Improved spatial fuzzy clustering traffic state estimation method and device based on multi-modal fusion

Abstract

The application relates to the technical field of traffic state estimation, and discloses a multi-modal fusion improved spatial fuzzy clustering traffic state estimation method. The method comprises the following steps: obtaining multi-modal road section data of each road section to be analyzed by clustering this time, constructing a graph structure based on the road section data and calculating spatial neighborhood mean values corresponding to the road section data; constructing an FCM objective function according to a weighted sum of the road section data and the spatial neighborhood mean values corresponding to the road section data, taking the minimum of the objective function as an optimization target, and solving to obtain a membership matrix and clustering centers; evaluating a clustering effect, re-determining weights according to the clustering effect, so as to facilitate the calculation of the weighted sum of the road section data and the spatial neighborhood mean values corresponding to the road section data in the next clustering analysis process; and determining a traffic state according to the membership matrix and the clustering centers. The method can make the FCM obtain a better clustering result. The application further discloses a multi-modal fusion improved spatial fuzzy clustering traffic state estimation device.

Description

Technical Field

[0001] This application relates to the field of traffic state estimation technology, and for example to an improved spatial fuzzy clustering traffic state estimation method and apparatus using multimodal fusion. Background Technology

[0002] Accurate understanding and monitoring of traffic conditions is beneficial for traffic management departments to develop timely congestion solutions and improve people's travel experience. However, there is no clear-cut standard between congested and smooth traffic conditions. For example, exceeding a certain standard might be considered congested, while remaining within that standard might be considered smooth. In other words, there is a certain degree of ambiguity between traffic congestion and smooth traffic, which hinders accurate understanding and monitoring of traffic conditions.

[0003] Fuzzy C-means (FCM) clustering algorithm can divide samples into different cluster centers based on their differences. Then, it classifies all samples according to their distance from the cluster centers and uses membership degrees to represent the similarity between a sample and a cluster center. In other words, FCM clustering results have a certain degree of fuzziness, similar to the fuzziness of traffic congestion. FCM provides a solution to the fuzziness of traffic conditions.

[0004] In the process of implementing the embodiments of this application, at least the following problems were found in the related technology:

[0005] Fuzzy weighted index in FCM The value of is related to the clarity of the clustering structure of the sample set. Generally, when the clustering structure of the sample set is clear, the fuzzy weighted index is more effective. Taking a larger value is more appropriate. However, traffic flow is affected by vehicles, pedestrians, and other factors, exhibiting a certain degree of randomness and uncertainty. Therefore, the significance of the clustering structure of traffic flow samples is random and uncertain, making it difficult to adjust the fuzzy weighting index. Obtaining better clustering results makes it difficult to further estimate the traffic conditions more accurately, which is not conducive to traffic management departments accurately understanding and grasping the traffic operation status. Summary of the Invention

[0006] To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments, but rather as a prelude to the detailed description that follows.

[0007] This application provides an improved spatial fuzzy clustering traffic state estimation method and apparatus based on multimodal fusion, so that FCM can still obtain better clustering results when facing traffic flow datasets with randomness and uncertainty, thereby obtaining more accurate estimation results, which makes it easier for traffic management departments to accurately understand and grasp the traffic operation status.

[0008] In some embodiments, the improved spatial fuzzy clustering traffic state estimation method using multimodal fusion includes:

[0009] Obtain multimodal road segment data for each road segment that needs to be clustered for this analysis. The road segment data includes traffic flow, occupancy rate, and speed.

[0010] Using each road segment data as a set element, all road segment data are used to form a vertex set of a graph structure, and the spatial neighborhood mean of each road segment data is calculated based on the adjacency matrix of the graph structure.

[0011] Using each road segment data and its corresponding spatial neighborhood mean as road segment sample points, the distance between the sample point and the cluster center is calculated as follows: Calculate the first distance between the road segment data and the cluster center, and the second distance between the spatial neighborhood mean of the road segment data and the cluster center; use the first weight as the weight of the first distance and the second weight as the weight of the second distance, calculate the weighted sum of the first distance and the second distance, and use the weighted sum as the distance between the sample point and the cluster center; wherein, the sum of the first weight and the second weight is 1;

[0012] The FCM objective function is constructed based on the distance between sample points and cluster centers. The goal is to minimize the objective function and solve for the membership matrix and cluster centers.

[0013] The clustering effect is evaluated, and the first weight is re-determined based on the clustering effect; the better the clustering effect, the smaller the first weight, and the worse the clustering effect, the larger the first weight; the re-determined first weight is used in the next clustering process;

[0014] Traffic status is determined based on the membership matrix and cluster centers.

[0015] Optionally, the objective function Specifically:

[0016]

[0017] in, This represents the total number of road segment data. Total number of cluster centers For the first The nth sample point pair Membership degree of each cluster center For fuzzy weighted index, , The larger the value, the more blurred the clustering effect; As the first weight, As the second weight, For the first The data for the first road segment and the first The first distance between the cluster centers For the first The spatial neighborhood mean of the data for the first road segment and the first The second distance between the cluster centers.

[0018] Optionally, the constraints include: ; .

[0019] Optionally, with the objective function Taking minimization as the optimization objective, the Lagrange multiplier method is used to solve the above constrained problem, and the update formula for the elements in the membership matrix is ​​obtained as follows:

[0020]

[0021] in, For the first The data for the first road segment and the first The first distance between the cluster centers For the first The spatial neighborhood mean of the data for the first road segment and the first... The second distance between the cluster centers.

[0022] Optionally, the update formula for cluster centers is:

[0023]

[0024] in, For the first Cluster centers, For the first Data for each road segment No. Data for each road segment The corresponding spatial neighborhood mean.

[0025] Optionally, the first weight can be re-determined based on the clustering results, including:

[0026] When the clustering effect is represented by the partition coefficient (PC), the first weight is determined as follows:

[0027]

[0028] in, As the first weight, The total number of cluster centers. , This is an adaptive adjustment factor.

[0029] Optionally, the first weight may be re-determined based on the clustering results, including:

[0030] When the clustering effect is represented by the Partition Entropy Coefficient (PE), the first weight is determined as follows:

[0031]

[0032] or,

[0033]

[0034] in, As the first weight, The total number of cluster centers. , This is an adaptive adjustment factor.

[0035] Optionally, the first weight can be re-determined based on the clustering results, including:

[0036] When both the partitioning coefficient PC and the partitioning entropy coefficient PE represent the clustering effect, the first weight is determined as follows:

[0037]

[0038] or,

[0039]

[0040] in, As the first weight, The total number of cluster centers. , These are the weighting coefficients. , This is an adaptive adjustment factor.

[0041] Optionally, the first weight is re-determined based on the clustering effect, including: if the clustering effect is better than the set threshold, the first weight is decreased in a step-by-step manner to obtain the re-determined first weight; if the clustering effect is worse than the set threshold, the first weight is increased in a step-by-step manner to obtain the re-determined first weight.

[0042] Optionally, when this method is executed for the first time, a first weight is initialized based on the predicted traffic condition; wherein, if the predicted traffic condition indicates that the traffic is relatively congested, a larger first weight is set; if the predicted traffic condition indicates that the traffic is relatively smooth, a smaller first weight is set.

[0043] Optionally, the mean of the spatial neighborhood corresponding to each road segment data is calculated based on the adjacency matrix of the graph structure, including:

[0044]

[0045] in, It is the mean matrix of the spatial neighborhood; Let be the degree matrix of the graph structure, and let the first diagonal line be the degree matrix of the graph. The element is related to the first element. The total number of adjacent road segments of each road segment Degree matrix The inverse operation; Given an adjacency matrix, if two road segments are passable, then in the adjacency matrix... The element "1" indicates that if two road segments are impassable, then in the adjacency matrix... Represented by the element "0"; Let be the set of vertices in the graph structure.

