Regional bridge group traffic flow prediction method based on full connection graph and double convolution

Abstract

The invention provides a regional bridge group traffic flow prediction method and device based on a full connection graph and double convolution, and relates to the technical field of time series data prediction. The method comprises the following steps: monitoring traffic flows of different bridge nodes in a regional bridge group to obtain original flow data of a two-dimensional data structure time sequence; performing data structure dimension raising processing on the original traffic flow data to obtain improved data of a three-dimensional data structure; training the model by using the improved data; according to the current improved data, performing feature extraction through a graph convolutional neural network to obtain space-time related features; capturing data periodicity through a two-dimensional convolutional network to obtain periodicity features; and splicing, fusing and inputting the time-space correlation features and the periodic features into a full-connection layer for prediction to obtain a traffic flow prediction result. The invention relates to a regional bridge group traffic flow prediction method based on a full connection graph and a double convolution multidimensional time sequence, and provides a basis for performance evolution prediction of bridges in a regional road network.

Description

Technical Field

[0001] The present invention relates to the technical field of time series data prediction, and in particular to a method and device for predicting regional bridge group traffic flow based on a fully connected graph and double convolution. Background Art

[0002] In modern data analysis, the prediction of time series data is crucial and is widely used in fields such as weather forecasting, financial analysis, and real-time monitoring of monitoring point data. With the complexity of application scenarios, traditional time series prediction methods have gradually exposed their limitations, especially in the detection and monitoring of regional bridge groups. Although traditional methods such as autoregressive models and long short-term memory networks have advantages in capturing time dependencies, they are difficult to effectively handle the complex dependencies between different dimensions when faced with multidimensional time series data in regional bridge groups.

[0003] The above methods only focus on single-dimensional time series data, ignoring the complexity of spatiotemporal interactions in multidimensional time series. In the monitoring of regional bridge groups, the traffic flow, stress and strain data at each monitoring point not only change over time, but are also closely related to the data at other monitoring points. Existing methods are difficult to fully capture these complex cross-dimensional correlations, resulting in prediction accuracy and robustness that cannot meet actual needs.

[0004] Fully connected graphs and double convolution can improve the accuracy and stability of multi-dimensional time series prediction. This method combines graph neural networks and convolutional neural networks through special processing of multi-dimensional time series data. It can not only effectively extract the time-dependent features of each monitoring point, but also use the fully connected graph to capture the spatiotemporal interaction relationship between different monitoring points, thus achieving more accurate traffic flow prediction.

[0005] In the prior art, there is a lack of an accurate and efficient regional bridge group traffic flow prediction method based on a fully connected graph and a multi-dimensional time series with double convolution. Summary of the invention

[0006] In order to solve the technical problem that it is difficult to fully capture the complex cross-dimensional correlations in the prior art, resulting in the prediction accuracy and robustness failing to meet actual needs, the embodiment of the present invention provides a method and device for predicting the traffic flow of regional bridge groups based on a fully connected graph and double convolution. The technical solution is as follows:

[0007] On the one hand, a method for predicting regional bridge group traffic flow based on a fully connected graph and dual convolution is provided, the method being implemented by a regional bridge group traffic flow prediction device, the method comprising:

[0008] At the preset monitoring points of the regional bridge group, the vehicle flow is monitored to obtain the original flow data of the two-dimensional data structure time series;

[0009] Performing data structure dimensionality upgrading processing on the original flow data to obtain improved data with a three-dimensional data structure;

[0010] Using the improved data, training the graph convolutional neural network to be trained and the two-dimensional convolutional neural network to be trained to obtain a graph convolutional neural network and a two-dimensional convolutional neural network;

[0011] Acquire current traffic flow data; perform dimension-upgrading processing on the current traffic flow data to obtain current improved data;

[0012] According to the current improved data, feature extraction is performed through a graph convolutional neural network to obtain spatiotemporal related features;

[0013] According to the current improved data, capturing the data periodicity through a two-dimensional convolutional network to obtain periodic features;

[0014] The spatiotemporal correlation features and the periodic features are concatenated and fused to obtain fused features; the fused features are input into a fully connected layer to perform traffic flow prediction to obtain a traffic flow prediction result.

