0052] Exemplarily, the above target state transition model error can be determined based on the following formula and an indirect method: (20) Wherein, is used to represent the target state transition model error at the k +1st moment; is used to represent the actual state value at the k +1st moment; is used to represent the state estimated value at the k +1st moment; is used to represent the state vector at the k +1st moment; Exemplarily, the above target measurement model error can be determined based on the following formula and a brief method: (21) Wherein, is used to represent the target measurement model error at the kth moment; used to represent the actual measurement value at the k-th moment; used to represent the measurement estimated value at the k-th moment; used to represent the state vector at the k-th moment.

[0053] Preferably, formulas (18) to (19) can be transformed into the following formulas (22) to (23): (22) (23) wherein, used to represent the state vector at the (k + 1)-th moment; used to represent the measurement estimated value at the k-th moment; used to represent the Gaussian process noise at the k-th moment; used to represent the target measurement model error at the k-th moment; used to represent the Gaussian process noise at the (k + 1)-th moment; used to represent the target measurement model error at the (k + 1)-th moment.

[0054] Thus, the above method can introduce the target state transition model error and / or the target measurement model error to determine the target motion state compensation result, which is beneficial to improving the determination accuracy of the target motion state compensation result, thereby improving the execution accuracy of the target tracking operation.

[0055] Step S23: Generate a target motion state compensation result according to the target model error and the target UKF motion state prediction result.

[0056] In some feasible implementation manners, the above target UKF motion state prediction result includes: The target UKF motion state vector prediction result, the target UKF motion measurement vector prediction result, the covariance matrix prediction result corresponding to the target UKF motion state vector prediction result, and / or the covariance matrix prediction result corresponding to the target UKF motion measurement vector prediction result.

[0057] It should be noted that the previous state can be normalized and concatenated with the previous memory state ; then send the target memory update gate to obtain and , and get , and send to the target state prediction gate and the target state prediction gate.

[0058] In some feasible implementation manners, the above target gated recurrent neural network includes: a target memory update gate, a target state update gate, and a target state prediction gate.

[0059] It should be noted that the above target state update gate, and / or, the target state prediction gate, are set with a target activation function; The above target activation function includes: (24) Wherein, is used to represent the target activation function; is used to represent the input of the gating network; is used to represent the weight of each layer of the neural network; is used to represent the bias of each layer of the neural network; is used to represent the weight of the first layer of the neural network; is used to represent the weight of the second layer of the neural network; is used to represent the bias of the first layer of the neural network; is used to represent the bias of the second layer of the neural network; ; represents the hyperbolic tangent activation function.

[0060] It should be noted that at the target state update gate, input to the two-layer neural network with the hyperbolic tangent of the above target activation function, and the corresponding vectors are output by the above two-layer neural network to determine the corresponding target measurement model error and the measurement model error covariance matrix , during the state prediction process of the UFK, for the above target measurement model error and the measurement model error covariance matrix compensation is performed to obtain the predicted value of the state vector and the predicted value of the state covariance matrix .

[0061] It should be noted that after normalization processing and are concatenated and then sent to the target state prediction gate, input to the two-layer neural network with the hyperbolic tangent of the above target activation function, and the corresponding measurement vectors are output by the above two-layer neural network to determine the target state transition model error and the measurement model error covariance matrix , during the state prediction process of the UFK, for the above target state transition model error and the above measurement model error covariance matrix compensation is performed to obtain the predicted value of the state vector corresponding to the target motion state compensation result and the predicted value of the state covariance matrix corresponding to the target motion state compensation result .

[0062] Thus, the above method can achieve improving the construction accuracy of the target gated recurrent neural network based on the above target activation function, thereby improving the construction accuracy of the target model to further improve the determination accuracy of the target motion state compensation result.

[0063] In some feasible embodiments, the above method further includes: Determine the predicted result of the target UKF motion state vector based on the following formula: (25) Wherein, is used to represent the predicted value of the state vector at the k th moment for the k +1th moment; is used to represent the number of points at the target sigma ; is used to represent the i th target sigma point weight; is used to represent the i th target sigma point state prediction value at the k +1th moment; is used to represent the target measurement model error at the k th moment; Determine the predicted result of the covariance matrix corresponding to the predicted result of the target UKF motion state vector based on the following formula: (26) Wherein, is used to represent the predicted value of the state covariance matrix at the kth moment for the k +1th moment; is used to represent the number of points at the target sigma ; is used to represent the i th target sigma point weight; is used to represent the i th target sigma point state prediction deviation vector at the k +1th moment; is used to represent the i th target sigma point state prediction deviation vector transpose at the k +1th moment; is used to represent the process noise covariance matrix; is used to represent the measurement model error covariance matrix; Determine the predicted result of the target UKF motion measurement vector based on the following formula: (27) Wherein, is used to represent at thek At all times, the predicted value of the measurement vector at the k +1 moment; Used to represent at the target sigma Number of points; Used to represent the i th target sigma Weight of the point; Used to represent the i th target sigma Point at the k +1 moment of the predicted value of the measurement vector; Used to represent the k Target state transition model error at the +1 moment; Determined based on the following formula to determine the predicted result of the covariance matrix corresponding to the target UKF motion measurement vector prediction result: (28) Where Used to represent the predicted value of the measurement covariance matrix at the kth moment for the k +1 moment; Used to represent at the target sigma Number of points; Used to represent the i th target sigma Weight of the point; Used to represent the i th target sigma Point at the k +1 moment of the measurement prediction deviation vector; Used to represent the i th target sigma Point at the k Transpose of the measurement prediction deviation vector at the +1 moment; Used to represent the measurement noise covariance matrix; Used to represent the k Measurement model error covariance matrix at the +1 moment.