[0046] Optionally, before constructing the vertex set of the graph structure using each road segment data as a set element, the improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion further includes normalizing the road segment data in the following manner:

[0047] exist In this case, ;

[0048] exist or In this case, ,or, ;

[0049] in, For a moment The The first of the road segment data One parameter; For the first The data for the first road segment The preset lower quantile of the historical data range for each parameter. , For the first The data for the first road segment The historical minimum value of each parameter For the first In the data of the first road segment The historical maximum value of each parameter; For the first In the data of the first road segment The preset upper limit quantile of the historical data range for each parameter. ; The time after normalization The The first of the road segment data One parameter; For the first The first of the road segment data The average value of the historical normalized results of each parameter; For the first The first of the road segment data The historical average value of each parameter.

[0050] Optionally, using each road segment data as a set element, all road segment data are used to form a vertex set of a graph structure, including: using each normalized road segment data as a set element, all road segment data are used to form a vertex set of a graph structure.

[0051] Optionally, before constructing a vertex set of the graph structure using each road segment data as a set element, the improved spatial fuzzy clustering traffic state estimation method of multimodal fusion further includes: obtaining the maximum and minimum thresholds of each class of parameters in the road segment data; determining the parameter as an anomalous parameter if the parameter in the obtained road segment data is greater than or equal to its corresponding maximum threshold, or less than or equal to its corresponding minimum threshold; and replacing the anomalous parameter with the historical mean of the sliding time window of the parameter.

[0052] Optionally, using each road segment data as a set element, the entire road segment data is used to form a vertex set of the graph structure, including: using each road segment data after replacing abnormal data as a set element, the entire road segment data is used to form a vertex set of the graph structure.

[0053] In some embodiments, the multimodal fusion-based improved spatial fuzzy clustering traffic state estimation device includes an acquisition module, a graph structure construction module, a distance calculation module, a fuzzy clustering and solving module, a feedback module, and an estimation module.

[0054] The acquisition module is used to obtain multimodal road segment data for each road segment that needs to be clustered in this analysis. The road segment data includes traffic flow, occupancy rate, and speed.

[0055] The graph structure construction module is used to construct a vertex set of a graph structure with each road segment data as a set element, and to calculate the spatial neighborhood mean of each road segment data according to the adjacency matrix of the graph structure.

[0056] The distance calculation module uses each road segment data and its corresponding spatial neighborhood mean as road segment sample points, and calculates the distance between the sample point and the cluster center in the following way: calculate the first distance between the road segment data and the cluster center, and the second distance between the spatial neighborhood mean of the road segment data and the cluster center; use the first weight as the weight of the first distance, use the second weight as the weight of the second distance, calculate the weighted sum of the first distance and the second distance, and use the weighted sum as the distance between the sample point and the cluster center; wherein, the sum of the first weight and the second weight is 1;

[0057] The fuzzy clustering and solution module is used to construct the FCM objective function based on the distance between sample points and cluster centers, and to obtain the membership matrix and cluster centers by minimizing the objective function.

[0058] The feedback module is used to evaluate the clustering effect and redetermine the first weight based on the clustering effect; the better the clustering effect, the smaller the first weight, and the worse the clustering effect, the larger the first weight; the redetermined first weight is used for the next clustering process;

[0059] The estimation module is used to determine traffic status based on the membership matrix and cluster centers.

[0060] In some embodiments, the improved spatial fuzzy clustering traffic state estimation apparatus with multimodal fusion includes a processor and a memory storing program instructions, the processor being configured to execute the improved spatial fuzzy clustering traffic state estimation method with multimodal fusion provided in the foregoing embodiments when executing the program instructions.

[0061] The improved spatial fuzzy clustering with multimodal fusion provided in this application can achieve the following technical effects:

[0062] The adjacency matrix of a graph structure can represent the adjacency relationship between vertices. The spatial neighborhood mean of the adjacency matrix is ​​the average value of the road segment data of multiple road segments adjacent to a certain road segment. The spatial neighborhood mean can represent the spatial information around a certain road segment.

[0063] The objective function of FCM is to calculate the distance between a sample point and the cluster center. In the process of calculating this distance, the spatial neighborhood mean is introduced so that in the process of clustering road segment data, not only the characteristics of the data of this road segment are considered, but also the data of the road segments adjacent to this road segment, that is, the spatial information of the road segment is considered.

[0064] In traffic flow, the driving status of each vehicle does not exist in isolation but is influenced by the driving status of surrounding vehicles. For example, in a fast-moving traffic flow, the speed of each vehicle will not be too low; in a slow-moving traffic flow, the speed of each vehicle will not be too high. That is, the road segment data of adjacent road segments have a certain degree of spatial consistency.

[0065] Furthermore, as time goes on, the road segment data of the current road segment will affect the road segment data of the next road segment. For example, if the next road segment is congested, and the current road segment has a high traffic volume, the next road segment will become even more congested over time. That is, there is a certain interaction relationship between the road segment data of adjacent road segments over time.

[0066] Therefore, considering the spatial information of road segments means taking into account the consistency and interaction between adjacent road segments. That is, in the scenario of estimating traffic conditions, the second distance between the spatial neighborhood mean of a road segment's data and the cluster center can reflect the consistency and interaction between adjacent road segments.

[0067] Generally speaking, good clustering results mean good classification results, indicating that the differences between different categories are relatively obvious; poor clustering results mean poor classification results, indicating that the differences between different categories are not obvious enough, and the sample points are more like a random distribution. However, the more obvious the differences between different categories, the more likely it is to lead to "hard" classification (the hardest case being fuzzy weighted index). FCM degenerates into K-means, ignoring the continuity of traffic flow and contradicting the ambiguity of traffic flow.

[0068] The distance between a sample point and the cluster center is calculated by using the weighted sum of the first distance between the road segment data and the cluster center, the spatial neighborhood mean and the second distance in the cluster, and then constructing and solving the FCM objective function.

[0069] The better the clustering effect, the smaller the first weight; the worse the clustering effect, the larger the first weight. If the current clustering effect is better than the previous one, the first weight will be decreased. Since the sum of the first weight and the second weight is 1, decreasing the first weight means increasing the second weight, which means reducing the influence of the road segment data itself on the clustering result and increasing the influence of the spatial neighborhood mean on the clustering result.

[0070] Whether the spatial mean represents the consistency or interaction between adjacent road segments, it also represents ambiguity in the road segment data itself. This reduces the specificity of the road segment data and makes the sample points of adjacent road segments, which are composed of the road segment data itself and the spatial neighborhood mean, exhibit consistency or gradual change. This increases the ambiguity of the sample points, so that the next clustering result of FCM conforms to the continuity and ambiguity of traffic flow, avoiding "hard" clustering and improving the accuracy of the clustering results.

[0071] Meanwhile, since the ambiguity of sample points actually reflects the consistency or gradual change of traffic flow as a whole, and the cluster centers in the FCM clustering results also ignore individual differences and reflect the characteristics of the data as a whole, increasing the ambiguity of sample points in this way will not lose information or reduce the accuracy of the clustering results.

[0072] If the clustering effect is worse than the previous one, it means that the sample points tend to be randomly distributed, and the clustering result is not conducive to classifying traffic conditions. In this case, the first weight is increased while the second weight is decreased. This means increasing the influence of the road segment data itself on the clustering result and reducing the influence of the spatial neighborhood mean on the clustering result. This improves the specificity of the sample points composed of the road segment data itself and the spatial neighborhood mean, making it easier for FCM to capture the commonalities of the sample points in the next clustering process and obtain a better clustering effect.

