Sparse trajectory recovery method based on generative adversarial network

Abstract

The invention discloses a sparse trajectory recovery method based on a generative adversarial network. The method comprises the following steps: for each round of iteration, acquiring a first track sample data set which comprises a real sparse track data sample and a real dense track data sample; feeding the first track sample data set to the GAN, so that a discriminator of the GAN outputs a first discrimination value; acquiring a second track sample data set, wherein the second track sample data set is generated by processing a real sparse track data sample by the generator of the GAN; feeding the second track sample data set to the GAN for the discriminator to output a second discrimination value; optimizing the general objective function of the GAN to obtain network parameters of the generator; and updating the generator according to the obtained generator network parameters. When the execution of the current iteration is finished, determining whether to stop the iteration; and in response to determining to stop iteration, taking the updated generator as a final generator to generate a complemented dense track data sample based on the to-be-complemented sparse track data sample.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a sparse trajectory recovery method based on a generative adversarial network. Background Art

[0002] With the development of information and science and technology, future operations will shift towards electronic and information-based operations. Especially in air defense and anti-missile operations, for high-speed moving objects such as missiles and fighter jets, operations will present the characteristics of strong real-time variability, strong uncertainty in combat information, and strong game confrontation. The prediction of target tracks is a key research direction in battlefield game confrontation. The target track information obtained in operations often has problems such as incomplete data and insignificant time correlation.

[0003] The methods described in this section are not necessarily methods that have been previously conceived or employed. Unless otherwise indicated, it should not be assumed that any method described in this section is considered to be prior art simply because it is included in this section. Similarly, unless otherwise indicated, the issues mentioned in this section should not be considered to have been recognized in any prior art. Summary of the invention

[0004] The present disclosure provides a track data processing method, device, electronic device and computer-readable storage medium.

[0005] According to one aspect of the present disclosure, a track data processing method is provided for recovering sparse track data samples with missing data points using a generative adversarial network, the generative adversarial network comprising a discriminator and a generator, the method comprising: for each round of one or more iterations, performing the following steps: obtaining a first track sample data set, the first track sample data set comprising real track data samples, the real track data samples comprising both real sparse track data samples with missing data points and real dense track data samples with complete data points; feeding the first track sample data set to the generative adversarial network so that the discriminator outputs a first discriminant value for the first track sample data set, the first discriminant value being used to calculate the function value of the discriminator's objective function; obtaining a second track sample data set, the second track sample data set being generated by the generator processing the real sparse track data samples with missing data points; feeding the second track sample data set to the generative adversarial network; The data set is fed to the generative adversarial network so that the discriminator outputs a second discriminant value for the second track sample data set, and the second discriminant value is used to calculate the function value of the generator objective function; the overall objective function of the generative adversarial network is optimized to obtain the network parameters of the generator when the overall objective function takes the optimal value, wherein the overall objective function is constructed based on the discriminator objective function and the generator objective function; and the generator is updated according to the obtained network parameters of the generator; when all steps of this round of iteration are executed, it is determined whether the iteration stop condition is met; and in response to determining that the iteration stop condition is met, the updated generator is used as the final generator, so that the final generator generates the completed dense track data samples with complete data points based on the sparse track data samples to be completed with missing data points; or in response to determining that the iteration stop condition is not met, the next round of iteration is performed based on the updated generator.

[0006] According to another aspect of the present disclosure, a track data processing device is provided for recovering sparse track data samples with missing data points using a generative adversarial network, the generative adversarial network including a discriminator and a generator, the device including: a single-round execution unit, for performing the following steps for each round of one or more iterations: obtaining a first track sample data set, the first track sample data set comprising real track data samples, the real track data samples comprising both real sparse track data samples with missing data points and real dense track data samples with complete data points; feeding the first track sample data set to the generative adversarial network so that the discriminator outputs a first discriminant value for the first track sample data set, the first discriminant value being used to calculate the function value of the discriminator's objective function; obtaining a second track sample data set, the second track sample data set being generated by the generator processing the real sparse track data samples with missing data points; feeding the second track sample data set to the generator An adversarial network is provided, so that the discriminator outputs a second discriminant value for the second track sample data set, and the second discriminant value is used to calculate the function value of the generator objective function; the overall objective function of the generated adversarial network is optimized to obtain the network parameters of the generator when the overall objective function takes the optimal value, wherein the overall objective function is constructed based on the discriminator objective function and the generator objective function; and the generator is updated according to the obtained network parameters of the generator; an iteration determination unit is used to determine whether the iteration stop condition is met when all steps of this round of iteration are executed; a final determination unit is used to respond to determining that the iteration stop condition is met, and use the updated generator as the final generator, so that the final generator generates the completed dense track data samples with complete data points based on the sparse track data samples to be completed with missing data points; and an iteration triggering unit is used to respond to determining that the iteration stop condition is not met, and perform the next round of iteration based on the updated generator.

[0007] According to another aspect of the present disclosure, there is provided an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program, and the computer program implements the above method when executed by the at least one processor.

[0008] According to yet another aspect of the present disclosure, a non-transitory computer-readable storage medium storing a computer program is provided, wherein the computer program implements the above method when executed by a processor.

[0009] According to yet another aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein the computer program implements the above method when executed by a processor.

[0010] According to one or more embodiments of the present disclosure, incomplete track data is restored into dense track data through a track recovery technology based on a generative adversarial network. On the one hand, the data volume requirement is significantly reduced compared with the baseline method, and on the other hand, the accuracy and completeness of track recovery are improved.

[0011] These and other aspects of the disclosure will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Further details, features and advantages of the present disclosure are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0013] Figure 1 is a schematic diagram illustrating an example system in which the various methods described herein may be implemented according to an exemplary embodiment;

[0014] Figure 2 is a flow chart illustrating a method of processing track data according to an exemplary embodiment;

[0015] Figure 3 is a schematic diagram illustrating the architecture of a generative adversarial network according to an exemplary embodiment;

[0016] Figure 4 is a diagram illustrating a comparison of visualization results of recovering a track using a method according to an exemplary embodiment and a baseline method;

[0017] Figure 5 is a schematic block diagram illustrating a track data processing apparatus according to an exemplary embodiment;

[0018] Figure 6 is a block diagram illustrating an exemplary computer device that can be applied to the exemplary embodiments. DETAILED DESCRIPTION

[0019] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0020] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, timing relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0021] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0022] With the development of information and science and technology, future operations will shift towards electronic and information-based operations. Especially in air defense and anti-missile operations, for high-speed moving objects such as missiles and fighter jets, operations will present characteristics such as strong real-time variability, strong uncertainty in combat information, and strong game confrontation. The prediction of target tracks is a key research direction in battlefield game confrontation. The target track information obtained in combat often has problems such as incomplete data and insignificant time correlation. How to obtain target tracks with the desired accuracy and completeness has become an urgent problem to be solved in the combat process.

