CN113643564B

CN113643564B - Parking data restoration method and device, computer equipment and storage medium

Info

Publication number: CN113643564B
Application number: CN202110851478.5A
Authority: CN
Inventors: 彭磊; 吴伟伟
Original assignee: Shenzhen Institute of Advanced Technology of CAS
Current assignee: Shenzhen Institute of Advanced Technology of CAS
Priority date: 2021-07-27
Filing date: 2021-07-27
Publication date: 2022-08-26
Anticipated expiration: 2041-07-27
Also published as: CN113643564A

Abstract

The embodiment of the application discloses a parking data restoration method, a parking data restoration device, computer equipment and a storage medium, wherein the method comprises the following steps: determining a target parking lot, determining a target area according to the target parking lot, and acquiring a parking lot topological graph of a middle parking lot in the target area; acquiring parking space occupation information of a parking lot in the target area in a target time period; constructing a confrontation neural network according to the parking lot topological graph and the parking space occupation information; generating historical parking space occupation information of the target parking lot by adopting the antagonistic neural network; the historical parking space occupation information is used for training a parking prediction system, and the parking prediction system is used for predicting the parking space occupation information of a parking lot. By the aid of the method and the device, accuracy of parking space occupation information forecasting of the parking guidance system can be improved.

Description

Parking data restoration method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of internet technologies, and in particular, to a parking data recovery method and apparatus, a computer device, and a storage medium.

Background

With the rapid development of economy and the acceleration of modernization process in China, the holding quantity and the traveling times of automobiles are continuously increased. The continuous increase of the automobile holding amount causes road traffic jam and brings the problem of difficult parking. Intensive research and practice in the field of intelligent transportation in recent years has confirmed its superiority in alleviating traffic congestion, improving road traffic capacity and service level. Traffic prediction is an important component of intelligent traffic, and can help managers to know traffic change in advance so as to make corresponding management and control strategies.

At present, a City-wide Parking Guidance System (CPGS) is an important means for alleviating the increasingly worsened Parking problem in urban areas, and Parking space prediction is used as a key function of the System, and a large amount of historical Parking data is needed for training a prediction model of the System. However, the loss of parking data due to equipment failure or other reasons cannot be avoided, and incomplete data can affect the training precision of the prediction model. Therefore, parking data repair (generation) is essential to construct a highly reliable parking guidance system. However, for parking lot data with geographic space dependence, the influence of other related parking lot data in the geographic space on a target parking lot is not considered, so that the similarity between the data generated by the existing data repairing method and real data is not enough, and the accuracy of the parking guidance system for predicting the parking space occupation information is influenced.

Disclosure of Invention

The embodiment of the application provides a parking data restoration method and device, computer equipment and a storage medium, which can improve the accuracy of a parking guidance system for predicting parking space occupation information.

An aspect of an embodiment of the present application provides a parking data recovery method, which may include:

determining a target parking lot, determining a target area according to the target parking lot, and acquiring a parking lot topological graph of a middle parking lot in the target area;

acquiring parking space occupation information of a parking lot in the target area in a target time period;

constructing a confrontation neural network according to the parking lot topological graph and the parking space occupation information;

generating historical parking space occupation information of the target parking lot by adopting the antagonistic neural network; the historical parking space occupation information is used for training a parking prediction system, and the parking prediction system is used for predicting the parking space occupation information of a parking lot.

In a possible implementation manner, the determining a target parking lot, determining a target area according to the target parking lot, and acquiring a parking lot topology map of a middle parking lot of the target area includes:

determining a target parking lot, and determining an area with the target parking lot as a center and a preset distance threshold as a radius as a target area; the target area comprises other parking lots except the target parking lot;

counting parking lot information of the target area, and determining an adjacency matrix and an attribute matrix corresponding to a parking lot in the target area according to the parking lot information;

and generating a parking lot topological graph of a middle parking lot of the target area according to the adjacency matrix and the attribute matrix.

In a possible implementation manner, the counting parking lot information of the target area, and determining an adjacency matrix and an attribute matrix corresponding to a parking lot in the target area according to the parking lot information includes:

counting parking lot information of the target area, wherein the parking lot information comprises position information and attribute information of a parking lot;

determining the shortest path between the parking lots according to the position information, and generating an adjacent matrix corresponding to the parking lots through the shortest path;

and converting the attribute information into a standard numerical value, and generating an attribute matrix corresponding to the parking lot according to the standard numerical value.

In a possible implementation manner, the acquiring the parking space occupation information of the parking lot in the target area in the target time period includes:

counting the number of vehicles in the parking lot in a target time period according to a vehicle counting rule;

carrying out normalization processing on the number of the vehicles to generate vehicle occupancy of the parking lot, and taking the maximum value of the vehicle occupancy as the full occupancy of the parking lot;

and determining the parking space occupation information of the parking lot in the target time period according to the full occupation amount and the vehicle occupation amount at each moment in the target time period.

In a possible embodiment, the constructing a countermeasure neural network according to the parking lot topological graph and the parking space occupation information includes:

acquiring an initial confrontation neural network, wherein the initial confrontation neural network comprises an initial discriminator and an initial generator;

generating synthetic parking space occupation information of the parking lot through the parking lot topological graph and the initial generator;

taking the synthesized parking space occupation information and the parking space occupation information as training parking space occupation information;

generating a judgment result of the training parking space occupation information through the parking lot topological graph, the training parking space occupation information and the initial judger;

and generating a loss function of the initial antagonistic neural network through the Wtherstein distance based on the synthetic parking space occupation information and the discrimination result, and adjusting the initial antagonistic neural network to generate the antagonistic neural network through the loss function.