[0073] In summary, by improving the objective function of FCM with the spatial neighborhood mean and employing a negative feedback mechanism to balance the road segment data itself and the spatial neighborhood mean, the clustering results of the dynamically adjusted FCM can still better reflect the real traffic conditions even when the traffic flow sample set undergoes random and uncertain changes. This leads to a more accurate understanding of traffic conditions, enabling traffic management departments to accurately understand and grasp the traffic operation status, formulate congestion solutions in a timely manner, effectively alleviate traffic congestion, and improve people's travel experience.

[0074] The above general description and the description below are exemplary and illustrative only and are not intended to limit this application. Attached Figure Description

[0075] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are considered similar elements, and wherein:

[0076] Figure 1a , Figure 1b and Figure 1c This is a schematic diagram illustrating the applicable scenarios of the improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in the embodiments of this application;

[0077] Figure 2 This is a flowchart illustrating an improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in an embodiment of this application.

[0078] Figure 3 This is a flowchart illustrating an improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in an embodiment of this application.

[0079] Figure 4 This is a flowchart illustrating an improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in an embodiment of this application.

[0080] Figure 5 This is a schematic diagram of an improved spatial fuzzy clustering traffic state estimation device based on multimodal fusion provided in an embodiment of this application;

[0081] Figure 6This is a schematic diagram of an improved spatial fuzzy clustering traffic state estimation device based on multimodal fusion provided in this application. Detailed Implementation

[0082] To provide a more detailed understanding of the features and technical content of the embodiments of this application, the implementation of the embodiments of this application will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this application. In the following technical description, for ease of explanation, several details are used to provide a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be simplified in their depiction to simplify the drawings.

[0083] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0084] Unless otherwise stated, the term "multiple" means two or more.

[0085] In this embodiment, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0086] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0087] Figure 1a , Figure 1b and Figure 1c This is a schematic diagram illustrating the applicable scenarios of the improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in this application embodiment.

[0088] exist Figure 1a , Figure 1b and Figure 1c In the middle, a detection device 12 is installed on lane 11, which is used to detect road segment data.

[0089] Figure 1a This serves as an example to illustrate the application scenario of a crossroads. Figure 1b This is used as an example to illustrate an application scenario where there are fewer lanes 11 and fewer detection devices 12. Figure 1c This is used as an example to illustrate an application scenario where there are many lanes 11 and many detection devices 12.

[0090] The term "few" can mean less than 5 items / items, less than 6 items / items, less than 7 items / items, less than 8 items / items, less than 9 items / items, less than 10 items / items, less than 20 items / items, etc. The part opposite to "few" is "more".

[0091] In application scenarios at intersections or with a small number of lanes 11 and detection devices 12, the road segment data for each road segment that needs cluster analysis typically includes real-time road segment data and historical road segment data for each road segment. In application scenarios with a large number of lanes 11 and detection devices 12, the data for each road segment that needs cluster analysis may only include real-time road segment data, or it may include both real-time and historical road segment data for each road segment.

[0092] In application scenarios at intersections or with a small number of lanes 11 and detection devices 12, the estimation result of this multimodal fusion improved spatial fuzzy clustering traffic state estimation method is usually the traffic state of each road segment; in application scenarios with a large number of lanes 11 and detection devices 12, the estimation result of this multimodal fusion improved spatial fuzzy clustering traffic state estimation method is usually the traffic state of multiple road segments and / or the overall traffic state.

[0093] Figure 2 This is a flowchart illustrating an improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion, provided in an embodiment of this application. This improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion can be executed on a local computer or a cloud server.

[0094] Combination Figure 2 As shown, the improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion includes:

[0095] S201. Obtain multimodal road segment data for each road segment that needs to be clustered for this analysis.

[0096] The road segment data includes traffic flow. (vehicles / 5 minutes), market share and speed .

[0097] S202. Using each road segment data as a set element, construct a vertex set of the graph structure from all road segment data, and calculate the mean of the spatial neighborhood corresponding to each road segment data according to the adjacency matrix of the graph structure.

[0098] Graph structures can be directed or undirected.

[0099] The adjacency matrix of a graph structure refers to the adjacency relationship between vertices in the graph structure. Each row of the adjacency matrix represents a vertex of the graph structure, and each column of the adjacency matrix also represents a vertex of the graph structure. If two vertices are adjacent, the element at the intersection of the two vertices in the matrix is ​​the first eigenvalue; if two vertices are not adjacent, the element at the intersection of the two vertices in the matrix is ​​the second eigenvalue.

[0100] Specifically, the first eigenvalue can be "1" and the second eigenvalue can be "0". Of course, these two eigenvalues ​​are only illustrative examples, and those skilled in the art can redefine the first and second eigenvalues ​​according to the improvements of the graph structure algorithm.

[0101] Furthermore, in graph structures, "adjacency" includes first-order neighbors and second-order and higher neighbors. Specifically, if two vertices are directly adjacent, they are first-order neighbors; if two vertices are separated by another vertex, they are second-order neighbors; and so on, thus obtaining third-order and higher neighbors.

[0102] In the embodiments of this application, "adjacent" in the adjacency matrix can be a first-order neighbor, a second-order neighbor, or a third-order or higher neighbor.

[0103] Optionally, the mean of the spatial neighborhood corresponding to each road segment data is calculated based on the adjacency matrix of the graph structure, including:

[0104]

[0105] in, It is the mean matrix of the spatial neighborhood; Let be the degree matrix of the graph structure, and let the first diagonal line be the degree matrix of the graph. The element is related to the first element. The total number of adjacent road segments of each road segment Degree matrix The inverse operation; Given an adjacency matrix, if two road segments are passable, then in the adjacency matrix... The element "1" indicates that if two road segments are impassable, then in the adjacency matrix... Represented by the element "0"; Let be the set of vertices in the graph structure.

[0106] Each element , The total number of modes in the road segment data, for example .

[0107] The fact that the two road sections mentioned above are open to traffic is a visual description of the aforementioned "adjacent" relationship. In practical applications, it can be set as follows:

[0108] If two road segments are directly connected, then in the adjacency matrix... Represented by element "1", otherwise represented by "0"; or, if two road segments are directly connected or separated by one or more connected road segments, then in the adjacency matrix... It is represented by the element "1" or "0".

[0109] S203. Using the data of each road segment and its corresponding spatial neighborhood mean as the road segment sample point, calculate the distance between the sample point and the cluster center.

[0110] The distance between a sample point and a cluster center is calculated as follows: the first distance between the road segment data and the cluster center is calculated, and the second distance between the mean of the spatial neighborhood of the road segment data and the cluster center is calculated; the first weight is used as the weight of the first distance, and the second weight is used as the weight of the second distance, and the weighted sum of the first distance and the second distance is used as the distance between the sample point and the cluster center; wherein, the sum of the first weight and the second weight is 1.

[0111] The first distance between the road segment data and the cluster center, and the second distance between the mean of the spatial neighborhood of the road segment data and the cluster center, can be Euclidean distance, Manhattan distance, or Mahalanobis distance.

[0112] During the initial cluster analysis, the first and second weights mentioned above need to be initialized. These weights can be initialized with random numbers; alternatively, after multiple executions of the improved spatial fuzzy clustering traffic state estimation method using multimodal fusion, the first and second weights can be initialized empirically; alternatively, the first weight can be initialized to 0.5 and the second weight to 0.5, or the first weight to 0.6 and the second weight to 0.4, or the first weight to 0.4 and the second weight to 0.6.

