A state space model driven track intelligent association method

Abstract

Embodiments of the present application relate to the technical field of multi-radar target association, and particularly relate to a state space model driven track intelligent association method, comprising: performing spatial registration on track data captured by a radar sensor, so that position information of the track data is converted from a radar self-coordinate system to a geocentric geodetic coordinate system; performing time registration on a pair of track data to be associated which have completed spatial registration, and using a cubic spline interpolation algorithm to time-align the two pieces of track data; inputting the two pieces of track data which have completed time and spatial registration into a feature extractor constituted by a state space model, and respectively performing sequence modeling on each piece of track data; inputting track features of the two pieces of track data into a double state space aggregation module, comparing the track features of the two pieces of track data, and generating a decision vector; and inputting the decision vector into a classifier to obtain an association result of the two pieces of track data. The method can well improve the accuracy of the association result.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of multi-radar target association, and particularly relate to a state space model driven track intelligent association method. BACKGROUND

[0002] In the technical field of multi-radar target association, in order to correctly determine whether the targets captured by multiple radars are the same individual, data association between “tracks-tracks” needs to be performed first. Therefore, as the basis of the entire decision system, the correctness of track association directly determines the quality of subsequent track fusion and target recognition, and plays a key role in subsequent decision tasks. In order to further improve the correctness of track association, many research teams have made a lot of in-depth research on various problems in the association scene, and have developed a variety of association algorithms with different characteristics.

[0003] The currently widely used track association algorithms can be roughly divided into three categories: probability statistics based, fuzzy mathematics based, and artificial intelligence and neural network based. The core idea of the association algorithm based on probability statistics is to calculate various statistical distances between tracks, and then verify whether the tracks are associated through hypothesis testing. However, with the increase of the number of targets in the observation area and the sensor error, this kind of algorithm often fails to achieve the expected results. The association algorithm based on fuzzy mathematics calculates the fuzzy membership degree between tracks through various fuzzy membership functions, and judges the association relationship between tracks by comparing the size of the fuzzy membership degree, which changes the association idea and further improves the association correctness of the algorithm in dense scenes. Although the fuzzy mathematics association algorithm solves the association problem in the case of target merging and crossing, the parameter setting in the algorithm is complex and not conducive to engineering implementation.

[0004] With the rapid development of machine learning, more and more research teams apply machine learning to the technical field of track association. The main idea is to extract features from track segment data first, and then classify tracks according to the features to complete the association. Although the association algorithm based on artificial intelligence and neural network well solves the problem of radar track association under homogeneous sensors, in the case of a large number of tracked targets, close distance, track repetition and crossing, and the influence of factors such as the decline of radar accuracy caused by the measurement error of a single multi-target radar propagating with the target distance, and the increasing complexity of track matching between multiple radars, the association speed and reliability are not good. SUMMARY

[0005] Therefore, embodiments of the present application propose a state space model driven track intelligent association method, which aims to use the modeling capability of long time series of state space model to solve the problem of track error association in dense scenes and improve the accuracy of association results.

[0006] In a first aspect, embodiments of the present application provide a state space model driven track intelligent association method, comprising the following steps: performing spatial registration on track data captured by a radar sensor, so that the position information of the track data is converted from a radar self-coordinate system to a geocentric geodetic coordinate system, to obtain track data that has completed spatial registration; performing time registration on a pair of track data that has completed spatial registration, using a cubic spline interpolation algorithm to time-align the two pieces of track data, to obtain two pieces of track data that have completed time and spatial registration; inputting the two pieces of track data that have completed time and spatial registration into a feature extractor composed of a state space model, performing sequence modeling on each piece of track data respectively, to obtain track features of the two pieces of track data; inputting the track features of the two pieces of track data into a double-state space aggregation module, comparing the track features of the two pieces of track data, to generate a decision vector; and inputting the decision vector into a classifier, to obtain an association result of the two pieces of track data.