In one possible embodiment, the initial arbiter and the initial generator each include a graph convolution module, a spatio-temporal fusion module, and a gated-loop module;

generating synthetic parking space occupation information of the parking lot through the parking lot topological graph and the initial generator, wherein the synthetic parking space occupation information comprises:

inputting an adjacent matrix and an attribute matrix in the parking lot topological graph into a graph volume module of the initial generator to generate a first graph characteristic matrix;

based on a space-time fusion module of the initial generator, fusing the first graph feature matrix and initial noise to generate a first fusion matrix;

and generating synthetic parking space occupation information of the parking lot through the first fusion matrix and a gating cycle module of the initial generator.

In a possible implementation manner, the generating, by the parking lot topological graph, the training parking space occupation information, and the initial arbiter, a determination result of the training parking space occupation information includes:

inputting the adjacent matrixes and the attribute matrixes in the parking lot topological graph into a graph convolution module of the initial discriminator to generate a second graph characteristic matrix;

fusing the second graph characteristic matrix and the training parking space occupation information through a space-time fusion module of the initial discriminator to generate a second fusion matrix;

and generating a judgment result of the training parking space occupation information through the second fusion matrix and a gating cycle module of the initial discriminator.

In a possible embodiment, the generating a loss function of the initial antagonistic neural network through the wotherstein distance based on the synthetic parking space occupation information and the discrimination result, and the adjusting the initial antagonistic neural network through the loss function to generate the antagonistic neural network includes:

generating a first target function of the initial generator according to the synthesized parking space occupation information and the Wtherstein distance, generating a second target function of the initial arbiter according to the discrimination result and the Wtherstein distance,

constructing a loss function of the initial antagonistic neural network based on the first objective function and the second objective function;

adjusting model parameters of the initial antagonistic neural network according to the loss function; the model parameters of the initial antagonistic neural network comprise model parameters of the initial arbiter and model parameters of the initial generator;

and when the adjusted initial antagonistic neural network meets the convergence condition, determining the initial antagonistic neural network containing the adjusted model parameters as the antagonistic neural network.

An aspect of an embodiment of the present application provides a parking data recovery apparatus, which may include:

the topological graph acquisition unit is used for determining a target parking lot, determining a target area according to the target parking lot and acquiring a parking lot topological graph of a middle parking lot in the target area;

the information acquisition unit is used for acquiring the parking space occupation information of the parking lot in the target area in the target time period;

the network construction unit is used for constructing a confrontation neural network according to the parking lot topological graph and the parking space occupation information;

the information generating unit is used for generating historical parking space occupation information of the target parking lot by adopting the countermeasure neural network; the historical parking space occupation information is used for training a parking prediction system, and the parking prediction system is used for predicting the parking space occupation information of a parking lot.

In a possible implementation manner, the topology obtaining unit includes:

the target area determining subunit is used for determining a target parking lot, and determining an area with the target parking lot as a center and a preset distance threshold as a radius as a target area; the target area comprises other parking lots except the target parking lot;

the matrix generation subunit is used for counting the parking lot information of the target area and determining an adjacent matrix and an attribute matrix corresponding to the parking lot in the target area according to the parking lot information;

and the topological graph acquiring subunit is used for generating a parking lot topological graph of a middle parking lot of the target area according to the adjacency matrix and the attribute matrix.

In a possible implementation, the matrix generation subunit is specifically configured to:

In a possible implementation manner, the information obtaining unit is specifically configured to:

counting the number of vehicles in the parking lot in the target time period according to the vehicle counting rule;

In a possible implementation, the network construction unit includes:

the synthetic information generating subunit is used for acquiring an initial confrontation neural network, and the initial confrontation neural network comprises an initial discriminator and an initial generator; generating synthetic parking space occupation information of the parking lot through the parking lot topological graph and the initial generator;

a judgment result generating subunit, configured to use the synthesized parking space occupation information and the parking space occupation information as training parking space occupation information; generating a judgment result of the training parking space occupation information through the parking lot topological graph, the training parking space occupation information and the initial judger;

and the network construction subunit is used for generating a loss function of the initial antagonistic neural network through the Wtherstein distance based on the synthetic parking space occupation information and the judgment result, and adjusting the initial antagonistic neural network through the loss function to generate the antagonistic neural network.

the synthesis information generation subunit is specifically configured to:

based on a space-time fusion module of the initial generator, fusing the first graph characteristic matrix and initial noise to generate a first fusion matrix;

In a possible implementation manner, the determination result generating subunit is specifically configured to:

inputting the adjacency matrix and the attribute matrix in the parking lot topological graph into a graph convolution module of the initial discriminator to generate a second graph characteristic matrix;

In a possible embodiment, the network construction subunit is specifically configured to:

An aspect of an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored, the computer program being adapted to be loaded by a processor and to perform the above-mentioned method steps.

An aspect of an embodiment of the present application provides a computer device, including: a processor, a memory, and a network interface; the processor is connected with the memory and the network interface, wherein the network interface is used for providing a network communication function, the memory is used for storing program codes, and the processor is used for calling the program codes to execute the method steps.

An aspect of an embodiment of the present application provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. The computer instructions are read by a processor of the computer device from a computer-readable storage medium, and the computer instructions are executed by the processor to cause the computer device to perform the method steps described above.

