A train operation path identification method, device, equipment and medium
By using a graph neural network model and a dual-index path selection method, the problem of train path mismatch in complex road networks was solved, achieving high-precision and high-efficiency train path identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN NAT HIGH-TECH IND INNOVATION CENT
- Filing Date
- 2026-04-27
- Publication Date
- 2026-07-21
AI Technical Summary
In complex road network scenarios with multiple parallel lines and densely branched hub stations, existing technologies are prone to train path mismatch issues.
By employing a graph neural network model, train operation data and train network vector GIS data are acquired, and an improved graph neural network model is used for feature extraction and path matching. The model introduces dual indicators of matching probability and confidence to filter paths, and combines data comparison and visualization for verification, thus forming a standardized train operation path identification process.
It improves the accuracy and efficiency of path identification, reduces the risk of path misjudgment, and achieves automated batch identification and reliable train operation path identification.
Smart Images
Figure CN122078472B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of rail transit technology, and in particular to a method, apparatus, equipment, and medium for identifying train routes. Background Technology
[0002] Accurate identification and map matching of train routes are fundamental to optimizing rail transit scheduling, assessing operational efficiency, guiding passenger travel, and handling emergencies. With the increasing density of train networks, the growing complexity of train operation plans, and the rising demand for accurate travel information from passengers, there is an urgent need for a technological method that can automatically, efficiently, and accurately match actual train routes based on multi-source data. Currently, in the rail transit industry, train route identification and map matching technologies are mainly divided into three categories: route positioning technology based on track circuits and signaling systems, train trajectory tracking technology based on satellite positioning, and route matching algorithms based on multi-source data fusion. The latter two are more widely used technologies, both using "data acquisition-preprocessing-map matching" as the core technical link. They rely on multi-source data fusion and algorithm calculation to complete train route matching, providing fundamental support for train scheduling and route visualization. Summary of the Invention
[0003] This invention provides a train route identification method to solve the problem of route mismatch that easily occurs in complex road network scenarios with multiple parallel lines and dense branching of hub stations in the existing technology.
[0004] In a first aspect, the present invention provides a train route identification method, the method comprising: Acquire train operation data and train network vector GIS data; the train operation data includes train number, stop sequence, departure and arrival time, and interval running time; the train network vector GIS data includes station latitude and longitude, station level, route alignment, route length, and network topology. The train operation data and train network vector GIS data are input into a pre-trained graph neural network model, which outputs feature data. The feature data includes the latitude and longitude trajectory of the train matching path, the operation parameters of each section, the path matching probability, and the confidence level. The feature data is associated and matched with a preset train network, and the train's corresponding operating path is determined based on the latitude and longitude trajectory with the highest matching probability and confidence level meeting a preset threshold.
[0005] Secondly, the present invention provides a device for identifying train routes based on graph neural networks, comprising: The data acquisition unit is used to acquire train operation data and train network vector GIS data; the train operation data includes train number, stop sequence, departure and arrival time, and interval running time; the train network vector GIS data includes station latitude and longitude, station level, line direction, line length, and road network topology. The feature extraction unit is used to input the train operation data and train network vector GIS data into a pre-trained graph neural network model and output feature data; the feature data includes the latitude and longitude trajectory of the train matching path, the operation parameters of each section, the path matching probability and confidence level; The path determination unit is used to associate and match the feature data with a preset train network, and determine the train's corresponding operating path based on the latitude and longitude trajectory with the highest matching probability and confidence level that meets a preset threshold.
[0006] Thirdly, an electronic device is provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a program stored in memory, it implements the steps of the train route identification method according to any embodiment of the first aspect.
[0007] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the train route identification method as described in any embodiment of the first aspect.