[0007] Optionally, the spatial registration on the track data captured by the radar sensor, so that the position information of the track data is converted from the radar self-coordinate system to the geocentric geodetic coordinate system, to obtain the track data that has completed spatial registration, comprises:

[0008] Let the coordinates of the radar sensor in the latitude-longitude coordinate system be l、 H respectively represent the longitude, latitude and altitude of the radar sensor in the latitude-longitude coordinate system;

[0009] Convert the coordinates of the radar sensor in the latitude-longitude coordinate system to the geocentric geodetic coordinate system, and the coordinates of the radar sensor in the geocentric geodetic coordinate system are:

[0010]

[0011] wherein N is the radius of the equator, (x ′ ,y ′ ,z ′ ) are the coordinates of the radar sensor in the geocentric geodetic coordinate system;

[0012] Rotate the coordinate axes of the radar self-coordinate system to be parallel to the coordinate axes of the geocentric geodetic coordinate system, and then attach the position quantities of the track data captured by the radar sensor to the coordinates of the radar sensor in the geocentric geodetic coordinate system, to complete the conversion of the track data from the radar self-coordinate system to the geocentric geodetic coordinate system, to obtain the track data that has completed spatial registration;

[0013] The conversion formula of the track data from the radar self-coordinate system to the geocentric geodetic coordinate system is:

[0014]

[0015] wherein (x 1i ,y1i , z 1i ) represents the coordinates of a point i in the radar self-coordinate system in the track data, (x ei , y ei , z ei ) represents the coordinates of a point i in the track data in the geocentric geodetic coordinate system, R T is a rotation matrix determined based on the coordinates of the radar sensor in the latitude and longitude coordinate system.

[0016] Optionally, the pair of track data to be associated is time-registered, and a cubic spline interpolation algorithm is used to time-align the two pieces of track data, so as to obtain two pieces of track data that are time-registered and spatially registered, including:

[0017] A pair of track data to be associated is denoted as first track data and second track data;

[0018] The latest M seconds of the common time of the first track data and the second track data are intercepted, and the two pieces of track data are sampled into time sequences of a preset length according to a cubic spline interpolation function, so as to obtain two pieces of track data that are time-registered and spatially registered;

[0019] The two pieces of track data are sampled into time sequences of a preset length according to a cubic spline interpolation function, so as to obtain two pieces of track data that are time-registered and spatially registered, which is expressed by a formula as follows:

[0020] A align,1 = S[I(A ori,1 )];

[0021] A align,2 = S[I(A ori,2 )];

[0022] Wherein, A ori,1 and A ori,2 represent the first track data and the second track data respectively, I(·) represents a cubic spline interpolation function, S(·) represents a sampling function, A align,1 and A align,2 represent the first track data that is time-registered and spatially registered and the second track data that is time-registered and spatially registered respectively.

[0023] Optionally, the two pieces of track data that are time-registered and spatially registered are input into a feature extractor constituted by a state space model, and sequence modeling is performed on each piece of track data respectively, so as to obtain track features of the two pieces of track data, including:

[0024] The two pieces of track data that are time-registered and spatially registered are input into a feature extractor constituted by a state space model, and normalization is performed first, and then the sequence feature dimensions of the two pieces of track data after normalization are projected to P through a linear projection layer;

[0025] The two track data after linear projection are input into the state space model to obtain track features of the two track data, and the track features of the two track data are expressed by a formula as follows:

[0026] F1=SSM(T1);

[0027] F2=SSM(T2);

[0028] T1=Linear(A align,1 );

[0029] T2=Linear(A align,2 );

[0030] Wherein, Linear(·) represents a linear projection layer, SSM(·) represents a state space model, F1 represents track features of a first track data in the two track data, and F2 represents track features of a second track data in the two track data.

[0031] Optionally, the track features of the two track data are input into a double state space aggregation module, the track features of the two track data are compared, a decision vector is generated, and the following formula is used to realize the comparison:

[0032] O=Linear[SSM(H1),SSM(H2)];

[0033] H1=Norm[Linear(F1)];

[0034] yQ2=Norm[Linear(F2)];

[0035] Wherein, Norm(·) represents a normalization layer, and O represents a decision vector output by the double state space aggregation module.

