CN114973640B - Traffic flow prediction method, device and system - Google Patents

Abstract

The invention discloses a traffic flow prediction method, a traffic flow prediction device and a traffic flow prediction system. Wherein the method comprises the following steps: acquiring a track transfer matrix, historical track data and a weighted road adjacency matrix in a preset road network range, wherein the weighted road adjacency matrix is used for representing adjacency relations among roads in the preset road network range; determining a traffic conversion state according to the track transfer matrix and the historical track data, wherein the traffic conversion state is used for representing the corresponding traffic flow after each track transfer when the traffic flow is subjected to track transfer for a plurality of times; determining a neighborhood traffic state according to the weighted road adjacency matrix and the historical track data, wherein the neighborhood traffic state is used for representing traffic flow of an adjacency road section; and predicting the future traffic flow according to the traffic conversion state and the neighborhood traffic state. The traffic flow prediction method and the traffic flow prediction device solve the technical problem that in the prior art, when traffic flow is predicted, the accuracy of non-periodic traffic prediction is low.

Description

Traffic flow prediction method, device and system

Technical Field

The present invention relates to the field of intelligent transportation, and in particular to a method, device and system for predicting traffic flow.

Background technique

In the field of transportation, road traffic is divided into two parts: periodic traffic and non-periodic traffic. Periodic traffic is usually caused by periodic changes in travel demand, such as morning and evening rush hour commuting; and non-periodic traffic is caused by some unexpected events, such as subway system failures, road traffic accidents, etc. In related technologies, by learning the spatiotemporal traffic correlation from the laws existing in historical training data, the prediction of future vehicle traffic conditions is achieved, such as predicting future traffic flow, speed, density, etc. Robust and accurate traffic prediction is necessary for many transportation services such as traffic control and route planning. However, related prediction schemes can only predict periodic traffic conditions. Due to insufficient observation data of similar laws in history, the accuracy of non-periodic traffic prediction is low.

With respect to the technical problem that the prediction accuracy of non-periodic traffic is low when predicting traffic flow in the above-mentioned prior art, no effective solution has been proposed yet.

Summary of the invention

The embodiments of the present invention provide a method, device and system for predicting traffic flow, so as to at least solve the technical problem in the prior art that the prediction accuracy of non-periodic traffic is low when predicting traffic flow.

According to one aspect of an embodiment of the present invention, a method for predicting traffic flow is provided, comprising: obtaining a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network, wherein the trajectory transfer matrix is used to represent the transfer probability between an upstream road section and a downstream road section in adjacent road sections within the preset road network, and the weighted road adjacency matrix is used to represent the adjacency relationship between roads within the preset road network; determining a traffic conversion state according to the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when multiple trajectory transfers are performed; determining a neighborhood traffic state according to the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of adjacent road sections; and predicting future traffic flow according to the traffic conversion state and the neighborhood traffic state.

According to another aspect of an embodiment of the present invention, a traffic flow prediction device is also provided, including: an acquisition module, used to acquire a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network range, wherein the trajectory transfer matrix is used to represent the transfer probability between an upstream road section and a downstream road section in adjacent road sections within the preset road network range, and the weighted road adjacency matrix is used to represent the adjacency relationship between roads within the preset road network range; a traffic conversion state determination module, used to determine the traffic conversion state according to the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when the traffic flow performs multiple trajectory transfers; a neighborhood traffic state determination module, used to determine the neighborhood traffic state according to the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of the adjacent road section; a prediction module, used to predict the future traffic flow according to the traffic conversion state and the neighborhood traffic state.

According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned traffic flow prediction methods.

According to another aspect of an embodiment of the present invention, a processor is further provided, and the processor is used to run a program, wherein when the program is run, any one of the above-mentioned traffic flow prediction methods is executed.

According to another aspect of an embodiment of the present invention, a traffic flow prediction system is also provided, characterized in that it includes: a processor; and a memory connected to the processor, for providing the processor with instructions for processing the following processing steps: obtaining a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network range, wherein the trajectory transfer matrix is used to represent the transfer probability between an upstream section and a downstream section in adjacent sections within the preset road network range, and the weighted road adjacency matrix is used to represent the adjacency relationship between roads within the preset road network range; determining a traffic conversion state according to the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when multiple trajectory transfers are performed; determining a neighborhood traffic state according to the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of adjacent sections; and predicting future traffic flow according to the traffic conversion state and the neighborhood traffic state.

In an embodiment of the present invention, a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network are obtained, a traffic conversion state is determined according to the trajectory transfer matrix and the historical trajectory data, and future traffic flow is predicted according to the traffic conversion state and the neighborhood traffic state. Since the trajectory transfer matrix, the historical trajectory data and the weighted road adjacency matrix can characterize the transfer relationship of traffic flow from one road segment to another, and the influence of the traffic flow of an upstream segment on the traffic flow of a downstream segment, the present solution can predict future traffic flow through historical driving trajectories, and thus can achieve accurate prediction of sudden traffic states (for example, traffic jams caused by traffic accidents), thereby improving the accuracy of the prediction results of future vehicle traffic conditions, especially greatly improving the accuracy of the prediction results of non-periodic traffic, and solving the technical problem of low accuracy of non-periodic traffic prediction in the prior art when predicting traffic flow.

BRIEF DESCRIPTION OF THE DRAWINGS

The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

FIG1 shows a schematic diagram of a hardware structure block diagram of a computing device for implementing a traffic flow prediction method;

FIG2 is a flow chart of a method for predicting traffic flow according to an embodiment of the present invention;

FIG3a is a schematic diagram of the historical trajectory from time 0 to time t in the historical time period T;

FIG3b is a schematic diagram of adjacent roads in a preset road network according to an embodiment of the present invention;

FIG3 c provides a schematic diagram of inferring the probability distribution of the downstream road segment based on the driving trajectory of the vehicle on the upstream road segment in the preset road network;

FIG4 is a schematic diagram of a method for predicting traffic flow according to an embodiment of the present invention;

FIG5 is a schematic diagram of a traffic flow prediction device according to an embodiment of the present invention;

FIG6 is a structural block diagram of a computer terminal according to an embodiment of the present invention.