[0008] The technical solutions provided in the embodiments of the present invention have the following advantages compared with the prior art: Employing a graph neural network model as the core computational carrier, this method is more effective than traditional graph theory algorithms and conventional machine learning algorithms in uncovering the correlation between train operation sequence features and road network spatial features, thereby improving the accuracy of path identification. Using multi-source data as input, feature data as intermediate carrier, and latitude and longitude trajectories as the final basis, a standardized train operation path identification process is formed. The operational logic is clear, enabling automated batch identification and improving path identification efficiency. Furthermore, the introduction of both matching probability and confidence level as dual indicators for path selection effectively avoids path misjudgment caused by single-indicator judgment, enhancing the reliability of train operation path identification results. Attached Figure Description
[0009] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0011] Figure 1 This is a flowchart illustrating a train route identification method provided in an embodiment of the present invention. Figure 2 An example diagram of train operation data provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a train route identification device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation
[0012] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0013] Example Figure 1 This is a flowchart illustrating a train route identification method provided in an embodiment of the present invention. Specifically, this embodiment proposes a train route identification method; see [link to documentation]. Figure 1 The train route identification method includes the following steps S101-S103.
[0014] S101, acquire train operation data and train network vector GIS data; the train operation data includes train number, stop sequence, departure and arrival time, and interval running time; the train network vector GIS data includes station latitude and longitude, station level, line direction, line length, and road network topology.
[0015] See Figure 2 , Figure 2This is an example diagram of train operation data provided in an embodiment of the present invention. Train operation data refers to relevant data on the basic status of train operation, including train number, stop sequence, departure and arrival times, and interval running time. It is the core temporal feature data for identifying train operation paths. Train network vector GIS data is train network vector data based on geographic information systems. It includes spatial feature data such as station latitude and longitude, station level, line direction, line length, and network topology, and is the basis for constructing network models and matching train physical paths.
[0016] S102, the train operation data and train network vector GIS data are input into a pre-trained graph neural network model, and feature data is output; the feature data includes the latitude and longitude trajectory of the train matching path, the operation parameters of each section, the path matching probability and confidence level.
[0017] Feature data is the output data of the graph neural network model after processing train operation data and train network vector GIS data. It includes the latitude and longitude trajectory of the train matching path, the operation parameters of each section, the path matching probability and confidence level, and is the direct basis for determining the actual train operation path.
[0018] In one embodiment, the construction and training steps of the graph neural network model are specifically steps S201-S204: S201, Construct an improved graph neural network model; the graph neural network model is an improvement on the traditional graph neural network structure, embedding a spatial-feature dual attention collaborative module, adding a cross-modal feature fusion module, and the output layer is a three-level structure of candidate path generation-probability calculation-confidence screening.
[0019] In a specific embodiment, the traditional graph neural network is structurally improved by embedding a spatial-feature dual attention collaborative module in the message passing layer, adding a cross-modal feature fusion module, and optimizing the model output layer into a three-level structure of "candidate path generation - probability calculation - confidence screening", thus obtaining an improved graph neural network model.
[0020] In one embodiment, the spatial-feature dual attention collaborative module includes a spatial attention mechanism and a feature attention mechanism; the spatial attention mechanism constructs a dynamic association weight matrix based on the station's latitude and longitude coordinates and the road network topology distance; the feature attention mechanism quantifies the contribution of node and edge features to path matching through an adaptive weight allocation algorithm, and sets a dynamic gain factor for the interval average travel speed feature.
[0021] In a specific embodiment, the spatial attention mechanism is used to construct a dynamic association weight matrix based on distance values, setting higher association weights for station nodes in hub stations and parallel line areas to strengthen the association strength between nodes in these areas, and setting lower association weights for isolated stations to weaken their interference with model calculations; the feature attention mechanism is used to input node features (station level, number of associated lines) and edge features (line length, average travel speed in sections, parallel line identifiers) into the algorithm. The algorithm automatically quantifies the contribution of each feature and assigns corresponding weights based on the degree of influence of the features on path matching; at the same time, a dynamic gain factor is set for the average travel speed in sections, and the weight coefficient of this feature is dynamically adjusted according to the characteristics of the line section in which the train runs, focusing on capturing the speed adaptation characteristics of the train in different line sections.