In the embodiment of the application, a target parking lot is determined, a target area is determined according to the target parking lot, a parking lot topological graph of a middle parking lot of the target area is obtained, parking space occupation information of the parking lot in the target area in a target time period is further obtained, an antagonistic neural network is constructed according to the parking lot topological graph and the parking space occupation information, and finally historical parking space occupation information of the target parking lot is generated by adopting the antagonistic neural network and used for training a parking prediction system, and the parking prediction system is used for predicting the parking space occupation information of the parking lot. By adopting the method, the influence of other related parking lot data in the geographic space on the target parking lot is considered, the historical parking space occupation information of the target parking lot generated by adopting the antagonistic neural network has extremely high similarity with the real parking space occupation information, and the accuracy of the parking space occupation information predicted by the parking guidance system is improved.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

FIG. 1 is a system architecture diagram for parking data recovery according to an embodiment of the present application;

fig. 2 is a schematic flow chart of a parking data recovery method according to an embodiment of the present application;

FIG. 3 is a schematic diagram illustrating an example of determining a target area according to an embodiment of the present disclosure;

fig. 4 is an exemplary schematic diagram of a parking data recovery method provided in an embodiment of the present application;

fig. 5 is a schematic flowchart of a parking data recovery method according to an embodiment of the present application;

FIG. 6 is a schematic structural diagram of an anti-neural network provided in an embodiment of the present application;

fig. 7 is an exemplary diagram of cosine similarity comparison provided in an embodiment of the present application;

fig. 8 is a schematic structural diagram of a parking data recovery apparatus according to an embodiment of the present application;

fig. 9 is a schematic structural diagram of a computer device according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

As shown in fig. 1, the network architecture diagram may include a service server 100 and a user terminal cluster, and the user terminal cluster may include a user terminal 10a, a user terminal 10b, …, and a user terminal 10c, wherein a communication connection may exist between the user terminal cluster, for example, a communication connection exists between the user terminal 10a and the user terminal 10b, a communication connection exists between the user terminal 10b and the user terminal 10c, and any user terminal in the user terminal cluster may exist a communication connection with the service server 100, for example, a communication connection exists between the user terminal 10a and the service server 100, and a communication connection exists between the user terminal 10b and the service server 100.

The user terminal cluster (including the

user terminals

10a, 10b, and 10c) may be installed with target applications. Optionally, the target application may include an application having a function of displaying data information such as text, images, and videos. Specifically, a user terminal determines a target parking lot, determines a target area according to the target parking lot, acquires a parking lot topological graph of a middle parking lot of the target area, further acquires parking space occupation information of the parking lot in the target area in a target time period, constructs an antagonistic neural network according to the parking lot topological graph and the parking space occupation information, finally generates historical parking space occupation information of the target parking lot by adopting the antagonistic neural network, trains a parking prediction system by adopting the historical parking space occupation information, and further predicts the parking space occupation information of the parking lot according to the parking prediction system. Optionally, the ue may be any one selected from the ue cluster in the embodiment corresponding to fig. 1, for example, the ue may be the ue 10 b.

It is to be understood that the method provided in the embodiment of the present application may be executed by a computer device, where the computer device includes, but is not limited to, a terminal or a server, the service server 100 in the embodiment of the present application may be a computer device, and a user terminal in a user terminal cluster may also be a computer device, which is not limited herein. The service server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing basic cloud computing services such as cloud service, a cloud database, cloud computing, cloud functions, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN, big data and artificial intelligence platforms. The terminal may include: the smart terminal may be, but is not limited to, a smart terminal having an image recognition function, such as a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart television, a smart speaker, a desktop computer, and a smart watch. The user terminal and the service server may be directly or indirectly connected through a wired or wireless communication manner, which is not limited herein.

Further, for convenience of understanding, please refer to fig. 2, and fig. 2 is a schematic flow chart of a parking data restoration method according to an embodiment of the present application. The method may be executed by a user terminal (e.g., the user terminal shown in fig. 1) or may be executed by both the user terminal and a service server (e.g., the service server 100 in the embodiment corresponding to fig. 1). For ease of understanding, the present embodiment is described as an example in which the method is executed by the user terminal described above. The parking data restoration method at least includes the following steps S101 to S104:

s101, determining a target parking lot, determining a target area according to the target parking lot, and acquiring a parking lot topological graph of a middle parking lot in the target area;

specifically, the user terminal may determine a target parking lot, which is a parking lot in which parking space occupation information needs to be repaired, and determine a target area according to the target parking lot. The specific process of determining the target area is as follows.

The method comprises the steps that a user terminal determines a target parking lot, the target parking lot is taken as a center, a preset area with a distance threshold as a radius is determined as a target area, and the target area comprises other parking lots except the target parking lot. For convenience of understanding, please refer to fig. 3, fig. 3 is an exemplary schematic diagram of determining a target area provided in the embodiment of the present application, and as shown in fig. 3, a parking lot whose parking space occupancy information needs to be repaired is first determined as a target parking lot, an area with a preset distance threshold as a radius is determined as the target area, the distance threshold may be set to 500 meters, and there are 6 parking lots in the target area in the drawing. For example, after the target parking lot is determined, an area formed by the target parking lot and a nearby parking lot having the shortest distance of less than 500 meters may be determined as the target area.

