CN113643532B

CN113643532B - Regional traffic prediction method and device

Info

Publication number: CN113643532B
Application number: CN202110829300.0A
Authority: CN
Inventors: 于家傲; 彭磊
Original assignee: Shenzhen Institute of Advanced Technology of CAS
Current assignee: Shenzhen Institute of Advanced Technology of CAS
Priority date: 2021-07-22
Filing date: 2021-07-22
Publication date: 2022-08-16
Anticipated expiration: 2041-07-22
Also published as: CN113643532A

Abstract

The embodiment of the application discloses a method and equipment for predicting regional traffic, wherein the method comprises the following steps: according to the road information and the parking lot information of a target area, constructing a road network topological graph of a road in the target area and a parking lot topological graph of a parking lot; acquiring first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to historical average speed information of roads in the target area and historical parking space occupation information of the parking lot; fusing the first characteristic information and the second characteristic information through a multi-channel space network, the road network topological graph and the parking lot topological graph to generate space fusion characteristics; and predicting the average speed information of each road and the parking space occupation information of each parking lot through the circulating gate control network and at least two space fusion characteristics. By the method and the device, the accuracy rate of predicting the road traffic condition can be improved.

Description

Regional traffic prediction method and device

Technical Field

The present application relates to the field of internet technologies, and in particular, to a method and an apparatus for predicting regional traffic.

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. Both road traffic and parking saturation contribute to overall traffic, which together determine the traffic conditions in an area.

The current traffic condition prediction is based on road traffic flow or parking conditions. The prediction methods of the traffic flow and the parking condition are basically communicated, and a statistical-based model is adopted for early prediction, but the method has low anti-interference capability and inaccurate prediction result. In recent years, due to the fact that a deep learning model has strong feature extraction capability and sample space fitting capability, research on traffic flow and parking space occupation prediction methods is started to be carried out based on the deep learning model, and the prediction effect is improved compared with that of an early statistical model. However, in the vicinity of some popular points of interest, such as hospitals, scenic spots, large shopping malls, etc., traffic conditions are relatively complex, and on the one hand, huge traffic flow puts pressure on parking lots. On the other hand, limited parking spaces result in low speed cruising on the road for vehicles that cannot find empty spaces. The road traffic flow and the parking condition may affect each other, and there may still be a certain deviation in predicting the traffic condition through the road traffic flow or the parking condition.

Disclosure of Invention

The embodiment of the application provides a regional traffic prediction method and device, which can improve the accuracy of prediction of road traffic conditions.

An aspect of an embodiment of the present application provides a method for predicting regional traffic, which may include:

according to the road information and the parking lot information of a target area, constructing a road network topological graph of a road in the target area and a parking lot topological graph of a parking lot;

acquiring first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to historical average speed information of roads in the target area and historical parking space occupation information of the parking lot; the first characteristic information is used for representing historical average speed information of a road in a target area, and the second characteristic information is used for representing historical parking space occupation information of a parking lot in the target area;

fusing the first characteristic information and the second characteristic information through a multi-channel space network, the road network topological graph and the parking lot topological graph to generate space fusion characteristics;

and predicting the average speed information of each road and the parking space occupation information of each parking lot through the circulating gate control network and at least two space fusion characteristics.

In a possible implementation manner, the constructing a road network topological graph of a road in a target area and a parking lot topological graph of a parking lot according to road information and parking lot information of the target area includes:

the method comprises the steps of counting road information of a target area, determining a first connection relation between each road in the road information according to a natural connection rule of the roads, and constructing a road network topological graph of the target area according to the first connection relation; the first connection relation is used for indicating whether the roads are connected in the topological graph or not;

counting parking lot information of a target area, determining a second connection relation between each parking lot in the parking lot information according to the shortest path between the parking lots, and constructing a parking lot topological graph of the target area according to the second connection relation; the second connection relation is used for indicating whether the parking lots are connected in the topological graph.

In a possible implementation manner, the acquiring first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to the historical average vehicle speed information of the roads in the target area and the historical parking space occupation information of the parking lot includes:

acquiring historical average vehicle speed information of each road in a target area at a target moment, generating an average vehicle speed vector corresponding to the road according to the historical average vehicle speed information, and taking the average vehicle speed vector as first characteristic information of a road network topological graph at the target moment;

obtaining historical parking space occupation information of each parking lot in a target area at a target moment, generating a parking space occupation vector corresponding to the parking lot according to the historical parking space occupation information, and taking the parking space occupation vector as second characteristic information of a topological graph of the parking lot at the target moment.

In a possible implementation manner, the fusing the first feature information and the second feature information through a multi-channel spatial network, the road network topology map and the parking lot topology map to generate a spatial fusion feature includes:

inputting the adjacent matrix of the road network topological graph and first characteristic information at a target moment into a first channel of a multi-channel spatial network, and acquiring a first spatial characteristic at the target moment through a graph convolution neural network in the first channel;

inputting the adjacency matrix of the parking lot topological graph and second characteristic information at the target moment into a second channel of a multi-channel space network, and acquiring a second space characteristic at the target moment through a graph convolution neural network in the second channel;

and fusing the first spatial feature and the second spatial feature to generate a spatial fusion feature at the target moment.