[0023] Based on this, the present disclosure proposes a track data processing method, firstly, by obtaining a first track sample data set including both real sparse track data samples and real dense track data samples and feeding it to a generative adversarial network (GAN), a score of the network's discriminator on the authenticity of the real track data samples is obtained, and then the real sparse track data samples are processed by the network's generator to generate a second track sample data set that is close to the real dense track data samples and the second track sample data set is also fed to the generative adversarial network to obtain a score of the discriminator on the authenticity of the non-real track data samples in the second track sample data set, and then the overall objective function of the network is optimized to make the ability of the discriminator (i.e., better distinguish between real track data and non-real track data) and the ability of the generator (i.e., generate more and more The non-real track data that is close to the real track data) compete with each other in the optimization process and gradually tend to the Nash equilibrium, and finally the final updated generator is obtained after all rounds of iterations are completed, so that the obtained generator can be used to generate completed dense track data samples with complete data points based on the sparse track data samples to be completed with missing data points. At this point, with the help of the method disclosed in the present invention, the incomplete track data can be restored to dense track data through the track recovery technology based on the generative adversarial network. On the one hand, compared with the baseline method, the data volume requirement is significantly reduced, and on the other hand, the accuracy and completeness of the track recovery are improved.

[0024] Exemplary embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0025] Figure 1 is a schematic diagram illustrating an example system 100 in which the various methods described herein may be implemented, according to an example embodiment.

[0026] refer to Figure 1 The system 100 includes a client device 110 , a server 120 , and a network 130 that communicatively couples the client device 110 and the server 120 .

[0027] The client device 110 includes a display 114 and a client application (APP) 112 that can be displayed via the display 114. The client application 112 can be an application that needs to be downloaded and installed before running or a small program (liteapp) as a lightweight application. In the case where the client application 112 is an application that needs to be downloaded and installed before running, the client application 112 can be pre-installed on the client device 110 and activated. In the case where the client application 112 is a small program, the user 102 can directly run the client application 112 on the client device 110 by searching for the client application 112 in the host application (for example, by the name of the client application 112, etc.) or scanning the graphic code of the client application 112 (for example, a bar code, a QR code, etc.), without installing the client application 112. In some embodiments, the client device 110 can be any type of mobile computer device, including a mobile computer, a mobile phone, a wearable computer device (for example, a smart watch, a head-mounted device, including smart glasses, etc.) or other types of mobile devices. In some embodiments, client device 110 may alternatively be a stationary computer device, such as a desktop computer, a server computer, or other type of stationary computer device.

[0028] The server 120 is typically a server deployed by an Internet Service Provider (ISP) or an Internet Content Provider (ICP). The server 120 may represent a single server, a cluster of multiple servers, a distributed system, or a cloud server that provides basic cloud services (such as cloud databases, cloud computing, cloud storage, and cloud communications). It will be understood that although Figure 1 The server 120 is shown in FIG. 1 in communicating with only one client device 110 , but the server 120 may provide background services for multiple client devices simultaneously.

[0029] Examples of network 130 include a local area network (LAN), a wide area network (WAN), a personal area network (PAN), and / or a combination of communication networks such as the Internet. Network 130 can be a wired or wireless network. In some embodiments, the data exchanged through network 130 is processed using technologies and / or formats including hypertext markup language (HTML), extensible markup language (XML), etc. In addition, encryption technologies such as secure socket layer (SSL), transport layer security (TLS), virtual private network (VPN), Internet protocol security (IPsec) can also be used to encrypt all or some links. In some embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.

[0030] For the purpose of the embodiments of this disclosure, Figure 1In the example, the client application 112 may be a track data processing program, which may provide various functions for recovering sparse track data samples with missing data points, such as collecting real track data samples, initializing the generative adversarial network and / or setting the iteration conditions of the generative adversarial network, etc. Correspondingly, the server 120 may be a server used together with the track data processing program. The server 120 may provide services related to the recovery of sparse track data samples to the client application 112 running in the client device 110. Alternatively, the localization service may also be provided by the client application 112 running in the client device 110. As a non-limiting example, the method described below may be performed by Figure 1 The system 100 is executed.

[0031] Figure 2 2 is a flow chart illustrating a method 200 for processing track data according to an exemplary embodiment. The method 200 may be performed on a client device (e.g., Figure 1 The execution of each step of the method 200 may be performed at the client device 110 shown in FIG. Figure 1 In some embodiments, the method 200 may be performed on a server (e.g., Figure 1 In some embodiments, method 200 may be performed by a client device (eg, client device 110) and a server (eg, server 120) in combination.

[0032] According to an embodiment of the present disclosure, method 200 may be used to recover sparse track data samples with missing data points using a generative adversarial network, and the generative adversarial network may include a discriminator and a generator.

[0033] refer to Figure 2 The track data processing method 200 includes the following steps S210 to S240, wherein step S210 includes sub-steps S211 to S216.

[0034] Step S210 may include executing sub-steps S211 to S216 for each of one or more iterations.

[0035] Sub-step S211, obtaining a first track sample data set, the first track sample data set comprising real track data samples, the real track data samples comprising both real sparse track data samples with missing data points and real dense track data samples with complete data points.

[0036] In the example, the real track data samples can be obtained from target scenarios such as actual combat, exercises, simulated combat, etc. That is, the target environment can be a real environment (e.g., an actual combat / exercise combat environment), a simulation environment (e.g., a simulated combat environment), or a mixed environment (e.g., including both the actual combat / exercise combat environment and the simulated combat environment as components), etc. As an example and not a limitation, the real track data samples can be associated with a movable unit, which refers to a target (e.g., a blue target, a red target, etc.) that can move (e.g., move in a random manner, move in a formation, etc.) within the target environment.

[0037] As used in this article, the term track data sample refers to a data sample that represents the trajectory of a movable unit such as a ship or aircraft during navigation. Since the movable unit may be affected and disturbed by various factors during navigation, resulting in navigation errors, the track is generally not a straight line, but exceptions are not excluded.

