Traffic flow anomaly detection method and device, computer device and storage medium

Abstract

The application relates to a traffic flow anomaly detection method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a current traffic flow time sequence corresponding to each of a plurality of grid areas included in a to-be-detected area and a current adjacency matrix corresponding to the plurality of grid areas; updating the current adjacency matrix according to the similarity between the current traffic flow time sequences of the grid areas to obtain an updated adjacency matrix; aggregating all current traffic flow time sequences corresponding to elements with a target value in the updated adjacency matrix to obtain an aggregated feature; inputting the aggregated feature into a current variational autoencoder to output a reconstructed aggregated feature; repeating the foregoing steps until the reconstruction error meets a training end condition, thereby obtaining a trained target adjacency matrix and a target variational autoencoder; and performing anomaly detection on the traffic flow of the to-be-detected area based on the target adjacency matrix and the target variational autoencoder. The method can improve the anomaly detection accuracy.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer equipment, storage medium, and computer program product for detecting traffic flow anomalies. Background Technology

[0002] Traffic flow anomaly detection refers to the identification of abnormal situations that deviate from normal traffic patterns by monitoring traffic data. Through traffic flow anomaly detection, traffic congestion, accidents, or other abnormal situations can be detected early.

[0003] Currently, common traffic flow anomaly detection techniques include: statistical methods that use historical traffic data statistics to set thresholds for normal traffic data for detection; spatiotemporal analysis methods that use spatiotemporal graphs to represent traffic flow changes; machine learning methods that use machine learning algorithms to model normal traffic patterns and detect anomalies that deviate from the model; and time series analysis methods that use time series analysis to detect anomalies.

[0004] However, current detection methods only consider abnormal traffic flow on a single road when monitoring the number of vehicles on a road segment. In large-scale and complex traffic networks, they are unable to capture complex nonlinear relationships and have limited processing capabilities, making it impossible to detect abnormal traffic flow across the entire area. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can detect abnormal traffic conditions in complex traffic networks with higher accuracy, in order to address the above-mentioned technical problems.

[0006] Firstly, this application provides a method for detecting abnormal traffic flow, including:

[0007] Obtain the current traffic flow time series corresponding to each of the multiple grid regions included in the area to be detected and the current adjacency matrix corresponding to the multiple grid regions; each element in the current adjacency matrix corresponds to two grid regions, and the element value of each element represents the spatiotemporal correlation between the two grid regions corresponding to each element;

[0008] The current adjacency matrix is ​​updated based on the similarity between the current traffic flow time series of each grid region to obtain the updated adjacency matrix.

[0009] Aggregate all current traffic flow time series corresponding to elements in the updated adjacency matrix whose element value is the target value to obtain aggregated features;

[0010] The aggregated features are input into the current variational autoencoder, and the reconstructed aggregated features are output.

[0011] If the reconstruction error between the reconstructed aggregated features and the aggregated features does not meet the training termination condition, the parameters of the current variational autoencoder are adjusted, and the adjusted variational autoencoder is used as the new current variational autoencoder. The updated adjacency matrix is ​​used as the new current adjacency matrix. The next traffic flow time series corresponding to each grid region is obtained as the new current traffic flow time series. The steps of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region are returned until the obtained reconstruction error meets the training termination condition, and the trained target adjacency matrix and target variational autoencoder are obtained.

[0012] Based on the target adjacency matrix and the target variational autoencoder, anomaly detection is performed on the traffic flow in the area to be detected.

[0013] In one embodiment, updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region to obtain the updated adjacency matrix includes:

[0014] Based on the first similarity between the representation vectors of the traffic flow time series of each grid region, multiple target grid regions are determined from each grid region;

[0015] Based on the multiple target grid regions, construct the first adjacency matrix;

[0016] Based on the second similarity between the temporal characteristics of traffic flow time series of each target grid region, the element values ​​of each element in the first adjacency matrix are determined, and the second adjacency matrix is ​​obtained.

[0017] The current adjacency matrix is ​​replaced with the second adjacency matrix to obtain the updated adjacency matrix.

[0018] In one embodiment, determining multiple target grid regions from the various grid regions based on the similarity between the representation vectors of the traffic flow time series of each grid region includes:

[0019] A target first similarity is determined from the first similarity among the representation vectors of traffic flow time series of each grid region; the target first similarity is greater than the other first similarities among the first similarities.

[0020] Each grid region associated with the first similarity of the target is determined as the target grid region.

[0021] In one embodiment, determining the element values ​​of each element in the first adjacency matrix based on the second similarity between the temporal characteristics of the traffic flow time series of each target grid region, to obtain the second adjacency matrix, includes:

[0022] A target second similarity is determined from the second similarity among the temporal features of the traffic flow time series of each target grid region; the target second similarity is greater than the other second similarities among the various second similarities.

[0023] The element values ​​corresponding to the target grid regions associated with the second similarity of the target in the first adjacency matrix are determined as target values; and the element values ​​corresponding to the other target grid regions associated with the second similarity in the first adjacency matrix are determined as non-target values, thus obtaining a second adjacency matrix.

[0024] In one embodiment, the current variational autoencoder includes an encoder and a decoder;

[0025] The step of inputting the aggregated features into the current variational autoencoder and outputting the reconstructed aggregated features includes:

[0026] The aggregated features are input into the encoder in the current variational autoencoder to obtain the latent variables of the aggregated features in the latent space.

[0027] The latent variables are input into the decoder in the current variational autoencoder to obtain the reconstructed aggregate features.