In a possible embodiment, the predicting the average vehicle speed information of each road and the parking space occupation information of each parking lot through the cyclic gate control network and at least two space fusion features includes:

will T ₁ -T _k Inputting the spatial fusion characteristics of the time into a cyclic gating network to generate T ₁ -T _k State information of each of the moments and prediction information of the target area; the state information is a hidden state at each moment and is used for generating prediction information, k is a positive integer greater than 1, and the cyclic gating network comprises k cyclic gating units;

and predicting the average speed information of each road and the parking space occupation information of each parking lot according to the prediction information.

In one possible embodiment, the method includes the step of comparing T ₁ -T _k Inputting the spatial fusion characteristics of the time into a cyclic gating network to generate T ₁ -T _k The state information of the time and the prediction information of the target area include:

will T ₁ Inputting the temporal spatial fusion feature into the first cyclic gate control unit of the cyclic gate control network to generate the T ₁ Status information of time of dayh ₁ ；

Will T ₂ Spatial fusion feature of time of day and the T ₁ Status information h of time of day ₁ Inputting a second cyclic gate control unit of the cyclic gate control network to generate the T ₂ Status information h of time of day ₂ ；

Will T _k Temporal spatial fusion of features and T _k-1 Status information h of time of day _k-1 Inputting a kth cyclic gate unit of a cyclic gate network, generating the T _k Status information h of time of day _k And prediction information of the target area.

In a possible embodiment, the predicting the average vehicle speed information of each road and the parking space occupation information of each parking lot according to the prediction information includes:

the prediction information comprises a first vector and a second vector; the first vector corresponds to the average speed information of each road, and the second vector corresponds to the parking space occupation information of each parking lot;

predicting the average speed information of each road according to the first vector and the corresponding relation between each dimension in the first vector and the road;

and predicting the parking space occupation information of each parking lot according to the second vector and the corresponding relation between each dimensionality in the second vector and the parking lot.

An aspect of an embodiment of the present application provides a prediction apparatus for regional traffic, which may include:

the topological graph constructing unit is used for constructing a road network topological graph of a road in a target area and a parking lot topological graph of a parking lot according to the road information and the parking lot information of the target area;

the characteristic information acquisition unit is used for acquiring first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to historical average speed information of roads in the target area and historical parking space occupation information of the parking lot; the first characteristic information is used for representing historical average speed information of a road in a target area, and the second characteristic information is used for representing historical parking space occupation information of a parking lot in the target area;

the feature fusion unit is used for fusing the first feature information and the second feature information through a multi-channel space network, the road network topological graph and the parking lot topological graph to generate space fusion features;

and the information prediction unit is used for predicting the average speed information of each road and the parking space occupation information of each parking lot through the circulating gate control network and at least two space fusion characteristics.

In a possible implementation manner, the topology graph building unit is specifically configured to:

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

historical parking space occupation information of each parking lot in the target area at the target moment is obtained, a parking space occupation vector corresponding to the parking lot is generated according to the historical parking space occupation information, and the parking space occupation vector is used as second characteristic information of the parking lot topological graph at the target moment.

In a possible embodiment, the feature fusion unit is specifically configured to:

In a possible implementation, the information prediction unit includes:

an information generation subunit for generating T ₁ -T _k Inputting the spatial fusion characteristics of the time into a cyclic gating network to generate T ₁ -T _k State information of each time in the time and prediction information of the target area; the state information is a hidden state at each moment and is used for generating prediction information, k is a positive integer greater than 1, and the cyclic gating network comprises k cyclic gating units;

and the information prediction subunit is used for predicting the average speed information of each road and the parking space occupation information of each parking lot according to the prediction information.

In a possible implementation, the information generating subunit is specifically configured to:

will T ₁ Inputting the temporal spatial fusion feature into the first cyclic gate control unit of the cyclic gate control network to generate the T ₁ Status information h of time of day ₁ ；

Will T _k Temporal spatial fusion of features and T _k-1 Status information h of time of day _k-1 Input cyclic gate control networkThe kth cyclic gate unit of (1), generating the T _k Status information h of time of day _k And prediction information of the target area.

In a possible embodiment, the information predictor 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 above-mentioned method steps.

In the embodiment of the application, a road network topological graph of a road in a target area and a parking lot topological graph of a parking lot are constructed according to road information and parking lot information of the target area, first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph are further obtained according to historical average vehicle speed information of the road in the target area and historical parking space occupation information of the parking lot, the first characteristic information and the second characteristic information are fused through a multi-channel space network, the road network topological graph and the parking lot topological graph to generate space fusion characteristics, and finally the average vehicle speed information of each road and the parking space occupation information of each parking lot are predicted through a circulating gating network and at least two space fusion characteristics. By adopting the method, the problem that the traffic condition is predicted to have deviation by adopting a single road traffic flow or a parking condition due to the mutual influence of the road traffic flow and the parking condition in a traffic condition complex section is avoided, and the accuracy of predicting the road traffic condition 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 of regional traffic prediction according to an embodiment of the present application;

fig. 2 is a schematic flowchart of a method for predicting regional traffic according to an embodiment of the present disclosure;