[0022] The dual attention mechanism works in tandem to achieve precise linkage and control of node association and feature weights, enabling the model to accurately capture the spatial topological features of the road network and focus on the key features of train operation, thus doubly improving the model's feature recognition capabilities.
[0023] In one embodiment, the cross-modal feature fusion module adopts a three-level architecture of feature alignment, deep fusion, and dimensionality normalization. The cross-modal feature fusion module completes the dimensionality calibration, feature extraction, and adaptive fusion of train running sequence features and road network spatial features through the three-level architecture, and generates a fused feature vector with a unified dimension.
[0024] In a specific embodiment, feature alignment is used to perform time-space dimension calibration between temporal features such as stop sequences and interval running times in train operation data and node / edge spatial features in train network vector GIS data, ensuring that heterogeneous features maintain dimensional consistency and eliminating the impact of dimensional bias on feature fusion. Deep fusion is used to input the calibrated train operation temporal features into a convolutional neural network (CNN) to extract local correlation information of the temporal features; and to input the calibrated road network spatial features into a graph convolutional layer to extract global topological information of the road network features. Subsequently, the two types of extracted features are input into an attention gating unit, which adaptively fuses them according to feature importance while filtering redundant noise in heterogeneous data and retaining core feature information. Dimensional normalization is used to batch normalize the fused features output by the attention gating unit, eliminating differences in feature dimensions, and then converting the fused features into a unified dimension using a dimensionality compression algorithm, ultimately generating a unified dimensional fused feature vector, which is then input into subsequent computational layers of the graph neural network.
[0025] A three-level architecture is adopted to achieve cross-modal feature fusion, breaking through the traditional fusion method of simple field association, realizing the deep fusion of train operation sequence features and road network spatial features, and fully exploring the deep correlation features of heterogeneous data; The feature alignment process eliminates the dimensional bias of heterogeneous features, laying the foundation for subsequent deep fusion and reducing fusion bias caused by dimensional inconsistency. CNN and graph convolutional layers are used to extract local information of temporal features and global information of spatial features, respectively. Then, attention gating units are used for adaptive fusion, so that the fused features contain both the local temporal patterns of train operation and the global topological features of the road network, thereby improving the completeness and effectiveness of the features. The dimension normalization step eliminates differences in dimensions and unifies feature dimensions, adapting to the subsequent computational needs of the graph neural network. At the same time, it filters redundant noise, improves the purity of fused features, and further enhances the accuracy of model computation.
[0026] In one embodiment, a path confidence evaluation factor is introduced into the three-level structure of the output layer; the path confidence evaluation factor is calculated based on the interval average speed fit and topology fit, and is used to correct the matching probability of candidate paths.
[0027] In a specific embodiment, the interval average speed fit is the similarity between the interval average speed of the model-matched path and the actual interval average speed of the train, and the topology fit is the fit between the road network topology of the model-matched path and the train stop sequence. Preset weights are assigned to the interval average speed fit and the topology fit (e.g., speed fit weight 0.6, topology fit weight 0.4), and the confidence evaluation factor of each candidate path is calculated by weighted summation, with the formula: Path confidence evaluation factor = Interval average speed fit × 0.6 + Topology fit × 0.4. Probability correction refers to multiplying the calculated path confidence evaluation factor by the initial matching probability of the candidate path to correct the initial matching probability and obtain the corrected matching probability. The model sorts the candidate paths according to the corrected matching probability.
[0028] The weights of the confidence assessment factors can be dynamically adjusted according to the actual application scenario, thereby improving the flexibility and scenario adaptability of the model.
[0029] S202, generate training sample pairs and divide the dataset; the training sample pairs are generated by combining historical high-speed rail train operation data, corresponding open-source high-speed rail network vector GIS data and standardized high-speed rail network map model, and the dataset is divided into training set and test set according to a preset ratio.