[0028] In one embodiment, the anomaly detection of traffic flow in the area to be detected based on the target adjacency matrix and the target variational autoencoder includes:

[0029] Identify the associated grid regions in the target adjacency matrix whose element values ​​are the target values, and obtain all traffic flow time series to be detected in the associated grid regions;

[0030] The traffic flow time series to be detected is aggregated to obtain the aggregated features to be detected;

[0031] The aggregated features to be detected are input into the target variational autoencoder to obtain the reconstructed aggregated features to be detected;

[0032] If the reconstruction error between the reconstructed aggregated feature to be detected and the aggregated feature to be detected is greater than a threshold, it is determined that there is an anomaly in the traffic flow of the area to be detected.

[0033] Secondly, this application also provides a traffic flow anomaly detection device, comprising:

[0034] The adjacency matrix acquisition module is used to acquire the current traffic flow time series corresponding to each of the multiple grid regions included in the area to be detected and the current adjacency matrix corresponding to the multiple grid regions; each element in the current adjacency matrix corresponds to two grid regions, and the element value of each element represents the spatiotemporal correlation between the two grid regions corresponding to each element;

[0035] The adjacency matrix update module is used to update the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region, so as to obtain the updated adjacency matrix.

[0036] The aggregation feature acquisition module is used to aggregate all current traffic flow time series corresponding to elements with target values ​​in the updated adjacency matrix to obtain aggregation features;

[0037] The aggregated feature reconstruction module is used to input the aggregated features into the current variational autoencoder and output the reconstructed aggregated features.

[0038] The encoder training module is used to adjust the parameters of the current variational autoencoder when the reconstruction error between the reconstructed aggregated features and the aggregated features does not meet the training termination condition. The adjusted variational autoencoder is used as the new current variational autoencoder, and the updated adjacency matrix is ​​used as the new current adjacency matrix. The module obtains the next traffic flow time series corresponding to each grid region as the new current traffic flow time series. The module returns the steps of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region until the obtained reconstruction error meets the training termination condition, and obtains the trained target adjacency matrix and target variational autoencoder.

[0039] The traffic flow anomaly detection module is used to detect traffic flow anomalies in the area to be detected based on the target adjacency matrix and the target variational autoencoder.

[0040] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0041] Obtain the current traffic flow time series corresponding to each of the multiple grid regions included in the area to be detected and the current adjacency matrix corresponding to the multiple grid regions; each element in the current adjacency matrix corresponds to two grid regions, and the element value of each element represents the spatiotemporal correlation between the two grid regions corresponding to each element;

[0042] The current adjacency matrix is ​​updated based on the similarity between the current traffic flow time series of each grid region to obtain the updated adjacency matrix.

[0043] Aggregate all current traffic flow time series corresponding to elements in the updated adjacency matrix whose element value is the target value to obtain aggregated features;

[0044] The aggregated features are input into the current variational autoencoder, and the reconstructed aggregated features are output.

[0045] If the reconstruction error between the reconstructed aggregated features and the aggregated features does not meet the training termination condition, the parameters of the current variational autoencoder are adjusted, and the adjusted variational autoencoder is used as the new current variational autoencoder. The updated adjacency matrix is ​​used as the new current adjacency matrix. The next traffic flow time series corresponding to each grid region is obtained as the new current traffic flow time series. The steps of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region are returned until the obtained reconstruction error meets the training termination condition, and the trained target adjacency matrix and target variational autoencoder are obtained.

[0046] Based on the target adjacency matrix and the target variational autoencoder, anomaly detection is performed on the traffic flow in the area to be detected.

[0047] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0048] Obtain the current traffic flow time series corresponding to each of the multiple grid regions included in the area to be detected and the current adjacency matrix corresponding to the multiple grid regions; each element in the current adjacency matrix corresponds to two grid regions, and the element value of each element represents the spatiotemporal correlation between the two grid regions corresponding to each element;

[0049] The current adjacency matrix is ​​updated based on the similarity between the current traffic flow time series of each grid region to obtain the updated adjacency matrix.

[0050] Aggregate all current traffic flow time series corresponding to elements in the updated adjacency matrix whose element value is the target value to obtain aggregated features;

[0051] The aggregated features are input into the current variational autoencoder, and the reconstructed aggregated features are output.

[0052] If the reconstruction error between the reconstructed aggregated features and the aggregated features does not meet the training termination condition, the parameters of the current variational autoencoder are adjusted, and the adjusted variational autoencoder is used as the new current variational autoencoder. The updated adjacency matrix is ​​used as the new current adjacency matrix. The next traffic flow time series corresponding to each grid region is obtained as the new current traffic flow time series. The steps of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region are returned until the obtained reconstruction error meets the training termination condition, and the trained target adjacency matrix and target variational autoencoder are obtained.

[0053] Based on the target adjacency matrix and the target variational autoencoder, anomaly detection is performed on the traffic flow in the area to be detected.

[0054] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0055] Obtain the current traffic flow time series corresponding to each of the multiple grid regions included in the area to be detected and the current adjacency matrix corresponding to the multiple grid regions; each element in the current adjacency matrix corresponds to two grid regions, and the element value of each element represents the spatiotemporal correlation between the two grid regions corresponding to each element;

[0056] The current adjacency matrix is ​​updated based on the similarity between the current traffic flow time series of each grid region to obtain the updated adjacency matrix.

[0057] Aggregate all current traffic flow time series corresponding to elements in the updated adjacency matrix whose element value is the target value to obtain aggregated features;

[0058] The aggregated features are input into the current variational autoencoder, and the reconstructed aggregated features are output.