FIG. 3a is a schematic illustration of an example of a target area provided by an embodiment of the present application;

fig. 3b is an exemplary diagram of a road and parking lot visualization provided by an embodiment of the present application;

fig. 3c is an exemplary schematic diagram of a road network topology provided in the embodiment of the present application;

fig. 3d is an exemplary schematic diagram of a parking lot topology provided in an embodiment of the present application;

FIG. 4 is an exemplary diagram of generating a spatial fusion feature according to an embodiment of the present application;

fig. 5 is a schematic flowchart of a method for predicting regional traffic according to an embodiment of the present disclosure;

FIG. 6a is a schematic diagram illustrating an example of generating prediction information according to an embodiment of the present application;

FIG. 6b is a diagram illustrating an example of the prediction accuracy of a model provided by an embodiment of the present application;

fig. 6c is an exemplary diagram of a relationship between prediction accuracy and mutual information provided in the embodiment of the present application;

fig. 7 is a schematic structural diagram of a prediction device for regional traffic according to an embodiment of the present application;

fig. 8 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, which may include a user terminal 10a, a user terminal 10b, … and a user terminal 10c, wherein there may be a communication connection between the user terminal cluster, for example, there may be a communication connection between the user terminal 10a and the user terminal 10b, there may be a communication connection between the user terminal 10b and the user terminal 10c, and any user terminal in the user terminal cluster may have a communication connection with the service server 100, for example, there may be a communication connection between the user terminal 10a and the service server 100, and there may be a communication connection 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 functions of acquiring map data, processing road network information, constructing a topological graph, and the like. The database 10d stores a multi-channel space network and a cyclic gate control network, and specifically, the user terminal constructs a road network topological graph of a road in a target area and a parking lot topological graph of a parking lot according to road information and parking lot information of the target area, further, acquires first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to historical average vehicle speed information of the road in the target area and historical parking lot occupancy information of the parking lot, the first characteristic information is used for representing the historical average vehicle speed information of the road in the target area, the second characteristic information is used for representing the historical parking lot occupancy information of the parking lot in the target area, and the first characteristic information and the second characteristic information are fused through the multi-channel space network in the database 10d, the road network topological graph and the parking lot topological graph, and generating space fusion characteristics, and finally predicting the average speed information of each road and the parking space occupation information of each parking lot through the circulating gate control network and at least two space fusion characteristics in the database 10 d. It should be noted that the multi-channel space network and the cyclic gate control network stored in the database 10d may be stored locally in the user terminal, and the prediction of the average vehicle speed information of each road and the parking space occupation information of each parking lot may be completed on the user terminal side. Optionally, the ue may be any one ue selected from the ue cluster in the embodiment corresponding to fig. 1, for example, the ue may be the ue 10 b.

It is 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 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, a cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN, big data and artificial intelligence platform. 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 wired or wireless communication, and the application is not limited herein.

Further, for convenience of understanding, please refer to fig. 2, and fig. 2 is a schematic flowchart of a method for predicting regional traffic according to an embodiment of the present disclosure. 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 prediction method of the regional traffic at least comprises the following steps S101-S104:

s101, constructing a road network topological graph of a road in a target area and a parking lot topological graph of a parking lot according to road information and parking lot information of the target area;

specifically, the user terminal constructs a road network topological graph of a road in the target area and a parking lot topological graph of a parking lot according to road information and parking lot information of the target area, and it can be understood that the target area may be any area having a certain range and including the road and the parking lot, and may be a certain parcel of a city, for example. The road information is the number of roads of the target area and the connection relationship between the roads, and the parking lot information is the number of parking lots of the target area and the distance information between the parking lots. The road network topological graph is a topological structure graph representing road information, the road network topological graph comprises nodes representing roads and relations among the nodes, the parking lot topological graph represents a topological structure graph representing parking lot information, and the parking lot topological graph comprises nodes representing parking lots and relations among the nodes.

The specific process of constructing the road network topological graph and the parking lot topological graph is as follows:

referring to fig. 3a, fig. 3a is an exemplary schematic view of a target area provided in an embodiment of the present application, as shown in fig. 3a, a solid frame in the diagram is the target area, the target area includes a plurality of roads and a plurality of parking lots, and the roads and the parking lots in the target area are visualized according to vector data of the roads and the parking lots, as shown in fig. 3b, fig. 3b is an exemplary schematic view of the visualization of the roads and the parking lots provided in an embodiment of the present application, as shown in fig. 3b, a straight line in the diagram is the road, and a circular point is the parking lot, where the road includes direction information, that is, two roads are understood as two lanes.

