Service area vehicle track monitoring system and method based on data analysis
By using a data analysis-based vehicle trajectory monitoring system, which employs LSTM and graph neural networks for trajectory and traffic prediction, the system addresses the issues of insufficient accuracy and predictive capability in vehicle trajectory monitoring within service areas, thereby achieving efficient traffic management and safe driving analysis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGSU EXPRESSWAY NETWORK OPERATION & MANAGEMENT CO LTD
- Filing Date
- 2026-03-20
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies for vehicle trajectory monitoring within service areas suffer from insufficient monitoring granularity, coverage, and predictive capabilities, resulting in large data errors and making it difficult to achieve accurate vehicle behavior prediction and traffic management.
A data-driven vehicle trajectory monitoring system is adopted, including a trajectory restoration module, a path feature module, a demand analysis module, and a service area management module. The system acquires vehicle trajectories through GPS signals and camera monitoring, uses LSTM and graph neural networks for trajectory prediction and traffic prediction, and constructs a fully connected graph for vehicle guidance.
It improved the efficiency and accuracy of trajectory calculation, optimized service area route planning, reduced the accident rate, accurately counted traffic flow and service supply and demand, predicted traffic pressure in advance, alleviated congestion, and enhanced safety management capabilities.
Smart Images
Figure CN122024486A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of vehicle monitoring, specifically to a service area vehicle trajectory monitoring system and method based on data analysis. Background Technology
[0002] Service areas are crucial nodes in the highway network, undertaking important functions such as safety management, material supply, and commercial services. Understanding the traffic conditions within service areas can prevent localized congestion and better ensure service delivery and traffic flow. Due to the short dwell time of vehicles within service areas, frequent pedestrian and vehicle traffic, and high resource demand, predicting vehicle behavior is necessary for proactive traffic management during peak periods. Studying vehicle entry rates, behavioral characteristics, and driving trajectories can significantly improve the layout of service area facilities and enhance service capacity.
[0003] Most service areas have few entrances and exits, complex internal roads, and a large number of vehicles seeking parking, charging, purchasing, and pedestrians. This high mobility easily leads to problems such as difficulty finding their way, long waiting times, and significant supply-demand imbalances, resulting in a higher traffic accident rate. Vehicle guidance and behavior prediction are needed to coordinate vehicle operations. However, due to limitations in sensor capabilities and deployment locations, commonly used satellite monitoring systems have significant limitations in monitoring granularity, coverage, and predictive capabilities, making it difficult to describe the dynamic driving environment of vehicles and resulting in substantial data errors.
[0004] In addition, in the process of regional traffic flow management, it is necessary to make short-term predictions of the trajectories of each vehicle to determine the future traffic flow in each area. Existing trajectory prediction methods based on GNN have problems such as long training time, high computational complexity, and poor dynamic modeling ability. They are difficult to describe spatial dependencies, and the operation of vehicles in the service area is driven by demand factors, which limits the stability of trajectory prediction. Summary of the Invention
[0005] The purpose of this invention is to provide a service area vehicle trajectory monitoring system and method based on data analysis to solve the problems mentioned in the background art.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a service area vehicle trajectory monitoring system based on data analysis, comprising: a trajectory restoration module, a path feature module, a demand analysis module, a vehicle prediction module, and a service area management module;
[0007] The trajectory restoration module is used to obtain the visible trajectory of a vehicle within the service area based on the vehicle's GPS signal and camera monitoring images, through target detection and tracking algorithms. The visible trajectory is then cleaned and segmented. The historical visible trajectory of each vehicle is encoded using a multi-layer LSTM. The network model is trained using the trajectory error as the loss function to obtain the trajectory prediction model. The trajectory prediction model takes the historical trajectory sequence as input and the predicted trajectory points as output. The trajectory prediction model is used to perform sequence prediction and supplement missing trajectory points.
[0008] The path feature module is used to calculate the speed distribution curve of each vehicle in the service area section based on the time stamp and change distance of the trajectory points, and to classify the vehicles by clustering the embedding features of the trajectory point sequence. A VGMM hybrid clustering algorithm is established for the statistical features of each type of vehicle to output the trajectory features of each vehicle. The statistical features include the starting point of travel, the ending point of travel, the path inflection point, the direction of travel, the trajectory length, and the stopping point.