[0059] If the reconstruction error between the reconstructed aggregated features and the aggregated features does not meet the training termination condition, the parameters of the current variational autoencoder are adjusted, and the adjusted variational autoencoder is used as the new current variational autoencoder. The updated adjacency matrix is ​​used as the new current adjacency matrix. The next traffic flow time series corresponding to each grid region is obtained as the new current traffic flow time series. The steps of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region are returned until the obtained reconstruction error meets the training termination condition, and the trained target adjacency matrix and target variational autoencoder are obtained.

[0060] Based on the target adjacency matrix and the target variational autoencoder, anomaly detection is performed on the traffic flow in the area to be detected.

[0061] The aforementioned traffic flow anomaly detection method, apparatus, computer equipment, storage medium, and computer program products acquire traffic flow time-series sequences corresponding to multiple grid regions of the area to be detected and construct a current adjacency matrix based on the relationships between these grid regions. By calculating the similarity between the traffic flow time-series sequences corresponding to each grid region, the current adjacency matrix is ​​updated in real time. This results in more refined traffic flow detection in the area to be detected and a more accurate grasp of global traffic flow changes in complex traffic networks. The updated adjacency matrix, with elements corresponding to the target value, is aggregated and input into a variational autoencoder for reconstruction. The reconstruction error is further obtained, and the adjacency matrix is ​​updated and optimized simultaneously with the goal of optimizing the reconstruction error. Finally, a trained target adjacency matrix and a target variational autoencoder are obtained, which are used for anomaly detection of traffic flow data in the area to be detected. The application of the variational autoencoder improves the accuracy of anomaly detection. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a diagram illustrating the application environment of a traffic flow anomaly detection method in one embodiment.

[0064] Figure 2 This is a flowchart illustrating a traffic flow anomaly detection method in one embodiment;

[0065] Figure 3 This is a flowchart illustrating the steps of updating the adjacency matrix in one embodiment;

[0066] Figure 4 This is a flowchart illustrating the steps for anomaly detection in one embodiment;

[0067] Figure 5 This is a detailed flowchart illustrating the traffic flow anomaly detection method in another embodiment;

[0068] Figure 6 This is a structural block diagram of a traffic flow anomaly detection device in one embodiment;

[0069] Figure 7This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0071] The traffic flow anomaly detection method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Server 104 acquires and processes traffic flow data of the area to be detected collected by terminal 102. Server 104 further acquires the current traffic flow time series corresponding to each of the multiple grid regions included in the processed detection area and the current adjacency matrix corresponding to the multiple grid regions; and updates the current adjacency matrix according to the similarity between the current traffic flow time series of each grid region, obtaining an updated adjacency matrix. Then, server 104 aggregates all current traffic flow time series corresponding to elements in the updated adjacency matrix whose element value is the target value, obtaining aggregated features, and inputs the aggregated features into the current variational autoencoder, outputting reconstructed aggregated features. If the reconstruction error between the reconstructed aggregated features does not meet the training termination condition, server 104 adjusts the parameters of the current variational autoencoder and uses the adjusted variational autoencoder as the new current variational autoencoder. It then uses the updated adjacency matrix as the new current adjacency matrix, obtains the next traffic flow time series corresponding to each grid region as the new current traffic flow time series, and returns the steps of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region. This process continues until the obtained reconstruction error meets the training termination condition, resulting in the trained target adjacency matrix and target variational autoencoder. In application, server 104 performs anomaly detection on the traffic flow in the area to be detected based on the target adjacency matrix and target variational autoencoder. Terminal 102 can be, but is not limited to, various personal computers and laptops. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0072] In one exemplary embodiment, such as Figure 2 As shown, a method for detecting traffic flow anomalies is provided, which can be applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps S202 to S212. Wherein:

[0073] Step 202: Obtain the current traffic flow time series corresponding to each of the multiple grid regions included in the area to be detected and the current adjacency matrix corresponding to the multiple grid regions; each element in the current adjacency matrix corresponds to two grid regions, and the element value of each element represents the spatiotemporal correlation between the two grid regions corresponding to each element.

[0074] The area to be monitored can be the area where traffic flow monitoring is required; the grid area is the area corresponding to the grid obtained after dividing the map of the area to be monitored; the traffic flow time series is the traffic flow data of the area to be monitored, including vehicle inflow and outflow, which is obtained by segmenting the sequence according to the preset time slice.

[0075] The current adjacency matrix can be a dynamic traffic flow relationship graph constructed based on the region to be detected and the current traffic flow data of the region to be detected. It is a data structure used to represent graphs.

[0076] The spatiotemporal correlation can include the temporal correlation and spatial correlation between grid regions. For example, the temporal correlation is obtained by calculating the similarity based on the temporal characteristics corresponding to different time periods of the time series of the grid region, and the spatial correlation is obtained by calculating the similarity based on the spatial characteristics corresponding to the time series of the grid region. It also includes whether the grid regions are adjacent or connected by roads.

[0077] Optionally, the server divides the map of the area to be detected into multiple grid areas, obtains the current traffic flow data corresponding to each grid, and performs sequence segmentation according to a preset time slice to obtain the time series corresponding to each grid area. Based on the corresponding time series, the similarity between each pair of grid areas is calculated to obtain the current adjacency matrix. Each element in the current adjacency matrix corresponds to two grid areas whose similarity has been calculated, and the assignment of the element represents the spatiotemporal relationship between the grid areas.