[0113] S204. Construct the FCM objective function based on the distance between the sample points and the cluster centers. Take the minimum objective function as the optimization objective and solve to obtain the membership matrix and cluster centers.

[0114] Based on the traditional FCM objective function framework, by incorporating the improved distance between sample points and cluster centers, the objective function can be obtained. for:

[0115]

[0116] This represents the total number of road segment data points, which is usually the same as the total number of road segments. However, if no data is collected for some road segments, the total number of road segment data points will be less than the total number of road segments. Total number of cluster centers For the first The nth sample point pair Membership degree of each cluster center For fuzzy weighted index, , The larger the value, the more blurred the clustering effect; As the first weight, As the second weight, For the first The data for the first road segment and the first The first distance between the cluster centers For the first The spatial neighborhood mean of the data for the first road segment and the first... The second distance between the cluster centers.

[0117] objective function The constraints include: ; .

[0118] Set the maximum number of iterations. and convergence threshold The cluster center matrix is ​​initialized to With objective function The goal is to minimize the number of clusters. The Lagrange multiplier method is used to solve the above-mentioned constrained problem. The solution methods include: updating the membership matrix while fixing the cluster centers, and updating the cluster centers while fixing the membership matrix.

[0119] The update formula for the elements in the membership matrix can be obtained as follows:

[0120]

[0121] in, For the first The data for the first road segment and the first The first distance between the cluster centers For the first The spatial neighborhood mean of the data for the first road segment and the first... The second distance between the cluster centers.

[0122] The formula for updating cluster centers is:

[0123]

[0124] in, For the first Cluster centers, For the first Data for each road segment No. Data for each road segment The corresponding spatial neighborhood mean.

[0125] When the maximum number of iterations is reached Or convergence threshold Then, the membership matrix and cluster centers are obtained.

[0126] S205. Evaluate the clustering effect and redetermine the first weight based on the clustering effect.

[0127] The clustering effect can be evaluated using existing indicators for evaluating FCM clustering.

[0128] The better the clustering effect, the smaller the first weight and the larger the second weight; the worse the clustering effect, the larger the first weight and the smaller the second weight.

[0129] If the clustering effect is good, the reduced first weight can be obtained through a finite number of trials; if the clustering effect is poor, the increased first weight can be obtained through a finite number of trials. Then, the second weight is obtained by subtracting the first weight from 1. In this way, the first and second weights are determined.

[0130] The redefined first and second weights are used in the next clustering process to achieve dynamic adjustment of FCM.

[0131] S206. Determine traffic status based on membership matrix and cluster centers.

[0132] The clustering effect can be evaluated first, and the first weight can be re-determined based on the clustering effect. Then, the traffic status can be determined based on the membership matrix and cluster centers. Alternatively, the traffic status can be determined first based on the membership matrix and cluster centers, and then the clustering effect can be evaluated, and the first weight can be re-determined based on the clustering effect. Figure 2 (Not shown in the image).

[0133] The adjacency matrix of a graph structure can represent the adjacency relationship between vertices. The spatial neighborhood mean of the adjacency matrix is ​​the average value of the road segment data of multiple road segments adjacent to a certain road segment. The spatial neighborhood mean can represent the spatial information around a certain road segment.

[0134] The objective function of FCM is to calculate the distance between a sample point and the cluster center. In the process of calculating this distance, the spatial neighborhood mean is introduced so that in the process of clustering road segment data, not only the characteristics of the data of this road segment are considered, but also the data of the road segments adjacent to this road segment, that is, the spatial information of the road segment is considered.

[0135] In traffic flow, the driving status of each vehicle does not exist in isolation but is influenced by the driving status of surrounding vehicles. For example, in a fast-moving traffic flow, the speed of each vehicle will not be too low; in a slow-moving traffic flow, the speed of each vehicle will not be too high. That is, the road segment data of adjacent road segments have a certain degree of spatial consistency.

[0136] Furthermore, as time goes on, the road segment data of the current road segment will affect the road segment data of the next road segment. For example, if the next road segment is congested, and the current road segment has a high traffic volume, the next road segment will become even more congested over time. That is, there is a certain interaction relationship between the road segment data of adjacent road segments over time.

[0137] Therefore, considering the spatial information of road segments means taking into account the consistency and interaction between adjacent road segments. That is, in the scenario of estimating traffic conditions, the second distance between the spatial neighborhood mean of a road segment's data and the cluster center can reflect the consistency and interaction between adjacent road segments.

[0138] Generally speaking, good clustering results mean good classification results, indicating that the differences between different categories are relatively obvious; poor clustering results mean poor classification results, indicating that the differences between different categories are not obvious enough, and the sample points are more like a random distribution. However, the more obvious the differences between different categories, the more likely it is to lead to "hard" classification (the hardest case being fuzzy weighted index). FCM degenerates into K-means, ignoring the continuity of traffic flow and contradicting the ambiguity of traffic flow.

[0139] The distance between a sample point and the cluster center is calculated by using the weighted sum of the first distance between the road segment data and the cluster center, the spatial neighborhood mean and the second distance in the cluster, and then constructing and solving the FCM objective function.

[0140] The better the clustering effect, the smaller the first weight; the worse the clustering effect, the larger the first weight. If the current clustering effect is better than the previous one, the first weight will be decreased. Since the sum of the first weight and the second weight is 1, decreasing the first weight means increasing the second weight, which means reducing the influence of the road segment data itself on the clustering result and increasing the influence of the spatial neighborhood mean on the clustering result.

[0141] Whether the spatial mean represents the consistency or interaction between adjacent road segments, it also represents ambiguity in the road segment data itself. This reduces the specificity of the road segment data and makes the sample points of adjacent road segments, which are composed of the road segment data itself and the spatial neighborhood mean, exhibit consistency or gradual change. This increases the ambiguity of the sample points, so that the next clustering result of FCM conforms to the continuity and ambiguity of traffic flow, avoiding "hard" clustering and improving the accuracy of the clustering results.

[0142] Meanwhile, since the ambiguity of sample points actually reflects the consistency or gradual change of traffic flow as a whole, and the cluster centers in the FCM clustering results also ignore individual differences and reflect the characteristics of the data as a whole, increasing the ambiguity of sample points in this way will not lose information or reduce the accuracy of the clustering results.

[0143] If the clustering effect is worse than the previous one, it means that the sample points tend to be randomly distributed, and the clustering result is not conducive to classifying traffic conditions. In this case, the first weight is increased while the second weight is decreased. This means increasing the influence of the road segment data itself on the clustering result and reducing the influence of the spatial neighborhood mean on the clustering result. This improves the specificity of the sample points composed of the road segment data itself and the spatial neighborhood mean, making it easier for FCM to capture the commonalities of the sample points in the next clustering process and obtain a better clustering effect.

[0144] In summary, by improving the objective function of FCM with the spatial neighborhood mean and employing a negative feedback mechanism to balance the road segment data itself and the spatial neighborhood mean, the clustering results of the dynamically adjusted FCM can still better reflect the real traffic conditions even when the traffic flow sample set undergoes random and uncertain changes. This leads to a more accurate understanding of traffic conditions, enabling traffic management departments to accurately understand and grasp the traffic operation status, formulate congestion solutions in a timely manner, effectively alleviate traffic congestion, and improve people's travel experience.

[0145] The following uses Euclidean distance, Manhattan distance, and Mahalanobis distance to evaluate the objective function. Further explanation of the membership matrix follows.