Detailed ways

In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

Example 1

According to an embodiment of the present invention, a method for predicting traffic flow is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computing device or a similar computing device. FIG1 shows a hardware structure block diagram of a computing device (or mobile device) for implementing a method for predicting traffic flow. As shown in FIG1 , a computing device 10 (or mobile device 10) may include one or more (102a, 102b, ..., 102n are used in the figure to show) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission module 106 for communication functions. In addition, it may also include: a display, an input/output interface (I/O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I/O interface), a network interface, a power supply and/or a camera. It can be understood by those skilled in the art that the structure shown in FIG1 is only for illustration and does not limit the structure of the above-mentioned electronic device. For example, the computing device 10 may also include more or fewer components than those shown in FIG1 , or have a configuration different from that shown in FIG1 .

It should be noted that the one or more processors 102 and/or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computing device 10 (or mobile device). As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

The memory 104 can be used to store software programs and modules of application software, such as the program instructions/data storage device corresponding to the traffic flow prediction method in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, to implement the vulnerability detection method of the above-mentioned application program. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computing device 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

The transmission module 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computing device 10. In one example, the transmission module 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission module 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

The display may be, for example, a touch screen liquid crystal display (LCD) that may enable a user to interact with a user interface of computing device 10 (or mobile device).

In the above operating environment, the present application provides a traffic flow prediction method as shown in Figure 2. Figure 2 is a flow chart of a traffic flow prediction method according to Embodiment 1 of the present invention. As shown in Figure 2, the method includes the following steps:

Step S201, obtaining a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network range, wherein the trajectory transfer matrix is used to represent the transfer probability between an upstream section and a downstream section in adjacent sections within the preset road network range, and the weighted road adjacency matrix is used to represent the adjacency relationship between roads within the preset road network range.

The preset road network is the road range for which traffic flow prediction is required. The preset road network range may include multiple adjacent or non-adjacent road segments, wherein adjacent roads represent road segments with common connection points, and non-adjacent road segments represent road segments without connection relationships. For example, FIG3b is a schematic diagram of adjacent roads of the preset road network. As shown in FIG3b, the preset road network range includes at least road segment i and road segment j. Road segment i and road segment j are connected by a connection point, so that a vehicle can drive from road segment i into road segment j. Then road segment i and road segment j are adjacent road segments.

Specifically, the elements in the trajectory transfer matrix are used to represent the transfer probability from the upstream section to the downstream section, and the transfer probability is used to characterize the ratio of the traffic flow from the upstream section to a certain downstream section to the traffic flow of the upstream section. For example, FIG3c provides a schematic diagram of inferring the probability distribution of a downstream section in a road network based on the driving trajectory of a vehicle on an upstream section. As shown in FIG3c, when a vehicle travels from an upstream section i to a fork in the road, there are three downstream sections ahead: downstream section k, downstream section l, and downstream section j. Based on the data of the historical driving trajectory of the vehicle from the upstream section i to the three downstream sections, the transition probability P _t,i,k from the upstream section i to the downstream section k within a time period t, the transition probability P _t,i,l from the upstream section i to the downstream section l, and the transition probability P _t,i,j from the upstream section i to the downstream section j can be calculated. Then, the above trajectory transfer matrix can include the transition probability P _t,i,k , the transition probability P _t,i,l, and the transition probability P _t,i,j . The transition probability P _t,i,k , the transition probability P _t,i,l and the transition probability P _t,i,j can macroscopically characterize the transfer ratio of the traffic flow from the upstream segment i to the downstream segments j, k and l.

In an optional embodiment, the trajectory transfer matrix P in the historical time period T can be:

P∈R ^T×|V|×|V| ;

Furthermore, the historical time T is divided into t time periods at fixed time intervals, and the trajectory transfer matrix _Pt at the tth moment (i.e., the collection time of the historical trajectory corresponding to the tth time period) is:

P _t ∈R ^|V|×|V| ;

Wherein, V＝{ _vi } _{i＝1，2，...M} , M is the number of road segments in the preset road network, _vi represents a road segment in the preset road network, and each element in the trajectory transfer matrix _Pt represents the probability of a vehicle transferring from road segment i to road segment j in the preset road network.

The above historical trajectory data is the estimated data of the vehicle traveling on the road segments in the preset road network during a period of history, which may include the traffic flow in each road segment, and may also include the trajectory of the vehicle traveling on different road segments. The historical trajectory data can be obtained through the driving trajectory data of different vehicles traveling to the same road segment, and the driving trajectory data of the same vehicle traveling to different road segments in the preset road network. In an optional embodiment, during a period of history, the trajectory data of the car can be recorded once every period of time. FIG3a is a schematic diagram of the historical trajectory from the 0th moment to the tth moment obtained at a fixed time interval in the historical time period T. As shown in FIG3a, the historical trajectory data graphs a, b, c and d are respectively the driving trajectories in the preset road network obtained by recording the same vehicle at fixed time intervals at different moments in the historical time period T, and the historical trajectory data graphs a-d correspond to different combinations of upstream road segments and downstream road segments. Specifically, the historical trajectory data graph 3a may be a historical trajectory data graph recorded at the first moment, the historical trajectory data graph 3b may be a historical trajectory data graph recorded at the second moment, and the historical trajectory data graph 3c may be a historical trajectory data graph recorded at the third moment. The above-mentioned fixed time intervals may be set according to sampling requirements. For example, the fixed time interval may be 15 minutes, that is, the time interval between the above-mentioned first moment, the second moment and the third moment is 15 minutes.

In an optional embodiment, the historical trajectory data X can be represented in the form of a matrix:

X∈R ^T×|V|

Where T is the total historical time of the statistical historical trajectory data, and each element in the above matrix X represents the traffic flow of a certain road section in a certain time period. Further, the historical time T is divided into t time periods at fixed time intervals, and the historical trajectory data _Xt at the tth moment is:

_Xt∈R ^|V| ;

Among them, t=1, 2,…T.

The weighted road adjacency matrix represents the adjacency relationship of roads in the form of a matrix. The composition of the weighted road adjacency matrix is related to the structure of the road segments in the preset road network. Specifically, the weighted road adjacency matrix is related to whether every two road segments in the preset road network are adjacent and the distance between the midpoints of the two road segments. In an optional embodiment, 1 can be used in the matrix to represent the adjacency between two road segments, and 0 can be used to represent the non-adjacent two road segments. The weighted road adjacency matrix A can be represented by the following matrix:

A∈[0，1] ^M×M ;

Among them, the weighted road adjacency matrix A is an M×M matrix, M is the number of road segments in the preset road network, if road segment i and road segment j in the preset road network have an adjacency relationship, then the i-th row and j-th column of the matrix A are recorded as 1, if road segment i and road segment j in the preset road network are not adjacent, then the i-th row and j-th column of the matrix A are recorded as 0, i=1, 2, 3...M, j=1, 2, 3...M.