[0038] As used herein, the term track sample data set refers to a data set consisting of track data samples. In an example, the first track sample data set may be a real track sample data set, which may consist of or include real track data samples.

[0039] As used herein, the term data point missing refers to the lack of one or more data points forming the navigation track of the movable unit represented by a certain track data sample, so that the track data sample visually shows undesirable features such as discontinuous sections and / or discontinuities compared to the navigation track with complete data points. The data point missing of the track data sample may be caused by terrain occlusion, data crosstalk, packet loss, etc.

[0040] In the present disclosure, a track data sample with missing data points in the real track data sample is referred to as a real sparse track data sample. In contrast, a track data sample with no missing data points (ie, complete data points) in the real track data sample is referred to as a real dense track data sample.

[0041] Sub-step S212, feeding the first track sample data set to the generative adversarial network, so that the discriminator outputs a first discriminant value for the first track sample data set, and the first discriminant value is used to calculate the function value of the discriminator objective function.

[0042] In the example, the first discriminant value of the first track sample data set output by the discriminator of the generative adversarial network can be the score of the discriminator on the authenticity of the real track data sample. In the example, from the perspective of the discriminator, the higher the authenticity of the track data sample, the higher the score, and the score value interval can be, for example, an interval [0, 1], etc. In the example, To represent the discriminator of the generative adversarial network, where θ d Represents the network parameters of the discriminator. Thus, the input-output mapping relationship of the discriminator can be established, that is, Among them, x is the input of the discriminator (for example, the track data sample fed to the generative adversarial network), and y is the discriminant value of the input x output by the discriminator, that is, the discriminator's score on the authenticity of the input x.

[0043] In an example, the first discriminant value may be used to calculate a function value of a discriminator objective function so that the function value of the discriminator objective function can be associated with the performance of the discriminator (ie, the ability to distinguish between real track data samples and non-real track data samples).

[0044] In the example, the discriminator objective function may be the optimization target for optimizing the discriminator, so that the discriminator can update the network parameters according to the network parameters corresponding to the optimal value of its objective function, thereby optimizing the discriminator. In the example, the optimal value of the discriminator objective function may be the maximum value (e.g., maximum value, minimum value) on all iterations, or may be the extreme value (e.g., maximum value, minimum value) on some iterations, etc. It can be understood that the selection of the optimal value of the objective function may vary depending on the actual optimization process and / or optimization requirements, etc., and the present disclosure does not impose any restrictions on this.

[0045] Sub-step S213, obtaining a second track sample data set, where the second track sample data set is generated by a generator by processing real sparse track data samples with missing data points.

[0046] In the example, since the real sparse track data sample lacks one or more data points that form the navigation track of the movable unit represented by the track data sample, the processing of the real sparse track data sample by the generator may include completing the missing data points in the track data sample, so that the track data sample generated by the generator by processing the track data sample may be, for example, a completed track data sample that is close to the real dense track data sample. Of course, the processing of the real sparse track data sample by the generator may also include partially completing the missing data points in the track data sample, filling all the missing data points in the track data sample with random values, filling some of the missing data points in the track data sample with random values, or a combination thereof, etc., and the present disclosure does not impose any restrictions on this.

[0047] In an example, the second track sample data set may include track data samples generated by processing all real sparse track data samples in the first track sample data set by the generator of the generative adversarial network. In another example, the second track sample data set may include track data samples generated by processing some real sparse track data samples in the first track sample data set by the generator.

[0048] In the example, you can use To represent the generator of the generative adversarial network, where θ g Represents the network parameters of the generator. Thus, the input-output mapping relationship of the generator can be established, that is, Among them, u is the input of the generator (for example, a real sparse track data sample with missing data points), and v is the generated sample output by the generator after processing the input data sample u.

[0049] Sub-step S214, feeding the second track sample data set to the generative adversarial network, so that the discriminator outputs a second discriminant value for the second track sample data set, and the second discriminant value is used to calculate the function value of the generator objective function.

[0050] In the example, the second discriminant value for the second track sample data set output by the discriminator of the generative adversarial network may be a score given by the discriminator regarding the authenticity of the non-real track data samples (e.g., close to the real track data samples). In the example, from the perspective of the discriminator, the higher the authenticity of the track data samples, the higher the score, and the score range may be, for example, the interval [0,1], etc. It can be understood that in the initial (multiple) rounds of iterations, the mathematical expectation (e.g., mean) of the second discriminant value of the discriminator for the second track data sample data set on the data set may be very low, which is attributed to the fact that the ability of the generator at the same time has not yet been improved to be sufficient to generate non-real track data samples that are increasingly close to the real track data samples. In the example, as described above, it can be used To represent the discriminator of the generative adversarial network. As a result, the second discriminant value output by the discriminator for any track data sample in the second track sample data set can be expressed as

[0051] In an example, the second discriminant value may be used to calculate a function value of a generator objective function so that the function value of the generator objective function can be associated with the performance of the generator (ie, the ability to generate non-real track data samples that are sufficiently close to real track data samples).

[0052] In the example, the generator objective function may be the optimization target for optimizing the generator, so that the generator can update the network parameters according to the network parameters corresponding to the optimal value of its objective function, thereby optimizing the generator. In the example, the optimal value of the generator objective function may be the maximum value (e.g., maximum value, minimum value) over all iterations, or may be the extreme value (e.g., maximum value, minimum value) over some iterations, etc. It can be understood that the selection of the optimal value of the objective function may vary depending on the actual optimization process and / or optimization requirements, etc., and the present disclosure does not impose any restrictions on this.

[0053] Sub-step S215, optimizes the overall objective function of the generative adversarial network to obtain the network parameters of the generator when the overall objective function takes the optimal value, wherein the overall objective function is constructed based on the discriminator objective function and the generator objective function.

[0054] As an example and not limitation, the overall objective function V of the generative adversarial network can be constructed as

[0055]

[0056] For convenience, the network parameters of the discriminator and the generator are omitted. The above formula (1) is the general form of the overall objective function of the generative adversarial network. The optimization of the generative adversarial network using this overall objective function can be divided into two optimization processes. The first step is to optimize the discriminator, that is, first consider the following formula (2):

[0057]

[0058] In the process of optimizing the discriminator, the generator G is fixed. The first part on the right side of the equal sign in equation (2) is The role of is to ensure the basic discrimination ability of the discriminator. x is the distribution of real data p data The samples obtained by sampling are The larger the value, the larger the D(x), which means that the discriminator can more accurately identify the real samples as real samples, so there is The second part on the right side of the equal sign in formula (2) The role of is to ensure that the discriminator can distinguish non-real samples. z is a value from a specific distribution p z The samples sampled in the above example are G(z), which is a non-real sample generated by the generator that is close to the real sample. The larger the value, the smaller D(G(z)) is (equivalent to the lower the score given by the discriminator on the authenticity of the non-real track data samples), that is, the more correctly the discriminator can distinguish the non-real samples, so there is In the example, stochastic gradient ascent can be performed on equation (2) to optimize the network parameters of the discriminator.