[0036] Optionally, the classifier is composed of a multilayer perceptron, and a classification effect of an arbitrary nonlinear function is realized by stacking a linear layer and an activation function. The decision vector is input into the classifier to obtain a correlation result of the two tracks, and the following formula is used to express the correlation result:

[0037] c=Mlp(O);

[0038] Wherein, Mlp(·) represents a multilayer perceptron, and c represents a correlation result of the two tracks.

[0039] Optionally, the feature extractor composed of the state space model, the double state space aggregation module and the classifier jointly constitute a track correlation model, and the track correlation model is obtained by the following steps:

[0040] The sample track data pair with the real association label is input into the track association model to obtain an association result of the sample track data pair output by the track association model;

[0041] Based on the cross-entropy loss function, a loss value between the association result of the sample track data pair output by the track association model and the real association label is calculated. The calculated loss value is represented by a formula as follows:

[0042]

[0043] wherein Batch represents a batch size, represents a real association label of the bth sample track data pair, c b represents an association result of the bth sample track data pair output by the track association model, and Loss represents the calculated loss value.

[0044] Based on the loss value, the network parameters of the track association model are updated by using the stochastic gradient descent method until the track association model converges, the optimal network parameters are saved, and the trained track association model is obtained.

[0045] The embodiment of the application proposes a state space model driven track intelligent association method. First, the track data captured by the radar sensor is spatially registered, so that the position information of the track data is converted from the radar self-coordinate system to the geocentric geodetic coordinate system. Then, a pair of track data to be associated after spatial registration is time-registered, so that the two track data are time-aligned, and two regular track data after time and space registration are obtained. The time and space registration process facilitates subsequent model processing and helps to improve the determination speed of the association result. The feature extractor composed of the state space model is used for feature extraction, and the double-state space aggregation module is used for comparison and decision. Finally, the classifier outputs the association result. As a component of the feature extractor and the feature aggregation network, the state space model has excellent performance in capturing local correlation and long sequence modeling, effectively solving the association problem under the condition of dense track. In addition, the state space model also has outstanding advantages in hardware acceleration processing, which can improve the association speed and is easier to deploy in actual scenarios.

[0046] In a second aspect, embodiments of the present application provide a state space model driven track intelligent association system, the system comprising: a spatial registration unit, a time registration unit, and a track association model, the track association model comprising a feature extractor constituted by a state space model, a dual state space aggregation module, and a classifier; the spatial registration unit is configured to perform spatial registration on track data captured by a radar sensor, so that the position information of the track data is converted from a radar self-coordinate system to a geocentric geodetic coordinate system, to obtain track data that has completed spatial registration; the time registration unit is configured to perform time registration on a pair of track data that has completed spatial registration, and uses a cubic spline interpolation algorithm to time-align the two pieces of track data, to obtain two pieces of track data that have completed time and spatial registration; the feature extractor constituted by the state space model is configured to perform sequence modeling on each piece of track data respectively, to obtain track features of the two pieces of track data; the dual state space aggregation module is configured to compare the track features of the two pieces of track data, to generate a decision vector; and the classifier is configured to input the decision vector into the classifier, to obtain an association result of the two pieces of track data.

[0047] It can be understood that the beneficial effects of the second aspect described above can be referred to the related description in the first aspect described above, and will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technical solutions, the drawings needed to be used in the embodiments of the present application or the related technical solutions will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort based on these drawings.

[0049] Figure 1 is a flowchart of a state space model driven track intelligent association method provided in an embodiment of the present application;

[0050] Figure 2 is a structural schematic diagram of a track association model provided in an embodiment of the present application;

[0051] Figure 3 is a structural schematic diagram of a feature extractor provided in an embodiment of the present application;

[0052] Figure 4 is a structural schematic diagram of a state space model driven track intelligent association system provided in another embodiment of the present application;

[0053] Figure 5 is a structural schematic diagram of an electronic device deployed in another embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below with reference to the drawings. In the various embodiments of the present application, many technical details are proposed in order to make the readers better understand the present application. However, the technical solutions claimed by the present application can be implemented even without these technical details and various changes and modifications based on the various embodiments below. The division of the various embodiments below is only for the convenience of description, and should not constitute any limitation on the specific implementation modes of the present application. The various embodiments can be combined and referenced to each other on the premise of not contradicting.