It should be noted that the trajectory transfer matrix and historical trajectory data are both related to historical time, while the weighted road adjacency matrix is only related to the road relationship and length in the preset road network, and has nothing to do with historical time.

Step S202, determining a traffic conversion state according to the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when multiple trajectory transfers are performed.

The traffic conversion state is related to the historical trajectory data within the above historical time T and the above transfer probability. Each trajectory transfer can be the movement of a vehicle from one road segment to another road segment in the preset road network. For example, the historical trajectory data X _t within the historical time T with a fixed time interval has been obtained. At the tth moment, the vehicle travels from road segment i to road segment j, which is equivalent to a trajectory transfer from road segment i to road segment j. The traffic conversion state can represent the traffic flow of road segment j as the vehicle's destination after completing the trajectory transfer.

In an optional embodiment, the traffic conversion state is represented in the form of a matrix. Specifically, FIG4 is a schematic diagram of a traffic flow prediction method according to an embodiment of the present application. As shown in FIG4, based on the trajectory transfer matrix P and the matrix X of historical trajectory data, the traffic conversion state matrix Dt at the t-th moment in the historical time period T is obtained through a graph convolutional neural network model (i.e., a GCNS model). Each element in the traffic conversion state matrix can represent the traffic flow of the road segment corresponding to the destination after each trajectory transfer. As shown in FIG4, in the time period corresponding to the t-th moment, the vehicle may have multiple trajectory transfers (for example, driving from road segment i to road segment j, and then to road segment k, etc.), and the vehicle's trajectory can be considered to be expanding outward on the basis of road segment i, and the traffic conversion state Dt can represent the impact of multiple trajectory transfers on the traffic flow of all road segments expanding outward from road segment i.

Step S203, determining the neighborhood traffic state according to the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of the adjacent road segment.

The traffic flow of the adjacent road segments may be the total traffic flow of the adjacent road segments. For example, if road segment i to road segment j are adjacent road segments, the traffic flow of the adjacent road segments includes the traffic flow of road segment i and the traffic flow of road segment j.

The traffic state of the neighborhood can be characterized by the speed of vehicle transfer in the adjacent road section through the traffic flow of the adjacent road section at the tth moment, and then the speed state of the traffic flow in the adjacent road section area. For example, if the traffic flow of the adjacent road section is large, it means that the traffic speed of the adjacent road section is slow; conversely, if the traffic flow of the adjacent road section is small, it means that the traffic speed of the adjacent road section is fast.

In an optional embodiment, the neighborhood traffic state can be represented in the form of a matrix. Specifically, as shown in FIG4 , based on the weighted road adjacency matrix A and the matrix X of historical trajectory data in each time period, the neighborhood traffic state matrix _St is obtained through a graph convolutional neural network model (i.e., GCNS model). Each element in the neighborhood traffic state matrix can represent the traffic flow of each adjacent road section in each time period.

Step S204, predicting future traffic flow according to the traffic conversion status and the neighborhood traffic status.

After obtaining the traffic conversion status and neighborhood traffic status in t time periods under historical time T, the traffic flow in the future t+1 time period can be predicted. The future traffic flow includes periodic traffic flow and non-periodic traffic flow.

In an optional embodiment, Figure 4 is a schematic diagram of a traffic flow prediction method according to an embodiment of the present application. As shown in Figure 4, a trajectory transfer matrix P, historical trajectory data X and a weighted road adjacency matrix A within a historical time T are obtained, and the historical time T is divided into t time periods (including 0-t time periods) according to a fixed time interval, and the trajectory transfer matrix _Pt and historical trajectory data _Xt within each time period are extracted, and multiple trajectory transfer matrices Pt and multiple historical trajectory data _Xt are input into a graph convolutional neural network model 41 to obtain a traffic conversion state Dt corresponding to each time period, and the weighted road adjacency matrix A and multiple historical trajectory data _Xt are input into a graph convolutional neural network model 42 to obtain a neighborhood traffic state _St corresponding to each time period, and the predicted traffic flow within 0-t time periods is obtained based on the traffic conversion state Dt and the neighborhood traffic state _St , and further based on the predicted traffic flow within 0-t time periods, the traffic flow within the future t+1 time period is predicted. The traffic flow prediction method based on the embodiment of the present application can reduce the prediction error of the traffic flow prediction results by more than 5% compared with the related technology. In particular, for the prediction results of non-periodic traffic flow, the prediction error can be reduced by 14%, thereby improving the accuracy of traffic flow prediction.

In this embodiment, a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network are obtained, and a traffic conversion state is determined according to the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when multiple trajectory transfers are performed, and a neighborhood traffic state is determined according to the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of adjacent road sections, and future traffic flow is predicted according to the traffic conversion state and the neighborhood traffic state. The trajectory transfer matrix, historical trajectory data and weighted road adjacency matrix can characterize the transfer relationship of traffic flow from one road segment to another, as well as the impact of traffic flow in an upstream segment on traffic flow in a downstream segment. The related technology only obtains historical traffic flow data of each segment, and can only predict periodic traffic flow based on historical traffic flow data. The present solution predicts future traffic flow based on the vehicle's driving trajectory, and achieves accurate prediction of sudden traffic conditions (for example, traffic jams caused by traffic accidents), thereby improving the accuracy of the prediction results of future vehicle traffic conditions, especially greatly improving the accuracy of the prediction results of non-periodic traffic, and solving the technical problem of low accuracy in predicting non-periodic traffic in the prior art when predicting traffic flow.

As an optional embodiment, obtaining a trajectory transfer matrix within a preset road network range includes: extracting the number of vehicles that transfer from the upstream section to each downstream section within a preset time interval from historical trajectory data; determining the transfer probability of transferring from the upstream section to each downstream section based on the number; and determining the trajectory transfer matrix based on the transfer probability between multiple sections within the preset road network range.

The preset time interval is the length of each time period with a fixed length into which the historical time period T is divided. The preset time interval can be determined according to the interval requirements for the observation and sampling of historical trajectory data. For example, if the historical trajectory data is the vehicle trajectory data in the past day, the preset time interval can be set to 15 minutes, that is, the historical trajectory data is extracted every 15 minutes.