[0030] In a specific embodiment, historical train operation data and corresponding open-source train network vector GIS data are selected, and combined with a pre-built standardized train network map model, train operation features and network map features are extracted. The actual train routes are manually labeled as tags to generate training sample pairs of "train operation features - network map features - actual route tags". The training sample pairs are divided into training set and test set according to a preset ratio of 7:3.
[0031] S203 iteratively trains the improved graph neural network model; initial model parameters are set during training, the cross-entropy loss function is used as the optimization objective, the Adam optimizer is used to iteratively update the parameters, and an early stopping mechanism and regularization are introduced to prevent the model from overfitting.
[0032] In a specific embodiment, the initial parameters of the improved graph neural network model are set, with the learning rate set to 0.001, the number of iterations set to 100 rounds, and the hidden layer dimension set to 256 dimensions. The cross-entropy loss function is used as the model optimization target, and the Adam optimizer is used to iteratively update the model parameters. During the training process, an early stopping mechanism (stopping if the test set loss does not decrease for 5 consecutive rounds) and L2 regularization are introduced to prevent the model from overfitting.
[0033] S204, iteratively optimize the model parameters to obtain the trained improved graph neural network model; optimize using path matching accuracy and average matching time as evaluation indicators until the model performance reaches the preset standard.
[0034] In a specific embodiment, path matching accuracy and average matching time are used as model performance evaluation indicators. Test set data is input into the trained model for performance testing. Training samples are supplemented for low-accuracy scenarios such as parallel lines and hub stations. The model attention weights and feature fusion coefficients are adjusted, and the model is trained iteratively again until the path matching accuracy and average matching time of the model reach the preset application standard, thus obtaining the improved graph neural network model after training.
[0035] Once trained, the graph neural network model can be directly called upon, and incremental training is only performed when there are significant changes in the road network or traffic operation plan. The model has low maintenance costs and high reusability.
[0036] S103, the feature data is associated and matched with the preset train network, and the train's corresponding operating path is determined based on the latitude and longitude trajectory with the highest matching probability and confidence level meeting the preset threshold.
[0037] In one embodiment, step S103 above includes a verification step of the opening path, specifically: The proposed route is compared with the train trajectory data of the preset train dispatching system, and the proposed route is overlaid on an open-source map for visualization.
[0038] In a specific embodiment, data comparison verification refers to comparing the train's trajectory with the data in the preset train dispatching system segment by segment to verify the consistency of the path's latitude and longitude trajectory, section operation parameters, and dispatching system data, and to investigate issues such as path deviation and section mismatch. Visualization verification refers to overlaying the determined train route onto an open-source map for spatial visualization. Through manual verification or automated image recognition, it is verified whether the route is consistent with the actual train network, and issues such as mismatch of parallel lines and mismatch of hub station branches are investigated. If both of the above verifications pass, the route is determined to be the final actual train route. If any of the above two verifications finds a problem, the scenario is marked and archived as a sample for subsequent incremental training of the model.
[0039] The addition of a dual verification step verifies path matching results from both data comparison and visualization, further improving the accuracy and reliability of train route identification and minimizing path mismatch issues. Visual verification makes path matching results more intuitive, facilitating manual review and problem localization, and reducing the difficulty of subsequent problem handling. Problem scenarios discovered during verification are marked and archived, providing accurate samples for subsequent incremental training of the model, realizing a closed loop of "identification-verification-optimization," and enabling the model's performance to continuously improve with application. The verification step can be automated, requiring only minimal manual intervention when problems are discovered, without significantly increasing the manual cost of path identification and ensuring the automation efficiency of the identification process.