[0009] The demand analysis module is used to construct a demand prediction model that integrates Monte Carlo simulation for each vehicle in the service area, based on the vehicle type and vehicle trajectory characteristics. This model predicts the vehicle's demand for various resources. Based on the location of supply points in the service area, a target prediction model is established that includes factors such as distance, congestion level, resource availability, and waiting time. The supply points include gas stations, charging piles, parking lots, restaurants, and restrooms. Vehicle information is input into the model, and the predicted probability of the vehicle going to each destination is output.
[0010] The vehicle prediction module is used to treat each vehicle in the service area as a type 1 node and the supply point as a type 2 node. It generates edge trajectories according to the trajectory characteristics of the vehicles, uses the predicted probability of going to each destination as the node connection probability, and uses an edge generation strategy to project each type 1 node to each type 2 node to generate lane trajectories between nodes. By projecting the trajectory into a feature representation vector, measuring the relative angle factor of the vehicle, and generating edge connections, a fully connected graph of the vehicle trajectory in the service area is obtained.
[0011] The service area management module is used to learn node representations using a graph neural network based on a fully connected graph. During each training iteration, the node features are randomly masked, and the cross-entropy loss of destination prediction is minimized through a dynamic information discarding mechanism. The trained graph neural network is used to build an LSTM-GRU prediction model, which predicts the expected traffic flow of each road segment in the service area in future time intervals. Road segment maintenance and vehicle guidance are carried out according to the expected traffic flow, and congestion is warned in advance.
[0012] Furthermore, the trajectory restoration module includes: an information acquisition unit and a trajectory encoding unit;
[0013] The information acquisition unit is used to access the vehicle's GPS via a network protocol, collect the vehicle's location coordinates, process camera images using computer vision technology, and obtain the vehicle's trajectory.
[0014] The trajectory encoding unit is used to dynamically weight the state weights of each time interval through an attention mechanism, and to generalize the data through an RMSprop optimizer to train the trajectory prediction model.
[0015] Furthermore, the path feature module includes: a velocity distribution unit and a feature classification unit;
[0016] The speed distribution unit is used to calculate speed from vehicle trajectory and draw speed distribution curves for each road segment of the service area.
[0017] The feature classification unit is used to classify vehicles according to their travel paths using an unsupervised clustering algorithm and output the trajectory features of each vehicle. The trajectory features include: vehicle speed, acceleration, path length, angular velocity, path complexity, curvature, and dwell probability.
[0018] Furthermore, the demand analysis module includes: a resource consumption unit, a model prediction unit, and a target supply unit;
[0019] The resource consumption unit is used to identify vehicle types by vehicle information for vehicles operating across service areas, and to simulate the distribution of vehicle resource consumption by coupling calculations based on vehicle type, weight, speed, flow rate and energy consumption.
[0020] The model prediction unit is used to simulate the vehicle's resource consumption based on the energy consumption model and dwell decision, aggregate the simulation results, and determine the intensity of the vehicle's demand for resupply.
[0021] The target supply unit is used to calculate the probability of a vehicle going to each supply point using multiple logistic regression based on the status of each supply point.
[0022] Furthermore, the vehicle prediction module includes: a vector metric unit and a network generation unit;
[0023] The vector metric unit is used to model vehicles and destinations in the service area as graph nodes, measure the relative angle factor vector, and connect edges between nodes.
[0024] The network generation unit is used to map trajectory features into edge weights using a learnable function to construct a fully connected network graph within the service area.
[0025] Furthermore, the service area management module includes: a trajectory connection unit, a traffic prediction unit, and a service area response unit;
[0026] The trajectory connection unit is used to aggregate graph node information of a fully connected network using a graph convolutional network, update the attention coefficients of the nodes, and minimize the predicted cross-entropy loss.
[0027] The traffic prediction unit is used to encode the historical trajectory of each vehicle using LSTM, and perform multi-step regression based on the output of the graph neural network to predict the traffic flow of each road segment within the service area.
[0028] The service area response unit is used to upload time-segmented traffic flow maps to the visualization platform and automatically perform lane diversion and maintenance warnings.
[0029] The service area vehicle trajectory monitoring method based on data analysis includes the following steps:
[0030] Step S1. Obtain the visible trajectory of the vehicle based on the vehicle's GPS signal and camera monitoring images. Use a multi-layer LSTM to encode the historical visible trajectory of each vehicle. Train the network model with the trajectory error as the loss function to obtain the trajectory prediction model. Use the trajectory prediction model to predict the trajectory point sequence and fill in the missing trajectory.