Specifically, the historical trajectory data may include the trajectory information of the vehicle on each road segment within the historical time period T. The historical time period T is divided into t time periods according to the preset time interval, and the tth moment is the data extraction time of the tth time period. At the tth moment, the trajectory P(τ|t) of a vehicle in the preset road network can be expressed as:

Wherein, v ₁ , v ₂ , …, v _I represent the road segments traveled by the vehicle at the tth moment.

Based on the driving trajectory of each vehicle, the number of vehicles turning from a certain upstream section to each downstream section adjacent to the upstream section at the tth moment can be obtained. According to the relationship between the number of vehicles turning into each downstream section and the total number of vehicles in the upstream section, the transfer probability from the upstream section to each downstream section can be calculated.

For example, in a preset road network, an upstream road segment i has three downstream road segments (downstream road segment k, downstream road segment l and downstream road segment j). In the tth time period, the trajectory of vehicle A includes traveling from road segment _vi to road segment v _j ), the trajectory of vehicle B includes traveling from road segment _vi to road segment v _j , and the trajectory of vehicle C includes traveling from road segment _vi to road segment v _k . In the tth time period, the number of vehicles turning from the upstream road segment _vi to the downstream road segment v _j is 2, the number of vehicles turning from the upstream road segment _vi to the downstream road segment v _k is 1, and the number of vehicles turning _from the upstream road segment _vi to the downstream road segment v ₁ is 0. Then, the transition probability P _t,i,k from the upstream road segment _vi to the downstream road segment v _k is 1/3, the transition probability P _t,i,l from the upstream road segment _vi to the downstream road segment v _l is 0, and the transition probability P _{t,i, j} from the upstream road segment vi to the downstream road segment v j is 2/3. The obtained P _t,i,k , P _t,i,j , P _t,i,l are the elements constituting the trajectory transfer matrix P _t .

As an optional embodiment, the transfer probability from the upstream section to each downstream section is determined based on the quantity, including: amplifying the number of vehicles that transfer from the upstream section to each downstream section within a preset time interval to densify the trajectory transfer matrix; determining the ratio of the amplified number of vehicles that transfer from the upstream section to each downstream section within the preset time interval to the number of all vehicles in the upstream section within the preset time interval as the transfer probability from the upstream section to each downstream section.

In the constructed trajectory transfer matrix P _t ∈R ^|V|×|V| , if there is no vehicle entering the downstream section corresponding to a certain upstream section, the corresponding transfer probability is 0. If there are many 0s in the trajectory transfer matrix, the trajectory transfer matrix will be too sparse. By increasing the number of upstream sections transferring to each downstream section, the existence of 0 elements in the trajectory transfer matrix can be avoided.

In an optional embodiment, in the transfer probability calculation formula, the number of vehicles entering the downstream section can be added by 1 to complete the amplification of the number of vehicles entering each downstream section from the upstream section. The transfer probability calculation formula for driving from the upstream section vi to the downstream section vj within the time period t can be:

Among them, #veh icles( _vi → _vj |t) represents the number of vehicles traveling from the upstream segment _vi to the downstream segment _vj at the tth time, #veh icles( _vi |t) represents the total number of vehicles on the upstream segment _vi , and N( _vi ) represents the set of all downstream segments corresponding to the upstream segment _vi .

As an optional embodiment, determining the traffic conversion state according to the trajectory transfer matrix and the historical trajectory data includes: obtaining a first graph propagation parameter, wherein the first graph propagation parameter is used to characterize the maximum number of trajectory transfers when the trajectory transfer matrix and the historical trajectory data are propagated through graph; inputting the trajectory transfer matrix and the historical trajectory data into a first graph neural network, performing graph propagation according to the first graph propagation parameter, and obtaining the traffic conversion state.

The first graph neural network may be a graph convolutional neural network model, the first graph propagation parameter may be a conversion state hop count (demand hop), and the first graph propagation parameter may be used to control the farthest possible destination of a vehicle in a preset road network during trajectory transfer. The first graph propagation parameter may be determined based on historical data, for example, the average trajectory transfer times or the maximum trajectory transfer times of vehicles within 15 minutes in historical data may be determined as the first graph propagation parameter.

In an optional embodiment, the following simplified graph convolutional neural network model can be used as the first graph neural network:

Graphprop(X,A;K) := [X||AX||A ² X||...||A ^K X||];

Among them, K is the number of hops in the graph convolutional neural network model, A is the weighted road adjacency matrix, and X is the historical trajectory data.

In an optional embodiment, as shown in FIG4 , the trajectory transfer matrix P of the historical time T includes t trajectory transfer matrices at t moments (the t-th moment corresponds to the trajectory transfer matrix P _t ), and the historical trajectory data X includes t historical trajectory data at t moments (the t-th moment corresponds to the historical trajectory data X _t ). The first graph neural network is the graph convolutional neural network model 41 in FIG4 , and the trajectory transfer matrix P and the historical traffic matrix X are input into the graph convolutional neural network model 41, and the traffic conversion state matrix D _t at the t-th moment in the 0-t moments within the time period T is output:

D _t ∈R ^|V|×(d+1) ;

Wherein, d is the first graph propagation parameter, V = { _vi } _i = _{1, 2, ... M} , M is the number of road segments in the preset road network, and _vi represents a road segment in the preset road network. The first graph propagation parameter can characterize the maximum number of trajectory transfers of the vehicle in the preset road network at the t-th moment. For example, as shown in FIG4, in the graph convolutional neural network model 41, the driving trajectory 411 at the t-1th moment is compared with the driving trajectory 412 at the t-th moment. Under d trajectory transfers, the range formed by the driving trajectory 412 at the t-th moment is larger than the trajectory range formed by the driving trajectory 411 at the t-1th moment. Therefore, the larger the value of the first graph propagation parameter d, the more times the vehicle transfers in the preset time period corresponding to the t-th moment, and the larger the trajectory range that the vehicle may form in the preset road network (that is, it can reach a farther destination).

As an optional embodiment, the trajectory transfer matrix and the historical trajectory data are input into the first graph neural network, and graph propagation is performed according to the first graph propagation parameter to obtain the traffic conversion state, including: multiplying the nth power of the transpose of the trajectory transfer matrix at the specified time by the historical trajectory data at the specified time to obtain the intermediate value of the traffic conversion state after the nth trajectory transfer at the specified time, where n is less than or equal to the maximum number of trajectory transfers; connecting the historical trajectory data at the specified time with the intermediate value of the traffic conversion state after each trajectory transfer at the specified time to obtain the traffic conversion state at the specified time.