The user terminal counts road information of a target area, determines a first connection relation between each road in the road information according to a natural connection rule of the road, and constructs a road network topological graph of the target area according to the first connection relation. The natural connection rule is whether roads intersect in the real world, and the first connection relation is whether roads are connected in the topological graph. For example, if the road R is ₁ And road R ₂ Is intersected in the real world, then the road R ₁ And road R ₂ Are connected in the topological graph if the road R is ₁ And road R ₂ Not intersecting in the real world, then the road R ₁ And road R ₂ Are not connected in the topology. Referring to fig. 3c, fig. 3c is an exemplary schematic diagram of a road network topology provided in the embodiment of the present application, and as shown in fig. 3c, a target area includes 11 roads, which are respectively denoted as R ₁ 、R ₂ 、...R ₁₁ Each road is regarded as a node, if two roads are intersected, the two nodes are connected, and the road network topological graph of the roads can be marked as G _r ＝(V _r ,E _r ) Wherein V is _r Representing a set of all nodes, V _r ＝{v ₁ ,v ₂ ,…,v _N N is the number of nodes, E _r Representing the set of all edges. In addition, the adjacency matrix a for the connection relationship between the nodes _r Is represented by A _r ∈R ^N×N 。

Further, the user terminal counts the parking lot information of the target area, determines a second connection relation between each parking lot in the parking lot information according to the shortest path between the parking lots, and constructs a parking lot topological graph of the target area according to the second connection relation. The shortest path is the distance of the parking lot in the real world, and a second connection relation is determined according to whether the shortest distance is smaller than a preset distance threshold value or not, wherein the second connection relation is whether the parking lot is connected in the topological graph or not. For example, if the parking lot P ₁ And parking lot P ₂ The actual distance in the real world is less than the preset distance threshold, the parking lot P ₁ And parking lot P ₂ Connected in the topological graph if the parking lot P is ₁ And parking lot P ₂ If the actual distance in the real world is greater than or equal to the preset distance threshold, the parking lot P ₁ And parking lot P ₂ Are not connected in the topological graph. Referring to fig. 3d, fig. 3d is an exemplary schematic diagram of a parking lot topology provided in the embodiment of the present application, and as shown in fig. 3c, a target area includes 7 parking lots, which are respectively denoted as P ₁ 、P ₂ 、...P ₇ And each parking lot is regarded as a node, the shortest road distance between two parking lot nodes is calculated, and if the distance is less than 600 meters, the two nodes are connected. Parking lot topological graph of parking lot can be marked as G _p ＝(V _p ,E _p ) Wherein V is _p Representing a set of all nodes, V _p ＝{v ₁ ,v ₂ ,…,v _M M is the number of parking lots, E _p Representing the set of all edges. In addition, A is used as an adjacency matrix for the connection between nodes _p Is represented by A _p ∈R ^M×M 。

S102, acquiring first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to historical average speed information of roads in the target area and historical parking space occupation information of the parking lot;

specifically, the user terminal acquires the road network topological graph according to the historical average speed information of the roads in the target area and the historical parking space occupation information of the parking lotThe method comprises the steps that first characteristic information and second characteristic information of a parking lot topological graph are obtained, the first characteristic information is used for representing historical average vehicle speed information of roads in a target area, the second characteristic information is used for representing historical parking space occupation information of the parking lot in the target area, it can be understood that a user terminal obtains the historical average vehicle speed information of each road in the target area at a target moment, the target moment is a sampling moment of the historical average vehicle speed information, an average vehicle speed vector corresponding to the road is generated according to the historical average vehicle speed information, and the average vehicle speed vector is used as first characteristic information of a road network topological graph at the target moment. For example, historical average vehicle speed information x for T target times is acquired ₁ ，x ₂ ，...x _T Then the average vehicle speed vector of each road node, i.e. the first characteristic information, is recorded as x _r ＝[x ₁ ,x ₂ ,…,x _t ,…,x _T ]T is the length of the historical time series, x _t The node characteristics at the time t are shown,

representing all the node characteristics at time t,

further, the user terminal obtains historical parking space occupation information of each parking lot in the target area at the target moment, generates parking space occupation vectors corresponding to the parking lots according to the historical parking space occupation information, and uses the parking space occupation vectors as second characteristic information of the parking lot topological graph at the target moment. For example, historical parking space occupation information x at T target moments is acquired ₁ ，x ₂ ，...x _T And then the parking space occupation vector of each parking lot node, namely the second characteristic information is recorded as x _p ＝[x ₁ ,x ₂ ,…,x _t ,…,x _T ]T is the length of the historical time series, x _t The node characteristics at the time t are shown,

all node characteristics representing time t，

S103, fusing the first characteristic information and the second characteristic information through a multi-channel space network, the road network topological graph and the parking lot topological graph to generate space fusion characteristics;

specifically, the user terminal fuses the first feature information and the second feature information through a multi-channel spatial network, the road network topological graph and the parking lot topological graph to generate a spatial fusion feature, it can be understood that the multi-channel spatial network is called as MCSN and includes a plurality of prediction channels, each prediction channel includes a two-layer graph convolution neural network (GCN), but each prediction channel is heterogeneous and can process different node topological graph data, and a two-layer GCN model is represented as follows:

where X is the feature matrix and A 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 A and the identity matrix I, i.e. by adding

Further in pair

Performing a normalization process, i.e.

Wherein

Is a matrix of degrees, and is,

W ₀ and W ₁ Is a weight matrix, σ (-) represents the activation function, and Relu () is generally used as the activation function.

In the following, a multi-channel spatial network MCSN comprising two channels is described, which can be expressed as follows:

wherein A is _r And A _p Respectively a road network topological graph and an adjacency matrix of a parking lot topological graph,

and

and respectively a road network topological graph and a parking lot topological graph at the moment t. f (-) represents a double layer GCN. FC (-) denotes a full connectivity layer.