[0031] Step S2. Calculate the speed distribution curve of each vehicle on each road segment based on the trajectory point sequence, cluster the vehicles using the embedding features of the trajectory point sequence, classify the vehicles, establish VGMM hybrid clustering for the statistical features of each vehicle class, and output the trajectory features.
[0032] Step S3. Based on vehicle type and trajectory characteristics, construct a demand prediction model that integrates Monte Carlo simulation to predict the vehicle's demand for various resources. Based on the location of supply points in the service area, establish a target prediction model. Input vehicle information into the model to predict the probability of the vehicle going to each destination.
[0033] Step S4. Model the vehicles and destinations in the service area as graph nodes, generate trajectories according to the trajectory characteristics of the vehicles, use the probability of going to each destination as the connection probability between nodes, use the edge generation strategy to generate edge connections between nodes, and output the fully connected graph of vehicle trajectories in the service area.
[0034] Step S5. Based on the fully connected graph, use a graph neural network to learn node representations, build an LSTM-GRU prediction model using the trained graph neural network, and predict the expected traffic flow of each road segment in the service area in future time intervals. Perform road segment maintenance and vehicle guidance according to the expected traffic flow.
[0035] Furthermore, step S1 includes:
[0036] Step S11. Access the vehicle's GPS via network protocol, collect the vehicle's location coordinates, process the camera images using computer vision technology, obtain the vehicle's trajectory based on target detection and tracking algorithms, and perform data cleaning and trajectory segmentation on the visible trajectory.
[0037] Step S12. Train a multi-layer LSTM network model using trajectory error as the loss function, dynamically weight the state weights of each time interval through the attention mechanism, and generalize the data through the RMSprop optimizer to obtain a trajectory prediction model. The trajectory prediction model takes the historical trajectory sequence as input and the predicted trajectory points as output, and outputs a complete trajectory point sequence.
[0038] Furthermore, step S2 includes:
[0039] Step S21. Based on the time stamp and change distance of the trajectory points, calculate the speed distribution curve of each vehicle in the service area section, use an unsupervised clustering algorithm to classify the vehicles according to the driving path, and output the trajectory features of each vehicle. The trajectory features include: vehicle speed, acceleration, path length, angular velocity, path complexity, curvature, and dwell probability.
[0040] Step S22. Establish a VGMM hybrid clustering algorithm based on the statistical characteristics of vehicles, and output the trajectory characteristics of each vehicle. The statistical characteristics include the starting point of travel, the ending point of travel, the inflection point of the path, the direction of travel, the trajectory length, and the stopping point.
[0041] Furthermore, step S3 includes:
[0042] Step S31. For vehicles operating across service areas, identify vehicle types through vehicle information, and construct a demand prediction model that integrates Monte Carlo simulation by coupling calculations of vehicle type, weight, speed, flow rate and energy consumption to simulate the distribution of vehicle resource consumption.
[0043] Step S32. Based on the location of supply points within the service area, establish a target prediction model that includes factors such as distance, congestion level, resource availability, and waiting time. The supply points include: gas stations, charging piles, parking lots, restaurants, and restrooms. Input vehicle information into the model and output the predicted probability of the vehicle reaching each destination.
[0044] Furthermore, step S4 includes:
[0045] Step S41. Each vehicle in the service area is regarded as a type 1 node and the supply point is regarded as a type 2 node. The edge trajectory is generated according to the trajectory characteristics of the vehicle. The predicted probability of going to each destination is used as the node connection probability. The edge generation strategy is used to project each type 1 node to each type 2 node to generate the lane trajectory between nodes.
[0046] Step S42. Use a learnable function to map trajectory features to edge weights, measure the relative angle factor vector, connect edges between nodes, and construct a fully connected network graph within the service area.
[0047] Furthermore, step S5 includes:
[0048] Step S51. Use a graph convolutional network to aggregate the graph node information of the fully connected network. In each training iteration, randomly mask the node features and minimize the cross-entropy loss of the destination prediction through a dynamic information discarding mechanism to obtain the trained graph neural network.
[0049] Step S52. Use LSTM to encode the historical trajectory of each vehicle, perform multi-step regression based on the graph neural network output, predict the traffic flow of each road segment in the service area, upload the time-segmented traffic flow map to the visualization platform, and automatically perform lane diversion and maintenance warnings.