[0064] It should be noted that the i th target sigma Point at the k +1 moment of the state prediction deviation vector Can be determined based on formula (9); the i th target sigma Point at the k +1 moment of the measurement prediction deviation vector can be determined based on formula (13).

[0065] Thus, the above method can accurately determine the predicted result of the target UKF motion state vector, the predicted result of the target UKF motion measurement vector, the predicted result of the covariance matrix corresponding to the predicted result of the target UKF motion state vector, and / or the predicted result of the covariance matrix corresponding to the predicted result of the target UKF motion measurement vector based on the above formula, thereby further improving the determination accuracy of the target motion state compensation result and further improving the execution accuracy of the target tracking operation.

[0066] Step S3: Perform a target tracking operation according to the target motion state compensation result.

[0067] In some feasible embodiments, the above target motion state compensation result includes: the target motion state compensation result corresponding to the target divergence section; and / or the target motion state compensation result corresponding to the target convergence section.

[0068] Exemplarily, if the above target divergence section corresponds to a highway diversion section, then the target motion state compensation result corresponding to the above target divergence section corresponds to the target motion state compensation result corresponding to the highway diversion section.

[0069] Exemplarily, if the above target convergence section corresponds to a highway confluence section, then the target motion state compensation result corresponding to the above target convergence section corresponds to the target motion state compensation result corresponding to the highway confluence section.

[0070] It should be noted that since the highway divergence section is also the highway diversion section and the convergence section is also the highway confluence section, and the vehicle motion state accelerates, decelerates, and / or changes lanes frequently, the above method can accurately track the target vehicle motion state in the highway divergence section and the convergence section, so as to improve the applicability of the present application to target tracking in complex vehicle driving scenarios and improve the target tracking accuracy.

[0071] Based on this, the radar-based highway target tracking method proposed in the present application includes: constructing a target model according to the target unscented Kalman filter framework and the target gated recurrent neural network; generating a target motion state compensation result according to the target model; and performing a target tracking operation according to the target motion state compensation result. The present application is beneficial to avoiding using a single prior model to estimate the target state, thereby improving the applicability to target tracking in complex vehicle driving scenarios and improving the target tracking accuracy.

[0072] It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0073] The above is the introduction of the method embodiments. The following further illustrates the solution of this application through system embodiments.

[0074] Figure 2 Fig. shows a structural schematic diagram of a radar-based highway target tracking device 200 proposed by an embodiment of this application, as Figure 2 The shown device 200 includes: a construction unit 210, a generation unit 220, and an execution unit 230.

[0075] The construction unit 210 is configured to construct a target model according to a target featureless Kalman filter framework and a target gated recurrent neural network; The generation unit 220 is configured to generate a target motion state compensation result according to the target model; The execution unit 230 is configured to perform a target tracking operation according to the target motion state compensation result.

[0076] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the described modules can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0077] In the third aspect of this application, an electronic device is provided, including a memory and a processor. A computer program is stored on the memory, and when the processor executes the computer program, the method described in any one of the above is implemented.

[0078] Figure 3 Fig. shows a structural schematic diagram of an electronic device suitable for implementing the embodiments of this application.

[0079] As Figure 3 shown, the electronic device includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 into the random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other through a bus 304. The input / output (I / O) interface 305 is also connected to the bus 304.

[0080] The following components are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. as well as a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as required. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is mounted on the drive 310 as required so that a computer program read therefrom is installed into the storage section 308 as required.

[0081] In particular, according to an embodiment of the present application, the above method flow steps can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product which includes a computer program carried on a machine-readable medium, and the computer program includes program code for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 309, and / or installed from the removable medium 311. When the computer program is executed by a central processing unit (CPU) 301, the above functions defined in the system of the present application are executed.

[0082] In a fourth aspect of the present application, there is provided a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method described in any one of the above is implemented.

[0083] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of a computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus, or device. And in this application, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0084] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram can represent a module, a program segment, or a part of code, and the aforementioned module, program segment, or part of code contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0085] The units or modules involved in the embodiments described in this application can be implemented in software or in hardware. The described units or modules can also be provided in a processor. Among them, the names of these units or modules do not, in some cases, constitute a limitation on the units or modules themselves.