[0059] The second step is to optimize the generator, that is, consider the following formula (3)

[0060]

[0061] In the process of optimizing the generator, the discriminator D is fixed. If G is more powerful, the discriminator D will make a wrong judgment, and D(G(z)) will become larger. will be closer to zero, that is, the value of the entire formula (3) will be smaller, so we have In the example, stochastic gradient descent can be performed on equation (3) to optimize the network parameters of the generator.

[0062] The optimization of the generative adversarial network is essentially a minimax game. In theory, the final result of the optimization will make p G(z) =p data , that is, the distribution p of non-real samples generated by the generator G(z) With the real data distribution p data In this case, the discriminator's score on the authenticity of the non-genuine samples generated by the generator tends to the probability of either side facing up in a coin tossing test, that is, 50%. In summary, through one or more rounds of iterations / optimization, the discriminator and the generator compete with each other in the optimization process and gradually tend to the Nash equilibrium.

[0063] It should be noted that in the present disclosure, the term optimization, the term training or the term iteration, etc. can be used interchangeably with each other, which is obvious to those skilled in the art, and these terms have general meanings well known in the art.

[0064] In the example, in the above example overall objective function V, p data (x) may be the distribution of the real track data samples contained in the first track sample data set, p z (z) may be the distribution of non-real track data samples generated by the generator by processing the real sparse track data samples (for example, part or all) in the first track sample data set.

[0065] In the prior art, the training of the Generative Adversarial Network (GAN) starts with training the discriminator, with the goal of allowing the discriminator to initially obtain a standard. Taking the application of GAN in the field of images as an example, the so-called standard is to let the discriminator recognize a series of good images (such as real images) so as to know what a good image looks like. These images can come from various places, such as databases, from which these good images are extracted and input into the GAN discriminator. Then, after the discriminator is trained, the discriminator already has a standard, and then the generator is trained. In the prior art, a specific distribution p z is a noise distribution, such as Gaussian distribution, uniform distribution, etc. Therefore, it is sufficient to generate a set of noises for the generator at random, such as generating noises that conform to Gaussian distribution, and the generator will then generate an image through these noises. It can be understood that this image (and the images generated by the generator in the initial multiple rounds of iterations) are undesirable, and such images cannot be discriminated by the discriminator at the same time as real images with higher authenticity scores. After the generator generates an image, the discriminator will determine whether the generated image is a real image from the database or a generated image derived from the generator. If it is judged to be a real image, a higher score is given to the image, for example, the highest score can be 1, and if it is judged to be a generated image (i.e., a non-real image), a lower score is given, for example, the lowest score can be 0. Subsequently, the overall objective function of GAN is optimized according to the discriminant value output by the discriminator. The network parameters of the discriminator and the generator are obtained when the overall objective function obtains the optimal value, and the discriminator and / or the generator are updated.

[0066] In the example, additionally, the overall objective function of the generative adversarial network can be optimized to obtain the network parameters of the discriminator when the objective function takes the optimal value.

[0067] In some embodiments, instead of z (z) takes a noise distribution such as Gaussian distribution or uniform distribution, but p z(z) is consistent with the distribution of the track data samples generated by the generator by processing the real sparse track data samples with missing data points. That is, z is taken from the group consisting of the real sparse track data samples in the real track data samples contained in the first track sample data set (for example, a subset of the first track sample data set), rather than from the group consisting of random data samples that conform to a certain noise distribution. Therefore, for the multiple rounds of adversarial training process from the first round of iteration to the last round of iteration, the generator does not learn the distribution of the real track data samples from the random data samples as the starting point, but derives the generated samples for the discriminator to judge the authenticity based on the real sparse track data samples with missing data points, which effectively reduces the number of iterations required for the objective function to converge (for example, to obtain the optimal value), and thus saves the data volume required for the adversarial training / optimization of the generative adversarial network (for example, the number of track data samples in the first and / or second track sample data sets).

[0068] Sub-step S216, updating the generator according to the obtained network parameters of the generator.

[0069] In the example, in addition, while updating the generator, the discriminator can be updated according to the network parameters of the discriminator obtained when the overall objective function of the generative adversarial network takes the optimal value, so that in the next round of iterations (if any), the network parameters of the discriminator and the generator are updated.

[0070] It can be understood that, as described above with respect to sub-step S215, the adversarial training / optimization of the generative adversarial network starts with the training of the discriminator, and each iteration round is performed in an alternating manner between the discriminator and the generator. That is, in the process of optimizing the generator according to the above formula (3), the object to be minimized is V(D, G) after the discriminator is optimized before, that is, V(D, G) corresponding to the maximum or maximum value of the above formula (2), wherein, in the process of optimizing the generator, the network parameters of the discriminator corresponding to the maximum or maximum value of the formula (2) are fixed. In other words, when optimizing / training the generator, the discriminator is considered to be fixed; vice versa, for example, when optimizing / training the discriminator in the next iteration, the updated generator network parameters obtained in the previous iteration remain unchanged.

[0071] Step S220 may include determining whether an iteration stop condition is satisfied when all steps of the current iteration are executed.

[0072] In the example, the iteration stopping condition may include that the overall objective function of the generative adversarial network reaches a preset threshold (the preset threshold may be different from the theoretical maximum or extreme value to prevent slow computational convergence, etc.), the mathematical expectation of the discriminator's discrimination value for the included non-real track data samples in the second track sample data set is 0.5 (50%) (assuming that the discrimination value ranges from the interval [0,1]) or the number of iteration rounds reaches a preset threshold, etc.

[0073] Step S230 may include, in response to determining that the iteration stop condition is satisfied, using the updated generator as the final generator, so that the final generator generates completed dense track data samples with complete data points based on the sparse track data samples to be completed with missing data points.

[0074] Step S240 may include, in response to determining that the iteration stop condition is not satisfied, performing a next round of iterations based on the updated generator.