Further, the user terminal counts the parking lot information of the target area, and determines an adjacent matrix and an attribute matrix corresponding to the parking lot in the target area according to the parking lot information. Specifically, the user terminal counts parking lot information of the target area, the parking lot information includes position information and attribute information of the parking lot, and the position information is coordinate information of the parking lot and represents an actual position of the parking lot. The attribute information is the owned attribute of the parking lot, and represents the actual characteristics of the parking lot, and the attribute information of the parking lot may include the total parking space of the parking lot, the parking rate, the main service type (public service, residential, commercial, or office) of the parking lot, and the like.

The specific process of acquiring the adjacency matrix corresponding to the parking lot is as follows: and the user terminal determines the shortest path between the parking lots according to the position information and generates an adjacent matrix corresponding to the parking lots through the shortest path. The shortest path is the distance of the parking lot in the real world. Referring to fig. 3, the target area includes 6 parking lots, which are respectively denoted as P ₁ 、P ₂ 、...P ₆ Each parking lot is regarded as a node, the shortest path between two parking lot nodes is calculated, and an adjacency matrix E of the parking lot is generated, wherein E is an n x n matrix, n is the number of the parking lots, and an element E in E _ij Indicating parking lot P _i And parking lot P _j The shortest path between them.

The specific process of obtaining the attribute matrix corresponding to the parking lot is as follows: and the user terminal converts the attribute information into a standard numerical value and generates an attribute matrix corresponding to the parking lot according to the standard numerical value. The standard value is a predetermined value range, for example, a number between 0 and 1. Each parking lot in the target area may have a plurality of attribute information, but attributes related to factors of vehicle parking are: mainly service range, capacity and price. The service range indicates which vehicles may be allowed to be parked there. For example, a parking lot in a shopping mall is open for all vehicles, while a residential parking lot serves only the owner, and a factor can be used to represent the service range of the parking lot, the larger the factor, the larger the service energyThe higher the force. The capacity corresponds the parking stall in parking area, and the parking stall is more, and service ability just is higher. The price is expressed in terms of hourly charges. Thus, the service range, capacity and price may be converted into standard values from which attribute matrices corresponding to parking lots are generated, and the attribute matrices are three-dimensional vectors, for example, attribute matrix V may be represented as V _i ＝[x _i ,y _i ,z _i ]Three dimensions in the vector correspond to service range, capacity and price, respectively.

Further, the user terminal generates a parking lot topological graph of a middle parking lot of the target area according to the adjacency matrix and the attribute matrix. Specifically, the adjacency matrix and the attribute matrix of the parking lot are E and V, respectively, and then the parking lot topology can be represented as G (V, E).

S102, acquiring parking space occupation information of a parking lot in the target area in a target time period;

specifically, the user terminal counts the number of vehicles in the parking lot in the target time period according to a vehicle counting rule, the vehicle counting rule is that counting is started from a timestamp of a first vehicle entering and exiting the parking lot, an initial value of counting is 0, one counting is subtracted from one counting, one counting is added, counting history of each time is kept, and an average value can be kept at a fixed time interval, for example, the average value is taken every 5 minutes.

Further, the vehicle quantity is normalized to generate the vehicle occupancy of the parking lot, and the maximum value of the vehicle occupancy is used as the full occupancy of the parking lot. Specifically, all vehicle data of the parking lot are traversed to obtain the historical minimum value of the vehicle data, the historical minimum value can be a positive number, zero or a negative number, and the vehicle number is normalized according to the historical minimum value to generate the vehicle occupancy of the parking lot. For example, if the minimum value of the history is a negative number, the negative number is added to each piece of vehicle data of all the count histories, and if the minimum value of the history is a positive number, the negative number is subtracted from each piece of vehicle data of all the count histories, that is, the minimum value of the history is corrected to 0.

Further, the user terminal determines the parking space occupation information of the parking lot in the target time period according to the full occupation amount and the vehicle occupation amount at each moment in the target time period. Specifically, the vehicle occupancy at each time in the target time period is divided by the full occupancy to obtain the parking space occupancy information of the parking lot in the target time period.

In order to understand the target area more directly, the topological graph and the parking space occupation information of the target area can be expressed as space-time tensors. Assuming that the target area has n parking lots, a space-time tensor (STT) of the target area is defined as follows:

STT＝<T,G(V,E)>

wherein, T is the time sequence data of the parking area, and is represented as parking space occupation information. T is a two-dimensional vector, the first dimension represents the ID of time, the second dimension represents the ID of the parking lot, i.e. T _ij The parking space occupation information of the jth parking lot at the ith time, and T _i And (4) representing the parking space occupation information of all the parking lots at the time i. G (V, E) is a parking lot topological graph of the target area, E is an adjacent matrix of the parking lot, and V is an attribute matrix of the parking lot.

S103, constructing a confronting neural network according to the parking lot topological graph and the parking space occupation information;

specifically, the user terminal obtains an initial confrontation neural network, and the initial confrontation neural network comprises an initial discriminator and an initial generator. And specifically, the user terminal inputs an adjacency matrix and an attribute matrix in the parking lot topological graph into the initial generator, and generates synthetic parking space occupation information through feature extraction and fusion, wherein the synthetic parking space occupation information is unreal parking space occupation information.

Further, the synthesized parking space occupation information and the parking space occupation information are used as training parking space occupation information, and a judgment result of the training parking space occupation information is generated through the parking lot topological graph, the training parking space occupation information and the initial judger. Specifically, the user terminal inputs the adjacency matrix and the attribute matrix in the parking lot topological graph into the initial discriminator, and generates a discrimination result of the training parking space occupation information through feature extraction and fusion, wherein the discrimination result is scalar data and is used for judging whether the training parking space occupation information is true or false and judging whether the training parking space occupation information is synthesized parking space occupation information or parking space occupation information.