[0055] One embodiment of the present application proposes a state space model driven track intelligent association method, which is applied to an electronic device. The electronic device can be a terminal or a server. In the embodiments of the present application and the various embodiments below, the electronic device is taken as an example of a server. The implementation details of the state space model driven track intelligent association method proposed in the embodiment of the present application will be described in detail below. The following content only provides related implementation details for the convenience of understanding, and is not necessary for implementing the present solution.

[0056] The specific process of the state space model driven track intelligent association method proposed in the embodiment of the present application can be as shown in Figure 1 , which includes:

[0057] In step 101, the track data captured by the radar sensor is spatially registered, so that the position information of the track data is converted from the radar self-coordinate system to the geocentric geodetic coordinate system, to obtain the track data after spatial registration.

[0058] In a specific implementation, the track data captured by the radar sensor is recorded based on the radar self-coordinate system, and needs to be converted to the geocentric geodetic coordinate system to be more scientific and reasonable. Therefore, the server first needs to spatially register the track data captured by the radar sensor, so that the position information of the track data is converted from the radar self-coordinate system to the geocentric geodetic coordinate system, to obtain the track data after spatial registration.

[0059] In one example, the coordinates of the radar sensor in the latitude-longitude coordinate system are l、 H respectively represent the longitude, latitude and altitude of the radar sensor in the latitude-longitude coordinate system. When the server performs spatial registration of the track data, it first needs to convert the coordinates of the radar sensor in the latitude-longitude coordinate system to the geocentric geodetic coordinate system.

[0060] The coordinates of the radar sensor in the geocentric geodetic coordinate system can be expressed as:

[0061]

[0062] wherein N is the radius of the lunar orbit, a represents the long radius of the earth, e represents the first eccentricity of the earth, (x ′ ,y ′ ,z ′ ) is the coordinate of the radar sensor in the geocentric geodetic coordinate system.

[0063] After obtaining the coordinate of the radar sensor in the geocentric geodetic coordinate system, the server can rotate the coordinate axis of the radar self-coordinate system to be parallel to the coordinate axis of the geocentric geodetic coordinate system, and then attach the position quantity of the track data captured by the radar sensor to the coordinate of the radar sensor in the geocentric geodetic coordinate system, to complete the conversion of the track data from the radar self-coordinate system to the geocentric geodetic coordinate system, and obtain the track data completed with spatial registration.

[0064] The conversion formula of the track data from the radar self-coordinate system to the geocentric geodetic coordinate system can be expressed as:

[0065]

[0066] wherein (x 1i ,y 1i ,z 1i ) represents the coordinate of a point i in the track data in the radar self-coordinate system (also known as the northern celestial east coordinate system), (x ei ,y ei ,z ei ) represents the coordinate of a point i in the track data in the geocentric geodetic coordinate system, and R T is a rotation matrix determined based on the coordinate of the radar sensor in the latitude and longitude coordinate system.

[0067] In step 102, a pair of track data completed with spatial registration to be associated is time-registered, and a cubic spline interpolation algorithm is used to time-align the two track data, to obtain two track data completed with spatial registration and time.

[0068] In a specific implementation, after completing the spatial registration of the track data captured by the radar sensor, the server can time-register a pair of track data completed with spatial registration to be associated, and a specific manner is to use a cubic spline interpolation algorithm to time-align the two track data, to obtain two track data completed with spatial registration and time.

[0069] In one example, the pair of track data to be associated for complete spatial registration are denoted as first track data and second track data respectively. The server intercepts the latest M seconds (usually set to 60 seconds) of common time of the first track data and the second track data, and samples the two track data into time series of preset length according to a cubic spline interpolation function, to obtain two track data of complete temporal-spatial registration. The preset length is usually set to 16, and the specific value can be set by a person skilled in the art according to actual needs.