[0075] In an example, before performing the next round of iterations, the method may include obtaining the network parameters of the discriminator when the overall objective function of the generative adversarial network takes the optimal value, so that the next round of iterations can also be performed based on the updated discriminator.

[0076] At this point, it should be understood that in each round of iteration, the generator is specifically used to generate a navigation trajectory with completed data points based on sparse track data samples with missing data points. Compared with the generator in the conventional generative adversarial network, which is trained to generate samples that are close to the distribution of real data samples based on random data samples, it can effectively avoid the undesirable situation where the generator generates a track that is close to the real track in appearance but may actually have fabricated tracks (for example, it is impossible for track data points to exist for some sections, but the data points completed in the generated samples fall into such sections). This can effectively ensure the accuracy and completeness of track recovery.

[0077] In the example, the first track sample data set in each round of iteration may remain unchanged, that is, the network parameters of the discriminator in all rounds of iterations will not be updated with each round of iteration, which is especially true when the representativeness and coverage of the real track data samples included in the first track data samples are good. Of course, the first track data samples in each round of iteration may also change or include different real track data samples, and the present disclosure does not impose any restrictions on this.

[0078] In the example, the real sparse track data samples constituting the second track sample data set in each iteration may be different subsets of all the real sparse track data samples in the first track sample data set. Of course, these different subsets may have intersections with each other, which is especially true when the number of real sparse track data samples is not so redundant, and the present disclosure does not impose any limitation on this.

[0079] As an example and not a limitation, the discriminator objective function may include only the first part on the right side of the equal sign of the above formula (2), or include both the first part and the second part on the right side of the equal sign of the above formula (2). It can be understood that the example overall objective function described in sub-step S215 is essentially to minimize the JS divergence of the distribution of both the first track sample data set and the second track sample data set. There are many other ways to measure the difference (such as distance) between the distributions of two data sets, and JS divergence is just one of the measurement methods. Therefore, if different measurement methods are adopted, corresponding different objective functions can be obtained. The measurement methods may include, for example, Inception Score, Mode Score, Kernel MMD (Maximum Mean Discrepancy Maximum Mean Difference), Wasserstein distance, Fréchet Inception Distance (FID), etc. The present disclosure does not impose any restrictions on this and the discriminator objective function constructed based on this. Similarly, the present disclosure does not impose any restrictions on the generator objective function.

[0080] According to an embodiment of the present disclosure, firstly, a first track sample data set including both real sparse track data samples and real dense track data samples is obtained and fed to a generative adversarial network to obtain a score of the network's discriminator on the authenticity of the real track data samples. Then, the real sparse track data samples are processed by the network's generator to generate a second track sample data set that is close to the real dense track data samples and the second track sample data set is also fed to the generative adversarial network to obtain a score of the discriminator on the authenticity of the non-real track data samples in the second track sample data set. Then, the overall objective function of the network is optimized to improve the ability of the discriminator (i.e., better distinguish the real track data from the real track data). The ability of the generator (i.e., generating non-real track data that is closer to the real track data) and the generator (i.e., generating non-real track data that is closer to the real track data) compete with each other in the optimization process and gradually tend to the Nash equilibrium, and finally the final updated generator is obtained after all rounds of iterations are completed, so that the obtained generator can be used to generate completed dense track data samples with complete data points based on the sparse track data samples to be completed with missing data points. At this point, with the help of the method disclosed in the present invention, the incomplete track data can be restored to dense track data through the track recovery technology based on the generative adversarial network. On the one hand, compared with the baseline method, the data volume requirement is significantly reduced, and on the other hand, the accuracy and completeness of the track recovery are improved.

[0081] In some embodiments, the generator may include an encoder and a decoder. The encoder may be used to receive and encode real sparse track data samples with missing data points to obtain an encoding vector corresponding to the real sparse track data samples with missing data points. The decoder may be used to receive and decode a concatenated vector concatenated by the encoding vector and a noise vector to obtain a completed dense track data sample with complete data points corresponding to the real sparse track data samples with missing data points.

[0082] In the example, the noise distribution may adopt Gaussian distribution, uniform distribution, etc., and the present disclosure does not impose any limitation on this.

[0083] In the example, concatenating the coding vector and the noise vector may include adding (i.e., equivalent to adding additive noise to the coding vector) or multiplying (i.e., equivalent to adding multiplicative noise to the coding vector) corresponding elements of the two, or concatenating the coding vector and the noise vector end to end, etc., and the present disclosure does not impose any limitations on this.

[0084] In some embodiments, the encoder may be a time-aware long short-term memory network T-LSTM.

[0085] Considering that the intervals between adjacent time points in the obtained (e.g., observed) single real track data (e.g., real sparse track data samples with missing data points) are not fixed, an LSTM that is suitable for processing data samples with irregular time intervals is needed to encode such track data. T-LSTM has the forget gate, input gate, and output gate of the standard LSTM, but the memory unit is adjusted. Specifically, T-LSTM uses a time decay function to convert the elapsed time into a weight. That is, T-LSTM converts the time Δt elapsed between two consecutive records into a weight to reduce the influence of short-term memory. T-LSTM adjusts short-term memory according to the time span between the records of time steps t to t-1. If the gap between time t and t-1 is large, it means that the movable unit (e.g., enemy target) has no new information records for a long time. Therefore, the reliance on short-term memory should not play an important role in the prediction of the current output. This processing for short-term memory can be expressed by the following mathematical formula: in, represents short-term memory, g(Δ t ) represents the t The discount function, Represents the processed short-term memory. Therefore, as a supplement to the adversarial training / optimization inherent in the generative adversarial network itself, by configuring the generator to have a structure consisting of an encoder and a decoder, the accuracy and completeness of the generated samples derived by the generator can be further ensured, which is conducive to the rapid convergence of multiple rounds of iterations of the generative adversarial network, which in turn saves data volume requirements and computing resources.

[0086] In some embodiments, the discriminator objective function may be E y～p(y) log(D(y)), where y is the input of the discriminator, p(y) is the distribution of the real track data sample, y~p(y) means that y is taken from the distribution p(y), D(·) represents the discriminant value output by the discriminator after receiving the input, which represents the degree of authenticity of the input, and E y～p(y) represents the mathematical expectation of all track data samples in the first track sample data set. Considering that the generator does not learn the distribution of real track data samples from random data samples, but derives generated samples for the discriminator to discriminate the authenticity based on real sparse track data samples with missing data points, focusing the discriminator objective function on ensuring the basic discrimination ability of the discriminator can improve the efficiency of each round of iteration and shorten the time for the overall objective function to converge (for example, to obtain the optimal value or suboptimal value), thereby saving computing resources, which is suitable for meeting the timeliness requirements of quickly completing data points for sparse tracks in combat scenarios.