[0015] On the other hand, a regional bridge group traffic flow prediction device based on a fully connected graph and dual convolution is provided, and the device is applied to a regional bridge group traffic flow prediction method based on a fully connected graph and dual convolution, and the device includes:

[0016] A training data acquisition module is used to monitor the vehicle flow at the preset monitoring points of the regional bridge group and obtain the original flow data of the two-dimensional data structure time series;

[0017] A data dimension-upgrading module, used for performing data structure dimension-upgrading processing on the original flow data to obtain improved data with a three-dimensional data structure;

[0018] A model training module, used to train the graph convolutional neural network to be trained and the two-dimensional convolutional neural network to be trained using the improved data, to obtain the graph convolutional neural network and the two-dimensional convolutional neural network;

[0019] The current data acquisition module is used to acquire the current vehicle flow data; the current vehicle flow data is processed by dimension upgrading to obtain the current improved data;

[0020] A first feature extraction module, configured to extract features through a graph convolutional neural network according to the current improved data to obtain spatiotemporal related features;

[0021] A second feature extraction module is used to capture data periodicity through a two-dimensional convolutional network according to the current improved data to obtain periodic features;

[0022] The vehicle flow prediction module is used to splice and fuse the spatiotemporal correlation features and the periodic features to obtain fused features; input the fused features into the fully connected layer to perform vehicle flow prediction to obtain a vehicle flow prediction result.

[0023] On the other hand, a regional bridge group traffic flow prediction device is provided, and the regional bridge group traffic flow prediction device includes: a processor; a memory, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, any one of the above-mentioned regional bridge group traffic flow prediction methods based on fully connected graphs and double convolutions is implemented.

[0024] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement any one of the above-mentioned methods for predicting regional bridge group traffic flow based on a fully connected graph and double convolution.

[0025] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0026] The present invention proposes a method for predicting the traffic flow of regional bridge groups based on a fully connected graph and dual convolution, converting multidimensional time series data into a three-dimensional model, using graph convolution to capture the spatiotemporal correlation characteristics between multiple measurement point information, and two-dimensional convolution to capture the periodic characteristics of single measurement point information and perform feature splicing, and using the spliced ​​features to complete the final time series prediction task. The present invention processes multivariate time data with high prediction accuracy, strong stability and high precision. The present invention is an accurate and efficient method for predicting the traffic flow of regional bridge groups based on a multidimensional time series of a fully connected graph and dual convolution. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0028] Figure 1 It is a flow chart of a method for predicting regional bridge group traffic flow based on a fully connected graph and double convolution provided by an embodiment of the present invention;

[0029] Figure 2 It is a block diagram of a regional bridge group vehicle flow prediction device based on a fully connected graph and double convolution provided by an embodiment of the present invention;

[0030] Figure 3 It is a structural schematic diagram of a regional bridge group traffic flow prediction device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0032] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0033] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0034] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are consistent.

[0035] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0036] The embodiment of the present invention provides a method for predicting regional bridge group traffic flow based on a fully connected graph and dual convolution, which can be implemented by a regional bridge group traffic flow prediction device, which can be a terminal or a server. Figure 1 The flowchart of the regional bridge group traffic flow prediction method based on the fully connected graph and double convolution is shown. The processing flow of the method may include the following steps:

[0037] S1. Monitor the vehicle flow at the preset monitoring points of the regional bridge group to obtain the original flow data of the two-dimensional data structure time series.

[0038] Optionally, the vehicle flow is monitored at preset monitoring points of the regional bridge group to obtain the original flow data of the two-dimensional data structure time series, including:

[0039] At the monitoring points of the regional bridge group, data of passing vehicles is collected based on the dynamic weighing system to obtain data on all types of vehicle flow;

[0040] Based on the traffic flow data of all types of vehicles, the number of vehicles with a total weight greater than a preset weight is recorded to obtain the traffic flow data of heavy-loaded vehicles.

[0041] In a feasible implementation, a dynamic weighing system is installed at the upper and lower parts of the bridge to detect the traffic flow data of all types of vehicles passing through. The traffic flow data of all types of vehicles includes information such as the speed, axle weight, axle type, gross vehicle weight, and traffic flow at the monitoring point. The number of vehicles with a gross weight greater than the preset total weight is recorded to obtain the traffic flow data of heavy-loaded vehicles.