Further, based on the synthetic parking space occupation information and the discrimination result, generating a loss function of the initial antagonistic neural network through the Wtherstein distance, and adjusting the initial antagonistic neural network through the loss function to generate the antagonistic neural network. The wotherstein distance is an index for measuring the similarity of two distributions, and the smaller the wotherstein distance, the higher the similarity.

S104, generating historical parking space occupation information of the target parking lot by adopting the countermeasure neural network; the historical parking space occupation information is used for training a parking prediction system, and the parking prediction system is used for predicting the parking space occupation information of a parking lot.

Specifically, the user terminal may generate historical parking space occupation information of the target parking lot by using the countermeasure neural network, so as to repair missing parking space occupation information caused by equipment failure or other reasons, and further train a parking prediction system by using the historical parking space occupation information, where the parking prediction system may predict parking space occupation information of the parking lot. Referring to fig. 4, fig. 4 is a schematic view illustrating an example of a parking data repairing method according to an embodiment of the present application, and as shown in fig. 4, a target parking lot whose parking space occupation information needs to be repaired is first determined, and then a parking lot topology map of a target area where the target parking lot is located is obtained. Further, the parking space occupation information of the parking lot in the target area in the target time period is counted, the initial confrontation neural network is trained according to the parking lot topological graph and the parking space occupation information, parameters of the network are adjusted according to a loss function of the initial confrontation neural network until the initial confrontation neural network converges, and the confrontation neural network is generated. And finally, generating historical parking space occupation information of the target parking lot by adopting the antagonistic neural network.

Referring to fig. 5 and fig. 6, fig. 5 is a schematic flow chart of a parking data restoration method according to an embodiment of the present application, and fig. 6 is a schematic structural diagram of an anti-neural network according to an embodiment of the present application. The method may be executed by a user terminal (e.g., the user terminal shown in fig. 1) or may be executed by both the user terminal and a service server (e.g., the service server 100 in the embodiment corresponding to fig. 1). For ease of understanding, the present embodiment is described as an example in which the method is executed by the user terminal described above. The parking data restoration method at least includes the following steps S201 to S206:

s201, determining a target parking lot, determining a target area according to the target parking lot, and acquiring a parking lot topological graph of a middle parking lot in the target area;

step S201 in the embodiment of the present invention refers to the detailed description of step S101 in the embodiment shown in fig. 2, which is not repeated herein.

S202, acquiring parking space occupation information of a parking lot in the target area in a target time period;

step S202 in the embodiment of the present invention refers to the detailed description of step S102 in the embodiment shown in fig. 2, which is not repeated herein.

S203, acquiring an initial confrontation neural network, and generating synthetic parking space occupation information of the parking lot through the parking lot topological graph and the initial generator; taking the synthesized parking space occupation information and the parking space occupation information as training parking space occupation information;

specifically, the initial confrontation neural network comprises an initial discriminator D and an initial generator G, and the initial discriminator and the initial generator both comprise a graph convolution module, a space-time fusion module and a gated loop module. The user terminal inputs the adjacency matrix and the attribute matrix in the parking lot topological Graph into a Graph convolution module of the initial generator to generate a first Graph characteristic matrix, wherein the Graph convolution module is a two-layer Graph Convolution Network (GCN), and a two-layer GCN model is represented as follows:

where V is the attribute matrix and E is the adjacency matrix. In order to retain self-information in the process of aggregating node characteristics, a self-loop generally needs to be added to each node. In particular, this can be achieved by adding the adjacency matrix E and the identity matrix I, i.e. by adding

Further in pair

Performing a normalization process, i.e.

Wherein

Is a matrix of the node degree, and is,

σ (-) represents an activation function, in generalRelu is used as the activation function. W ₀ And W ₁ Is a weight matrix, W ₀ ∈R ^N×H Representing a weight matrix from an input layer to a hidden layer, N being the number of parking lots, H being the number of hidden units, W ₁ ∈R ^H×T Representing a weight matrix from the hidden layer to the output layer. Finally PSN _ SPE belongs to R ^N×T The parking space information is expressed as a low-dimensional space characteristic, and the length of the parking space information is the same as that of the parking space occupation information of the target area.

Further, the user terminal fuses the first graph feature matrix and the initial noise based on a space-time fusion module of the initial generator to generate a first fusion matrix. Specifically, the initial noise is preset, for example, the initial noise may be random noise. The dimensionality of the initial noise is the same as the dimensionality of the parking space occupation information. The user terminal performs dot product fusion on the first graph feature matrix and the initial noise to generate a first fusion matrix, for example, the dimension of the first graph feature matrix is N × N, the dimension of the initial noise is T × N, where N is the number of graph nodes, T is the length of the parking space occupation information, and the dimension of the first fusion matrix is T × N.

Further, the user terminal generates synthetic parking space occupation information of the parking lot through the first fusion matrix and the gate control cycle module of the initial generator. Specifically, the first fusion matrix is input into a gating circulation module of the initial generator, time characteristics are extracted, fusion characteristics are extracted through two layers of full-connection layer networks in the gating circulation module, the T vectors are output through the full-connection layer at last, and the synthetic parking space occupation information and the parking space occupation information serve as training parking space occupation information.