Specifically, the designated time may be any time from the 1st time to the tth time within the above historical time T, then the historical trajectory data at the designated time may be the historical trajectory data X _t corresponding to the tth time, the trajectory transfer matrix at the designated time may be the trajectory transfer matrix P _t corresponding to the tth time, and the calculation formula of the traffic conversion state matrix D _t may be:

in, represents the transposed matrix of the trajectory transfer matrix P _t , || represents the connection of matrices, /> They are the intermediate values of traffic conversion states corresponding to when n is 1, 2...d respectively.

As an optional embodiment, determining the neighborhood traffic status based on the weighted road adjacency matrix and historical trajectory data includes: obtaining a second graph propagation parameter, wherein the second graph propagation parameter is used to characterize the maximum transfer range when the weighted road adjacency matrix and the historical trajectory data are propagated through graph; inputting the weighted road adjacency matrix at a specified time and the historical trajectory data at a specified time into a second graph neural network, performing graph propagation according to the second graph propagation parameter, and obtaining the neighborhood traffic status at the specified time.

The second graph neural network may be a graph convolutional neural network model, and the second graph propagation parameter may be a neighborhood state hop count. The specified time may be any time from the 0th time to the tth time within the historical time T.

In an optional embodiment, the neighborhood traffic state can be represented in matrix form. As shown in FIG4 , the second graph neural network is a graph convolutional neural network model 42. The weighted road adjacency matrix A is the same in the historical time T. The historical trajectory data at a specified time may be the historical trajectory data X _t at the t-th time. The weighted road adjacency matrix A and the historical trajectory data X _t are input into the graph convolutional neural network model 42. The graph convolutional neural network model 42 outputs the neighborhood traffic state matrix S _t at the t-th time:

Wherein, s is the second graph propagation parameter, V={ _vi } _{i=1, 2, ...M} , M is the number of road segments in the preset road network, _vi represents a road segment in the preset road network, as shown in FIG4, the second graph propagation parameter s can represent the range of the farthest adjacent road segment that the vehicle can reach in the preset road network at the t-th moment. For example, as shown in FIG4, in the graph convolutional neural network model 42, the driving trajectory 421 at the t-1th moment can represent the range of the designated road segment that the vehicle can reach after s trajectory transfers (that is, the vehicle within the range can reach the designated road segment after s or less than s transfers), and the driving trajectory 422 at the t-th moment represents the designated adjacent road segment reached after s trajectory transfers, that is, the vehicle within the circle range of the driving trajectory 421 at the t-1th moment in FIG4 can reach the road segment pointed to by the driving trajectory 422 at the t-1th moment after s trajectory transfers. Therefore, the larger the value of the second graph propagation parameter s, the larger the range of the initial position of the vehicle with the designated adjacent road segment as the destination.

As an optional embodiment, the weighted road adjacency matrix at a specified time and the historical trajectory data at a specified time are input into a second graph neural network, and graph propagation is performed according to the second graph propagation parameter to obtain the neighborhood traffic state at the specified time, including: multiplying the mth power of the transpose of the weighted road adjacency matrix at the specified time by the historical trajectory data at the specified time to obtain the intermediate value of the traffic conversion state of the m road sections at the specified time, where m is less than or equal to the maximum transfer range; connecting the historical trajectory data at the specified time with the intermediate value of the traffic conversion state of each m road section at the specified time to obtain the neighborhood traffic state at the specified time.

Specifically, the neighborhood traffic state matrix _St at the tth moment can be obtained by the following formula:

in, is the normalized matrix of the weighted road adjacency matrix A,/> The intermediate values of the traffic conversion state corresponding to when m is 1, 2...s respectively.

As an optional embodiment, predicting future traffic flow based on traffic conversion status and neighborhood traffic status includes: using an attention mechanism, weighting the traffic conversion status obtained by each trajectory transfer at the specified time based on the neighborhood traffic status at the specified time, to obtain the traffic flow at the time to be predicted predicted at the specified time, wherein the neighborhood traffic status is a query parameter in the attention mechanism, the traffic conversion status is a value in the attention mechanism, and the key parameter in the attention mechanism is a preset value; determining multiple specified time periods from a preset initial time to a previous time of the time to be predicted according to a preset time interval; fusing the traffic flow at the time to be predicted obtained by predicting multiple specified time periods through a fully connected layer to obtain the final predicted traffic flow at the time to be predicted.

The designated time may be any time from the 0th time to the tth time within the above historical time T. In the traffic flow at the time to be predicted predicted by the designated time, the time to be predicted is from the 0th time to the tth time, and the traffic flow at the time to be predicted in the above final prediction is at the t+1th time after the tth time.

The attention mechanism model can be:

Attention(Q, K, V) := softmax(QK ^T )V;

Among them, Q is the query parameter, K is the key parameter, and V is the value in the attention mechanism.

Specifically, as shown in FIG4 , the traffic conversion state Dt at the t-th moment is weighted based on the neighborhood traffic state St at the t-th moment through the attention mechanism 43, that is, the weighted average value of the conversion state hops d at different distances can be obtained, and the traffic conversion state Dt can be used as the value V of the above-mentioned attention mechanism model, and the neighborhood traffic state St can be used as the query parameter Q of the above-mentioned attention mechanism model. According to the above-mentioned attention mechanism model, the predicted traffic flow H within the historical time T is obtained. The predicted traffic flow H includes the traffic flow at the t+1 moment predicted from the 0th moment to the t-th moment. The predicted traffic flow _Ht at the t-th moment is:

Among them, the key parameter The weight value α∈[0,1] ^|V|×(d+1) .

As shown in FIG4 , the predicted traffic flow H from the 0th time to the tth time is input into the fully connected layer 44, and the fully connected layer 44 performs multi-step fusion processing on the predicted traffic flow H from the 0th time to the tth time, and outputs the predicted traffic flow y _v at the t+1th time:

y _v :=X _t+1,v =FulyConnected(H _:,v ; Θ)=Θ ^T H _:,v ;

The elements in the matrix Xt _+1,v represent the predicted traffic flow of the road segment v in the road network at the future time t+1.