[0146] For the objective function :

[0147]

[0148] and each element in the membership matrix :

[0149]

[0150] When using Euclidean distance:

[0151] ;

[0152] in, For the first Data for each road segment; For the first Cluster centers; No. The spatial neighborhood mean of each road segment data point.

[0153] When using Manhattan distance:

[0154] ;

[0155] in, This represents the total number of modes in the road segment data. For the first Data for each road segment The first in One modal parameter, For the first Cluster centers The first in One modal parameter; For the first The spatial neighborhood mean of each road segment data point The first in One modal parameter.

[0156] When using Mahalanobis distance:

[0157] ;

[0158] in, The covariance matrix is ​​the matrix formed by all road segment data and all cluster centers. Covariance matrix The reverse, Let be the covariance matrix of the matrix formed by the mean of all spatial neighborhoods corresponding to the road segment data and the total number of cluster centers. Covariance matrix The reverse.

[0159] In the aforementioned explanation of S205 regarding the re-determination of the first weight based on the clustering effect, it was mentioned that the first weight can be determined through a limited number of trials, and there are other methods besides this.

[0160] Optionally, the first weight can be re-determined based on the clustering results, including:

[0161] If the clustering effect is better than the set threshold, the first weight is decreased stepwise to obtain a newly determined first weight; if the clustering effect is worse than the set threshold, the first weight is increased stepwise to obtain a newly determined first weight.

[0162] This approach balances the mean of the road segment data itself with the mean of its corresponding spatial neighborhood at the expected clustering effect, thereby achieving better clustering results and estimating traffic conditions more accurately.

[0163] Specifically, if the clustering effect is better than the set threshold, the difference between the current first weight and the first preset step value is used as the redefined first weight; if the clustering effect is worse than the set threshold, the sum between the current first weight and the second preset step value is used as the redefined first weight; wherein the first preset step value and the second preset step value may be equal or unequal.

[0164] The magnitudes of the first and second preset step values ​​are related to the average rate of change of traffic flow. The faster the average rate of change of traffic flow, the larger the first and second preset step values; the slower the average rate of change of traffic flow, the smaller the first and second preset step values.

[0165] Specifically, the average rate of change of traffic flow refers to the average rate of change of traffic volume, occupancy rate, and speed.

[0166] Optionally, the first weight is re-determined based on the clustering effect, including: when the clustering effect is represented by the partition coefficient PC, the first weight is determined based on the partition coefficient PC, and the first weight is positively correlated with the partition coefficient PC.

[0167] Optionally, the first weight is re-determined based on the clustering effect, including: when the clustering effect is represented by the partition entropy coefficient PE, the first weight is determined based on the partition entropy coefficient PE, and the first weight is negatively correlated with the partition entropy coefficient PE.

[0168] Optionally, when the clustering effect is represented by both the partition coefficient PC and the partition entropy coefficient PE, the first weight is determined based on the partition coefficient PC and the partition entropy coefficient PE. The first weight is positively correlated with the partition coefficient PC and negatively correlated with the partition entropy coefficient.

[0169] Furthermore, when the clustering effect is represented by the partition coefficient PC, the first weight is determined as follows:

[0170]

[0171] Where, is the first weight. The total number of cluster centers. , The adaptive adjustment coefficients were obtained through specific experiments.

[0172] When the clustering effect is represented by the partition entropy coefficient PE, the first weight is determined as follows:

[0173]

[0174] or,

[0175]

[0176] Where, is the first weight. The total number of cluster centers. , The adaptive adjustment coefficients were obtained through specific experiments.

[0177] This approach allows for the direct balancing of the road segment data itself with the mean of its corresponding spatial neighborhood at the desired clustering effect, enabling FCM to follow traffic flow changes in real time and obtain the clustering results of the desired clustering effect. This makes it easier to accurately estimate traffic conditions.

[0178] Furthermore, when the clustering effect is represented by both the partition coefficient PC and the partition entropy coefficient PE, the first weight is determined as follows:

[0179]

[0180] or,

[0181]

[0182] Where, is the first weight. The total number of cluster centers. , These are the weighting coefficients. , The adaptive adjustment coefficient was obtained experimentally. , and .

[0183] The following is a supplementary explanation of the partition coefficient PC and the partition entropy coefficient PE.

[0184] A larger partition coefficient (PC) indicates a better clustering effect. The partition coefficient PC is calculated as follows:

[0185]

[0186] in, The total number of samples in the cluster. The total number of cluster centers. For the first The nth sample point pair The membership degree of each cluster center.

[0187] A smaller partitioning entropy coefficient (PE) indicates better clustering performance. The clustering entropy coefficient (PE) is calculated as follows:

[0188] ,or, ;

[0189] in, The total number of samples in the cluster. The total number of cluster centers. For the first The nth sample point pair The membership degree of each cluster center.

[0190] The above examples illustrate how to obtain the first weight in continuous cluster analysis. In the first cluster analysis, i.e. the first execution of this method, several implementation methods are also listed in the foregoing embodiments. In addition, the first weight can also be initialized according to the predicted traffic status. If the predicted traffic status indicates that the traffic is relatively congested, a larger first weight is set; if the predicted traffic status indicates that the traffic is relatively smooth, a smaller first weight is set.

[0191] The terms "relatively congested," "relatively smooth," "significantly congested," and "relatively congested" are used to describe a comparison between at least two different traffic conditions. For example, if the first traffic condition is more congested than the second, then the first weight corresponding to the first traffic condition is greater than the first weight corresponding to the second traffic condition.

[0192] Of course, in practical applications, traffic conditions are generally divided into several typical conditions, which can be based on relevant standards and / or expert experience.

[0193] The advantages of adopting the above technical solution are: when traffic is relatively congested, the differences in road segment data are small, and setting a larger first weight is conducive to improving the specificity of road segment data, so as to obtain better clustering results.

[0194] In some application scenarios, a higher primary weight can be set during peak commuting hours and a lower primary weight during off-peak hours.

[0195] The following is an example of the steps to determine traffic status based on the membership matrix and cluster centers.

[0196] Traffic conditions can be divided into four typical states: free-flowing state, steady flow state, congested flow state, and severe congestion state.

[0197] Among them, in free-driving mode, occupancy rate Lower speed Relatively fast, traffic volume Lower.

[0198] Under stable traffic flow conditions, traffic volume occupancy rate in this state Medium speed Relatively fast, traffic volume Relatively high.

[0199] In a congested flow state, as occupancy increases... With the increase in distance, the distance between vehicles decreases, and the speed... Traffic volume dropped sharply Relatively large.

[0200] In a state of severe congestion, occupancy rate High, speed Lowest, traffic volume Relatively large.

[0201] In this way, occupancy rates under typical traffic conditions can be set. ,speed and traffic flow After obtaining the membership matrix and cluster centers, the similarity between the cluster center and the typical traffic state is determined based on the membership matrix and cluster centers. Finally, the current traffic state is estimated based on the similarity and the typical traffic state.

[0202] Alternatively, determining traffic status based on the membership matrix and cluster centers can include:

[0203] The membership matrix and cluster centers are input into the trained neural network, and the output of the neural network is used as the estimated result of traffic state.

[0204] Furthermore, the neural network can be a neural network with Long Short-Term Memory (LSTM) as its core, so as to extract the temporal features of the membership matrix and cluster centers, so as to adapt to the changes in traffic flow over time and thus obtain better estimation results.

[0205] Figure 3 This is a flowchart illustrating an improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in an embodiment of this application.

[0206] Combination Figure 3 As shown, the improved spatial fuzzy clustering traffic state estimation method with multimodal characteristics includes:

[0207] S301. Obtain multimodal road segment data for each road segment that needs to be clustered for this analysis.