[0050] Compared with the prior art, the beneficial effects achieved by the present invention are:
[0051] 1. This invention can obtain the visible trajectory of a vehicle based on GPS tracking and camera monitoring, construct an LSTM trajectory prediction model with an attention mechanism, encode the vehicle's historical trajectory, and recover the missing parts of the trajectory through a trajectory generation algorithm. This improves the efficiency of trajectory calculation and prediction accuracy, solves the oversmoothing problem, enhances the model's generalization ability, optimizes service area route planning, evaluates driver driving habits, realizes safe driving behavior analysis, and reduces the accident rate.
[0052] 2. This invention can construct a demand forecasting model that integrates Monte Carlo simulation. By coupling calculations of vehicle type, weight, speed, flow rate and energy consumption, it simulates the distribution of vehicle resource consumption, calculates the probability of vehicles going to various destinations based on demand, and accurately counts traffic flow and service supply and demand in different areas by analyzing vehicle trajectories, thereby reducing congestion within the area and improving the quality of public services.
[0053] 3. This invention can model the lane trajectories between vehicles and various destinations using an edge generation strategy, build a graph neural network with dynamic information discarding using a fully connected graph, construct an LSTM-GRU prediction model, and predict the expected traffic flow of each road segment within the service area. It is suitable for vehicle trajectory prediction tasks with dependencies, can predict the traffic pressure of the service area in advance, alleviate mainline congestion, enhance safety management capabilities, and improve the operational efficiency and service value of the service area. Attached Figure Description
[0054] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:
[0055] Figure 1 This is a schematic diagram of the service area vehicle trajectory monitoring system based on data analysis according to the present invention;
[0056] Figure 2 This is a schematic diagram illustrating the steps of the service area vehicle trajectory monitoring method based on data analysis according to the present invention. Detailed Implementation
[0057] 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, and 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.
[0058] Please see Figures 1 to 2 The present invention provides a technical solution: a service area vehicle trajectory monitoring system based on data analysis, comprising: a trajectory restoration module, a path feature module, a demand analysis module, a vehicle prediction module, and a service area management module;
[0059] The trajectory restoration module is used to obtain the visible trajectory of a vehicle within the service area based on the vehicle's GPS signal and camera monitoring images, through target detection and tracking algorithms. The visible trajectory is then cleaned and segmented. The historical visible trajectory of each vehicle is encoded using a multi-layer LSTM. The network model is trained using the trajectory error as the loss function to obtain the trajectory prediction model. The trajectory prediction model takes the historical trajectory sequence as input and the predicted trajectory points as output. The trajectory prediction model is used to perform sequence prediction and supplement missing trajectory points.
[0060] The trajectory restoration module includes: an information acquisition unit and a trajectory encoding unit;
[0061] The information acquisition unit is used to access the vehicle's GPS via a network protocol, collect the vehicle's location coordinates, process camera images using computer vision technology, and obtain the vehicle's trajectory.
[0062] The trajectory encoding unit is used to dynamically weight the state weights of each time interval through an attention mechanism, and to generalize the data through an RMSprop optimizer to train the trajectory prediction model.
[0063] The path feature module is used to calculate the speed distribution curve of each vehicle in the service area section based on the time stamp and change distance of the trajectory points, and to classify the vehicles by clustering the embedding features of the trajectory point sequence. A VGMM hybrid clustering algorithm is established for the statistical features of each type of vehicle to output the trajectory features of each vehicle. The statistical features include the starting point of travel, the ending point of travel, the path inflection point, the direction of travel, the trajectory length, and the stopping point.
[0064] The path feature module includes: a velocity distribution unit and a feature classification unit;
[0065] The speed distribution unit is used to calculate speed from vehicle trajectory and draw speed distribution curves for each road segment of the service area.
[0066] The feature classification unit is used to classify vehicles according to their travel paths using an unsupervised clustering algorithm and output the trajectory features of each vehicle. The trajectory features include: vehicle speed, acceleration, path length, angular velocity, path complexity, curvature, and dwell probability.
[0067] The demand analysis module is used to construct a demand prediction model that integrates Monte Carlo simulation for each vehicle in the service area, based on the vehicle type and vehicle trajectory characteristics. This model predicts the vehicle's demand for various resources. Based on the location of supply points in the service area, a target prediction model is established that includes factors such as distance, congestion level, resource availability, and waiting time. The supply points include gas stations, charging piles, parking lots, restaurants, and restrooms. Vehicle information is input into the model, and the predicted probability of the vehicle going to each destination is output.