Referring to fig. 4, fig. 4 is an exemplary schematic diagram of generating a spatial fusion feature according to an embodiment of the present application, and as shown in fig. 4, the multi-channel spatial network includes two channels, i.e., a first channel and a second channel, and each channel includes a two-layer graph convolutional neural network GCN.

And the user terminal inputs the adjacency matrix of the road network topological graph and the first characteristic information of the target moment into a first channel of the multi-channel space network, the adjacency matrix of the road network topological graph represents the connection relation between each road node, and the first channel processes the road network topological graph of N nodes. Further, a first spatial feature at the target time is obtained through a graph convolution neural network in the first channel, the first spatial feature is feature information of the road node extracted through the first channel, specifically, feature extraction is performed through convolution check of the adjacency matrix and the first feature information in the graph convolution neural network, and the first spatial feature at the target time is generated through a full connection layer.

And the user terminal inputs the adjacency matrix of the parking lot topological graph and the second characteristic information of the target time into a second channel of the multi-channel space network, the adjacency matrix of the parking lot topological graph represents the connection relation among all the parking lot nodes, and the second channel processes the road network topological graph of the M nodes. Further, a second spatial feature at the target time is obtained through a graph convolution neural network in the second channel, the second spatial feature is feature information of the parking lot node extracted through the second channel, specifically, feature extraction is performed through convolution check in the graph convolution neural network on the adjacent matrix and the second feature information, and the second spatial feature at the target time is generated through a full connection layer.

And finally, the user terminal fuses the first spatial feature and the second spatial feature through the multi-channel spatial network to generate a spatial fusion feature at the target moment, specifically, the first spatial feature and the second spatial feature are input into a splicing layer of the multi-channel spatial network to perform vector splicing, for example, if the first spatial feature is an n-dimensional vector and the second spatial feature is an m-dimensional vector, a vector with a dimension of n + m is generated through the splicing layer. And further, generating the space fusion characteristics of the spliced vectors at the target moment through a full connection layer.

It should be noted that, by using the multi-channel spatial network, the spatial fusion feature at

time

1, 2.

And S104, predicting the average speed information of each road and the parking space occupation information of each parking lot through the circulating gate control network and at least two space fusion characteristics.

In a possible embodiment, the cyclic gating network may predict the output of the current time by inputting and memorizing information of a time sequence through a gating mechanism, and the like, and specifically, the cyclic gating network may include a plurality of cyclic gating units (GRUs), and the space fusion feature may be generated by a multi-channel space network MCSN, so that the cyclic gating network is combined with the MCSN to generate an MCSTN model, the MCSTN model may predict the average vehicle speed information of each road and the parking space occupancy information of each parking lot, the MCSTN model is a multi-input multi-output prediction model, the MCSTN may include a plurality of MCSNs and the same number of GRUs as the MCSN, each MCSN corresponds to one GRU, the input data of the MCSN is the first feature information and the second feature information of the target time, the output data of the MCSN is the space fusion feature of the target time, the input data of the GRU is the space fusion feature of the target time output of the MCSN corresponding to the GRU and the output of the GRU at the previous time, the output data of the GRU is the state information of the target time, and is also used as the input data of the GRU at the next time.

Specifically, if the cyclic gate control network includes k cyclic gate control units and k MCSNs, the user terminal will transmit T ₁ -T _k Inputting the spatial fusion characteristics of the time into a cyclic gating network to generate T ₁ -T _k State information of each time in the time and prediction information of the target area; the state information is a hidden state at each moment and is used for generating prediction information, and k is a positive integer greater than 1.

And further, predicting the average speed information of each road and the parking space occupation information of each parking lot according to the prediction information. Specifically, the prediction information includes vectors corresponding to the average vehicle speed information and the parking space occupation information, and the average vehicle speed information of each road and the parking space occupation information of each parking lot are predicted according to the vectors.

Referring to fig. 5, fig. 5 is a flowchart illustrating a method for predicting regional traffic according to an embodiment of the present disclosure. 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 prediction method of the regional traffic at least comprises the following steps S201-S205:

s201, constructing a road network topological graph of a road in a target area and a parking lot topological graph of a parking lot according to road information and parking lot information of 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. 1, which is not repeated herein.

S202, acquiring first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to historical average speed information of roads in the target area and historical parking space occupation information of the parking lot;

step S202 in the embodiment of the present invention refers to the detailed description of step S102 in the embodiment shown in fig. 1, which is not repeated herein.

S203, fusing the first characteristic information and the second characteristic information through a multi-channel space network, the road network topological graph and the parking lot topological graph to generate space fusion characteristics;

step S203 in the embodiment of the present invention refers to the detailed description of step S103 in the embodiment shown in fig. 1, which is not repeated herein.

S204, adding T ₁ -T _k Inputting the spatial fusion characteristics of the time into a cyclic gating network to generate T ₁ -T _k State information of each time in the time and prediction information of the target area; the state information is a hidden state at each moment for generating a predictionAnd information, wherein k is a positive integer greater than 1.

Referring to fig. 6a, fig. 6a is a schematic diagram illustrating an example of generating prediction information according to an embodiment of the present application, as shown in fig. 6a, which is an MCSTN model generated by combining a cyclic gating network and a multi-channel spatial network MCSN, where the cyclic gating network includes k cyclic gating units (GRUs), and the MCSTN model is a multi-input multi-output prediction model.