It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described order of actions, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

Example 2

According to an embodiment of the present invention, a device for implementing the above-mentioned traffic flow prediction method is also provided. FIG. 5 is a schematic diagram of a traffic flow prediction device according to Embodiment 2 of the present invention. As shown in FIG. 5 , the device 500 includes:

The acquisition module 51 is used to acquire the trajectory transfer matrix, historical trajectory data and weighted road adjacency matrix within the preset road network range, wherein the trajectory transfer matrix is used to represent the transfer probability between the upstream section and the downstream section in the adjacent sections within the preset road network range, and the weighted road adjacency matrix is used to represent the adjacency relationship between the roads within the preset road network range; the traffic conversion state determination module 52 is used to determine the traffic conversion state according to the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when multiple trajectory transfers are performed; the neighborhood traffic state determination module 53 is used to determine the neighborhood traffic state according to the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of the adjacent sections; the prediction module 54 is used to predict the future traffic flow according to the traffic conversion state and the neighborhood traffic state.

It should be noted that the acquisition module 51, the traffic conversion state determination module 52, the neighborhood traffic state determination module 53 and the prediction module 54 correspond to steps S201 to S202 in Example 1, and the four modules and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules as part of the device can be run in the computing device 10 provided in Example 1.

As an optional embodiment, the above-mentioned acquisition module includes: an extraction submodule, which is used to extract the number of vehicles turning from the upstream section to each downstream section within a preset time interval from the historical trajectory data; a transfer probability determination submodule, which is used to determine the transfer probability of transferring from the upstream section to each downstream section according to the number; and a trajectory transfer matrix determination submodule, which is used to determine the trajectory transfer matrix based on the transfer probability between multiple sections within a preset road network range.

As an optional embodiment, the transfer probability determination submodule includes: an amplification submodule, which is used to amplify the number of vehicles turning from the upstream section to each downstream section within a preset time interval to densify the trajectory transfer matrix; and a determination submodule, which is used to determine the ratio of the amplified number of vehicles turning from the upstream section to each downstream section within the preset time interval to the number of all vehicles in the upstream section within the preset time interval as the transfer probability from the upstream section to each downstream section.

As an optional embodiment, the traffic conversion state determination module includes: a first parameter acquisition submodule, used to obtain a first graph propagation parameter, wherein the first graph propagation parameter is used to characterize the maximum number of trajectory transfers when the trajectory transfer matrix and the historical trajectory data are propagated through the graph; a first input submodule, used to input the trajectory transfer matrix and the historical trajectory data into the first graph neural network, perform graph propagation according to the first graph propagation parameter, and obtain the traffic conversion state.

As an optional embodiment, the first input submodule includes: a traffic conversion state intermediate value acquisition submodule, which is used to multiply the nth power of the transpose of the trajectory transfer matrix at the specified time by the historical trajectory data at the specified time to obtain the intermediate value of the traffic conversion state after the nth trajectory transfer at the specified time, where n is less than or equal to the maximum number of trajectory transfers; a first connection submodule, which is used to connect the historical trajectory data at the specified time with the intermediate value of the traffic conversion state after each trajectory transfer at the specified time to obtain the traffic conversion state at the specified time.

As an optional embodiment, the neighborhood traffic status determination module includes: a second parameter acquisition submodule, used to obtain a second graph propagation parameter, wherein the second graph propagation parameter is used to characterize the maximum transfer range when the weighted road adjacency matrix and the historical trajectory data are propagated through the graph; a second input submodule, used to input the weighted road adjacency matrix at a specified time and the historical trajectory data at a specified time into a second graph neural network, perform graph propagation according to the second graph propagation parameter, and obtain the neighborhood traffic status at the specified time.

As an optional embodiment, the second input submodule includes: a traffic conversion state intermediate value acquisition submodule, which is used to multiply the mth power of the transpose of the weighted road adjacency matrix at the specified time by the historical trajectory data at the specified time, and obtain the intermediate value of the traffic conversion state of the m road sections at the specified time, where m is less than or equal to the maximum transfer range; a second connection submodule, which is used to connect the historical trajectory data at the specified time with the intermediate value of the traffic conversion state of each m road section at the specified time, to obtain the neighborhood traffic state at the specified time.

As an optional embodiment, the prediction module includes: a weighted submodule, which is used to weight the traffic conversion state obtained by each trajectory transfer at the specified time based on the neighborhood traffic state at the specified time through an attention mechanism to obtain the traffic flow at the time to be predicted predicted at the specified time, wherein the neighborhood traffic state is the query parameter in the attention mechanism, the traffic conversion state is the value in the attention mechanism, and the key parameter in the attention mechanism is a preset value; a designated time determination submodule, which is used to determine multiple designated times from a preset initial time to a previous time of the time to be predicted according to a preset time interval; a fusion submodule, which is used to fuse the traffic flow at the time to be predicted obtained by predicting multiple designated times through a fully connected layer to obtain the final predicted traffic flow at the time to be predicted.

Example 3

The embodiment of the present invention further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the traffic flow prediction method provided in the first embodiment.

Optionally, in this embodiment, the above storage medium may be located in any one of the computing devices in the computing device group in the computer network, or in any one of the mobile terminals in the mobile terminal group.

The storage medium may be a computer-readable storage medium, and the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned traffic flow prediction method.

Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps: obtaining a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network range, wherein the trajectory transfer matrix is used to represent the transfer probability between an upstream section and a downstream section in adjacent sections within the preset road network range, and the weighted road adjacency matrix is used to represent the adjacency relationship between roads within the preset road network range; determining a traffic conversion state based on the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when multiple trajectory transfers are performed; determining a neighborhood traffic state based on the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of adjacent sections; and predicting future traffic flow based on the traffic conversion state and the neighborhood traffic state.

Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: obtaining a trajectory transfer matrix within a preset road network range, including: extracting the number of vehicles turning from the upstream section to each downstream section within a preset time interval from historical trajectory data; determining the transfer probability of transferring from the upstream section to each downstream section based on the number; and determining the trajectory transfer matrix based on the transfer probability between multiple sections within the preset road network range.

Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: determining the transfer probability from the upstream section to each downstream section based on the quantity, including: amplifying the number of vehicles transferring from the upstream section to each downstream section within a preset time interval to densify the trajectory transfer matrix; determining the ratio of the number of vehicles transferring from the upstream section to each downstream section within the amplified preset time interval to the number of all vehicles in the upstream section within the preset time interval as the transfer probability from the upstream section to each downstream section.

Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: determining a traffic conversion state based on a trajectory transfer matrix and historical trajectory data, including: obtaining a first graph propagation parameter, wherein the first graph propagation parameter is used to characterize the maximum number of trajectory transfers when the trajectory transfer matrix and historical trajectory data are subjected to graph propagation; inputting the trajectory transfer matrix and the historical trajectory data into a first graph neural network, performing graph propagation according to the first graph propagation parameter, and obtaining a traffic conversion state.

Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: inputting the trajectory transfer matrix and the historical trajectory data into the first graph neural network, performing graph propagation according to the first graph propagation parameter, and obtaining the traffic conversion state, including: multiplying the nth power of the transpose of the trajectory transfer matrix at the specified time by the historical trajectory data at the specified time to obtain the intermediate value of the traffic conversion state after the nth trajectory transfer at the specified time, where n is less than or equal to the maximum number of trajectory transfers; connecting the historical trajectory data at the specified time with the intermediate value of the traffic conversion state after each trajectory transfer at the specified time to obtain the traffic conversion state at the specified time.

Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: determining the neighborhood traffic status based on the weighted road adjacency matrix and historical trajectory data, including: obtaining a second graph propagation parameter, wherein the second graph propagation parameter is used to characterize the maximum transfer range when the weighted road adjacency matrix and the historical trajectory data are propagated through the graph; inputting the weighted road adjacency matrix at a specified time and the historical trajectory data at a specified time into a second graph neural network, performing graph propagation according to the second graph propagation parameter, and obtaining the neighborhood traffic status at the specified time.

Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: inputting the weighted road adjacency matrix at a specified time and the historical trajectory data at a specified time into a second graph neural network, performing graph propagation according to the second graph propagation parameter, and obtaining the neighborhood traffic state at the specified time, including: multiplying the mth power of the transpose of the weighted road adjacency matrix at the specified time by the historical trajectory data at the specified time to obtain the intermediate value of the traffic conversion state of the m road sections at the specified time, where m is less than or equal to the maximum transfer range; connecting the historical trajectory data at the specified time with the intermediate value of the traffic conversion state of each m road section at the specified time to obtain the neighborhood traffic state at the specified time.

Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: predicting future traffic flow based on the traffic conversion state and the neighborhood traffic state, including: through an attention mechanism, weighting the traffic conversion state obtained by each trajectory transfer at the specified time based on the neighborhood traffic state at the specified time, to obtain the traffic flow at the time to be predicted predicted at the specified time, wherein the neighborhood traffic state is a query parameter in the attention mechanism, the traffic conversion state is a value in the attention mechanism, and the key parameter in the attention mechanism is a preset value; determining multiple specified times from a preset initial time to a previous time of the time to be predicted according to a preset time interval; fusing the traffic flow at the time to be predicted obtained by predicting multiple specified times through a fully connected layer to obtain the final predicted traffic flow at the time to be predicted.

Example 4

According to an embodiment of the present application, an embodiment of a computer terminal is also provided, and the computer terminal can be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the above-mentioned computer terminal can also be replaced by a terminal device such as a mobile terminal.

Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of the computer network.

In this embodiment, the computer terminal can execute the following steps of the program code in the method for processing the video of the application: obtaining a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network range, wherein the trajectory transfer matrix is used to represent the transfer probability between an upstream section and a downstream section in adjacent sections within the preset road network range, and the weighted road adjacency matrix is used to represent the adjacency relationship between roads within the preset road network range; determining a traffic conversion state according to the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when multiple trajectory transfers are performed; determining a neighborhood traffic state according to the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of adjacent sections; and predicting future traffic flow according to the traffic conversion state and the neighborhood traffic state.

Optionally, Figure 6 is a structural block diagram of a computer terminal according to Example 6 of the present application. As shown in Figure 6, the computer terminal 700 may include: one or more (only one is shown in the figure) processors 702, a memory 704, and a peripheral interface 706.

Among them, the memory can be used to store software programs and modules, such as the program instructions/modules corresponding to the video interpolation method and device in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned video processing method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal 700 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

The processor is used to run the program, and can call the information and application program stored in the memory through the transmission device to perform the following steps: obtaining a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network range, wherein the trajectory transfer matrix is used to represent the transfer probability between an upstream road section and a downstream road section in adjacent road sections within the preset road network range, and the weighted road adjacency matrix is used to represent the adjacency relationship between roads within the preset road network range; determining a traffic conversion state according to the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when multiple trajectory transfers are performed; determining a neighborhood traffic state according to the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of adjacent road sections; and predicting future traffic flow according to the traffic conversion state and the neighborhood traffic state.

Those skilled in the art will appreciate that the structure shown in FIG6 is for illustration only, and the computer terminal may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a palm computer, a mobile Internet device (Mobile Internet Devices, MID), a PAD, and other terminal devices. FIG6 does not limit the structure of the above electronic devices. For example, the computer terminal 700 may also include more or fewer components (such as a network interface, a display device, etc.) than those shown in FIG6, or have a configuration different from that shown in FIG6.

A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

Example 5

According to an embodiment of the present application, a traffic flow prediction system is also provided, the system comprising: a processor; and a memory connected to the processor, for providing the processor with instructions for processing the following processing steps: obtaining a trajectory transfer matrix, historical trajectory data and a weighted road adjacency matrix within a preset road network range, wherein the trajectory transfer matrix is used to represent the transfer probability between an upstream section and a downstream section in adjacent sections within the preset road network range, and the weighted road adjacency matrix is used to represent the adjacency relationship between roads within the preset road network range; determining a traffic conversion state according to the trajectory transfer matrix and the historical trajectory data, wherein the traffic conversion state is used to represent the traffic flow corresponding to each trajectory transfer when multiple trajectory transfers are performed; determining a neighborhood traffic state according to the weighted road adjacency matrix and the historical trajectory data, wherein the neighborhood traffic state is used to represent the traffic flow of adjacent sections; predicting future traffic flow according to the traffic conversion state and the neighborhood traffic state.