[0068] The demand analysis module includes: a resource consumption unit, a model prediction unit, and a target supply unit;
[0069] The resource consumption unit is used to identify vehicle types by vehicle information for vehicles operating across service areas, and to simulate the distribution of vehicle resource consumption by coupling calculations based on vehicle type, weight, speed, flow rate and energy consumption.
[0070] The model prediction unit is used to simulate the vehicle's resource consumption based on the energy consumption model and dwell decision, aggregate the simulation results, and determine the intensity of the vehicle's demand for resupply.
[0071] The target supply unit is used to calculate the probability of a vehicle going to each supply point using multiple logistic regression based on the status of each supply point.
[0072] The vehicle prediction module is used to treat each vehicle in the service area as a type 1 node and the supply point as a type 2 node. It generates edge trajectories according to the trajectory characteristics of the vehicles, uses the predicted probability of going to each destination as the node connection probability, and uses an edge generation strategy to project each type 1 node to each type 2 node to generate lane trajectories between nodes. By projecting the trajectory into a feature representation vector, measuring the relative angle factor of the vehicle, and generating edge connections, a fully connected graph of the vehicle trajectory in the service area is obtained.
[0073] The vehicle prediction module includes: a vector metric unit and a network generation unit;
[0074] The vector metric unit is used to model vehicles and destinations in the service area as graph nodes, measure the relative angle factor vector, and connect edges between nodes.
[0075] The network generation unit is used to map trajectory features into edge weights using a learnable function to construct a fully connected network graph within the service area.
[0076] The service area management module is used to learn node representations using a graph neural network based on a fully connected graph. During each training iteration, the node features are randomly masked, and the cross-entropy loss of destination prediction is minimized through a dynamic information discarding mechanism. The trained graph neural network is used to build an LSTM-GRU prediction model, which predicts the expected traffic flow of each road segment in the service area in future time intervals. Road segment maintenance and vehicle guidance are carried out according to the expected traffic flow, and congestion is warned in advance.
[0077] The service area management module includes: a trajectory connection unit, a traffic prediction unit, and a service area response unit;
[0078] The trajectory connection unit is used to aggregate graph node information of a fully connected network using a graph convolutional network, update the attention coefficients of the nodes, and minimize the predicted cross-entropy loss.
[0079] The traffic prediction unit is used to encode the historical trajectory of each vehicle using LSTM, and perform multi-step regression based on the output of the graph neural network to predict the traffic flow of each road segment within the service area.
[0080] The service area response unit is used to upload time-segmented traffic flow maps to the visualization platform and automatically perform lane diversion and maintenance warnings.
[0081] The service area vehicle trajectory monitoring method based on data analysis includes the following steps:
[0082] Step S1. Obtain the visible trajectory of the vehicle based on the vehicle's GPS signal and camera monitoring images. Use a multi-layer LSTM to encode the historical visible trajectory of each vehicle. Train the network model with the trajectory error as the loss function to obtain the trajectory prediction model. Use the trajectory prediction model to predict the trajectory point sequence and fill in the missing trajectory.
[0083] Step S1 includes:
[0084] Step S11. Access the vehicle's GPS via network protocol, collect the vehicle's location coordinates, process the camera images using computer vision technology, obtain the vehicle's trajectory based on target detection and tracking algorithms, and perform data cleaning and trajectory segmentation on the visible trajectory.
[0085] Step S12. Train a multi-layer LSTM network model using trajectory error as the loss function, dynamically weight the state weights of each time interval through the attention mechanism, and generalize the data through the RMSprop optimizer to obtain a trajectory prediction model. The trajectory prediction model takes the historical trajectory sequence as input and the predicted trajectory points as output, and outputs a complete trajectory point sequence.
[0086] Step S2. Calculate the speed distribution curve of each vehicle on each road segment based on the trajectory point sequence, cluster the vehicles using the embedding features of the trajectory point sequence, classify the vehicles, establish VGMM hybrid clustering for the statistical features of each vehicle class, and output the trajectory features.
[0087] Step S2 includes:
[0088] Step S21. Based on the time stamp and change distance of the trajectory points, calculate the speed distribution curve of each vehicle in the service area section, use an unsupervised clustering algorithm to classify the vehicles according to the driving path, and output the trajectory features of each vehicle. The trajectory features include: vehicle speed, acceleration, path length, angular velocity, path complexity, curvature, and dwell probability.