[0040] In one embodiment, the train route identification method further includes the step of constructing a standardized train network map model, specifically: using the station latitude and longitude in the train network vector GIS data as nodes and the route of adjacent two stations as edges, the nodes are configured with feature vectors containing spatial attributes and network attributes, the edges are configured with feature vectors containing spatial attributes and network attributes, and the standardized train network map model is an optimized undirected graph model, which serves as the network input basis for the graph neural network model.
[0041] In a specific embodiment, an initial undirected graph model is constructed based on nodes, edges, and corresponding feature vectors. Then, isolated nodes and redundant edges in the model are deleted, and the topological relationships of line intersections and hub stations are calibrated to complete model optimization. Finally, a standardized train network map model is generated, which is used as the network input basis for the graph neural network model.
[0042] The embodiments of the present invention can achieve the following advantages: Employing a graph neural network model as the core computational carrier, this method is more effective than traditional graph theory algorithms and conventional machine learning algorithms in uncovering the correlation between train operation sequence features and road network spatial features, thereby improving the accuracy of path identification. Using multi-source data as input, feature data as intermediate carrier, and latitude and longitude trajectories as the final basis, a standardized train operation path identification process is formed. The operational logic is clear, enabling automated batch identification and improving path identification efficiency. Furthermore, the introduction of both matching probability and confidence level as dual indicators for path selection effectively avoids path misjudgment caused by single-indicator judgment, enhancing the reliability of train operation path identification results.
[0043] See Figure 3 This invention also provides a train route identification device 400, which includes a data acquisition unit 401, a feature extraction unit 402, and a route determination unit 403.
[0044] The data acquisition unit 401 is used to acquire train operation data and train network vector GIS data. The train operation data includes train number, stop sequence, departure and arrival time, and interval running time. The train network vector GIS data includes station latitude and longitude, station level, line direction, line length, and road network topology.
[0045] The feature extraction unit 402 is used to input the train operation data and train network vector GIS data into a pre-trained graph neural network model and output feature data; the feature data includes the latitude and longitude trajectory of the train matching path, the operation parameters of each section, the path matching probability and confidence level.
[0046] The path determination unit 403 is used to associate and match the feature data with a preset train network, and determine the train's corresponding operating path based on the latitude and longitude trajectory with the highest matching probability and confidence level that meets the preset threshold.
[0047] In one embodiment, a construction unit is further included for constructing a standardized train network map model. Specifically, the model uses the latitude and longitude of stations in the train network vector GIS data as nodes and the route directions of adjacent stations as edges. The nodes are configured with feature vectors containing spatial attributes and network attributes, and the edges are configured with feature vectors containing spatial attributes and network attributes. The standardized train network map model is an optimized undirected graph model, which serves as the network input basis for the graph neural network model.
[0048] In one embodiment, a training unit is further included for constructing and training the graph neural network model, specifically: An improved graph neural network model is constructed; the graph neural network model is an improvement on the traditional graph neural network structure, which embeds a spatial-feature dual attention collaborative module, adds a cross-modal feature fusion module, and the output layer is a three-level structure of candidate path generation-probability calculation-confidence screening; Generate training sample pairs and divide the dataset; the training sample pairs are generated by combining historical high-speed rail train operation data, corresponding open-source high-speed rail network vector GIS data and standardized high-speed rail network map model, and the dataset is divided into training set and test set according to a preset ratio; The improved graph neural network model is iteratively trained. During training, initial model parameters are set, the cross-entropy loss function is used as the optimization objective, the Adam optimizer is used to iteratively update the parameters, and an early stopping mechanism and regularization are introduced to prevent the model from overfitting. Iterative optimization of model parameters yields a trained improved graph neural network model; optimization is performed using path matching accuracy and average matching time as evaluation metrics until the model performance reaches the preset standard.
[0049] In one embodiment, the spatial-feature dual attention collaborative module includes a spatial attention mechanism and a feature attention mechanism; The spatial attention mechanism is based on a dynamic correlation weight matrix constructed by the station's latitude and longitude coordinates and the road network topology distance; The feature attention mechanism quantifies the contribution of node and edge features to path matching through an adaptive weight allocation algorithm, and sets a dynamic gain factor for the interval average travel speed feature.