Specifically, T is ₁ Inputting the temporal spatial fusion feature into the first cyclic gate control unit of the cyclic gate control network to generate the T ₁ Status information h of time of day ₁ ；

Wherein, T ₁ -T _k Temporal spatial fusion features from historical average vehicle speed information

And historical parking space occupation information

The MCSTN model can be taken as a whole, and the input of the model is historical average vehicle speed information

And historical parking space occupation information

And the MCSN is used for completing the multi-channel feature extraction and fusion, and the GRU is used for completing the time sequence prediction. The output of the model is

And

combining the expressions of MCSN and GRU can obtain the expression of MCSTN:

wherein h is _t-1 The hidden state at the time t-1 contains the relevant state of the previous node. r is _t The reset gate is used to control the extent to which status information at a previous time is ignored. z is a radical of _t For updating the gate, it is used to control the extent to which the state information from the previous moment is brought into the current state.

The candidate hidden state is the memory information of the current moment. h is _t The output state at time t will be passed to the next node. W _z To update the weight of the gate, W _r To reset the weight of the gate, W is the weight of the candidate hidden state.

And S205, predicting the average speed information of each road and the parking space occupation information of each parking lot according to the prediction information.

Specifically, the prediction information includes a first vector and a second vector, the first vector corresponds to the average vehicle speed information of each road, and the second vector corresponds to the parking space occupation information of each parking lot. Referring to FIG. 6a, the output of the model is

And

in order to be the first vector, the vector is,

is the second vector.

Further, according to the first vector and the corresponding relation between each dimension in the first vector and the road, the average speed information of each road is predicted, and specifically, the first vector

The vector is an N-dimensional vector and corresponds to N roads, namely, the first dimension of the first vector corresponds to the average vehicle speed information of the first road, and the Nth dimension of the first vector corresponds to the average vehicle speed information of the Nth road.

And further, predicting the parking space occupation information of each parking lot according to the second vector and the corresponding relation between each dimensionality in the second vector and the parking lot. In particular, the second vector

The vector is an M-dimension vector and corresponds to M parking lots, namely, the first dimension of the second vector corresponds to the parking space occupation information of the first parking lot, and the Mth dimension of the first vector corresponds to the parking space occupation information of the Mth parking lot.

By adopting the method, the target area traffic condition integrated prediction based on the MCSTN is realized. Whether the traffic in the target area is smooth or congested is influenced not only by the traffic flow but also by the parking saturation of the same area. The existing prediction model is single-channel compared with the MCSTN, only a single datum is concerned in the prediction process, and other traffic behaviors related to the data are ignored, such as only road traffic conditions or parking conditions. In some cases, there is a strong correlation between road traffic and parking. Especially, the traffic situation around some hot interest points, such as scenic spots, hospitals and large shopping malls, is very complicated.

In the MCSTN model, traffic conditions in a region including road traffic conditions and parking conditions are synchronously predicted, and the MCSTN has a wider visual field, so that the prediction effect on the road traffic and the parking lot conditions is better than that of the conventional model.

The method in the scheme is compared with the method in the prior art according to the actual scene to predict the average speed information and the parking space occupation information. The prior art adopts a T-GCN model for explanation, and the T-GCN model is a single-channel space-time model.

And in the comparison experiment, a plurality of parking lots in the area B of the city A and a plurality of roads around the parking lots are selected as experiment scenes for carrying out the experiment. And collecting the average speed information of each road and the parking space occupation information of each parking lot within 30 days, and specifically, selecting a target parking lot and a target road from a plurality of parking lots and roads as prediction objects.

Referring to fig. 6b, fig. 6b is a schematic diagram illustrating an example of the prediction accuracy of the model provided in the embodiment of the present application, and as shown in fig. 6b, a curve in the diagram is a change of the prediction accuracy of the model in one day, a curve 1 is the prediction accuracy of the MCSTN model, and a curve 2 is the prediction accuracy of the T-GCN model, specifically, the prediction accuracy of each 15 time slices is calculated on the test set, and a curve changing with time is obtained. The prediction Accuracy is shown by the following formula.

Wherein, Y _r As the true average vehicle speed information,

for predicted average vehicle speed information, Y _p The parking space occupation information is real parking space occupation information,

for predicted parking space occupation information | · | _F Is the F norm.

The accuracy of the two models is closer between 8 pm and 6 am, while at other times the prediction accuracy of the MCSTN model is significantly higher than that of the T-GCN model. It is possible to presume that the change in the prediction accuracy of the model is related to the traffic condition. Between 8 pm and 6 am, there are relatively few vehicles on the road and sufficient parking spaces, and the prediction accuracy of the two models is very close. However, as the number of stops increases dramatically from 8:00 a.m., the accuracy of the prediction of T-GCN decreases, and in this case the accuracy of MCSTN increases slightly. Therefore, the difference in prediction accuracy of the two models may be caused by the correlation between road traffic and parking.