[0089] Step S22. Establish a VGMM hybrid clustering algorithm based on the statistical characteristics of vehicles, and output the trajectory characteristics of each vehicle. The statistical characteristics include the starting point of travel, the ending point of travel, the inflection point of the path, the direction of travel, the trajectory length, and the stopping point.
[0090] Step S3. Based on vehicle type and trajectory characteristics, construct a demand prediction model that integrates Monte Carlo simulation to predict the vehicle's demand for various resources. Based on the location of supply points in the service area, establish a target prediction model. Input vehicle information into the model to predict the probability of the vehicle going to each destination.
[0091] Step S3 includes:
[0092] Step S31. For vehicles operating across service areas, identify vehicle types through vehicle information, and construct a demand prediction model that integrates Monte Carlo simulation by coupling calculations of vehicle type, weight, speed, flow rate and energy consumption to simulate the distribution of vehicle resource consumption.
[0093] Step S32. Based on the location of supply points within the service area, establish a target prediction model that includes factors such as distance, congestion level, resource availability, and waiting time. The supply points include: gas stations, charging piles, parking lots, restaurants, and restrooms. Input vehicle information into the model and output the predicted probability of the vehicle reaching each destination.
[0094] Step S4. Model the vehicles and destinations in the service area as graph nodes, generate trajectories according to the trajectory characteristics of the vehicles, use the probability of going to each destination as the connection probability between nodes, use the edge generation strategy to generate edge connections between nodes, and output the fully connected graph of vehicle trajectories in the service area.
[0095] Step S4 includes:
[0096] Step S41. Each vehicle in the service area is regarded as a type 1 node and the supply point is regarded as a type 2 node. The edge trajectory is generated according to the trajectory characteristics of the vehicle. The predicted probability of going to each destination is used as the node connection probability. The edge generation strategy is used to project each type 1 node to each type 2 node to generate the lane trajectory between nodes.
[0097] Step S42. Use a learnable function to map trajectory features to edge weights, measure the relative angle factor vector, connect edges between nodes, and construct a fully connected network graph within the service area.
[0098] Step S5. Based on the fully connected graph, use a graph neural network to learn node representations, build an LSTM-GRU prediction model using the trained graph neural network, and predict the expected traffic flow of each road segment in the service area in future time intervals. Perform road segment maintenance and vehicle guidance according to the expected traffic flow.
[0099] Step S5 includes:
[0100] Step S51. Use a graph convolutional network to aggregate the graph node information of the fully connected network. In each training iteration, randomly mask the node features and minimize the cross-entropy loss of the destination prediction through a dynamic information discarding mechanism to obtain the trained graph neural network.
[0101] Step S52. Use LSTM to encode the historical trajectory of each vehicle, perform multi-step regression based on the graph neural network output, predict the traffic flow of each road segment in the service area, upload the time-segmented traffic flow map to the visualization platform, and automatically perform lane diversion and maintenance warnings.
[0102] Example: Trajectory data collection, data cleaning, identification of missing values, encoding of historical trajectories using multi-layer LSTM, prediction of future trajectory point sequences, trajectory generation algorithm to recover missing trajectory parts, calculation of speed distribution, vehicle path classification, prediction of VGMM clustered trajectories, simulation of vehicle resource consumption distribution, prediction of multi-factor destination probabilities, construction of service area spatial map, generation of trajectory feature representation vectors, construction of LSTM-GRU hybrid prediction model, prediction of traffic flow for each road segment, and feedback of resource scheduling scheme.
[0103] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0104] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A service area vehicle trajectory monitoring method based on data analysis, characterized in that, The method includes the following steps: Step S1. Obtain the visible trajectory of the vehicle based on the vehicle's GPS signal and camera monitoring images. Use a multi-layer LSTM to encode the historical visible trajectory of each vehicle. Train the network model with the trajectory error as the loss function to obtain the trajectory prediction model. Use the trajectory prediction model to predict the trajectory point sequence and fill in the missing trajectory. Step S2. Calculate the speed distribution curve of each vehicle on each road segment based on the trajectory point sequence, cluster the vehicles using the embedding features of the trajectory point sequence, classify the vehicles, establish VGMM hybrid clustering for the statistical features of each vehicle class, and output the trajectory features. Step S3. Based on vehicle type and trajectory characteristics, construct a demand prediction model that integrates Monte Carlo simulation to predict the vehicle's demand for various resources. Based on the location of supply points in the service area, establish a target prediction model. Input vehicle information into the model to predict the probability of the vehicle going to each destination. Step S4. Model the vehicles and destinations in the service area as graph nodes, generate trajectories according to the trajectory characteristics of the vehicles, use the probability of going to each destination as the connection probability between nodes, use the edge generation strategy to generate edge connections between nodes, and output the fully connected graph of vehicle trajectories in the service area. Step S5. Based on the fully connected graph, use a graph neural network to learn node representations, build an LSTM-GRU prediction model using the trained graph neural network, and predict the expected traffic flow of each road segment in the service area in future time intervals. Perform road segment maintenance and vehicle guidance according to the expected traffic flow.