[0078] Step 204: Update the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region to obtain the updated adjacency matrix.

[0079] In graph theory and network analysis, similarity refers to the degree of similarity between nodes, calculated by examining the similarity between nodes in the adjacency matrix. The adjacency matrix is ​​a matrix representing the structure of a graph, where each element represents the connection between nodes. The similarity of the adjacency matrix can be used to compare the similarity between different nodes, thereby understanding and analyzing the relationships between nodes in a graph structure.

[0080] Optionally, the server calculates the similarity between the current traffic flow time series corresponding to each grid region, and updates the current adjacency matrix by adding or deleting nodes, adding or deleting edges, or updating weights to obtain an updated adjacency matrix.

[0081] Step 206: Aggregate all current traffic flow time series corresponding to elements in the updated adjacency matrix whose element value is the target value to obtain aggregated features.

[0082] The target value can be a specific numerical value of an element in the adjacency matrix, for example, the target value is 1.

[0083] Aggregation can be the process of merging or combining multiple features of original data into a new feature. This new feature can provide higher-level information or better represent the characteristics of the original data.

[0084] Optionally, the server performs aggregation processing on all current traffic flow time series corresponding to elements in the updated adjacency matrix that have the target value, such as elements with a value of 1 in the adjacency matrix, to merge or combine them into a new feature, thus obtaining the aggregated feature.

[0085] Step S208: Input the aggregated features into the current variational autoencoder and output the reconstructed aggregated features.

[0086] Variational Autoencoder (VAE) is a generative model that combines the ideas of autoencoders and probabilistic graphical models. It uses an encoder to map input data to latent variables in a latent space, and a decoder to map the latent variables back to the original data space to generate reconstructed data.

[0087] Optionally, the server inputs the aggregated features into the currently constructed variational autoencoder, and after steps such as mapping, sampling, and reconstruction, outputs the reconstructed aggregated features.

[0088] Step S210: If the reconstruction error between the reconstructed aggregated features does not meet the training termination condition, adjust the parameters of the current variational autoencoder, and use the adjusted variational autoencoder as the new current variational autoencoder. Use the updated adjacency matrix as the new current adjacency matrix, obtain the next traffic flow time series corresponding to each grid region as the new current traffic flow time series, and return to the step of updating the current adjacency matrix according to the similarity between the current traffic flow time series of each grid region until the obtained reconstruction error meets the training termination condition, and obtain the trained target adjacency matrix and target variational autoencoder.

[0089] Reconstruction error, in an autoencoder, refers to the difference or loss between the reconstructed data and the original input data after the latent variables obtained by the encoder are mapped back to the original data space by the decoder. Reconstruction error is a metric for measuring the reconstruction capability of an autoencoder and is typically calculated using a loss function.

[0090] The training termination condition can be that the reconstructed aggregate features output by the variational autoencoder after training and the reconstruction error calculated from the aggregate features are lower than a preset threshold.

[0091] Optionally, if the reconstruction error calculated between the reconstructed aggregated features does not meet the training termination condition, the server calculates the gradient based on the reconstruction error and adjusts the model parameters of the variational autoencoder using the backpropagation algorithm. The adjusted variational autoencoder is used as the new current variational autoencoder, and the updated adjacency matrix is ​​used as the new current adjacency matrix. The traffic flow time series of each grid region corresponding to the next time step at the current time is obtained as the new current traffic flow time series. The server returns to the step above of updating the current adjacency matrix based on the similarity calculated between the new current traffic flow time series. The adjacency matrix and variational autoencoder are iteratively updated until the obtained reconstruction error meets the training termination condition, thus obtaining the trained target adjacency matrix and target variational autoencoder.

[0092] Step S212: Based on the target adjacency matrix and the target variational autoencoder, anomaly detection is performed on the traffic in the area to be detected.

[0093] Among them, anomaly detection can determine whether traffic congestion, traffic accidents, etc. have occurred based on the traffic flow data of the current area to be detected.

[0094] Optionally, the server obtains the current traffic flow data of the area to be detected, and performs anomaly detection on the traffic flow data of the area to be detected based on the trained target adjacency matrix and target variational autoencoder.

[0095] In the aforementioned traffic flow anomaly detection method, traffic flow time series corresponding to multiple grid areas of the area to be detected and a current adjacency matrix constructed based on the relationships between each grid area are obtained. The adjacency matrix is ​​then updated in real time by calculating the similarity between the traffic flow time series corresponding to each grid area. This makes the traffic flow detection of the area to be detected more refined and provides a more accurate grasp of global traffic flow changes in complex traffic network situations. The traffic flow time series corresponding to the target values ​​in the updated adjacency matrix are aggregated and input into a variational autoencoder for reconstruction. The reconstruction error is further obtained, and the adjacency matrix is ​​updated and optimized simultaneously with the goal of optimizing the reconstruction error. Finally, a trained target adjacency matrix and a target variational autoencoder are obtained, which are used for anomaly detection of traffic flow data in the area to be detected. The application of the variational autoencoder improves the accuracy of anomaly detection.

[0096] In one exemplary embodiment, such as Figure 3 As shown, step S204 updates the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region, resulting in the updated adjacency matrix, which includes steps S302 to S308. Wherein:

[0097] Step 302: Based on the first similarity between the representation vectors of traffic flow time series of each grid region, determine multiple target grid regions from each grid region.