[0042] Each measuring point will simultaneously record two types of traffic flow data, namely, traffic flow data of all types of vehicles and heavy-loaded vehicles. Each type of traffic flow data is a one-dimensional data structure, and all the original traffic flow data collected in the present invention are considered as a two-dimensional data structure as a whole.

[0043] S2. Perform data structure dimensionality upgrade processing on the original traffic data to obtain improved data with a three-dimensional data structure.

[0044] Optionally, the original traffic data is processed by data structure dimensionality upgrade to obtain improved data with a three-dimensional data structure, including:

[0045] Perform data preprocessing on the original traffic data to obtain processed traffic data;

[0046] Divide the time window of the processed traffic flow data according to the preset shortest time period to obtain the time window length;

[0047] Based on the time window length, the processed traffic flow data is segmented according to the preset shortest time period to obtain a set of time series data segments;

[0048] According to the set of time series data fragments, the data are stacked in the same node alignment manner to obtain improved data.

[0049] In a feasible implementation manner, in the present invention, a mathematical expression of a set of time series data X is as follows (1):

[0050] (1);

[0051] Wherein, N is the number of monitoring points, L is the data length, and R represents the entire real number space. The present invention uses 24 hours as the shortest period S of vehicle flow data, which can also be adjusted according to actual conditions.

[0052] In the case of individual missing data during the acquisition process, the average of the previous and next data at the missing position is used to fill the missing data; the local feature enhancement processing is performed on the filled multidimensional time series data, and the normalization method is used to reduce the influence of noise. The mathematical expression of this process is as follows (2):

[0053] (2);

[0054] in, Represents monitoring point data, Represents the coordinates on the time axis, Indicates the serial number of the monitoring point.

[0055] Downsample the data with high sampling frequency. By selecting sampling points within a fixed interval, ensure that the downsampled data can cover the entire period while maintaining the same time step as other time series.

[0056] For the input data in each time window, the input data is segmented according to the determined shortest period S, and the time series is cut into segments of the same length. , the mathematical expression of the fragment is as follows (3):

[0057] (3);

[0058] in is the length of the input data, express The total number is rounded up. These fragments are stacked in the time dimension in a way that the same nodes are aligned to construct the dimensional data. The mathematical expression of this process is as follows (4):

[0059] (4);

[0060] in, represents a slice. Formula (4) shows the dimensionality-enhanced data from different perspectives.

[0061] S3. Use the improved data to train the graph convolutional neural network to be trained and the two-dimensional convolutional neural network to be trained to obtain a graph convolutional neural network and a two-dimensional convolutional neural network.

[0062] In a feasible implementation, the multidimensional time series traffic flow data is divided into windows according to the preset shortest cycle length S. The length of each time window is set to (n1+n2)S, where n1 and n2 are positive integers, ensuring that each window contains a complete periodic pattern. In each time window, the data of n1 cycles are used as input, and the data of n2 cycles are used as actual values ​​for comparison with the predicted results. In order to generate more training or test samples, the sliding window method is used to gradually move the window on the time series to generate new input and corresponding true value pairs, thereby increasing the number and diversity of training samples.

[0063] The first 60% of the time window is set as the training set, the second 20% of the time window is set as the validation set, and the last 20% of the time window is set as the test set. In each round of training, the loss function is used to measure the difference between the predicted result and the true value, and the network structure parameters are optimized through the back propagation algorithm, including the one-dimensional convolution kernel, the two-dimensional convolution kernel, the fully connected graph edge weight value, the attenuation matrix, and the weight and bias parameters in the fully connected layer. After each round of training, the validation set is used to test the training results of that round. When testing, only the predicted value and the true value need to be compared. If the error is less than the saved optimal parameter, the optimal parameter is updated. Use the test set for testing. Use the test set to test the optimal parameters, obtain the test results, and finally put it into use.

[0064] S4. Obtain current traffic flow data; perform dimension-upgrading processing on the current traffic flow data to obtain current improved data.

[0065] In a feasible implementation manner, the vehicle flow data to be measured is collected at a monitoring point, and the data is processed from two dimensions to three dimensions in terms of data structure.