2. The service area vehicle trajectory monitoring method based on data analysis according to claim 1, characterized in that: Step S1 includes: Step S11. Access the vehicle's GPS via network protocol, collect the vehicle's location coordinates, process the camera images using computer vision technology, obtain the vehicle's trajectory based on target detection and tracking algorithms, and perform data cleaning and trajectory segmentation on the visible trajectory. Step S12. Train a multi-layer LSTM network model using trajectory error as the loss function, dynamically weight the state weights of each time interval through the attention mechanism, and generalize the data through the RMSprop optimizer to obtain a trajectory prediction model. The trajectory prediction model takes the historical trajectory sequence as input and the predicted trajectory points as output, and outputs a complete trajectory point sequence.
3. The service area vehicle trajectory monitoring method based on data analysis according to claim 2, characterized in that: Step S2 includes: Step S21. Based on the time stamp and change distance of the trajectory points, calculate the speed distribution curve of each vehicle in the service area section, use an unsupervised clustering algorithm to classify the vehicles according to the driving path, and output the trajectory features of each vehicle. The trajectory features include: vehicle speed, acceleration, path length, angular velocity, path complexity, curvature, and dwell probability. Step S22. Establish a VGMM hybrid clustering algorithm based on the statistical characteristics of vehicles, and output the trajectory characteristics of each vehicle. The statistical characteristics include the starting point of travel, the ending point of travel, the inflection point of the path, the direction of travel, the trajectory length, and the stopping point. Step S3 includes: Step S31. For vehicles operating across service areas, identify vehicle types through vehicle information, and construct a demand prediction model that integrates Monte Carlo simulation by coupling calculations of vehicle type, weight, speed, flow rate and energy consumption to simulate the distribution of vehicle resource consumption. Step S32. Based on the location of supply points within the service area, establish a target prediction model that includes factors such as distance, congestion level, resource availability, and waiting time. The supply points include: gas stations, charging piles, parking lots, restaurants, and restrooms. Input vehicle information into the model and output the predicted probability of the vehicle reaching each destination.
4. The service area vehicle trajectory monitoring method based on data analysis according to claim 3, characterized in that: Step S4 includes: Step S41. Each vehicle in the service area is regarded as a type 1 node and the supply point is regarded as a type 2 node. The edge trajectory is generated according to the trajectory characteristics of the vehicle. The predicted probability of going to each destination is used as the node connection probability. The edge generation strategy is used to project each type 1 node to each type 2 node to generate the lane trajectory between nodes. Step S42. Use a learnable function to map trajectory features to edge weights, measure the relative angle factor vector, connect edges between nodes, and construct a fully connected network graph within the service area.
5. The service area vehicle trajectory monitoring method based on data analysis according to claim 4, characterized in that: Step S5 includes: Step S51. Use a graph convolutional network to aggregate the graph node information of the fully connected network. In each training iteration, randomly mask the node features and minimize the cross-entropy loss of the destination prediction through a dynamic information discarding mechanism to obtain the trained graph neural network. Step S52. Use LSTM to encode the historical trajectory of each vehicle, perform multi-step regression based on the graph neural network output, predict the traffic flow of each road segment in the service area, upload the time-segmented traffic flow map to the visualization platform, and automatically perform lane diversion and maintenance warnings.