S204, taking the synthesized parking space occupation information and the parking space occupation information as training parking space occupation information; generating a judgment result of the training parking space occupation information through the parking lot topological graph, the training parking space occupation information and the initial judger;

specifically, the user terminal inputs the adjacency matrix and the attribute matrix in the parking lot topological graph into the graph convolution module of the initial arbiter, and generates the second graph feature matrix, the graph convolution module is a two-layer GCN, the graph convolution module of the initial arbiter and the initial generator have the same structure, and the process of generating the second graph feature matrix is the same as the process of generating the first graph feature matrix by the initial generator.

Further, the user terminal fuses the second graph feature matrix and the training parking space occupation information through a space-time fusion module of the initial discriminator to generate a second fusion matrix. Specifically, the user terminal performs dot product fusion on the second graph feature matrix and the training parking space occupation information to generate a second fusion matrix, for example, the dimension of the second graph feature matrix is N × N, the dimension of the training parking space occupation information is T × N, where N is the number of graph nodes, T is the length of the training parking space occupation information, and the dimension of the second fusion matrix is T × N.

Further, the user terminal generates a judgment result of the training parking space occupation information through the second fusion matrix and a gating cycle module of the initial judger. Specifically, the second fusion matrix is input into a gating circulation module of the initial discriminator to extract time characteristics, fusion characteristics are extracted through two layers of full-connection layer networks in the gating circulation module, and finally, the result of judgment of the occupation information of the training parking spaces is output through the full-connection layer.

S205, based on the synthetic parking space occupation information and the judgment result, generating a loss function of the initial confrontation neural network through the Wtherstein distance, and adjusting the initial confrontation neural network through the loss function to generate the confrontation neural network.

Specifically, the user terminal generates a first objective function of the initial generator G according to the synthesized parking space occupation information and the wotherstein distance, where the first objective function is

Generating a second objective function of the initial discriminator D from the discrimination result and the Wtherstein distanceTwo objective functions of

Wtherstein distance is an index that measures the similarity of two distributions, the smaller the Wtherstein distance, the higher the similarity, where w and θ are the parameters of an initial discriminator D and an initial generator G, D _w (. is) the output of the discriminator, G _w Output of (h) generator, P _data And P _z Respectively, the distribution spaces of the parking space occupation information and the initial noise, p _{psn_spe} Is a distribution space of spatial features, where x obeys P _data Distribution space of (y) obeys p _{psn_spe} Z obeys P _z The distribution space of (2).

Further, a loss function of the initial antagonistic neural network is constructed based on the first objective function and the second objective function. The loss function being such that the first and second objective functions reach a maximum at the same time, i.e. the loss function STGAN _obj Can be expressed as follows:

adjusting model parameters of the initial antagonistic neural network according to the loss function; the model parameters of the initial antagonistic neural network comprise the model parameters of the initial arbiter and the model parameters of the initial generator; and when the adjusted initial antagonistic neural network meets the convergence condition, determining the initial antagonistic neural network containing the adjusted model parameters as the antagonistic neural network. The existing training of generators and discriminators in the GAN-based antagonistic neural network adopts JS or KL divergence loss functions, but the JS or KL divergence loss functions have the problem of gradient disappearance under the condition of no cross of data distribution, so that the problems of slow training and possible non-convergence of a model are caused. By adopting the loss function based on the Wtherstein distance in the scheme, the problem of gradient disappearance during training can be avoided, and the training speed and the convergence effect are improved.

The training of the initial antagonistic neural network includes the adjustment of hyper-parameters including the Learning Rate (LR) of the Adam optimizer, the exponential decay rate (Adam _ Beta _1) of 1st-moment estimators, and the discard rate after the final significance layers of the initial generator G and the initial arbiter D. The hyper-parameters are sensitive to specific tasks and an attempt to find the appropriate hyper-parameter setting is required. Through experimental comparison, LR is finally set to 2e-4, Adam _ Beta _1 is set to 0.8 to ensure that the initial confrontation network can stably converge, and the discarding rate is set to 30% to avoid overfitting. Better convergence can be obtained in the experiment after about 2000-3000 training rounds.

S206, generating historical parking space occupation information of the target parking lot by adopting the countermeasure neural network; the historical parking space occupation information is used for training a parking prediction system, and the parking prediction system is used for predicting the parking space occupation information of a parking lot.

Step S206 in the embodiment of the present invention refers to the detailed description of step S104 in the embodiment shown in fig. 2, which is not repeated herein.

Referring to fig. 7, fig. 7 is an exemplary schematic diagram of cosine similarity comparison provided in an embodiment of the present application, and as shown in fig. 7, a curve 1 in the diagram represents cosine similarity between historical parking space occupancy information and actual parking space occupancy information generated by using the method in the present application, and a curve 2 in the diagram represents cosine similarity between historical parking space occupancy information and actual parking space occupancy information generated by using the RCGAN method in the prior art. As can be seen from the figure, the cosine similarity between the historical parking space occupation information finally generated by the method in the scheme and the real parking space occupation information is 98%, and is obviously improved compared with 96% in the prior art.

Please refer to fig. 8, fig. 8 is a schematic structural diagram of a parking data recovery apparatus according to an embodiment of the present application. The parking data recovery means may be a computer program (comprising program code) running on a computer device, for example an application software; the apparatus may be used to perform the corresponding steps in the methods provided by the embodiments of the present application. As shown in fig. 8, the parking data restoration device 1 according to the embodiment of the present application may include: a topological diagram acquisition unit 11, an information acquisition unit 12, a network construction unit 13, and an information generation unit 14.