In order to verify the prediction precision difference, mutual information is adopted to measure the correlation between the average speed information and the parking space occupation information on the road. According to the definition of mutual information, the larger the mutual information value is, the stronger the correlation between two variables is. Referring to fig. 6c, fig. 6c is a schematic diagram illustrating an example of a relationship between prediction accuracy and mutual information provided by the embodiment of the present application, where as shown in fig. 6c, "> is prediction accuracy of the MCSTN model," x "is prediction accuracy of the T-GCN model, and in the diagram, an abscissa is mutual information between average vehicle speed information and parking space occupancy information, and an ordinate is prediction accuracy of the model. It can be seen from the figure that when the mutual information is small, the precision of the two types of models is not very different. When mutual information is increased, the prediction precision of the MCSTN model is obviously superior to that of the T-GCN model. Therefore, the prediction of single data is only suitable for processing the situation with low relevance of traffic activities, and in the face of complex real traffic environment, the integrated prediction can better solve the problem of potential and subtle relevance between different traffic activities in the same space-time environment.

Therefore, the MCSTN model in the scheme has higher accuracy rate for predicting the road traffic and the parking conditions.

Referring to fig. 7, fig. 7 is a schematic structural diagram of a prediction device for regional traffic according to an embodiment of the present application. The prediction device of regional traffic may be a computer program (including program code) running on a computer device, for example an application software; the device can be used for executing the corresponding steps in the method provided by the embodiment of the application. As shown in fig. 7, the prediction apparatus 1 of regional traffic according to the embodiment of the present application may include: the device comprises a topological graph constructing unit 11, a feature information acquiring unit 12, a feature fusing unit 13 and an information predicting unit 14.

The topological graph constructing unit 11 is configured to construct a road network topological graph of a road in a target area and a parking lot topological graph of a parking lot according to road information and parking lot information of the target area;

a feature information obtaining unit 12, configured to obtain first feature information of the road network topological graph and second feature information of the parking lot topological graph according to historical average vehicle speed information of roads in the target area and historical parking space occupation information of the parking lot; the first characteristic information is used for representing historical average speed information of a road in a target area, and the second characteristic information is used for representing historical parking space occupation information of a parking lot in the target area;

the feature fusion unit 13 is configured to fuse the first feature information and the second feature information through a multi-channel spatial network, the road network topological graph and the parking lot topological graph to generate a spatial fusion feature;

and the information prediction unit 14 is used for predicting the average speed information of each road and the parking space occupation information of each parking lot through the circulating gate control network and at least two space fusion characteristics.

In a possible implementation manner, the topology graph building unit 11 is specifically configured to:

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

In a possible embodiment, the feature fusion unit 13 is specifically configured to:

Referring to fig. 7, the information prediction unit 14 according to the embodiment of the present application may include: an information generation subunit 141 and an information prediction subunit 142;

an information generation subunit 141 for generating T ₁ -T _k Inputting the spatial fusion characteristics of the time into a cyclic gating network to generate T ₁ -T _k State information of each time in the time and prediction information of the target area; the state information is a hidden state at each moment and is used for generating prediction information, k is a positive integer greater than 1, and the cyclic gating network comprises k cyclic gating units;

and the information prediction subunit 142 is configured to predict, according to the prediction information, average vehicle speed information of each road and parking space occupation information of each parking lot.

In a possible implementation, the information generating subunit 141 is specifically configured to:

In a possible implementation, the information prediction subunit 142 is specifically configured to:

predicting the average vehicle speed information of each road according to the first vector and the corresponding relation between each dimension in the first vector and the road;

Referring to fig. 8, fig. 8 is a schematic structural diagram of a computer device according to an embodiment of the present disclosure. As shown in fig. 8, the computer apparatus 1000 may include: at least one processor 1001, e.g., 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 further 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. 8, a 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. 8, 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 invoke a data processing application stored in the memory 1005, so as to implement the description of the method for predicting the regional traffic in the embodiment corresponding to any one of fig. 2 to 6c, 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 method for predicting the area traffic in the embodiment corresponding to any one of fig. 2 to fig. 6c, and may also perform the description of the device for predicting the area traffic in the embodiment corresponding to fig. 7, which is not described herein again. In addition, the beneficial effects of the same method are not described in detail.

Further, here, it is to be noted that: an embodiment of the present application further provides a computer-readable storage medium, where a computer program executed by the aforementioned regional traffic prediction device 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 regional traffic prediction method in any one of the embodiments corresponding to fig. 2 to 6c can be performed, 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 may be implemented by a computer program, which may be stored in a computer readable storage medium and includes the processes of the embodiments of the methods described above when the program is executed. The computer-readable storage medium may be a device for predicting regional traffic provided in any of the foregoing embodiments, or an internal storage unit of the device, such as 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, in the description and in the 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 a non-exclusive inclusion. 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 can be realized in a form of hardware, and can also be realized in a 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 method for predicting regional traffic, comprising:

acquiring first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to historical average speed information of each road at each target time in T target times and historical parking space occupation information of each parking lot at each target time in the target area; the first characteristic information is used for representing historical average vehicle speed information of each road in a target area at each target moment, and the second characteristic information is used for representing historical parking space occupation information of each parking lot in the target area at each target moment;

fusing the first characteristic information and the second characteristic information of each target moment through a multi-channel space network, the road network topological graph and the parking lot topological graph to generate a space fusion characteristic of each target moment in the T target moments;

will T ₁ -T _k Inputting the space fusion characteristics of the target moment into a cyclic gating network to generate T ₁ -T _k State information of each target time and prediction information of the target area in the target time; the state information is a hidden state of each target moment and is used for generating prediction information, k is a positive integer greater than 1,the cyclic gate network comprises k cyclic gate units, T ₁ -T _k The space fusion characteristics of the target time are generated by historical average vehicle speed information of all roads and historical parking space occupation information of all parking lots at each target time in the T target times, the prediction information comprises a first vector and a second vector, the first vector corresponds to the average vehicle speed information of each road, and the second vector corresponds to the parking space occupation information of each parking lot;