[0087] In some embodiments, the generator objective function may be E z～γ，x～q(x) log(1-D(G(z,x))), where z and x are both inputs of the generator, z~γ means z is taken from the noise distribution γ, x~q(x) means x is taken from the distribution q(x) of the real sparse track data samples with missing data points, G(·) represents the output generated by the generator after receiving and processing the input, D(·) represents the discriminant value output by the discriminator after receiving the input, and E z～γ，x～q(x) represents the mathematical expectation over all track data samples of the second track sample data set.

[0088] In some embodiments, the discriminator is pre-trained, and the track training data set used for pre-training includes: real track training data samples, distorted track training data samples, and labels corresponding to the real track training data samples and the distorted track training data samples, respectively. Thus, the discriminator's discriminative ability with respect to the real track data samples can be ensured from the beginning to prevent the adverse effects on the generator training / optimization caused by the fluctuation of the discriminative ability of the discriminator (for example, the generator continues to derive generated samples that are not close to the real dense track, because the scoring feedback of the discriminator shows that such generated samples are sufficiently real).

[0089] In an example, different true track training data samples used for pre-training the discriminator may be associated with the same true label (eg, 1), and different distorted track training data samples used for pre-training the discriminator may be associated with the same distorted label (eg, 0).

[0090] In an example, the distorted track training data sample may be obtained by partially modifying the real track data sample. In another example, the distorted track training data sample may be a data sample with randomized data point values, etc. The present disclosure does not impose any limitation on this.

[0091] In some embodiments, before feeding the first track sample data set to the generative adversarial network, the track data samples in the first track sample data set may be scaled as follows:

[0092]

[0093] Where p represents a track data sample in the first track sample data set, p new represents the track data sample obtained after scaling the track data sample p, p max represents the maximum track data sample in the first track sample data set, and p min represents the minimum track data sample in the first track sample data set. In this way, the differences between different tracks and the influence of the original track feature dimension on the model performance can be eliminated.

[0094] Figure 3 is a schematic diagram illustrating the architecture of a generative adversarial network 300 according to an exemplary embodiment.

[0095] As shown in the figure, the generative adversarial network 300 includes a generator 310 and a discriminator 320, wherein the generator 310 includes an encoder 311 and a decoder 312. Optionally, the encoder 311 includes a T-LSTM 314. The input 301 of the generative adversarial network includes both real sparse track data samples and real dense track data samples. As described above, the real sparse track data samples with missing data points (e.g., part or all) are encoded into a coding vector 302 after the encoding operation of the encoder 311, and the coding vector 302 is concatenated with the noise vector 303 taken from the noise distribution 313 to obtain a concatenated vector 304, which is then decoded by the decoder 312 into a decoded vector 305, which is a completed dense track data sample (i.e., a generated sample) with complete data points corresponding to the real sparse track data samples with missing data points. After being discriminated by the discriminator 320, the decoded vector 305 obtains the adversarial loss 306 (the adversarial loss is based on the second discriminant value as described above), and the adversarial loss 306 is then back-propagated to the generator 310, thereby updating the network parameters of the generator 310.

[0096] Please note that Figure 3 The components shown in the figure are schematic, and they are illustrated only for the purpose of facilitating the understanding of the present disclosure, and are not intended to limit the present disclosure in any way.

[0097] Figure 4 FIG. 4 is a diagram illustrating a comparison of visualization results of recovering a track using a method according to an exemplary embodiment and a baseline method.

[0098] As shown in the figure, the baseline method includes a linear prediction method, a long short-term memory network LSTM, a bidirectional LSTM, and a graph aggregation reasoning network GAIN. It can be seen that, relative to the ground truth (i.e., the real track data sample), the dense track data obtained by iteratively training / optimizing the generator of GAN using the method disclosed in the present invention to complete the track data to be completed is more consistent with the real track data in visual perception, and it can be expected that the discriminator's judgment on whether the dense track data obtained in this way is the real track data will tend to be a random 50% judgment.

[0099] Figure 5 is a schematic block diagram illustrating a track data processing apparatus 500 according to an exemplary embodiment. According to an embodiment of the present disclosure, the apparatus 500 may be used to recover sparse track data samples with missing data points using a generative adversarial network, wherein the generative adversarial network includes a discriminator and a generator. Figure 5As shown, the device 500 includes: a single-round execution unit 510, which is used to perform the following steps for each round of iteration in one or more rounds of iterations: obtaining a first track sample data set, the first track sample data set containing real track data samples, the real track data samples including both real sparse track data samples with missing data points and real dense track data samples with complete data points; feeding the first track sample data set to a generative adversarial network, so that the discriminator outputs a first discriminant value for the first track sample data set, and the first discriminant value is used to calculate the function value of the discriminator objective function; obtaining a second track sample data set, the second track sample data set is generated by the generator processing the real sparse track data samples with missing data points; feeding the second track sample data set to the generative adversarial network, so that the discriminator outputs a second discriminant value for the second track sample data set, and the second discriminant value is used to Calculate the function value of the generator objective function; optimize the overall objective function of the generative adversarial network to obtain the network parameters of the generator when the overall objective function takes the optimal value, wherein the overall objective function is constructed based on the discriminator objective function and the generator objective function; and update the generator according to the obtained network parameters of the generator; an iteration determination unit 520 is used to determine whether the iteration stop condition is met when all steps of this round of iteration are executed; a final determination unit 530 is used to respond to determining that the iteration stop condition is met, and use the updated generator as the final generator, so that the final generator generates a completed dense track data sample with complete data points based on the sparse track data sample to be completed with missing data points; and an iteration triggering unit 540 is used to respond to determining that the iteration stop condition is not met, and perform the next round of iteration based on the updated generator.

[0100] It should be understood that Figure 5 The various modules / units of the apparatus 500 shown in FIG. 5 can be used in conjunction with the reference Figure 2 The steps in the method 200 described above correspond to each other. Therefore, the operations, features and advantages described above for the method 200 are also applicable to the apparatus 500 and the modules included therein. For the sake of brevity, some operations, features and advantages are not described in detail here.