The topological graph acquiring unit 11 is configured to determine a target parking lot, determine a target area according to the target parking lot, and acquire a parking lot topological graph of a middle parking lot in the target area;

the information acquisition unit 12 is configured to acquire parking space occupation information of a parking lot in the target area in a target time period;

the network construction unit 13 is configured to construct a confrontation neural network according to the parking lot topological graph and the parking space occupation information;

an information generating unit 14, configured to generate historical parking space occupation information of the target parking lot by using the countermeasure neural network; the historical parking space occupation information is used for training a parking prediction system, and the parking prediction system is used for predicting the parking space occupation information of a parking lot.

Referring to fig. 8, the topology obtaining unit 11 according to the embodiment of the present application may include: a target area determining subunit 111, a matrix generating subunit 112, and a topological graph obtaining subunit 113.

A target area determining subunit 111, configured to determine a target parking lot, and determine an area with the target parking lot as a center and a preset distance threshold as a radius as a target area; the target area comprises other parking lots except the target parking lot;

a matrix generation subunit 112, configured to count parking lot information of the target area, and determine an adjacency matrix and an attribute matrix corresponding to a parking lot in the target area according to the parking lot information;

and a topological map obtaining subunit 113, configured to generate a parking lot topological map of a middle parking lot in the target area according to the adjacency matrix and the attribute matrix.

In a possible implementation, the matrix generation subunit 112 is specifically configured to:

In a possible implementation, the information obtaining unit 12 is specifically configured to:

Referring to fig. 8, the network construction unit 13 according to the embodiment of the present application may include: a composite information generating subunit 131, a discrimination result generating subunit 132, and a network constructing subunit 133.

A synthetic information generating subunit 131, configured to obtain an initial antagonistic neural network, where the initial antagonistic neural network includes an initial discriminator and an initial generator; generating synthetic parking space occupation information of the parking lot through the parking lot topological graph and the initial generator;

a determination result generation subunit 132, configured to use the synthesized parking space occupation information and the parking space occupation information as training parking space occupation information; generating a judgment result of the training parking space occupation information through the parking lot topological graph, the training parking space occupation information and the initial judger;

a network constructing subunit 133, configured to generate a loss function of the initial antagonistic neural network through the wotherstein distance based on the synthetic parking space occupancy information and the determination result, and adjust the initial antagonistic neural network through the loss function to generate an antagonistic neural network.

In one possible implementation, the initial arbiter and the initial generator each include a graph convolution module, a space-time fusion module, and a gated round robin module;

the synthesis information generation subunit 131 is specifically configured to:

In a possible implementation manner, the determination result generating subunit 132 is specifically configured to:

In a possible implementation, the network construction subunit 133 is specifically configured to:

Referring to fig. 9, fig. 9 is a schematic structural diagram of a computer device according to an embodiment of the present application. As shown in fig. 9, the computer apparatus 1000 may include: at least one processor 1001, such as a CPU, at least one network interface 1004, a user interface 1003, memory 1005, at least one communication bus 1002. Wherein a communication bus 1002 is used to enable connective communication between these components. The user interface 1003 may include a Display screen (Display), and the optional user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface). The Memory 1005 may be a Random Access Memory (RAM) or a non-volatile Memory (NVM), such as at least one disk Memory. The memory 1005 may optionally be at least one memory device located remotely from the processor 1001. As shown in fig. 9, the memory 1005, which is a kind of computer storage medium, may include therein an operating system, a network communication module, a user interface module, and a data processing application program.

In the computer apparatus 1000 shown in fig. 9, a network interface 1004 may provide a network communication function, and a user interface 1003 is mainly used as an interface for providing input for a user; the processor 1001 may be configured to call a data processing application stored in the memory 1005, so as to implement the description of the parking data recovery method in the embodiment corresponding to any one of fig. 2 to fig. 7, which is not described herein again.

It should be understood that the computer device 1000 described in this embodiment of the present application may perform the description of the parking data recovery method in the embodiment corresponding to any one of fig. 2 to fig. 7, and may also perform the description of the parking data recovery apparatus in the embodiment corresponding to fig. 8, which is not described herein again. In addition, the beneficial effects of the same method are not described in detail.

Furthermore, it is to be noted here that: an embodiment of the present application further provides a computer-readable storage medium, where a computer program executed by the aforementioned parking data recovery apparatus is stored in the computer-readable storage medium, and the computer program includes program instructions, and when the processor executes the program instructions, the description of the parking data recovery method in any one of the embodiments corresponding to fig. 2 to fig. 7 can be executed, so that details are not repeated here. In addition, the beneficial effects of the same method are not described in detail. For technical details not disclosed in embodiments of the computer-readable storage medium referred to in the present application, reference is made to the description of embodiments of the method of the present application. As an example, program instructions may be deployed to be executed on one computing device or on multiple computing devices at one site or distributed across multiple sites and interconnected by a communication network, which may constitute a block chain system.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The computer-readable storage medium may be a parking data recovery apparatus provided in any of the foregoing embodiments or an internal storage unit of the foregoing device, for example, a hard disk or a memory of an electronic device. The computer readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a Smart Memory Card (SMC), a Secure Digital (SD) card, a flash card (flash card), and the like, which are provided on the electronic device. The computer readable storage medium may further include a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (ram), or the like. Further, the computer readable storage medium may also include both an internal storage unit and an external storage device of the electronic device. The computer-readable storage medium is used for storing the computer program and other programs and quantities required by the electronic device. The computer readable storage medium may also be used to temporarily store data that has been output or is to be output.