2. The method according to claim 1, wherein the constructing a road network topological graph of roads in a target area and a parking lot topological graph of parking lots according to road information and parking lot information of the target area comprises:

3. The method according to claim 1, wherein the obtaining first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to historical average vehicle speed information of each road in the target area at each target time of T target times and historical parking space occupation information of each parking lot at each target time comprises:

acquiring historical average vehicle speed information of each road in a target area at each target time in T target times, generating an average vehicle speed vector corresponding to each road at each target time according to the historical average vehicle speed information, and taking the average vehicle speed vector as first characteristic information of a road network topological graph at each target time;

obtaining historical parking space occupation information of each parking lot in a target area at each target moment, generating a parking space occupation vector corresponding to each target moment of the parking lot according to the historical parking space occupation information, and taking the parking space occupation vector as second characteristic information of a topological graph of the parking lot at each target moment.

4. The method according to claim 3, wherein the fusing the first feature information and the second feature information of each target time through a multi-channel spatial network, the road network topology map and the parking lot topology map to generate a spatial fusion feature of each target time of the T target times comprises:

inputting the adjacent matrix of the road network topological graph and the first characteristic information of each target moment into a first channel of a multi-channel space network, and acquiring the first space characteristic of each target moment through a graph convolution neural network in the first channel;

inputting the adjacency matrix of the parking lot topological graph and the second characteristic information of each target moment into a second channel of a multi-channel space network, and acquiring a second space characteristic of each target moment through a graph convolution neural network in the second channel;

and fusing the first spatial feature and the second spatial feature at each target moment to generate a spatial fusion feature at each target moment in the T target moments.

5. The method of claim 1, wherein said comparing T ₁ -T _k Inputting the space fusion characteristics of the target moment into a cyclic gating network to generate T ₁ -T _k The state information of each target time and the prediction information of the target area at the target time include:

will T ₁ Inputting the space fusion characteristics of the target moment into a first cyclic gating unit of a cyclic gating network to generate the T ₁ Status information h of target time ₁ ；

Will T ₂ Spatial fusion features of target moments and the T ₁ Status information h of target time ₁ Inputting a second cyclic gate control unit of the cyclic gate control network to generate the T ₂ Status information h of target time ₂ ；

Will T _k Spatial fusion of features and T at target time _k-1 Status information h of target time _k-1 Inputting a kth cyclic gate unit of a cyclic gate network, generating the T _k Status information h of target time _k And prediction information of the target area.

6. The method according to claim 1, wherein the predicting the average vehicle speed information of each road and the parking space occupation information of each parking lot according to the prediction information comprises:

predicting the average speed information of each road according to the first vector and the corresponding relation between each dimension in the first vector and the road, wherein one dimension in the first vector corresponds to one road;

and predicting the parking space occupation information of each parking lot according to the second vector and the corresponding relation between each dimension in the second vector and the parking lot, wherein one dimension in the second vector corresponds to one parking lot.

7. An apparatus for predicting regional traffic, comprising:

the characteristic information acquisition unit is used for acquiring first characteristic information of the road network topological graph and second characteristic information of the parking lot topological graph according to historical average speed information of each road at each target time in T target times in the target area and historical parking space occupation information of each parking lot at each target time; the first characteristic information is used for representing historical average vehicle speed information of each road in a target area at each target moment, and the second characteristic information is used for representing historical parking space occupation information of each parking lot in the target area at each target moment;

a feature fusion unit, configured to fuse the first feature information and the second feature information at each target time through a multi-channel spatial network, the road network topology map, and the parking lot topology map, and generate a spatial fusion feature at each target time in the T target times;

the information prediction unit is used for predicting the average speed information of each road and the parking space occupation information of each parking lot through the circulating gate control network and at least two space fusion characteristics;

wherein the information prediction unit includes:

an information generation subunit for generating T ₁ -T _k Inputting the space fusion characteristics of the target moment into a cyclic gating network to generate T ₁ -T _k State information of each target time and prediction information of the target area in the target time; the state information is a hidden state of each target moment and is used for generating prediction information, k is a positive integer greater than 1, the cyclic gating network comprises k cyclic gating units, and T is ₁ -T _k The space fusion characteristics of the target time are generated by historical average vehicle speed information of all roads and historical parking space occupation information of all parking lots at each target time in the T target times, the prediction information comprises a first vector and a second vector, the first vector corresponds to the average vehicle speed information of each road, and the second vector corresponds to the parking space occupation information of each parking lot;

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.