[0208] The road segment data includes traffic flow, occupancy rate, and speed.

[0209] S302. Normalize the data for each road segment.

[0210] Normalization methods include:

[0211] exist In this case, ;

[0212] exist or In this case, ,or, ;

[0213] in, For a moment The The first of the road segment data The parameters are the latest parameters that need to be normalized. For the first The data for the first road segment The preset lower quantile of the historical data range for each parameter. , For the first The data for the first road segment The historical minimum value of each parameter For the first The data for the first road segment The historical maximum value of each parameter; For the first The data for the first road segment The preset upper limit quantile of the historical data range for each parameter. ; The time after normalization The The first of the road segment data The parameters are those obtained after the latest normalization process; For the first The first of the road segment data The average value of the historical normalized results of each parameter; For the first The first of the road segment data The historical average of each parameter.

[0214] This normalization process reduces the negative impact of nearby maximum or minimum values, which have a low probability of occurring, on the overall road segment data sample, making the overall road segment data more clustered and more conducive to cluster analysis. At the same time, compared to traditional normalization methods, the denominator... The reduction in the size means that the more clustered overall road segment data is stretched to a larger scale, which further helps FCM to obtain better cluster centers and thus obtain more accurate traffic conditions.

[0215] In some application scenarios, ,For example, It can be 99, 98, 97, 96, 95, 94, 93, 92, 91 or 90. ,For example, It can be 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10.

[0216] S303. Using each normalized road segment data as a set element, construct a vertex set of the graph structure from all road segment data, and calculate the spatial neighborhood mean of each road segment data according to the adjacency matrix of the graph structure.

[0217] S304. Using each normalized road segment data and its corresponding spatial neighborhood mean as road segment sample points, calculate the distance between the sample points and the cluster center.

[0218] The distance between a sample point and a cluster center is calculated as follows: the first distance between the road segment data and the cluster center is calculated, and the second distance between the mean of the spatial neighborhood of the road segment data and the cluster center is calculated; the first weight is used as the weight of the first distance, and the second weight is used as the weight of the second distance, and the weighted sum of the first distance and the second distance is used as the distance between the sample point and the cluster center; wherein, the sum of the first weight and the second weight is 1.

[0219] S305. Construct the FCM objective function based on the distance between the sample points and the cluster centers. Take the minimum objective function as the optimization objective and solve to obtain the membership matrix and cluster centers.

[0220] S306. Evaluate the clustering effect and redetermine the first weight based on the clustering effect.

[0221] The better the clustering effect, the smaller the first weight and the larger the second weight; the worse the clustering effect, the larger the first weight and the smaller the second weight.

[0222] S307. Determine traffic status based on membership matrix and cluster centers.

[0223] Figure 4 This is a flowchart illustrating an improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in an embodiment of this application.

[0224] Combination Figure 4 As shown, the improved spatial fuzzy clustering traffic state estimation method with multimodal characteristics includes:

[0225] S401. Obtain multimodal road segment data for each road segment that needs to be clustered for this analysis.

[0226] The road segment data includes traffic flow, occupancy rate, and speed.

[0227] S402. Obtain the maximum and minimum threshold values ​​for each type of parameter in the road segment data.

[0228] S403. If a parameter in the road segment data obtained this time is greater than or equal to its corresponding maximum threshold, or less than or equal to its corresponding minimum threshold, the parameter shall be identified as an abnormal parameter.

[0229] S404 Replace the outlier parameter with the historical mean of the sliding time window of this parameter.

[0230] For example, the speed in the data of a certain road segment Or, speed Then determine the speed. This is an abnormal parameter, based on the speed data for this road segment. Replace the outlier parameter with the historical mean of the sliding time window.

[0231] Or, traffic flow data for a certain road segment. Then determine the traffic flow. This is an outlier parameter, based on the traffic flow data for that road segment. Replace the outlier parameter with the historical mean of the sliding time window.

[0232] Or, the occupancy rate in the data of a certain road segment. Then determine the occupancy rate. An outlier parameter, defined by its occupancy rate in the data for that road segment. Replace the outlier parameter with the historical mean of the sliding time window.

[0233] S405. Using each road segment data after replacing the abnormal data as a set element, construct a vertex set of the graph structure from all road segment data, and calculate the mean of the spatial neighborhood corresponding to each road segment data according to the adjacency matrix of the graph structure.

[0234] S406. Using the road segment data after replacing the abnormal data and its corresponding spatial neighborhood mean as the road segment sample point, calculate the distance between the sample point and the cluster center.

[0235] The distance between a sample point and a cluster center is calculated as follows: the first distance between the road segment data and the cluster center is calculated, and the second distance between the mean of the spatial neighborhood of the road segment data and the cluster center is calculated; the first weight is used as the weight of the first distance, and the second weight is used as the weight of the second distance, and the weighted sum of the first distance and the second distance is used as the distance between the sample point and the cluster center; wherein, the sum of the first weight and the second weight is 1.

[0236] S407. Construct the FCM objective function based on the distance between the sample points and the cluster centers. Take the minimum objective function as the optimization objective and solve to obtain the membership matrix and cluster centers.

[0237] S408. Evaluate the clustering effect and redetermine the first weight based on the clustering effect.

[0238] The better the clustering effect, the smaller the first weight and the larger the second weight; the worse the clustering effect, the larger the first weight and the smaller the second weight.

[0239] S409. Determine traffic status based on membership matrix and cluster centers.

[0240] This can avoid the adverse effects of abnormal data on the overall road segment data and help to obtain better clustering results.

[0241] By replacing outlier data before normalization, the adverse effects of outlier data constituting extreme values ​​on the normalization process can be avoided, thus leading to better clustering results.

[0242] Figure 5 This is a schematic diagram of an improved spatial fuzzy clustering traffic state estimation device based on multimodal fusion, provided in an embodiment of this application. This improved spatial fuzzy clustering traffic state estimation device can be implemented via software, hardware, or a combination of both.

[0243] Combination Figure 5 As shown, the improved spatial fuzzy clustering traffic state estimation device with multimodal fusion includes an acquisition module 51, a graph structure construction module 52, a distance calculation module 53, a fuzzy clustering and solution module 54, a feedback module 55, and an estimation module 56.

[0244] The acquisition module 51 is used to obtain multimodal road segment data for each road segment that needs to be clustered in this analysis. The road segment data includes traffic flow, occupancy rate and speed.

[0245] The graph structure construction module 52 is used to construct a vertex set of a graph structure by taking each road segment data as a set element, and to calculate the spatial neighborhood mean of each road segment data according to the adjacency matrix of the graph structure.

[0246] The distance calculation module 53 is used to calculate the distance between the sample point and the cluster center by taking each road segment data and its corresponding spatial neighborhood mean as the road segment sample point, and calculating the first distance between the road segment data and the cluster center, and the second distance between the spatial neighborhood mean of the road segment data and the cluster center; using the first weight as the weight of the first distance and the second weight as the weight of the second distance, calculating the weighted sum of the first distance and the second distance, and using the weighted sum as the distance between the sample point and the cluster center; wherein, the sum of the first weight and the second weight is 1.

[0247] The fuzzy clustering and solution module 54 is used to construct the FCM objective function based on the distance between sample points and cluster centers, and to obtain the membership matrix and cluster centers by minimizing the objective function.

[0248] Feedback module 55 is used to evaluate the clustering effect and redetermine the first weight based on the clustering effect; the better the clustering effect, the smaller the first weight, and the worse the clustering effect, the larger the first weight; the redetermined first weight is used for the next clustering process.