[0070] In one example, the server samples the two track data into time series of preset length according to a cubic spline interpolation function to obtain two track data of complete temporal-spatial registration, which can be expressed by a formula as follows:

[0071] A align,1 = S [I (A ori,1 )];

[0072] A align,2 = S [I (A ori,2 )];

[0073] wherein A ori,1 and A ori,2 denote the first track data and the second track data respectively, I(·) denotes a cubic spline interpolation function, S(·) denotes a sampling function, A align,1 and A align,2 denote the first track data of complete temporal-spatial registration and the second track data of complete temporal-spatial registration respectively.

[0074] Step 103: input the two track data of complete temporal-spatial registration into a feature extractor composed of a state space model, and perform sequence modeling on each track data respectively to obtain track features of the two track data.

[0075] In a specific implementation, after the temporal-spatial registration of the track data is completed, the track association model can be used to determine whether the two track data are associated. The network structure of the track association model can be as shown in Figure 2 The track association model is composed of a feature extractor, a double state space aggregation module and a classifier. The feature extractor is composed of a state space model, and the server inputs the two track data of complete temporal-spatial registration into the feature extractor composed of the state space model, and performs sequence modeling on each track data respectively to obtain track features of the two track data.

[0076] In one example, the network structure of the feature extractor can be as shown in Figure 3As shown, the server inputs the two pieces of time-spatially registered track data into a feature extractor constituted by a state space model, normalizes the two pieces of track data, projects the sequence feature dimensions of the two pieces of track data from 3 to P through a linear projection layer, and then inputs the two pieces of track data after linear projection into the state space model to obtain track features of the two pieces of track data.

[0077] In one example, the track features of the two pieces of track data are expressed by a formula as follows:

[0078] F1 = SSM(T1);

[0079] F2 = SSM(T2);

[0080] T1 = Linear(A align,1 );

[0081] T2 = Linear(A align,2 );

[0082] wherein Linear(·) represents a linear projection layer, SSM(·) represents a state space model, T1 represents the first piece of track data after linear projection, T2 represents the second piece of track data after linear projection, T1 ∈ R 16×P , T2 ∈ R 16×P , F1 represents the track feature of the first piece of track data of the two pieces of track data, and F2 represents the track feature of the second piece of track data of the two pieces of track data.

[0083] In one example, the specific calculation process of the state space model can be expressed as:

[0084]

[0085] wherein A, B, and C are all network parameters learnable in the state space model, h t represents a hidden state, h -1 = 0, x t and y t represent the input and output of the state space model respectively, and Δ represents the change step of x t .

[0086] Step 104: inputting the track features of the two pieces of track data into a double-state space aggregation module, comparing the track features of the two pieces of track data, and generating a decision vector.

[0087] In a specific implementation, the dual-state space aggregation module is also built based on the state space model. The track feature of the two track data is input into the dual-state space aggregation module by the feature extractor. After the normalization layer and the projection layer, the track feature of the two track data is compared in the state space model, and finally the decision vector is generated.

[0088] In one example, the track feature of the two track data is input into the dual-state space aggregation module, the track feature of the two track data is compared, and the decision vector is generated. This can be achieved by the following formula:

[0089] O = Linear[SSM(H1), SSM(H2)];

[0090] H1 = Norm[Linear(F1)];

[0091] H2 = Norm[Linear(F2)];

[0092] Wherein, Norm(·) represents the normalization layer, and O represents the decision vector output by the dual-state space aggregation module.

[0093] Step 105, input the decision vector into the classifier to obtain the association result of the two tracks.

[0094] In a specific implementation, the dual-state space aggregation module inputs the decision vector into the classifier, and the classifier classifies whether to associate to obtain the association result of the two tracks.

[0095] In one example, the classifier is composed of a multilayer perceptron, which realizes the classification effect of any nonlinear function by stacking linear layers and activation functions. The decision vector is input into the classifier to obtain the association result of the two tracks, which can be expressed by the formula:

[0096] c = Mlp(O);

[0097] Wherein, Mlp(·) represents the multilayer perceptron, and c represents the association result of the two tracks.