[0066] S5. Based on the current improved data, feature extraction is performed through graph convolutional neural network to obtain spatiotemporal related features.

[0067] Optionally, based on the current improved data, feature extraction is performed through a graph convolutional neural network to obtain spatiotemporal related features, including:

[0068] Based on the current improved data, multiple fully connected graphs are constructed using the sliding window method;

[0069] Based on multiple fully connected graphs, the spatiotemporal correlation characteristics are captured through graph convolutional neural networks to obtain spatiotemporal correlation features.

[0070] In a feasible implementation, the present invention improves the accuracy and stability of multi-dimensional time series prediction based on a regional bridge group traffic flow prediction method based on a fully connected graph and double convolution. By special processing of multi-dimensional time series data, combined with graph neural network and convolutional neural network, it can not only effectively extract the time-dependent features of each monitoring point, but also use the fully connected graph to capture the spatiotemporal interaction relationship between different monitoring points, so as to achieve more accurate traffic flow prediction.

[0071] Optionally, based on the current improved data, a sliding window method is used to construct a fully connected graph, including:

[0072] Input the current improved data into a one-dimensional convolutional neural network for feature extraction to obtain the current traffic flow characteristics;

[0073] The similarity of the traffic flow between monitoring points is calculated based on the current improved data to obtain the traffic flow similarity matrix;

[0074] According to the current traffic flow characteristics, the traffic flow similarity matrix and the preset learnable time decay matrix, multiple fully connected graphs are constructed using the sliding window method.

[0075] In one feasible implementation, for each layer The monitoring points in the image are extracted using a one-dimensional convolutional neural network as follows:

[0076] (5);

[0077] Among them, ReLU is an activation function, which is used to perform nonlinear transformation on the extracted features. Its formula is as follows (6):

[0078] (6);

[0079] in, It is a regular one-dimensional convolution operation.

[0080] Use position coding to ensure the relative position information of each monitoring point. The mathematical expression of the relative position information is as follows:

[0081] (7);

[0082] in, Represents the location index of the monitoring point feature, Indicates time information. Representation and Related frequencies.

[0083] The final feature of each monitoring point can be expressed as follows (8):

[0084] (8);

[0085] Transform a multidimensional time series data segment The image is divided into s time slices, each of which contains all the monitoring point data at a time point. The sliding window method is used to select some time slices for composition, and different window sizes and strides are used to capture richer spatiotemporal correlations.

[0086] The traffic flow at the monitoring points has similar attributes in the feature space, so the similarity can represent the correlation between the traffic flow at the monitoring points. The similarity is measured by dot product, and the formula is as follows (9):

[0087] (9);

[0088] in, , represents the index of the time slice; , represents the index of the monitoring point; here T represents transposition. Function It is used for monitoring point feature transformation, and the softmax function normalizes the value to range.

[0089] The mathematical expression of the fully connected graph is as follows (10):

[0090] (10);

[0091] in, Represents the characteristics of all monitoring points, The adjacency matrix of the FC graph is represented by its elements, which represent the correlation between each monitoring point. In terms of time, the correlation between closely spaced points is large, while the correlation between far-spaced points is small.

[0092] Considering the time distance between monitoring points in different time slices, a learnable time decay matrix is ​​added as follows (11):

[0093] (11);

[0094] in, and One to one correspondence.

[0095] Combining formula (11), the fully connected graph is finally expressed as follows (12):

[0096] (12);

[0097] in, .

[0098] Optionally, based on multiple fully connected graphs, spatiotemporal correlation characteristics are captured by a graph convolutional neural network to obtain spatiotemporal correlation features, including:

[0099] Based on graph convolutional neural networks, monitoring point features are aggregated in multiple fully connected graphs to obtain multiple fully connected graph features;

[0100] Integrate multiple fully connected graph features to obtain spatiotemporal correlation features.