[0249] The estimation module 56 is used to determine traffic status based on the membership matrix and cluster centers.

[0250] Optionally, the feedback module 55 includes a first determining unit, a second determining unit, or a third determining unit.

[0251] The first determining unit is used to determine the first weight in the case where the clustering effect is represented by the partition coefficient PC, by means of the following method:

[0252]

[0253] in, As the first weight, The total number of cluster centers. , This is an adaptive adjustment factor.

[0254] The second determining unit is used to determine the first weight in the case where the clustering effect is represented by the partition entropy coefficient PE, by means of the following method:

[0255]

[0256] or,

[0257]

[0258] in, As the first weight, The total number of cluster centers. , This is an adaptive adjustment factor.

[0259] The third determining unit is used to determine the first weight in the case where the clustering effect is simultaneously represented by the partition coefficient PC and the partition entropy coefficient PE, by means of the following:

[0260]

[0261] or,

[0262]

[0263] in, , These are the weighting coefficients. , This is an adaptive adjustment factor.

[0264] Optionally, the feedback module 55 includes a fourth determining unit and a fifth determining unit.

[0265] The fourth determining unit is used to reduce the first weight in a stepwise manner when the clustering effect is better than the set threshold, so as to obtain a newly determined first weight.

[0266] The fifth determining unit is used to increase the first weight in a stepwise manner on the basis of the current first weight when the clustering effect is less than the set threshold, so as to obtain a redefined first weight.

[0267] Optionally, when this method is executed for the first time, a first weight is initialized based on the predicted traffic condition; wherein, if the predicted traffic condition indicates that the traffic is relatively congested, a larger first weight is set; if the predicted traffic condition indicates that the traffic is relatively smooth, a smaller first weight is set.

[0268] Optionally, the graph structure construction module 52 includes a calculation unit, which calculates the spatial neighborhood mean of each road segment data according to the following formula:

[0269]

[0270] in, It is the mean matrix of the spatial neighborhood; Let be the degree matrix of the graph structure, and let the first diagonal line be the degree matrix of the graph. The element is related to the first element. The total number of adjacent road segments of each road segment Degree matrix The inverse operation; Given an adjacency matrix, if two road segments are passable, then in the adjacency matrix... The element "1" indicates that if two road segments are impassable, then in the adjacency matrix... Represented by the element "0"; Let be the set of vertices in the graph structure.

[0271] Optionally, the improved spatial fuzzy clustering traffic state estimation device based on multimodal fusion also includes a normalization processing module.

[0272] The normalization module is used to normalize the data of each road segment before constructing the vertex set of the graph structure by treating each road segment data as a set element, through the following methods:

[0273] exist In this case, ;

[0274] exist or In this case, ,or, ;

[0275] in, For a moment The The first of the road segment data One parameter; For the first The data for the first road segment The preset lower quantile of the historical data range for each parameter. , For the first The data for the first road segment The historical minimum value of each parameter For the first The data for the first road segment The historical maximum value of each parameter; For the first The data for the first road segment The preset upper limit quantile of the historical data range for each parameter. ; The time after normalization The The first of the road segment data One parameter; For the first The first of the road segment data The average value of the historical normalized results of each parameter; For the first The first of the road segment data The historical average of each parameter.

[0276] Optionally, using each road segment data as a set element, the entire road segment data is used to form a vertex set of the graph structure, including: using each normalized road segment data as a set element, the entire road segment data is used to form a vertex set of the graph structure.

[0277] Optionally, the improved spatial fuzzy clustering traffic state estimation device based on multimodal fusion also includes an anomaly data replacement module.

[0278] The abnormal data replacement module includes an acquisition unit, a sixth determination unit, and a replacement unit.

[0279] The acquisition unit is used to obtain the maximum and minimum threshold values ​​of each type of parameter in the road segment data before constructing the vertex set of the graph structure with each road segment data as a set element.

[0280] The sixth determining unit is used to determine a parameter as an abnormal parameter if the parameter in the road segment data obtained this time is greater than or equal to its corresponding maximum threshold, or less than or equal to its corresponding minimum threshold.

[0281] The replacement unit is used to replace the outlier parameter with the historical mean of the sliding time window of that parameter.

[0282] Optionally, using each road segment data as a set element, the entire road segment data is used to form a vertex set of the graph structure, including: using each road segment data after replacing abnormal data as a set element, the entire road segment data is used to form a vertex set of the graph structure.

[0283] In some embodiments, the improved spatial fuzzy clustering traffic state estimation apparatus with multimodal fusion includes a processor and a memory storing program instructions, wherein the processor is configured to execute the improved spatial fuzzy clustering traffic state estimation method with multimodal fusion provided in the foregoing embodiments when executing the program instructions.

[0284] Figure 6 This is a schematic diagram of an improved spatial fuzzy clustering traffic state estimation device based on multimodal fusion provided in this application.

[0285] like Figure 6 As shown, the improved spatial fuzzy clustering traffic state estimation device based on multimodal fusion includes:

[0286] The processor 61 and memory 62 may also include a communication interface 63 and a bus 64. The processor 61, communication interface 63, and memory 62 can communicate with each other via the bus 64. The communication interface 63 can be used for information transmission. The processor 61 can call logical instructions in the memory 62 to execute the improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in the foregoing embodiments.

[0287] Furthermore, the logical instructions in the aforementioned memory 62 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0288] The memory 62, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this application. The processor 61 executes functional applications and data processing by running the software programs, instructions, and modules stored in the memory 62, thereby implementing the methods in the above-described method embodiments.

[0289] The memory 62 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 62 may include high-speed random access memory and may also include non-volatile memory.

[0290] This application provides an intelligent traffic assistance management system, which includes the improved spatial fuzzy clustering traffic state estimation device with multimodal fusion provided in the foregoing embodiments.

[0291] This application provides a computer-readable storage medium storing computer-executable instructions configured to execute the improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in the foregoing embodiments.

[0292] This application provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, which, when executed by a computer, cause the computer to perform the improved spatial fuzzy clustering traffic state estimation method based on multimodal fusion provided in the foregoing embodiments.

[0293] The aforementioned computer-readable storage medium may be a transient computer-readable storage medium or a non-transitory computer-readable storage medium.

[0294] The technical solutions of this application embodiment can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in this application embodiment. The aforementioned storage medium can be a non-transitory storage medium, including: USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, and other media capable of storing program code; it can also be a transient storage medium.

[0295] The foregoing description and accompanying drawings fully illustrate embodiments of this application to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operations may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Additionally, when used in this application, the terms “comprise” and its variations “comprises” and / or “comprising” refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Unless otherwise specified, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes that element. In this document, each embodiment may focus on describing the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, then the relevant parts can be referred to the description of the method section.

[0296] Those skilled in the art will recognize that the units and algorithm steps of the various examples 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 implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0297] The methods and products (including but not limited to devices and equipment) disclosed in the embodiments herein can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units may be merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed units may be through some interfaces, and the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to implement this embodiment according to actual needs. Furthermore, the functional units in the embodiments of this application 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.