[0098] In one example, the feature extractor, the dual-state space aggregation module and the classifier composed of the state space model jointly constitute the track association model. The track association model needs to be trained before use. The server needs to perform supervised iterative training on the track association model.

[0099] The server inputs the sample track data pair with the real association label into the track association model to obtain an association result of the sample track data pair output by the track association model. Then, based on a cross-entropy loss function, a loss value between the association result of the sample track data pair output by the track association model and the real association label is calculated. Finally, based on the loss value, the network parameters of the track association model are updated by using a stochastic gradient descent method until the track association model converges, the optimal network parameters are saved, and a trained track association model is obtained.

[0100] The calculated loss value can be represented by a formula as follows:

[0101]

[0102] where Batch represents a batch size, represents a real association label of the bth sample track data pair, c b represents an association result of the bth sample track data pair output by the track association model, and Loss represents the calculated loss value.

[0103] In an example, when the track association model is trained, an Adam optimizer with a learning rate of 1e-3 is used, the learning rate reduction rate is set to 0.1 per 1000 epochs, and the learning (training) process is completed after 10000 epochs.

[0104] The track intelligent association method driven by the state space model proposed in the embodiment performs spatial registration on the track data captured by the radar sensor, converts the position information of the track data from the radar self-coordinate system to the geocentric geodetic coordinate system, and then performs time registration on a pair of track data to be associated, so that the time of the two track data is aligned, and two track data that are regularly time and space registered are obtained. The time and space registration process facilitates the subsequent model processing and helps to improve the determination speed of the association result. The feature extractor composed of the state space model is used for feature extraction, and the double-state space aggregation module is used for comparison and decision-making. Finally, the classifier outputs the association result. As a component of the feature extractor and the feature aggregation network, the state space model has excellent performance in capturing local correlation and long sequence modeling, effectively solving the association problem in the case of dense track. In addition, the state space model also has outstanding advantages in hardware acceleration processing, which can improve the association speed and is easier to deploy in actual scenarios.

[0105] The step division of the above various methods is only for the purpose of clear description, and can be combined into one step or split into multiple steps in implementation, as long as the same logical relationship is included, and all are within the protection scope of the present application; adding insignificant modifications or introducing insignificant designs in the algorithm or flow, but not changing the core design of the algorithm and flow, are within the protection scope of the present application.

[0106] Another embodiment of the present application proposes a state space model driven track intelligent association system, and the implementation details of the state space model driven track intelligent association system proposed in the embodiment are specifically described below. The following details are provided for the convenience of understanding, and are not essential for implementing the embodiment, Figure 4 is a structural schematic diagram of the state space model driven track intelligent association system proposed in the embodiment. The system includes a space registration unit 201, a time registration unit 202, and a track association model 300. The track association model 300 includes a feature extractor 301 composed of a state space model, a double state space aggregation module 302, and a classifier 303.

[0107] The space registration unit 201 is configured to perform space registration on track data captured by a radar sensor, so that the position information of the track data is converted from a radar self-coordinate system to a geocentric geodetic coordinate system, and track data that has completed space registration is obtained.

[0108] The time registration unit 202 is configured to perform time registration on a pair of track data that has completed space registration, and uses a cubic spline interpolation algorithm to align the time of the two pieces of track data, so that two pieces of track data that have completed time registration are obtained.

[0109] The feature extractor 301 composed of a state space model is configured to perform sequence modeling on each piece of track data respectively, so that track features of the two pieces of track data are obtained.

[0110] The double state space aggregation module 302 is configured to compare the track features of the two pieces of track data, and generate a decision vector.

[0111] The classifier 303 is configured to input the decision vector into the classifier, so that an association result of the two pieces of track data is obtained.

[0112] It is worth mentioning that each module involved in the embodiment is a logical module. In actual application, one logical unit can be one physical unit, or a part of one physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of the present application, units not closely related to solving the technical problems proposed in the present application are not introduced in the embodiment, but this does not mean that there are no other units in the embodiment.