[0101] In one feasible implementation, a message passing network is used to capture the spatiotemporal dependencies of each node in the fully connected graph. The message passing network is mainly divided into two steps: message passing and message update. In message passing, for the Layer Central monitoring points on a fully connected graph , which is represented by a set of neighboring monitoring points Surrounded by, these monitoring points are from the same fully connected graph, and the edges between them can be expressed as The message passing process can be expressed by the following formula (13):

[0102] (13);

[0103] Use the RELU function to update the message and get the updated Layer features. It is expressed as follows:

[0104] (14);

[0105] In a single By using different scales and strides to build multiple fully connected graphs in parallel, and using graph convolution operations on these fully connected graphs in parallel, the features obtained in each fully connected graph are integrated. The extracted features are summarized to obtain . Represents the spatiotemporal correlation of monitoring points in multidimensional time series data over the entire period.

[0106] S6. According to the current improved data, the data periodicity is captured through a two-dimensional convolutional network to obtain periodic features.

[0107] Optionally, according to the current improved data, the data periodicity is captured by a two-dimensional convolutional network to obtain periodic features, including:

[0108] Based on the Inception module, multi-scale convolution processing is performed on the current improved data to obtain multi-scale periodic features;

[0109] In the channel dimension, multi-scale periodic features are spliced ​​to obtain periodic features.

[0110] In one possible implementation, for each slice of the dimension-increased data , is a The two-dimensional tensor of is used to extract periodic features using a two-dimensional convolutional neural network. The two-dimensional convolution kernel can capture the changes between adjacent time points in the same slice (intra-cycle changes) and the changes between different time slices (inter-cycle changes). The result of the convolution operation Y can be expressed as follows (15):

[0111] (15);

[0112] in, Is the position in the convolution result tensor Y The value at (here represents two-dimensional coordinates), is the position of the convolution kernel K The value of Yes Input Middle position The value of .

[0113] Using the Inception module Multi-scale convolution processing is performed, and convolution kernels of different sizes are used to extract features in parallel. The convolution results of different convolution kernels are concatenated in the channel dimension to obtain the final feature representation as follows (16):

[0114] (16);

[0115] in, Represents a splicing operation, Represents the convolution kernel size.

[0116] S7. Concatenate and fuse the spatiotemporal correlation features and the periodic features to obtain fused features; input the fused features into the fully connected layer to perform traffic flow prediction to obtain traffic flow prediction results.

[0117] In one possible implementation, the spatiotemporal correlation features captured by the fully connected graph And the periodic features extracted by 2D convolutional neural network The fused feature vector Input to the fully connected layer. In the fully connected layer, the input features will be further processed and integrated through linear transformation and nonlinear activation function. The fully connected layer is shown in the following formula (17):

[0118] (17);

[0119] in, represents the predicted value, Represents a fully connected layer operation.

[0120] The present invention uses mean square error and mean absolute error as loss functions to evaluate the accuracy of prediction results and optimize the prediction results. The loss function is defined as follows (18) and (19):

[0121] (18);

[0122] (19);

[0123] in, represents the predicted value, Represents the true value. Use the back-propagation algorithm to update the parameters through the gradient descent method. During the back-propagation process, the weights and biases in the layer are updated layer by layer according to the gradient of the loss function to reduce the prediction error.

[0124] The present invention proposes a method for predicting the traffic flow of regional bridge groups based on a fully connected graph and dual convolution, converting multidimensional time series data into a three-dimensional model, using graph convolution to capture the spatiotemporal correlation characteristics between multiple measurement point information, and two-dimensional convolution to capture the periodic characteristics of single measurement point information and perform feature splicing, and using the spliced ​​features to complete the final time series prediction task. The present invention processes multivariate time data with high prediction accuracy, strong stability and high precision. The present invention is an accurate and efficient method for predicting the traffic flow of regional bridge groups based on a multidimensional time series of a fully connected graph and dual convolution.