[0098] Here, the representation vector refers to the temporal features obtained after extracting the temporal features of the traffic flow data sequence of each grid area, and converting them into an embedded vector representation; the first similarity refers to the similarity calculated pairwise between the embedded vectors corresponding to the traffic flow time series of each grid area.

[0099] Optionally, the server determines multiple target grid regions whose corresponding similarity meets preset conditions based on the similarity calculated between the representation vectors represented by embedded vectors obtained after feature extraction of traffic flow data sequences of each grid region.

[0100] Step 304: Construct the first adjacency matrix based on multiple target grid regions.

[0101] Constructing the first adjacency matrix means determining the target grid region as a node and constructing the first adjacency matrix by using the relationships between the target grid regions as edges.

[0102] Optionally, the server constructs a first adjacency matrix based on multiple determined target grid regions, using the target grid regions as nodes and obtaining the relationships between the target grid regions as edges.

[0103] Step S306: Based on the second similarity between the temporal characteristics of the traffic flow time series of each target grid region, determine the element values ​​of each element in the first adjacency matrix to obtain the second adjacency matrix.

[0104] The second similarity can be calculated by using a graph attention feature extraction algorithm to extract the traffic flow time series of the target grid area, and then updating the embedded vector with the extracted features.

[0105] Optionally, the server uses a graph attention algorithm to extract temporal features based on traffic flow data corresponding to the target grid region, updates the embedded vectors with the extracted temporal features, calculates the similarity between each embedded vector to obtain the second similarity between the target grid regions, and determines the element value of each element in the first adjacency matrix based on the second similarity to obtain the second adjacency matrix.

[0106] Step S308: Replace the current adjacency matrix with the second adjacency matrix to obtain the updated adjacency matrix.

[0107] Optionally, the server replaces the current adjacency matrix corresponding to the previous time step with the second adjacency matrix after assigning values ​​to the elements, to obtain the updated adjacency matrix.

[0108] In this embodiment, by extracting features from the traffic flow time series and calculating the similarity between features, the nodes and edges of the adjacency matrix are determined through two similarity calculations, and the extracted features are converted into embedded vector representations. This enables flexible modeling of the relationships in the entire area to be detected, thereby better capturing the coupling relationship between the area and the road and reducing the reliance on prior knowledge.

[0109] In an exemplary embodiment, step S302 determines multiple target grid regions from each grid region based on the similarity between the representation vectors of traffic flow time series of each grid region, including:

[0110] The target first similarity is determined from the first similarity between the representation vectors of traffic flow time series of each grid area; the target first similarity is greater than the other first similarities in each first similarity; the grid areas associated with the target first similarity are determined as the target grid areas.

[0111] The target grid region can be a grid region with a time relationship.

[0112] Optionally, the server calculates the first similarity between the representation vectors of traffic flow time series of each grid area, sorts them in descending order, and takes the grid areas corresponding to the top N groups of similar traffic flow time series as the target grid areas.

[0113] In this embodiment, the temporal relationship between grid regions is determined by calculating similarity, thereby realizing graph structure modeling corresponding to the temporal relationship.

[0114] In an exemplary embodiment, step S306 determines the element values ​​of each element in the first adjacency matrix based on the second similarity between the temporal characteristics of the traffic flow time series of each target grid region, thereby obtaining the second adjacency matrix, including:

[0115] The target second similarity is determined from the second similarity between the time series features of traffic flow time series of each target grid region; the target second similarity is greater than the other second similarities in each second similarity; the element value corresponding to the target grid region associated with the target second similarity in the first adjacency matrix is ​​determined as the target value; and the element value corresponding to the target grid region associated with other second similarities in the first adjacency matrix is ​​determined as the non-target value, thus obtaining the second adjacency matrix.

[0116] Among them, the target grid regions associated with the second similarity of the target are grid regions with spatial relationships, such as being spatially adjacent or connected by traffic roads.

[0117] Optionally, the server calls a graph attention feature extraction algorithm from the traffic flow time series of each target grid area. First, it uses the basic unit of a recurrent neural network to extract time-series features, and then uses the extracted time-series features to update the embedding vectors of the time-series features. The similarity between the updated embedding vectors is then calculated to obtain a second similarity. The second similarity is sorted to determine the grid areas with spatial relationships. The elements corresponding to the grid areas with spatial relationships in the first adjacency matrix are assigned target values, for example, a target value of 1. The elements corresponding to the grid areas without spatial relationships are assigned non-target values, for example, a non-target value of 0, to obtain the second adjacency matrix.

[0118] In this embodiment, by calculating the similarity between traffic flow time series of grid areas, the temporal and spatial relationships between each grid area are learned, thereby constructing a second adjacency matrix as the spatiotemporal dynamic structure of traffic flow. This enables the modeling of nonlinear relationships in complex road networks, thereby better detecting abnormal traffic conditions in the entire area to be detected.

[0119] In an exemplary embodiment, the current variational autoencoder includes an encoder and a decoder; step S206 inputs the aggregated features into the current variational autoencoder and outputs the reconstructed aggregated features, including:

[0120] The aggregated features are input into the encoder in the current variational autoencoder to obtain the latent variables of the aggregated features in the latent space; the latent variables are input into the decoder in the current variational autoencoder to obtain the reconstructed aggregated features.

[0121] In this context, the latent space refers to the low-dimensional space to which the encoder maps input samples. The latent space is typically a continuous, multi-dimensional space, where each dimension corresponds to a latent variable. A latent variable is a random variable introduced into the encoder to model the input samples. Different values ​​of the latent variable in the latent space correspond to different sample features or attributes. By sampling the latent variables, new samples can be generated, thus enabling the generative capability of the variational autoencoder.