[0113] It can be found that the embodiment is a system embodiment corresponding to the method embodiments, and the embodiment can be implemented in cooperation with the method embodiments. The related technical details and technical effects mentioned in the above embodiments are still valid in this embodiment. In order to reduce repetition, they will not be described here. Accordingly, the related technical details mentioned in the embodiment can also be applied in the above embodiments.

[0114] Another embodiment of the present application deploys an electronic device, the specific structure of which is shown as Figure 5 The electronic device includes at least one processor 401 and a memory 402 connected with the at least one processor 401, wherein the memory 402 stores instructions executable by the at least one processor 401, and the instructions are executed by the at least one processor 401 to enable the at least one processor 401 to perform the state space model driven track intelligent association method described in the above embodiments.

[0115] The memory and the processor can be connected in a bus manner, the bus can include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripheral devices, voltage stabilizers, and power management circuits together, which are well known in the art, and will not be further described herein. The bus interface is responsible for providing an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, which provide a unit for communicating with various other devices on the transmission medium. The data processed by the processor is transmitted on the wireless medium through the antenna, and further, the antenna also receives data and transmits the data to the processor.

[0116] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interface, voltage regulation, power management, and other control functions. The memory can be used to store data used by the processor during execution.

[0117] Those skilled in the art can understand that all or part of the steps of the methods in the above embodiments can be completed by programs instructing related hardware, the programs are stored in a storage medium, and include a plurality of instructions for causing a device (such as a single-chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a ROM (Read-Only Memory), a RAM (Random Access Memory), a magnetic disk or an optical disk, and various storage medium capable of storing program codes.

[0118] Those skilled in the art can understand that the above-mentioned embodiments are specific embodiments for implementing the present application, and various changes can be made in form and details in actual application without departing from the spirit and scope of the present application.

Claims

1. A state-space model-driven intelligent trajectory association method, characterized in that, include: Spatial registration is performed on the track data captured by the radar sensor to transform the position information of the track data from the radar self-coordinate system to the geocentric coordinate system, thus obtaining spatially registered track data. For a pair of spatially registered track data to be associated, time registration is performed. A cubic spline interpolation algorithm is used to align the time of the two track data, resulting in two spatially registered track data. Two space-time registered track data are input into a feature extractor composed of a state-space model. Sequence modeling is performed on each track data to obtain the track features of the two track data. The trajectory features of the two trajectory data are input into the dual-state space aggregation module, and the trajectory features of the two trajectory data are compared to generate a decision vector; The decision vector is input into the classifier to obtain the association result of the two tracks; The process involves inputting two space-time registered track data sets into a feature extractor composed of a state-space model, performing sequence modeling on each track data set to obtain the track features of the two track data sets, including: Two space-time registered track data sets are input into a feature extractor composed of a state-space model. Normalization is performed first, and then the sequence feature dimensions of the two normalized track data sets are projected onto a linear projection layer. ; The two trajectory data points, after linear projection, are input into the state-space model to obtain the trajectory features of the two trajectory data points. The trajectory features of the two trajectory data points are expressed by the following formula: ； ； ； ； in, and These represent the first track data and the second track data after spatial registration at completion, respectively. Indicates a linear projection layer. Representing a state-space model, This represents the track characteristics of the first track data in the two track data sets. This indicates the track characteristics of the second track data in two track data sets. The process involves inputting the trajectory features of two trajectory data points into a dual-state space aggregation module, comparing the trajectory features of the two trajectory data points, and generating a decision vector, which is achieved through the following formula: ； ； ； in, Indicates the normalization layer. This represents the decision vector output by the dual-state space aggregation module.