[0125] Figure 2 The present invention is a block diagram of a regional bridge group vehicle flow prediction device based on a fully connected graph and dual convolution according to an exemplary embodiment. The device is used in a regional bridge group vehicle flow prediction method based on a fully connected graph and dual convolution. Figure 2 The device includes a training data acquisition module 210, a data dimension upgrade module 220, a model training module 230, a current data acquisition module 240, a first feature extraction module 250, a second feature extraction module 260 and a vehicle flow prediction module 270. Among them:

[0126] The training data acquisition module 210 is used to monitor the vehicle flow at the preset monitoring points of the regional bridge group to obtain the original flow data of the two-dimensional data structure time series;

[0127] The data dimension-upgrading module 220 is used to perform data structure dimension-upgrading processing on the original traffic data to obtain improved data with a three-dimensional data structure;

[0128] A model training module 230 is used to train the graph convolutional neural network to be trained and the two-dimensional convolutional neural network to be trained using the improved data to obtain the graph convolutional neural network and the two-dimensional convolutional neural network;

[0129] The current data acquisition module 240 is used to acquire the current vehicle flow data; the current vehicle flow data is processed by dimension upgrading to obtain the current improved data;

[0130] A first feature extraction module 250 is used to extract features through a graph convolutional neural network according to the current improved data to obtain spatiotemporal related features;

[0131] A second feature extraction module 260 is used to capture data periodicity through a two-dimensional convolutional network according to the current improved data to obtain periodic features;

[0132] The traffic flow prediction module 270 is used to concatenate and fuse the spatiotemporal correlation features and the periodic features to obtain fused features; and input the fused features into the fully connected layer to perform traffic flow prediction to obtain a traffic flow prediction result.

[0133] Optionally, the training data acquisition module 210 is further used to:

[0134] At the monitoring points of the regional bridge group, data of passing vehicles is collected based on the dynamic weighing system to obtain data on all types of vehicle flow;

[0135] Based on the traffic flow data of all types of vehicles, the number of vehicles with a total weight greater than a preset weight is recorded to obtain the traffic flow data of heavy-loaded vehicles.

[0136] Optionally, the data dimension upgrading module 220 is further used to:

[0137] Perform data preprocessing on the original traffic data to obtain processed traffic data;

[0138] Divide the time window of the processed traffic flow data according to the preset shortest time period to obtain the time window length;

[0139] Based on the time window length, the processed traffic flow data is segmented according to the preset shortest time period to obtain a set of time series data segments;

[0140] According to the set of time series data fragments, the data are stacked in the same node alignment manner to obtain improved data.

[0141] Optionally, the first feature extraction module 250 is further used to:

[0142] Based on the current improved data, multiple fully connected graphs are constructed using the sliding window method;

[0143] Based on multiple fully connected graphs, the spatiotemporal correlation characteristics are captured through graph convolutional neural networks to obtain spatiotemporal correlation features.

[0144] Optionally, the first feature extraction module 250 is further used to:

[0145] Input the current improved data into a one-dimensional convolutional neural network for feature extraction to obtain the current traffic flow characteristics;

[0146] The similarity of the traffic flow between monitoring points is calculated based on the current improved data to obtain the traffic flow similarity matrix;

[0147] According to the current traffic flow characteristics, the traffic flow similarity matrix and the preset learnable time decay matrix, multiple fully connected graphs are constructed using the sliding window method.

[0148] Optionally, the first feature extraction module 250 is further used to:

[0149] Based on graph convolutional neural networks, monitoring point features are aggregated in multiple fully connected graphs to obtain multiple fully connected graph features;

[0150] Integrate multiple fully connected graph features to obtain spatiotemporal correlation features.

[0151] Optionally, the second feature extraction module 260 is further used to:

[0152] Based on the Inception module, multi-scale convolution processing is performed on the current improved data to obtain multi-scale periodic features;

[0153] In the channel dimension, multi-scale periodic features are spliced ​​to obtain periodic features.

[0154] The present invention proposes a method for predicting the traffic flow of regional bridge groups based on a fully connected graph and dual convolution, converting multidimensional time series data into a three-dimensional model, using graph convolution to capture the spatiotemporal correlation characteristics between multiple measurement point information, and two-dimensional convolution to capture the periodic characteristics of single measurement point information and perform feature splicing, and using the spliced ​​features to complete the final time series prediction task. The present invention processes multivariate time data with high prediction accuracy, strong stability and high precision. The present invention is an accurate and efficient method for predicting the traffic flow of regional bridge groups based on a multidimensional time series of a fully connected graph and dual convolution.

[0155] Figure 3 is a schematic diagram of the structure of a regional bridge group traffic flow prediction device provided by an embodiment of the present invention, such as Figure 3 As shown, the regional bridge group traffic flow prediction device may include the above Figure 2 The regional bridge group traffic flow prediction device based on the fully connected graph and double convolution is shown. Optionally, the regional bridge group traffic flow prediction device 310 may include a first processor 2001.