[0101] Although specific functions are discussed above with reference to specific modules, it should be noted that the functions of the various modules discussed herein may be divided into multiple modules, and / or at least some functions of multiple modules may be combined into a single module. The specific module discussed herein performing an action includes the specific module itself performing the action, or alternatively the specific module calling or otherwise accessing another component or module that performs the action (or performs the action in conjunction with the specific module). Therefore, the specific module that performs an action may include the specific module itself that performs the action and / or another module that the specific module calls or otherwise accesses to perform the action.

[0102] It should also be understood that various techniques may be described herein in the general context of software hardware elements or program modules. Figure 5 The various modules and / or units described can be implemented in hardware or in hardware in combination with software and / or firmware. For example, these modules and / or units can be implemented as computer program codes / instructions, which are configured to be executed in one or more processors and stored in a computer-readable storage medium. Alternatively, these modules and / or units can be implemented as hardware logic / circuits. One or more of the modules and / or units can be implemented together in a system on chip (System on Chip, SoC). SoC can include an integrated circuit chip (which includes a processor (e.g., a central processing unit (CPU), a microcontroller, a microprocessor, a digital signal processor (DSP), etc.), a memory, one or more communication interfaces, and / or one or more components in other circuits), and can optionally execute the received program code and / or include embedded firmware to perform functions.

[0103] According to one aspect of the present disclosure, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory. The processor is configured to execute the computer program to implement the steps of any method embodiment described above.

[0104] According to one aspect of the present disclosure, a non-transitory computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of any method embodiment described above are implemented.

[0105] According to one aspect of the present disclosure, a computer program product is provided, which includes a computer program. When the computer program is executed by a processor, the steps of any method embodiment described above are implemented.

[0106] In the following, combined Figure 6 Illustrative examples of such a computer device, non-transitory computer-readable storage medium, and computer program product are described.

[0107] Figure 6 An example configuration of a computer device 600 that can be used to implement the methods described herein is shown. For example, Figure 1 The server 120 and / or the client device 110 shown in FIG. 5 may include an architecture similar to the computer device 600. The above-mentioned apparatus 500 may also be implemented in whole or at least in part by the computer device 600 or a similar device or system.

[0108] Computer device 600 can be a variety of different types of devices. Examples of computer device 600 include, but are not limited to, desktop computers, server computers, laptop or netbook computers, mobile devices (e.g., tablet computers, cellular or other wireless phones (e.g., smart phones), notepad computers, mobile stations), wearable devices (e.g., glasses, watches), entertainment devices (e.g., entertainment appliances, set-top boxes communicatively coupled to display devices, game consoles), televisions or other display devices, automotive computers, and the like.

[0109] The computer device 600 may include at least one processor 602, memory 604, communication interface(s) 606, a display device 608, other input / output (I / O) devices 610, and one or more mass storage devices 612, all capable of communicating with one another, such as via a system bus 614 or other appropriate connections.

[0110] The processor 602 may be a single processing unit or multiple processing units, all of which may include a single or multiple computing units or multiple cores. The processor 602 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuits, and / or any device that manipulates signals based on operating instructions. Among other capabilities, the processor 602 may be configured to obtain and execute computer-readable instructions stored in the memory 604, mass storage device 612, or other computer-readable media, such as program code of an operating system 616, program code of an application program 618, program code of other programs 620, and the like.

[0111] The memory 604 and the mass storage device 612 are examples of computer-readable storage media for storing instructions, which are executed by the processor 602 to implement the various functions described above. For example, the memory 604 can generally include both volatile memory and non-volatile memory (e.g., RAM, ROM, etc.). In addition, the mass storage device 612 can generally include a hard disk drive, a solid-state drive, a removable medium, including external and removable drives, memory cards, flash memory, a floppy disk, an optical disk (e.g., CD, DVD), a storage array, a network attached storage, a storage area network, etc. The memory 604 and the mass storage device 612 can all be collectively referred to as memory or computer-readable storage media in this article, and can be a non-transitory medium capable of storing computer-readable, processor-executable program instructions as computer program code, which can be executed by the processor 602 as a specific machine configured to implement the operations and functions described in the examples of this article. Multiple programs can be stored on the mass storage device 612. These programs include an operating system 616 , one or more application programs 618 , other programs 620 , and program data 622 , and they may be loaded into memory 604 for execution.

[0112] Although in Figure 6 604 of computer device 600, but blocks 616, 618, 620, and 622, or portions thereof, may be implemented using any form of computer-readable media accessible by computer device 600. As used herein, "computer-readable media" includes at least two types of computer-readable media, namely, computer-readable storage media and communication media.

[0113] Computer-readable storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information, such as computer-readable instructions, data structures, program modules or other data. Computer-readable storage media include but are not limited to RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD), or other optical storage device, magnetic cassette, magnetic tape, magnetic disk storage device or other magnetic storage device, or any other non-transmission medium that can be used to store information for access by a computer device. In contrast, communication media can embody computer-readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transmission mechanism. Computer-readable storage media as defined herein do not include communication media.

[0114] One or more communication interfaces 606 are used to exchange data with other devices, such as through a network, a direct connection, etc. Such communication interfaces can be one or more of the following: any type of network interface (e.g., a network interface card (NIC)), a wired or wireless (such as IEEE 802.11 wireless LAN (WLAN)) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a cellular network ... TM The communication interface 606 may facilitate communication within a variety of network and protocol types, including wired networks (e.g., LAN, cable, etc.) and wireless networks (e.g., WLAN, cellular, satellite, etc.), the Internet, etc. The communication interface 606 may also provide for communication with external storage devices (not shown) such as in a storage array, network attached storage, storage area network, etc.

[0115] In some examples, a display device 608 such as a monitor may be included for displaying information and images to the user. Other I / O devices 610 may be devices that receive various inputs from the user and provide various outputs to the user, and may include a touch input device, a gesture input device, a camera, a keyboard, a remote control, a mouse, a printer, an audio input / output device, and the like.

[0116] The technology described herein can be supported by these various configurations of the computer device 600, and is not limited to the specific examples of the technology described herein. For example, the functionality can also be implemented in whole or in part on the "cloud" by using a distributed system. The cloud includes and / or represents a platform for resources. The platform abstracts the underlying functionality of the hardware (e.g., server) and software resources of the cloud. Resources can include applications and / or data that can be used when performing computing processing on a server away from the computer device 600. Resources can also include services provided over the Internet and / or through a subscriber network such as a cellular or Wi-Fi network. The platform can abstract resources and functions to connect the computer device 600 to other computer devices. Therefore, the implementation of the functions described herein can be distributed throughout the cloud. For example, the functions can be implemented partially on the computer device 600 and partially through a platform that abstracts the functions of the cloud.