[0122] Optionally, the server inputs the aggregated features into the encoder in the current variational autoencoder. By mapping the aggregated features to latent variables in the latent space, the encoder consists of multiple fully connected layers. The last layer outputs the mean and variance of the latent variables. The latent variables are then input into the decoder of the current variational autoencoder to reconstruct the latent variables and obtain the reconstructed aggregated features.

[0123] In this embodiment, a variational autoencoder is introduced after the input layer of the adjacency matrix to reconstruct the aggregated features, laying the groundwork for subsequent anomaly detection and enhancing the interpretability of the technology.

[0124] In one exemplary embodiment, such as Figure 4 As shown, step S212 performs anomaly detection on traffic flow in the area to be detected based on the target adjacency matrix and the target variational autoencoder, including:

[0125] Step S402: Determine the associated grid regions in the target adjacency matrix whose element values ​​are the target values, and obtain all traffic flow time series to be detected in the associated grid regions.

[0126] Optionally, the server determines the associated grid regions in the target matrix after training whose element values ​​correspond to the target value 1, and obtains the traffic flow time series to be detected for each associated grid region.

[0127] Step S404: Aggregate the traffic flow time series to be detected to obtain the aggregated features to be detected.

[0128] Optionally, the server extracts features from the traffic flow time series to be detected and aggregates the extracted features to obtain aggregated features to be detected.

[0129] Step S406: Input the aggregated features to be detected into the target variational autoencoder to obtain the reconstructed aggregated features to be detected.

[0130] Optionally, the server inputs the aggregated features to be detected into the trained target variational autoencoder, samples the latent variables in the encoder, and then the decoder maps and reconstructs the latent variables to obtain the reconstructed aggregated features to be detected.

[0131] Step S408: If the reconstruction error between the reconstructed aggregated feature to be detected and the aggregated feature to be detected is greater than the threshold, it is determined that there is an anomaly in the traffic flow of the area to be detected.

[0132] Optionally, the server compares and calculates the reconstructed aggregated features to be detected with the reconstructed aggregated features to be detected, and compares the reconstructed error with a preset threshold. If the error is greater than the threshold, it is determined that there is an anomaly in the traffic flow of the area to be detected.

[0133] In this embodiment, anomaly detection is performed on the traffic flow of the region to be detected using the trained adjacency matrix and variational autoencoder. The adjacency matrix enables the model to adaptively adjust the connections between regions, thereby better reflecting changes in regional relationships. Combined with variational autoencoder for anomaly detection, the detection accuracy is further enhanced and the reliability of anomaly detection is improved.

[0134] In one exemplary embodiment, such as Figure 5 As shown, a detailed implementation method for detecting traffic flow anomalies is provided, including steps S502 to S516. Wherein:

[0135] Step S502: Obtain the target area to be detected, divide the map of the target area into several grid areas, each grid representing a sub-area, and obtain traffic flow data for each grid area, including vehicle inflow and outflow. Slice the traffic flow data according to time, establish time series relationships, and abstract the data into a multivariate time series.

[0136] Step S504: Introduce embedding vectors to represent the features of each time series, randomly initialize the embedding vectors, calculate the similarity between the embedding vectors of the multivariate time series, and then select the top P time series with high similarity as the relational sequences (indicating that there is a connection between regions). Here, P can be a hyperparameter, which can be used to adjust the sparsity of the adjacency matrix.

[0137] Step S506: Using the graph attention feature extraction module, the temporal features of the multivariate time series are first extracted using the basic unit of GRU (Gated Recurrent Unit, a variant of recurrent neural network (RNN)). Then, the spatial relationship between the multivariate time series is modeled to learn the potential relationship between regions.

[0138] In step S508, after updating the embedding vectors with the newly extracted temporal features, the similarity between traffic areas is recalculated. These similarities will be used to construct an updated adjacency matrix in the next round of training. In each round of training, the graph structure will be continuously updated, and the embedding vectors will be gradually adjusted to better capture the relationships between traffic areas.

[0139] Step S510: The aggregated representation of the embedding vectors of the multivariate time series is used as the input to the VAE encoder. The VAE encoder maps the input to the mean and variance parameters of the latent space. Latent variables are sampled from the mean and variance parameters obtained from the VAE encoder. These latent variables will contain information from graph structure learning, representing abstract features of traffic area relationships.

[0140] Step S512: Using the latent variables as input, construct the VAE decoder. The decoder's task is to map the latent variables back to the reconstructed aggregate representation of the multivariate time series. Specifically, the reconstruction error between the original aggregate representation of the multivariate time series and the decoded aggregate representation is calculated. This reconstruction error can serve as an indicator of anomaly detection, measuring whether the decoder can recover the original multivariate time series features.

[0141] Step S514: During the training of the VAE encoder, the goal is to minimize the reconstruction error so that the decoder can generate a representation close to the original multivariate time series. Simultaneously, a KL divergence term is added to ensure that the distribution of the latent variables approximates the standard normal distribution.

[0142] Step S516: After training, the trained VAE encoder is used as input to collect the traffic flow time series aggregated features of the area to be detected. The reconstruction result generated by the VAE decoder is then compared with the original data, and anomaly detection is performed based on the reconstruction error. A larger reconstruction error indicates that the data point has anomalous features in the latent space.