[0156] Optionally, the regional bridge group vehicle flow prediction device 310 may further include a memory 2002 and a transceiver 2003 .

[0157] The first processor 2001, the memory 2002 and the transceiver 2003 may be connected via a communication bus.

[0158] Combine the following Figure 3 The components of the regional bridge group traffic flow prediction device 310 are specifically introduced:

[0159] The first processor 2001 is the control center of the regional bridge group vehicle flow prediction device 310, which can be a processor or a general term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention, such as one or more microprocessors (digital signal processors, DSPs), or one or more field programmable gate arrays (field programmable gate arrays, FPGAs).

[0160] Optionally, the first processor 2001 may execute various functions of the regional bridge group vehicle flow prediction device 310 by running or executing a software program stored in the memory 2002 and calling data stored in the memory 2002 .

[0161] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 3 CPU0 and CPU1 are shown in FIG.

[0162] In a specific implementation, as an embodiment, the regional bridge group vehicle flow prediction device 310 may also include multiple processors, such as Figure 3 The first processor 2001 and the second processor 2004 are shown in FIG. Each of these processors may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0163] The memory 2002 is used to store the software program for executing the solution of the present invention, and is controlled to be executed by the first processor 2001. The specific implementation method can refer to the above method embodiment, which will not be repeated here.

[0164] Optionally, the memory 2002 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001, or may exist independently, and may be accessed through the interface circuit ( Figure 3 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0165] The transceiver 2003 is used to communicate with a network device or a terminal device.

[0166] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 3 The receiver is used to implement a receiving function, and the transmitter is used to implement a sending function.

[0167] Optionally, the transceiver 2003 may be integrated with the first processor 2001, or may exist independently, and communicate with the first processor 2001 through the interface circuit ( Figure 3 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0168] It should be noted that Figure 3 The structure of the regional bridge group traffic flow prediction device 310 shown in the figure does not constitute a limitation on the router, and the actual knowledge structure recognition device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0169] In addition, the technical effects of the regional bridge group traffic flow prediction device 310 can refer to the technical effects of the regional bridge group traffic flow prediction method based on the fully connected graph and double convolution described in the above method embodiment, which will not be repeated here.

[0170] It should be understood that the first processor 2001 in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0171] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0172] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0173] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0174] In the present invention, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0175] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0176] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0177] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0178] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, 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, such as multiple units or components can be combined or integrated into another device, 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 devices or units, which can be electrical, mechanical or other forms.

[0179] 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.

[0180] 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.

[0181] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0182] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for predicting regional bridge group traffic flow based on fully connected graph and double convolution, characterized in that: The method comprises: At the preset monitoring points of the regional bridge group, the vehicle flow is monitored to obtain the original flow data of the two-dimensional data structure time series; Performing data structure dimensionality upgrading processing on the original flow data to obtain improved data with a three-dimensional data structure; Using the improved data, training the graph convolutional neural network to be trained and the two-dimensional convolutional neural network to be trained to obtain a graph convolutional neural network and a two-dimensional convolutional neural network; Acquire current traffic flow data; perform dimension-upgrading processing on the current traffic flow data to obtain current improved data; According to the current improved data, feature extraction is performed through a graph convolutional neural network to obtain spatiotemporal related features; According to the current improved data, capturing the data periodicity through a two-dimensional convolutional network to obtain periodic features; The spatiotemporal correlation features and the periodic features are concatenated and fused to obtain fused features; the fused features are input into a fully connected layer to perform traffic flow prediction to obtain a traffic flow prediction result.

2. The method for predicting regional bridge group traffic flow based on fully connected graph and double convolution according to claim 1 is characterized in that: The vehicle flow is monitored at the preset monitoring points of the regional bridge group to obtain the original flow data of the two-dimensional data structure time series, including: At the monitoring points of the regional bridge group, data of passing vehicles is collected based on the dynamic weighing system to obtain data on all types of vehicle flow; According to the traffic flow data of all types of vehicles, the number of vehicles with a total weight greater than a preset weight is recorded to obtain the traffic flow data of heavy-loaded vehicles.