[0050] In one embodiment, the cross-modal feature fusion module adopts a three-level architecture of feature alignment, deep fusion, and dimensionality normalization. The cross-modal feature fusion module completes the dimensionality calibration, feature extraction, and adaptive fusion of train running sequence features and road network spatial features through the three-level architecture, and generates a fused feature vector with a unified dimension.
[0051] In one embodiment, a path confidence evaluation factor is introduced into the three-level structure of the output layer; the path confidence evaluation factor is calculated based on the interval average speed fit and topology fit, and is used to correct the matching probability of candidate paths.
[0052] In one embodiment, the step of associating and matching the feature data with a preset train network, and determining the train's corresponding operating path based on the latitude and longitude trajectory with the highest matching probability and a confidence level meeting a preset threshold, includes a verification step for the operating path, specifically: The proposed route is compared with the train trajectory data of the preset train dispatching system, and the proposed route is overlaid on an open-source map for visualization.
[0053] like Figure 4 As shown, Figure 4 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a terminal or a server. The terminal can be an electronic device with communication functions, such as a smartphone, tablet, laptop, desktop computer, personal digital assistant, or wearable device. The server can be a standalone server or a server cluster composed of multiple servers.
[0054] The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.
[0055] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, it causes the processor 502 to execute a train route identification method.
[0056] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.
[0057] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a train route identification method.
[0058] The network interface 505 is used for network communication with other devices. Those skilled in the art will understand that the above structure 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 500 to which the present application is applied. A specific computer device 500 may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements.
[0059] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.
[0060] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program may be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.
[0061] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program.
[0062] The storage medium is a physical, non-transient storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), magnetic disk, or optical disk, or any other physical storage medium capable of storing program code. The computer-readable storage medium can be non-volatile or volatile.
[0063] 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, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. 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 implementations should not be considered beyond the scope of this invention.
[0064] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.
[0065] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0066] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.
[0067] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.
[0068] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Since these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.
[0069] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for identifying train routes, characterized in that, The method includes: Acquire train operation data and train network vector GIS data; the train operation data includes train number, stop sequence, departure and arrival time, and interval running time; the train network vector GIS data includes station latitude and longitude, station level, route alignment, route length, and network topology. The train operation data and train network vector GIS data are input into a pre-trained graph neural network model, which outputs feature data. The feature data includes the latitude and longitude trajectory of the train matching path, the operation parameters of each section, the path matching probability, and the confidence level. The feature data is associated and matched with a preset train network, and the train's corresponding operating path is determined based on the latitude and longitude trajectory with the highest matching probability and confidence level that meets the preset threshold. The method also includes the construction and training steps of a graph neural network model, specifically: An improved graph neural network model is constructed; the graph neural network model is an improvement on the traditional graph neural network structure, which embeds a spatial-feature dual attention collaborative module, adds a cross-modal feature fusion module, and the output layer is a three-level structure of candidate path generation-probability calculation-confidence screening; Generate training sample pairs and divide the dataset; the training sample pairs are generated by combining historical high-speed rail train operation data, corresponding open-source high-speed rail network vector GIS data and standardized train network map models, and the dataset is divided into training set and test set according to a preset ratio; The specific steps for constructing a standardized train network map model are as follows: Using the station latitude and longitude coordinates in the train network vector GIS data as nodes and the route directions of adjacent stations as edges, the nodes are configured with feature vectors containing spatial attributes and road network attributes, and the edges are configured with feature vectors containing spatial attributes and road network attributes. The standardized train network map model is an optimized undirected graph model, which serves as the road network input basis for the graph neural network model. The improved graph neural network model is iteratively trained. During training, initial model parameters are set, the cross-entropy loss function is used as the optimization objective, the Adam optimizer is used to iteratively update the parameters, and an early stopping mechanism and regularization are introduced to prevent the model from overfitting. Iterative optimization of model parameters yields a trained improved graph neural network model; optimization is performed using path matching accuracy and average matching time as evaluation metrics until the model performance reaches the preset standard.