[0143] This embodiment enables a more comprehensive capture of traffic flow anomalies in large-scale and complex transportation networks. Through dynamic graph structures, complex nonlinear relationships within the road network can be modeled, thereby better monitoring abnormal traffic flow across the entire region. Furthermore, by learning through embedding vectors and dynamic graph structures, this technique can flexibly learn the complex correlations between traffic areas. This flexibility better captures the coupling relationships between areas and roads, reducing reliance on prior knowledge. Moreover, the dynamic graph structure allows the model to adaptively adjust the connections between areas, thus better reflecting changes in regional relationships. Embedding vectors provide richer feature information; combined with graph attention feature extraction and VAE for anomaly detection, this further enhances detection accuracy and improves the reliability of anomaly detection. By incorporating VAE structures, this technique not only performs anomaly detection but also provides anomaly explanations. VAE structures can map latent variables back to reconstructed multivariate time series representations, thus more clearly explaining why certain areas are detected as anomalies, enhancing the interpretability of the technique.

[0144] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0145] Based on the same inventive concept, this application also provides a traffic flow anomaly detection device 600 for implementing the traffic flow anomaly detection method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more embodiments of the traffic flow anomaly detection device 600 provided below can be found in the limitations of the traffic flow anomaly detection method described above, and will not be repeated here.

[0146] In one exemplary embodiment, such as Figure 6 As shown, a traffic flow anomaly detection device 600 is provided, including: an adjacency matrix acquisition module 602, an adjacency matrix update module 604, an aggregation feature acquisition module 606, an aggregation feature reconstruction module 608, an encoder training module 610, and a traffic flow anomaly detection module 612, wherein:

[0147] The adjacency matrix acquisition module 602 is used to acquire the current traffic flow time series corresponding to each of the multiple grid regions included in the area to be detected and the current adjacency matrix corresponding to the multiple grid regions; each element in the current adjacency matrix corresponds to two grid regions, and the element value of each element represents the spatiotemporal correlation between the two grid regions corresponding to each element.

[0148] The adjacency matrix update module 604 is used to update the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region, so as to obtain the updated adjacency matrix.

[0149] The aggregation feature acquisition module 606 is used to aggregate all current traffic flow time series corresponding to elements with target values ​​in the updated adjacency matrix to obtain aggregate features;

[0150] The aggregated feature reconstruction module 608 is used to input aggregated features into the current variational autoencoder and output reconstructed aggregated features.

[0151] The encoder training module 610 is used to adjust the parameters of the current variational autoencoder when the reconstruction error between the reconstructed aggregate features does not meet the training termination condition. The adjusted variational autoencoder is used as the new current variational autoencoder, and the updated adjacency matrix is ​​used as the new current adjacency matrix. The next traffic flow time series corresponding to each grid region is obtained as the new current traffic flow time series. The steps of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region are returned until the obtained reconstruction error meets the training termination condition, and the trained target adjacency matrix and target variational autoencoder are obtained.

[0152] The traffic flow anomaly detection module 612 is used to detect anomalies in traffic flow in the area to be detected based on the target adjacency matrix and the target variational autoencoder.

[0153] Furthermore, in one embodiment, the adjacency matrix update module 604 is further configured to: determine multiple target grid regions from each grid region based on a first similarity between the representation vectors of the traffic flow time series of each grid region; construct a first adjacency matrix based on the multiple target grid regions; determine the element value of each element in the first adjacency matrix based on a second similarity between the temporal features of the traffic flow time series of each target grid region to obtain a second adjacency matrix; and replace the current adjacency matrix with the second adjacency matrix to obtain an updated adjacency matrix.

[0154] Furthermore, in one embodiment, the adjacency matrix update module 604 is also used to determine a target first similarity from the first similarity between the representation vectors of traffic flow time series of each grid region; the target first similarity is greater than the other first similarities among the first similarities; and each grid region associated with the target first similarity is determined as the target grid region.

[0155] Furthermore, in one embodiment, the adjacency matrix update module 604 is further configured to determine a target second similarity from the second similarity between the temporal features of the traffic flow time series of each target grid region; the target second similarity is greater than the other second similarities among the second similarities; the element value corresponding to the target grid region associated with the target second similarity in the first adjacency matrix is ​​determined as the target value; and the element value corresponding to the target grid region associated with the other second similarities in the first adjacency matrix is ​​determined as the non-target value, thereby obtaining the second adjacency matrix.

[0156] Furthermore, in one embodiment, the aggregated feature reconstruction module 608 is further configured to input the aggregated features into the current variational autoencoder and output the reconstructed aggregated features, including:

[0157] The aggregated features are input into the encoder in the current variational autoencoder to obtain the latent variables of the aggregated features in the latent space; the latent variables are input into the decoder in the current variational autoencoder to obtain the reconstructed aggregated features.

[0158] Furthermore, in one embodiment, the traffic flow anomaly detection module 612 is also used to determine the associated grid regions in the target adjacency matrix whose element values ​​are the target values, obtain all traffic flow time series to be detected in the associated grid regions; aggregate the traffic flow time series to be detected to obtain aggregated features to be detected; input the aggregated features to be detected into the target variational autoencoder to obtain the reconstructed aggregated features to be detected; if the reconstruction error between the reconstructed aggregated features to be detected and the aggregated features to be detected is greater than a threshold, it is determined that there is an anomaly in the traffic flow of the area to be detected.

[0159] Each module in the aforementioned traffic flow anomaly detection device 600 can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0160] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores traffic flow data for the area to be detected. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a traffic flow anomaly detection method.