3. The method for predicting regional bridge group traffic flow based on fully connected graph and double convolution according to claim 1 is characterized in that: The step of performing data structure dimension-upgrading processing on the original flow data to obtain improved data with a three-dimensional data structure includes: Performing data preprocessing on the original traffic flow data to obtain processed traffic flow data; Dividing the time window of the processed traffic flow data according to a preset shortest time period to obtain a time window length; Based on the time window length, segmenting the processed traffic flow data according to a preset shortest time period to obtain a set of time series data segments; According to the set of time series data fragments, data stacking is performed in a same node alignment manner to obtain improved data.

4. The method for predicting regional bridge group traffic flow based on fully connected graph and double convolution according to claim 1 is characterized in that: The step of extracting features through a graph convolutional neural network based on the current improved data to obtain spatiotemporal related features includes: According to the current improved data, a plurality of fully connected graphs are constructed using a sliding window method; According to the multiple fully connected graphs, the spatiotemporal correlation characteristics are captured through a graph convolutional neural network to obtain spatiotemporal correlation features.

5. The method for predicting regional bridge group traffic flow based on fully connected graph and double convolution according to claim 4 is characterized in that: The step of constructing a fully connected graph using a sliding window method according to the current improved data includes: Inputting the current improved data into a one-dimensional convolutional neural network for feature extraction to obtain current traffic flow features; Calculating the similarity of vehicle flow between monitoring points based on the current improved data to obtain a vehicle flow similarity matrix; According to the current traffic flow characteristics, the traffic flow similarity matrix and a preset learnable time decay matrix, a plurality of fully connected graphs are constructed using a sliding window method.

6. The method for predicting regional bridge group traffic flow based on fully connected graph and double convolution according to claim 4 is characterized in that: The step of capturing spatiotemporal correlation characteristics through a graph convolutional neural network according to the multiple fully connected graphs to obtain spatiotemporal correlation features includes: Based on a graph convolutional neural network, monitoring point features are aggregated in the multiple fully connected graphs to obtain multiple fully connected graph features; The multiple fully connected graph features are integrated to obtain spatiotemporal correlation features.

7. The method for predicting regional bridge group traffic flow based on fully connected graph and double convolution according to claim 1 is characterized in that: According to the current improved data, capturing the data periodicity through a two-dimensional convolutional network to obtain periodic features includes: Based on the Inception module, multi-scale convolution processing is performed on the current improved data to obtain multi-scale periodic features; In the channel dimension, the multi-scale periodic features are spliced ​​to obtain periodic features.

8. A regional bridge group vehicle flow prediction device based on a fully connected graph and dual convolution, the regional bridge group vehicle flow prediction device based on a fully connected graph and dual convolution is used to implement the regional bridge group vehicle flow prediction method based on a fully connected graph and dual convolution as claimed in any one of claims 1 to 7, characterized in that: The device comprises: The training data acquisition module is used to monitor the vehicle flow at the preset monitoring points of the regional bridge group and obtain the original flow data of the two-dimensional data structure time series; A data dimension-upgrading module is used to perform data structure dimension-upgrading processing on the original traffic data to obtain improved data with a three-dimensional data structure; A model training module, used to train the graph convolutional neural network to be trained and the two-dimensional convolutional neural network to be trained using the improved data, to obtain the graph convolutional neural network and the two-dimensional convolutional neural network; The current data acquisition module is used to acquire the current traffic flow data; the current traffic flow data is processed by dimension upgrading to obtain the current improved data; A first feature extraction module, configured to extract features through a graph convolutional neural network according to the current improved data to obtain spatiotemporal related features; A second feature extraction module is used to capture data periodicity through a two-dimensional convolutional network according to the current improved data to obtain periodic features; The vehicle flow prediction module is used to splice and fuse the spatiotemporal correlation features and the periodic features to obtain fused features; input the fused features into the fully connected layer to perform vehicle flow prediction to obtain a vehicle flow prediction result.

9. A regional bridge group traffic flow prediction device, characterized in that: The regional bridge group traffic flow prediction device includes: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program codes, which can be called by a processor to execute the method according to any one of claims 1 to 7.

Images