[0117] Although the present disclosure has been illustrated and described in detail in the drawings and the foregoing description, such illustration and description should be considered illustrative and schematic, not restrictive; the present disclosure is not limited to the disclosed embodiments. By studying the drawings, the disclosure and the appended claims, those skilled in the art will be able to understand and implement variations to the disclosed embodiments when practicing the claimed subject matter. In the claims, the word "comprising" does not exclude other elements or steps that are not listed, the indefinite article "a" or "an" does not exclude a plurality, the term "plurality" means two or more, and the term "based on" should be interpreted as "based at least in part on". The mere fact that certain measures are recorded in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Claims

1. A method for processing track data, for recovering sparse track data samples with missing data points using a generative adversarial network, wherein the generative adversarial network includes a discriminator and a generator, and the method includes: For each of one or more iterations, perform the following steps: Acquire a first track sample data set, wherein the first track sample data set includes real track data samples, and the real track data samples include both real sparse track data samples with missing data points and real dense track data samples with complete data points; Feeding the first track sample data set to the generative adversarial network, so that the discriminator outputs a first discriminant value for the first track sample data set, wherein the first discriminant value is used to calculate a function value of a discriminator objective function; Acquire a second track sample data set, where the second track sample data set is generated by the generator by processing the real sparse track data samples with missing data points; Feeding the second track sample data set to the generative adversarial network, so that the discriminator outputs a second discriminant value for the second track sample data set, wherein the second discriminant value is used to calculate a function value of a generator objective function; Optimizing the overall objective function of the generative adversarial network to obtain the network parameters of the generator when the overall objective function takes an optimal value, wherein the overall objective function is constructed based on the discriminator objective function and the generator objective function; and Updating the generator according to the obtained network parameters of the generator; When all steps of this iteration are completed, determine whether the iteration stop condition is met; and In response to determining that the iteration stop condition is satisfied, using the updated generator as a final generator, so that the final generator generates completed dense track data samples with complete data points based on the sparse track data samples to be completed with missing data points; or In response to determining that the iteration stop condition is not satisfied, performing a next round of iterations based on the updated generator.

2. The method according to claim 1, wherein: The generator includes an encoder and a decoder, wherein the encoder is used to receive and encode the real sparse track data samples with missing data points to obtain the encoding vector corresponding to the real sparse track data samples with missing data points, and the decoder is used to receive and decode the spliced ​​vector obtained by splicing the encoding vector and the noise vector to obtain the completed dense track data samples with complete data points corresponding to the real sparse track data samples with missing data points.

3. The method according to claim 2, wherein: The encoder is a time-aware long short-term memory network T-LSTM.

4. The method according to any one of claims 1 to 3, wherein: The discriminator objective function is E y～p(y) log(D(y)) Where y is the input of the discriminator, p(y) is the distribution of the real track data sample, y~p(y) means that y is taken from the distribution p(y), D(·) represents the discriminant value output by the discriminator after receiving the input, which represents the authenticity of the input, and E y～p(y) represents the mathematical expectation over all track data samples of the first track sample data set.

5. The method according to any one of claims 1 to 3, wherein: The generator objective function is E z～γ，x～q(x) log(1-D(G(z,x))), Wherein, z and x are both inputs of the generator, z~γ means that z is taken from the noise distribution γ, x~q(x) means that x is taken from the distribution q(x) of the real sparse track data samples with missing data points, G(·) represents the output generated by the generator after receiving and processing the input, D(·) represents the discriminant value output by the discriminator after receiving the input, which represents the authenticity of the input, and E z～γ，x～q(x) represents the mathematical expectation over all track data samples of the second track sample data set.

6. The method according to any one of claims 1 to 3, wherein: The discriminator is pre-trained, and the track training data set used for pre-training includes: real track training data samples, distorted track training data samples, and labels corresponding to the real track training data samples and the distorted track training data samples respectively.

7. The method according to any one of claims 1 to 3, further comprising, before the step of feeding the first track sample dataset to the generative adversarial network: The track data samples in the first track sample data set are scaled according to the following formula: in, p represents a track data sample in the first track sample data set, p new represents the track data sample obtained after scaling the track data sample p, p max represents the maximum track data sample in the first track sample data set, and p min Represents the minimum track data sample in the first track sample data set.

8. A track data processing device, used to restore sparse track data samples with missing data points using a generative adversarial network, the generative adversarial network comprising a discriminator and a generator, the device comprising: The single-round execution unit is used to perform the following steps for each round of one or more rounds of iterations: Acquire a first track sample data set, wherein the first track sample data set includes real track data samples, and the real track data samples include both real sparse track data samples with missing data points and real dense track data samples with complete data points; Feeding the first track sample data set to the generative adversarial network, so that the discriminator outputs a first discriminant value for the first track sample data set, wherein the first discriminant value is used to calculate a function value of a discriminator objective function; Acquire a second track sample data set, where the second track sample data set is generated by the generator by processing the real sparse track data samples with missing data points; Feeding the second track sample data set to the generative adversarial network, so that the discriminator outputs a second discriminant value for the second track sample data set, wherein the second discriminant value is used to calculate a function value of a generator objective function; Optimizing the overall objective function of the generative adversarial network to obtain the network parameters of the generator when the overall objective function takes an optimal value, wherein the overall objective function is constructed based on the discriminator objective function and the generator objective function; and Updating the generator according to the obtained network parameters of the generator; An iteration determination unit, used to determine whether an iteration stop condition is satisfied when all steps of this round of iteration are executed; a final determination unit, configured to, in response to determining that the iteration stop condition is satisfied, use the updated generator as a final generator, so that the final generator generates a completed dense track data sample with complete data points based on the sparse track data sample to be completed with missing data points; and An iteration triggering unit is used to perform a next round of iteration based on the updated generator in response to determining that the iteration stop condition is not satisfied.

9. An electronic device, comprising: at least one processor; as well as a memory communicatively coupled to the at least one processor; in The memory stores a computer program, which, when executed by the at least one processor, implements the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing a computer program, wherein: The computer program implements the method according to any one of claims 1 to 7 when executed by a processor.

Images