2. The method according to claim 1, characterized in that, The spatial-feature dual-attention collaborative module includes a spatial attention mechanism and a feature attention mechanism; The spatial attention mechanism is based on a dynamic correlation weight matrix constructed by the station's latitude and longitude coordinates and the road network topology distance; The feature attention mechanism quantifies the contribution of node and edge features to path matching through an adaptive weight allocation algorithm, and sets a dynamic gain factor for the interval average travel speed feature.
3. The method according to claim 1, characterized in that, The cross-modal feature fusion module adopts a three-level architecture of feature alignment, deep fusion, and dimension normalization. The cross-modal feature fusion module completes the dimension calibration, feature extraction, and adaptive fusion of train running sequence features and road network spatial features through the three-level architecture, and generates a fused feature vector with a unified dimension.
4. The method according to claim 1, characterized in that, In the three-level structure of the output layer, a path confidence evaluation factor is introduced. The path confidence evaluation factor is calculated based on the interval average speed fit and topology fit, and is used to correct the matching probability of candidate paths.
5. The method according to claim 1, characterized in that, The step of associating and matching the feature data with a preset train network, and determining the train's corresponding operating path based on the latitude and longitude trajectory with the highest matching probability and a confidence level meeting a preset threshold, includes a verification step for the operating path, specifically: The proposed route is compared with the train trajectory data of the preset train dispatching system, and the proposed route is overlaid on an open-source map for visualization.
6. A train route identification device, characterized in that, include: The data acquisition unit is used to acquire train operation data and train network vector GIS data; the train operation data includes train number, stop sequence, departure and arrival time, and interval running time; the train network vector GIS data includes station latitude and longitude, station level, line direction, line length, and road network topology. The feature extraction unit is used to input the train operation data and train network vector GIS data into a pre-trained graph neural network model and output feature data. The feature data includes the latitude and longitude trajectory of the train matching path, the operating parameters of each section, the path matching probability, and the confidence level. The path determination unit is used to associate and match the feature data with a preset train network, and determine the train's corresponding operating path based on the latitude and longitude trajectory with the highest matching probability and confidence level that meets a preset threshold. The specific steps for constructing and training the graph neural network model are as follows: An improved graph neural network model is constructed; the graph neural network model is an improvement on the traditional graph neural network structure, which embeds a spatial-feature dual attention collaborative module, adds a cross-modal feature fusion module, and the output layer is a three-level structure of candidate path generation-probability calculation-confidence screening; Generate training sample pairs and divide the dataset; the training sample pairs are generated by combining historical high-speed rail train operation data, corresponding open-source high-speed rail network vector GIS data and standardized train network map models, and the dataset is divided into training set and test set according to a preset ratio; The specific steps for constructing a standardized train network map model are as follows: Using the station latitude and longitude coordinates in the train network vector GIS data as nodes and the route directions of adjacent stations as edges, the nodes are configured with feature vectors containing spatial attributes and road network attributes, and the edges are configured with feature vectors containing spatial attributes and road network attributes. The standardized train network map model is an optimized undirected graph model, which serves as the road network input basis for the graph neural network model. The improved graph neural network model is iteratively trained. During training, initial model parameters are set, the cross-entropy loss function is used as the optimization objective, the Adam optimizer is used to iteratively update the parameters, and an early stopping mechanism and regularization are introduced to prevent the model from overfitting. Iterative optimization of model parameters yields a trained improved graph neural network model; optimization is performed using path matching accuracy and average matching time as evaluation metrics until the model performance reaches the preset standard.
7. A computer device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the steps of the method according to any one of claims 1-5.
8. 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 as described in any one of claims 1-5.