The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims (12)

1. A method for predicting traffic flow, comprising:

Acquiring a track transfer matrix, historical track data and a weighted road adjacency matrix in a preset road network range, wherein the track transfer matrix is used for representing the transfer probability between an upstream road section and a downstream road section of adjacent road sections in the preset road network range, and the weighted road adjacency matrix is used for representing the adjacency relationship between roads in the preset road network range;

Determining a traffic conversion state according to the track transfer matrix and the historical track data, wherein the traffic conversion state is used for representing the corresponding traffic flow after each track transfer when the traffic flow is subjected to track transfer for a plurality of times;

Determining a neighborhood traffic state according to the weighted road adjacency matrix and the historical track data, wherein the neighborhood traffic state is used for representing traffic flow of an adjacency road section;

And predicting future traffic flow according to the traffic conversion state and the neighborhood traffic state.

2. The method of claim 1, wherein obtaining a trajectory transfer matrix within a predetermined road network range comprises:

extracting the number of vehicles from the upstream road sections to each downstream road section within a preset time interval from the historical track data;

Determining a transition probability of transitioning from the upstream road segment to each of the downstream road segments based on the number;

and determining the track transition matrix based on transition probabilities among a plurality of road sections in the preset road network range.

3. The method of claim 2, wherein determining a transition probability for transitioning from the upstream road segment to the each downstream road segment based on the number comprises:

amplifying the number of vehicles transferred from an upstream road section to each downstream road section within the preset time interval so as to densify the track transfer matrix;

and determining that the ratio of the number of the amplified vehicles in the preset time interval from the upstream road section to each downstream road section to the number of all the vehicles in the upstream road section in the preset time interval is the transition probability of the upstream road section to each downstream road section.

4. The method of claim 1, wherein determining traffic conversion status from the trajectory transfer matrix and the historical trajectory data comprises:

Acquiring a first graph propagation parameter, wherein the first graph propagation parameter is used for representing the maximum track transfer times when the track transfer matrix and the historical track data are subjected to graph propagation;

And inputting the track transfer matrix and the historical track data into a first graph neural network, and performing graph propagation according to the first graph propagation parameters to obtain the traffic conversion state.

5. The method of claim 4, wherein inputting the trajectory transfer matrix and the historical trajectory data to a first graph neural network, graph propagation according to the first graph propagation parameters, and obtaining the traffic conversion state comprises:

multiplying n-th power of transpose of a track transfer matrix at a specified time with historical track data at the specified time to obtain a traffic conversion state intermediate value after n-th track transfer at the specified time, wherein n is less than or equal to the maximum track transfer times;

And connecting the historical track data at the appointed time with the intermediate value of the traffic conversion state after each track transfer at the appointed time to obtain the traffic conversion state at the appointed time.

6. The method of claim 1, wherein the neighborhood traffic state determined from the weighted roadway adjacency matrix and the historical trajectory data comprises:

Acquiring a second graph propagation parameter, wherein the second graph propagation parameter is used for representing the maximum transfer range of the weighted road adjacency matrix and the historical track data when graph propagation is carried out;

And inputting the weighted road adjacency matrix at the appointed time and the historical track data at the appointed time into a second graph neural network, and performing graph propagation according to the second graph propagation parameters to obtain the neighborhood traffic state at the appointed time.

7. The method of claim 6, wherein inputting the weighted road adjacency matrix at the specified time and the historical track data at the specified time to a second graph neural network, performing graph propagation according to the second graph propagation parameter, and obtaining the neighborhood traffic state at the specified time comprises:

Multiplying the m-th power of the transpose of the weighted road adjacent matrix at the appointed moment by the historical track data at the appointed moment, wherein m is smaller than or equal to the maximum transition range;

And connecting the historical track data at the appointed time with the intermediate value of the traffic conversion state of each m road section track at the appointed time to obtain the neighborhood traffic state at the appointed time.

8. The method of claim 1, wherein predicting future traffic flow based on the traffic conversion state and the neighborhood traffic state comprises:

Weighting a traffic conversion state obtained by each track transfer at a specified time based on a neighborhood traffic state at the specified time through an attention mechanism to obtain a traffic flow at a time to be predicted, which is predicted at the specified time, wherein the neighborhood traffic state is a query parameter in the attention mechanism, the traffic conversion state is a value in the attention mechanism, and a key parameter in the attention mechanism is a preset value;

Determining a plurality of designated moments from a preset initial moment to a moment before the moment to be predicted according to a preset time interval;

And carrying out fusion processing on the traffic flow of the predicted time obtained by predicting the plurality of specified times through the full connection layer to obtain the finally predicted traffic flow of the predicted time.

9. A traffic flow prediction apparatus, comprising:

The acquisition module is used for acquiring a track transfer matrix, historical track data and a weighted road adjacency matrix in a preset road network range, wherein the track transfer matrix is used for representing the transfer probability between an upstream road section and a downstream road section in an adjacent road section in the preset road network range, and the weighted road adjacency matrix is used for representing the adjacency relationship between roads in the preset road network range;

The traffic conversion state determining module is used for determining a traffic conversion state according to the track transfer matrix and the historical track data, wherein the traffic conversion state is used for representing the traffic flow corresponding to each track transfer when the track transfer is carried out for a plurality of times;

the neighborhood traffic state determining module is used for determining neighborhood traffic states according to the weighted road adjacency matrix and the historical track data, wherein the neighborhood traffic states are used for representing traffic flow of adjacent road sections;

and the prediction module is used for predicting the future traffic flow according to the traffic conversion state and the neighborhood traffic state.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium comprises a stored program, wherein the program, when run, controls a device in which the computer-readable storage medium is located to perform the traffic flow prediction method according to any one of claims 1 to 8.

11. A processor, characterized in that the processor is configured to run a program, wherein the program, when run, performs the traffic flow prediction method according to any one of claims 1 to 8.

12. A traffic flow prediction system, comprising:

A processor; and

A memory, coupled to the processor, for providing instructions to the processor to process the following processing steps:

Acquiring a track transfer matrix, historical track data and a weighted road adjacency matrix in a preset road network range, wherein the track transfer matrix is used for representing the transfer probability between an upstream road section and a downstream road section of adjacent road sections in the preset road network range, and the weighted road adjacency matrix is used for representing the adjacency relationship between roads in the preset road network range;

Determining a traffic conversion state according to the track transfer matrix and the historical track data, wherein the traffic conversion state is used for representing the corresponding traffic flow after each track transfer when the traffic flow is subjected to track transfer for a plurality of times;

Determining a neighborhood traffic state according to the weighted road adjacency matrix and the historical track data, wherein the neighborhood traffic state is used for representing traffic flow of an adjacency road section;

And predicting future traffic flow according to the traffic conversion state and the neighborhood traffic state.