The terms "first", "second", and the like in the claims and in the description and drawings of the present invention are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements listed, but may alternatively include other steps or elements not listed, or inherent to such process, method, article, or apparatus. Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the invention. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments. The term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit may be implemented in the form of hardware, or may also be implemented in the form of a software functional unit.

The above disclosure is only for the purpose of illustrating the preferred embodiments of the present application and is not to be construed as limiting the scope of the present application, so that the present application is not limited thereto, and all equivalent variations and modifications can be made to the present application.

Claims

1. A parking data restoration method, comprising:

according to the parking lot topological graph and the parking space occupation information, a confrontation neural network is constructed, and the confrontation neural network comprises the following steps: acquiring an initial confrontation neural network, wherein the initial confrontation neural network comprises an initial discriminator and an initial generator; generating synthetic parking space occupation information of the parking lot through the parking lot topological graph and the initial generator; taking the synthesized parking space occupation information and the parking space occupation information as training parking space occupation information; generating a judgment result of the training parking space occupation information through the parking lot topological graph, the training parking space occupation information and the initial judger; generating a loss function of the initial antagonistic neural network through the Wtherstein distance based on the synthetic parking space occupation information and the discrimination result, and adjusting the initial antagonistic neural network through the loss function to generate an antagonistic neural network;

wherein the generating a loss function of the initial confrontation neural network through the Wtherstein distance based on the synthetic parking space occupation information and the discrimination result, and adjusting the initial confrontation neural network through the loss function to generate the confrontation neural network comprises: generating a first target function of the initial generator according to the synthetic parking space occupation information and the Wtherstein distance, generating a second target function of the initial discriminator according to the discrimination result and the Wtherstein distance, constructing a loss function of the initial antagonistic neural network based on the first target function and the second target function, and adjusting model parameters of the initial antagonistic neural network according to the loss function; the model parameters of the initial antagonistic neural network comprise the model parameters of the initial discriminator and the model parameters of the initial generator, and when the adjusted initial antagonistic neural network meets the convergence condition, the initial antagonistic neural network containing the adjusted model parameters is determined as the antagonistic neural network;

generating historical parking space occupation information of the target parking lot by adopting the antagonistic neural network; the historical parking space occupation information is used for training a parking prediction system, and the parking prediction system is used for predicting the parking space occupation information of the parking lot.

2. The method of claim 1, wherein the determining a target parking lot, determining a target area according to the target parking lot, and obtaining a parking lot topology map of a middle parking lot of the target area comprises:

counting parking lot information of the target area, and determining an adjacent matrix and an attribute matrix corresponding to a parking lot in the target area according to the parking lot information;

3. The method according to claim 2, wherein the counting parking lot information of the target area, and determining an adjacency matrix and an attribute matrix corresponding to a parking lot in the target area according to the parking lot information comprises:

4. The method according to claim 1, wherein the obtaining of the parking space occupation information of the parking lot in the target area in the target time period comprises:

5. The method of claim 1, wherein the initial arbiter and the initial generator each comprise a graph convolution module, a spatio-temporal fusion module, and a gated-loop module;

6. The method of claim 1, wherein generating a determination of the training parking space occupancy information from the parking lot topology map, the training parking space occupancy information, and the initial determiner comprises:

7. A parking data restoration device, comprising:

the topological graph acquiring unit is used for determining a target parking lot, determining a target area according to the target parking lot and acquiring a parking lot topological graph of a middle parking lot in the target area;

a network construction unit for constructing a confrontation neural network according to the parking lot topological graph and the parking space occupation information, wherein the network construction unit is further configured to obtain an initial antagonistic neural network, the initial antagonistic neural network comprising an initial discriminator and an initial generator, generating synthetic parking space occupation information of the parking lot through the parking lot topological graph and the initial generator, taking the synthetic parking space occupation information and the parking space occupation information as training parking space occupation information, generating a discrimination result of the training parking space occupation information through the parking lot topological graph, the training parking space occupation information and the initial discriminator, and based on the synthetic parking space occupation information and the discrimination result, generating a loss function of the initial antagonistic neural network by the Wtherstein distance, adjusting the initial antagonistic neural network by the loss function to generate an antagonistic neural network;

the network construction unit is further configured to generate a first objective function of the initial generator according to the synthetic parking space occupation information and the wotherstein distance, generate a second objective function of the initial discriminator according to the discrimination result and the wotherstein distance, construct a loss function of the initial antagonistic neural network based on the first objective function and the second objective function, and adjust model parameters of the initial antagonistic neural network according to the loss function; the model parameters of the initial antagonistic neural network comprise the model parameters of the initial discriminator and the model parameters of the initial generator, and when the adjusted initial antagonistic neural network meets the convergence condition, the initial antagonistic neural network containing the adjusted model parameters is determined as the antagonistic neural network;

8. A computer device, comprising: a processor, memory, and a network interface;

the processor is connected to the memory and the network interface, wherein the network interface is configured to provide a network communication function, the memory is configured to store program code, and the processor is configured to call the program code to perform the method of any one of claims 1 to 6.

9. A computer-readable storage medium, in which a computer program is stored which is adapted to be loaded by a processor and to carry out the method of any one of claims 1 to 6.