[0298] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

Claims

1. An improved spatial fuzzy clustering traffic state estimation method using multimodal fusion, characterized in that, include: Obtain multimodal road segment data for each road segment that needs to be clustered for this analysis. The road segment data includes traffic flow, occupancy rate, and speed. Using each road segment data as a set element, all road segment data are used to form a vertex set of a graph structure, and the spatial neighborhood mean of each road segment data is calculated based on the adjacency matrix of the graph structure. Using each road segment data and its corresponding spatial neighborhood mean as road segment sample points, the distance between the sample point and the cluster center is calculated as follows: Calculate the first distance between the road segment data and the cluster center, and the second distance between the spatial neighborhood mean of the road segment data and the cluster center; use the first weight as the weight of the first distance and the second weight as the weight of the second distance, calculate the weighted sum of the first and second distances, and use the weighted sum as the distance between the sample point and the cluster center; where the sum of the first weight and the second weight is 1; when this method is first executed, the first weight is initialized according to the predicted traffic state; if the predicted traffic state indicates relatively congested traffic, a larger first weight is set; if the predicted traffic state indicates relatively smooth traffic, a smaller first weight is set. The FCM objective function is constructed based on the distance between sample points and cluster centers. The goal is to minimize the objective function and solve for the membership matrix and cluster centers. The clustering effect is evaluated, and the first weight is re-determined based on the clustering effect; the better the clustering effect, the smaller the first weight, and the worse the clustering effect, the larger the first weight; the re-determined first weight is used in the next clustering process; Traffic status is determined based on the membership matrix and cluster centers; Among them, the first weight is re-determined based on the clustering effect, including: When the clustering effect is represented by the partition coefficient PC, the first weight is determined as follows: in, As the first weight, The total number of cluster centers. , For adaptive adjustment coefficients; When the clustering effect is represented by the partition entropy coefficient PE, the first weight is determined as follows: or, in, As the first weight, The total number of cluster centers. , For adaptive adjustment coefficients; When both the partition coefficient PC and the partition entropy coefficient PE represent the clustering effect, the first weight is determined as follows: or, in, As the first weight, The total number of cluster centers. , These are the weighting coefficients. , This is an adaptive adjustment factor.

2. The improved spatial fuzzy clustering traffic state estimation method according to claim 1, characterized in that, objective function Specifically: in, This represents the total number of road segment data. Total number of cluster centers For the first The nth sample point pair Membership degree of each cluster center For fuzzy weighted index, , The larger the value, the more blurred the clustering effect; As the first weight, As the second weight, For the first The data for the first road segment and the first The first distance between the cluster centers, For the first The spatial neighborhood mean of the data for the first road segment and the first The second distance between the cluster centers; The constraints include: ; ; With objective function Taking minimization as the optimization objective, the Lagrange multiplier method is used to solve the above constrained problem, and the update formula for the elements in the membership matrix is ​​obtained as follows: in, For the first The data for the first road segment and the first The first distance between the cluster centers, For the first The spatial neighborhood mean of the data for the first road segment and the first The second distance between the cluster centers; The formula for updating cluster centers is: in, For the first Cluster centers, For the first Data for each road segment No. Data for each road segment The corresponding spatial neighborhood mean.

3. The improved spatial fuzzy clustering traffic state estimation method according to claim 1 or 2, characterized in that, Calculate the spatial neighborhood mean of each road segment data based on the adjacency matrix of the graph structure, including: in, It is the mean matrix of the spatial neighborhood; Let be the degree matrix of the graph structure, and let the first diagonal line be the degree matrix of the graph. The element is related to the first element. The total number of adjacent road segments of each road segment Degree matrix The inverse operation; Given an adjacency matrix, if two road segments are passable, then in the adjacency matrix... The element "1" indicates that if two road segments are impassable, then in the adjacency matrix... Represented by the element "0"; Let be the set of vertices in the graph structure.

4. The improved spatial fuzzy clustering traffic state estimation method according to claim 1 or 2, characterized in that, Before constructing the vertex set of the graph structure using each road segment data as a set element, the process also includes: The data for each road segment were normalized using the following method: exist In this case, ; exist or In this case, ,or, ; in, For a moment The The first of the road segment data One parameter; For the first The data for the first road segment The preset lower quantile of the historical data range for each parameter. , For the first The data for the first road segment The historical minimum value of each parameter For the first The data for the first road segment The historical maximum value of each parameter; For the first The data for the first road segment The preset upper limit quantile of the historical data range for each parameter. ; The time after normalization The The first of the road segment data One parameter; For the first The first of the road segment data The average value of the historical normalized results of each parameter; For the first The first of the road segment data Historical average values ​​of each parameter; Using each road segment data as a set element, the entire road segment data is used to form a vertex set of the graph structure, including: using each normalized road segment data as a set element, the entire road segment data is used to form a vertex set of the graph structure.

5. The improved spatial fuzzy clustering traffic state estimation method according to claim 1 or 2, characterized in that, Before constructing the vertex set of the graph structure using each road segment data as a set element, the process also includes: Obtain the maximum and minimum thresholds for each type of parameter in the road segment data; if a parameter in the obtained road segment data is greater than or equal to its corresponding maximum threshold, or less than or equal to its corresponding minimum threshold, the parameter is identified as an abnormal parameter; replace the abnormal parameter with the historical average of the sliding time window of the parameter. Using each road segment data as a set element, construct a vertex set of the graph structure from all road segment data, including: using each road segment data after replacing abnormal data as a set element, construct a vertex set of the graph structure from all road segment data.

6. A multimodal fusion-based improved spatial fuzzy clustering traffic state estimation device, employing the multimodal fusion-based improved spatial fuzzy clustering traffic state estimation method as described in any one of claims 1 to 5; characterized in that, include: The acquisition module is used to obtain multimodal road segment data for each road segment that needs to be clustered in this analysis. The road segment data includes traffic flow, occupancy rate, and speed. The graph structure construction module is used to construct a vertex set of a graph structure with each road segment data as a set element, and to calculate the spatial neighborhood mean of each road segment data according to the adjacency matrix of the graph structure. The distance calculation module is used to calculate the distance between a sample point and a cluster center, using each road segment data and its corresponding spatial neighborhood mean as road segment sample points, in the following way: Calculate the first distance between the road segment data and the cluster center, and the second distance between the spatial neighborhood mean of the road segment data and the cluster center; use a first weight as the weight of the first distance and a second weight as the weight of the second distance, calculate the weighted sum of the first and second distances, and use this weighted sum as the distance between the sample point and the cluster center; wherein, the sum of the first and second weights is 1; when this method is executed for the first time, the first weight is initialized based on the predicted traffic state; if the predicted traffic state indicates relatively congested traffic, a larger first weight is set; if the predicted traffic state indicates relatively smooth traffic, a smaller first weight is set; The fuzzy clustering and solution module is used to construct the FCM objective function based on the distance between sample points and cluster centers, and to obtain the membership matrix and cluster centers by minimizing the objective function. The feedback module is used to evaluate the clustering effect and redetermine the first weight based on the clustering effect; the better the clustering effect, the smaller the first weight, and the worse the clustering effect, the larger the first weight; the redetermined first weight is used for the next clustering process; The estimation module is used to determine traffic status based on the membership matrix and cluster centers; Among them, the first weight is re-determined based on the clustering effect, including: When the clustering effect is represented by the partition coefficient PC, the first weight is determined as follows: in, As the first weight, The total number of cluster centers. , For adaptive adjustment coefficients; When the clustering effect is represented by the partition entropy coefficient PE, the first weight is determined as follows: or, in, As the first weight, The total number of cluster centers. , For adaptive adjustment coefficients; When both the partition coefficient PC and the partition entropy coefficient PE represent the clustering effect, the first weight is determined as follows: or, in, As the first weight, The total number of cluster centers. , These are the weighting coefficients. , This is an adaptive adjustment factor.

7. An improved spatial fuzzy clustering traffic state estimation device based on multimodal fusion, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to, when executing the program instructions, perform the improved spatial fuzzy clustering traffic state estimation method of multimodal fusion as described in any one of claims 1 to 5.

Images