[0086] As another aspect, this application also provides a computer-readable storage medium, which may be included in the electronic device described in the foregoing embodiments; or may exist separately without being assembled into the electronic device. The foregoing computer-readable storage medium stores one or more programs, and when the foregoing programs are executed by one or more processors, they implement the methods described in this application.

[0087] The above description is only for the preferred embodiments of this application and the description of the technical principles applied. Those skilled in the art should understand that the scope of the application involved in this application is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the foregoing application concept. For example, the technical solutions formed by mutually replacing the above features with other technical features having similar functions (but not limited to) claimed in this application.

Claims

1. A radar-based highway target tracking method, characterized in that, Including: Construct a target model according to the target unscented Kalman filter framework and the target gated recurrent neural network; Generate a target motion state compensation result according to the target model; Execute a target tracking operation according to the target motion state compensation result.

2. The radar-based highway target tracking method according to claim 1, characterized in that, The generating a target motion state compensation result according to the target model includes: Identify target hidden state information according to a target data set; Determine a target model error according to the target hidden state information; Generate the target motion state compensation result according to the target model error and the target UKF motion state prediction result.

3. The radar-based highway target tracking method according to claim 2, wherein The target model error includes: a target state transition model error, and / or, a target measurement model error.

4. The radar-based highway target tracking method according to claim 3, characterized in that, The target UKF motion state prediction result includes: A target UKF motion state vector prediction result, a target UKF motion measurement vector prediction result, a covariance matrix prediction result corresponding to the target UKF motion state vector prediction result, and / or, a covariance matrix prediction result corresponding to the target UKF motion measurement vector prediction result.

5. The radar-based highway target tracking method according to claim 4, wherein, Also included: Determine the predicted result of the target UKF motion state vector based on the following formula: where is used to represent the predicted value of the state vector at the k th moment for the k +1th moment; is used to represent the number of points at the target sigma ; is used to represent the weight of the i th target sigma point; is used to represent the state prediction value of the i th target sigma point at the k +1th moment; is used to represent the target measurement model error at the k th moment. It is determined based on the following formula to determine the predicted covariance matrix result corresponding to the predicted result of the target UKF motion state vector: where, is used to represent the predicted value of the state covariance matrix at the (k + 1)-th moment; k +1 moment state covariance matrix prediction value; is used to represent the number of points at the target sigma point quantity; is used to represent the i th target sigma point weight; is used to represent the i th target sigma point at the ( k + 1)-th moment state prediction deviation vector; is used to represent the i th target sigma point at the ( k + 1)-th moment state prediction deviation vector transpose; is used to represent the process noise covariance matrix; is used to represent the measurement model error covariance matrix; Determined based on the following formula to determine the predicted result of the target UKF motion measurement vector: where is used to represent the predicted value of the measurement vector at the k th moment for the k +1 moment; is used to represent the number of points at the target sigma ; is used to represent the i th target sigma point weight; is used to represent the i th target sigma point's predicted value of the measurement vector at the k +1 moment; is used to represent the target state transition model error at the k +1 moment. Determine the predicted covariance matrix corresponding to the predicted result of the target UKF motion measurement vector based on the following formula: where is used to represent the predicted value of the measurement covariance matrix at the (k + 1)-th moment k ; is used to represent the number of points at the target sigma ; is used to represent the weight of the i -th target sigma point; is used to represent the measurement prediction deviation vector of the i -th target sigma point at the (k + 1)-th moment k ; is used to represent the transpose of the measurement prediction deviation vector of the i -th target sigma point at the (k + 1)-th moment k ; is used to represent the measurement noise covariance matrix; is used to represent the measurement model error covariance matrix at the (k + 1)-th moment k .

6. The radar-based highway target tracking method according to any one of claims 1 to 5, characterized in that The target gated recurrent neural network includes: a target memory update gate, a target state update gate, and a target state prediction gate; Wherein, the target state update gate, and / or, the target state prediction gate, is provided with a target activation function; The target activation function includes: Wherein, is used to represent the target activation function; is used to represent the input of the gating network; is used to represent the weight of each layer of the neural network; is used to represent the bias of each layer of the neural network; is used to represent the weight of the first layer of the neural network; is used to represent the weight of the second layer of the neural network; is used to represent the bias of the first layer of the neural network; is used to represent the bias of the second layer of the neural network; ; represents the hyperbolic tangent activation function.

7. The radar-based highway target tracking method according to claim 6, characterized in that, The target motion state compensation result includes: A target motion state compensation result corresponding to a target divergence section; And / or, a target motion state compensation result corresponding to a target convergence section.

8. A radar-based highway target tracking device, characterized in that, Including: A construction unit for constructing a target model according to the target unscented Kalman filter framework and the target gated recurrent neural network; A generation unit for generating a target motion state compensation result according to the target model; An execution unit for executing a target tracking operation according to the target motion state compensation result.

9. An electronic device, comprising a memory and a processor, wherein a computer program is stored on the memory, characterized in that, When the processor executes the computer program, the method described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, the method described in any one of claims 1 to 7 is implemented.

Images