[0161] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0162] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0163] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0164] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0165] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0166] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0167] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0168] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for detecting abnormal traffic flow, characterized in that, The method includes: Obtain the current traffic flow time series corresponding to each of the multiple grid regions included in the area to be detected and the current adjacency matrix corresponding to the multiple grid regions; each element in the current adjacency matrix corresponds to two grid regions, and the element value of each element represents the spatiotemporal correlation between the two grid regions corresponding to each element; The current adjacency matrix is ​​updated based on the similarity between the current traffic flow time series of each grid region to obtain the updated adjacency matrix. Aggregate all current traffic flow time series corresponding to elements in the updated adjacency matrix whose element value is the target value to obtain aggregated features; The aggregated features are input into the current variational autoencoder, and the reconstructed aggregated features are output. If the reconstruction error between the reconstructed aggregated features and the aggregated features does not meet the training termination condition, the parameters of the current variational autoencoder are adjusted, and the adjusted variational autoencoder is used as the new current variational autoencoder. The updated adjacency matrix is ​​used as the new current adjacency matrix. The next traffic flow time series corresponding to each grid region is obtained as the new current traffic flow time series. The steps of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region are returned until the obtained reconstruction error meets the training termination condition, and the trained target adjacency matrix and target variational autoencoder are obtained. Based on the target adjacency matrix and the target variational autoencoder, anomaly detection is performed on the traffic flow in the area to be detected.

2. The method according to claim 1, characterized in that, The step of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region to obtain the updated adjacency matrix includes: Based on the first similarity between the representation vectors of the traffic flow time series of each grid region, multiple target grid regions are determined from each grid region; Based on the multiple target grid regions, a first adjacency matrix is ​​constructed; Based on the second similarity between the temporal characteristics of traffic flow time series of each target grid region, the element values ​​of each element in the first adjacency matrix are determined, and the second adjacency matrix is ​​obtained. The current adjacency matrix is ​​replaced with the second adjacency matrix to obtain the updated adjacency matrix.

3. The method according to claim 2, characterized in that, The step of determining multiple target grid regions from the various grid regions based on the first similarity between the representation vectors of the traffic flow time series of each grid region includes: A target first similarity is determined from the first similarity among the representation vectors of traffic flow time series of each grid region; the target first similarity is greater than the other first similarities among the first similarities. Each grid region associated with the first similarity of the target is determined as the target grid region.

4. The method according to claim 2, characterized in that, The step of determining the element values ​​of each element in the first adjacency matrix based on the second similarity between the temporal characteristics of the traffic flow time series of each target grid region, and obtaining the second adjacency matrix, includes: A target second similarity is determined from the second similarity among the temporal features of the traffic flow time series of each target grid region; the target second similarity is greater than the other second similarities among the various second similarities. The element values ​​corresponding to the target grid regions associated with the second similarity of the target in the first adjacency matrix are determined as target values; and the element values ​​corresponding to the other target grid regions associated with the second similarity in the first adjacency matrix are determined as non-target values, thus obtaining a second adjacency matrix.

5. The method according to claim 1, characterized in that, The current variational autoencoder includes an encoder and a decoder; The step of inputting the aggregated features into the current variational autoencoder and outputting the reconstructed aggregated features includes: The aggregated features are input into the encoder in the current variational autoencoder to obtain the latent variables of the aggregated features in the latent space. The latent variables are input into the decoder in the current variational autoencoder to obtain the reconstructed aggregate features.

6. The method according to claim 1, characterized in that, The anomaly detection of traffic flow in the area to be detected based on the target adjacency matrix and the target variational autoencoder includes: Identify the associated grid regions in the target adjacency matrix whose element values ​​are the target values, and obtain all traffic flow time series to be detected in the associated grid regions; The traffic flow time series to be detected is aggregated to obtain the aggregated features to be detected; The aggregated features to be detected are input into the target variational autoencoder to obtain the reconstructed aggregated features to be detected; If the reconstruction error between the reconstructed aggregated feature to be detected and the aggregated feature to be detected is greater than a threshold, it is determined that there is an anomaly in the traffic flow of the area to be detected.

7. A traffic flow anomaly detection device, characterized in that, The device includes: The adjacency matrix acquisition module is used to acquire the current traffic flow time series corresponding to each of the multiple grid regions included in the area to be detected and the current adjacency matrix corresponding to the multiple grid regions; each element in the current adjacency matrix corresponds to two grid regions, and the element value of each element represents the spatiotemporal correlation between the two grid regions corresponding to each element; The adjacency matrix update module is used to update the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region, so as to obtain the updated adjacency matrix. The aggregation feature acquisition module is used to aggregate all current traffic flow time series corresponding to elements with target values ​​in the updated adjacency matrix to obtain aggregation features; The aggregated feature reconstruction module is used to input the aggregated features into the current variational autoencoder and output the reconstructed aggregated features. The encoder training module is used to adjust the parameters of the current variational autoencoder when the reconstruction error between the reconstructed aggregated features and the aggregated features does not meet the training termination condition. The adjusted variational autoencoder is used as the new current variational autoencoder, and the updated adjacency matrix is ​​used as the new current adjacency matrix. The module obtains the next traffic flow time series corresponding to each grid region as the new current traffic flow time series. The module returns the steps of updating the current adjacency matrix based on the similarity between the current traffic flow time series of each grid region until the obtained reconstruction error meets the training termination condition, and obtains the trained target adjacency matrix and target variational autoencoder. The traffic flow anomaly detection module is used to detect traffic flow anomalies in the area to be detected based on the target adjacency matrix and the target variational autoencoder.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images