2. The state-space model-driven intelligent trajectory association method as described in claim 1, characterized in that, The process of spatially registering the track data captured by the radar sensor, transforming the position information of the track data from the radar's self-coordinate system to the geocentric coordinate system, to obtain spatially registered track data includes: Let the coordinates of the radar sensor in the latitude and longitude coordinate system be... , , , These represent the longitude, latitude, and altitude of the radar sensor in the latitude-longitude coordinate system, respectively. Transform the radar sensor's coordinates from the latitude and longitude coordinate system to the geocentric coordinate system. The radar sensor's coordinates in the geocentric coordinate system are: ； in, The radius of the circle is the area between the east and west. These are the coordinates of the radar sensor in the geocentric geodetic coordinate system. The radar's own coordinate system axes are rotated to be parallel to the coordinate system of the geocentric coordinate system. Then, the position data of the track data captured by the radar sensor is added to the radar sensor's coordinates in the geocentric coordinate system. This completes the transformation of the track data from the radar's own coordinate system to the geocentric coordinate system, resulting in track data with complete spatial registration. The formula for converting track data from the radar self-coordinate system to the geocentric coordinate system is as follows: ； ； in, Represents a point in the track data Coordinates in the radar's self-coordinate system Represents a point in the track data Coordinates in the geocentric geodetic coordinate system It is a rotation matrix determined based on the coordinates of the radar sensor in the latitude and longitude coordinate system.

3. The state-space model-driven intelligent trajectory association method as described in claim 2, characterized in that, The process involves temporal registration of two spatially registered track data sets to be associated, using a cubic spline interpolation algorithm to align the time of the two track data sets, resulting in two spatially registered track data sets. This includes: Let the track data that have been spatially registered and are to be associated be the first track data and the second track data; The latest common time of the first and second track data is extracted. Seconds, the two track data are sampled into a time series of a preset length according to the cubic spline interpolation function to obtain two track data with completed time-space registration; The two track data are sampled into time series of a preset length using a cubic spline interpolation function to obtain two track data sets with completed time-space registration, which can be expressed by the formula: ； ； in, and These represent the first track data and the second track data, respectively. This represents the cubic spline interpolation function. Represents the sampling function. and These represent the first track data and the second track data after spatial registration at completion, respectively.

4. The state-space model-driven intelligent trajectory association method as described in claim 1, characterized in that, The classifier consists of a multilayer perceptron, which achieves the classification effect of arbitrary nonlinear functions by stacking linear layers and activation functions. The decision vector is input into the classifier to obtain the association result between two tracks, expressed by the formula: ； in, This represents a multilayer perceptron. This indicates the correlation result between two flight paths.

5. A state-space model-driven intelligent trajectory association method as described in any one of claims 1 to 4, characterized in that, The trajectory association model is composed of a feature extractor (based on a state-space model), a dual-state-space aggregation module, and a classifier. The trajectory association model is trained through the following steps: Input sample track data pairs with real association labels into the track association model to obtain the association results of the sample track data pairs output by the track association model; Based on the cross-entropy loss function, the loss value between the association result of the sample track data pairs output by the track association model and the true association label is calculated; the calculated loss value is expressed by the formula: ； in, Indicates batch size, Indicates the first The true association labels for sample track data pairs, The output of the trajectory association model represents the first... The correlation results of sample track data pairs This represents the calculated loss value; Based on the loss value, the network parameters of the trajectory association model are updated using the stochastic gradient descent method until the trajectory association model converges. The optimal network parameters are then saved to obtain the trained trajectory association model.

6. A state-space model-driven intelligent trajectory association system, characterized in that, The system is based on a state-space model-driven intelligent trajectory association method as described in any one of claims 1 to 5. The system includes: a spatial registration unit, a temporal registration unit, and a trajectory association model. The trajectory association model includes a feature extractor composed of a state-space model, a dual-state-space aggregation module, and a classifier. The spatial registration unit is used to spatially register the track data captured by the radar sensor, so that the position information of the track data is transformed from the radar self-coordinate system to the geocentric geoordinate system, and the spatially registered track data is obtained. The time registration unit is used to perform time registration on two spatially registered track data to be associated. The cubic spline interpolation algorithm is used to align the time of the two track data to obtain two spatially registered track data. A feature extractor composed of a state-space model is used to perform sequence modeling on each track data to obtain the track features of the two track data. The dual-state space aggregation module is used to compare the track features of two track data sets and generate a decision vector; A classifier is used to input decision vectors into the classifier to obtain the association result between two tracks.

Images