6. A service area vehicle trajectory monitoring system based on data analysis, characterized in that, The system includes the following modules: trajectory restoration module, path feature module, demand analysis module, vehicle prediction module, and service area management module; The trajectory restoration module is used to obtain the visible trajectory of a vehicle within the service area based on the vehicle's GPS signal and camera monitoring images, through target detection and tracking algorithms. The visible trajectory is then cleaned and segmented. The historical visible trajectory of each vehicle is encoded using a multi-layer LSTM. The network model is trained using the trajectory error as the loss function to obtain the trajectory prediction model. The trajectory prediction model takes the historical trajectory sequence as input and the predicted trajectory points as output. The trajectory prediction model is used to perform sequence prediction and supplement missing trajectory points. The path feature module is used to calculate the speed distribution curve of each vehicle in the service area section based on the time stamp and change distance of the trajectory points, and to classify the vehicles by clustering the embedding features of the trajectory point sequence. A VGMM hybrid clustering algorithm is established for the statistical features of each type of vehicle to output the trajectory features of each vehicle. The statistical features include the starting point of travel, the ending point of travel, the path inflection point, the direction of travel, the trajectory length, and the stopping point. The demand analysis module is used to construct a demand prediction model that integrates Monte Carlo simulation for each vehicle in the service area, based on the vehicle type and vehicle trajectory characteristics. This model predicts the vehicle's demand for various resources. Based on the location of supply points in the service area, a target prediction model is established that includes factors such as distance, congestion level, resource availability, and waiting time. The supply points include gas stations, charging piles, parking lots, restaurants, and restrooms. Vehicle information is input into the model, and the predicted probability of the vehicle going to each destination is output. The vehicle prediction module is used to treat each vehicle in the service area as a type 1 node and the supply point as a type 2 node. It generates edge trajectories according to the trajectory characteristics of the vehicles, uses the predicted probability of going to each destination as the node connection probability, and uses an edge generation strategy to project each type 1 node to each type 2 node to generate lane trajectories between nodes. By projecting the trajectory into a feature representation vector, measuring the relative angle factor of the vehicle, and generating edge connections, a fully connected graph of the vehicle trajectory in the service area is obtained. The service area management module is used to learn node representations using a graph neural network based on a fully connected graph. During each training iteration, the node features are randomly masked, and the cross-entropy loss of destination prediction is minimized through a dynamic information discarding mechanism. The trained graph neural network is used to build an LSTM-GRU prediction model, which predicts the expected traffic flow of each road segment in the service area in future time intervals. Road segment maintenance and vehicle guidance are carried out according to the expected traffic flow, and congestion is warned in advance.
7. The service area vehicle trajectory monitoring system based on data analysis according to claim 6, characterized in that: The trajectory restoration module includes: an information acquisition unit and a trajectory encoding unit; The information acquisition unit is used to access the vehicle's GPS via a network protocol, collect the vehicle's location coordinates, process camera images using computer vision technology, and obtain the vehicle's trajectory. The trajectory encoding unit is used to dynamically weight the state weights of each time interval through an attention mechanism, and to generalize the data through the RMSprop optimizer to train the trajectory prediction model. The path feature module includes: a velocity distribution unit and a feature classification unit; The speed distribution unit is used to calculate speed from vehicle trajectory and draw speed distribution curves for each road segment of the service area. The feature classification unit is used to classify vehicles according to their travel paths using an unsupervised clustering algorithm and output the trajectory features of each vehicle. The trajectory features include: vehicle speed, acceleration, path length, angular velocity, path complexity, curvature, and dwell probability.
8. The service area vehicle trajectory monitoring system based on data analysis according to claim 7, characterized in that: The demand analysis module includes: a resource consumption unit, a model prediction unit, and a target supply unit; The resource consumption unit is used to identify vehicle types by vehicle information for vehicles operating across service areas, and to simulate the distribution of vehicle resource consumption by coupling calculations based on vehicle type, weight, speed, flow rate and energy consumption. The model prediction unit is used to simulate the vehicle's resource consumption based on the energy consumption model and dwell decision, aggregate the simulation results, and determine the intensity of the vehicle's demand for supplies. The target supply unit is used to calculate the probability of a vehicle going to each supply point using multinomial logistic regression based on the status of each supply point.
9. The service area vehicle trajectory monitoring system based on data analysis according to claim 8, characterized in that: The vehicle prediction module includes: a vector metric unit and a network generation unit; The vector metric unit is used to model vehicles and destinations in the service area as graph nodes, measure the relative angle factor vector, and connect edges between nodes. The network generation unit is used to map trajectory features into edge weights using a learnable function to construct a fully connected network graph within the service area.
10. The service area vehicle trajectory monitoring system based on data analysis according to claim 9, characterized in that: The service area management module includes: a trajectory connection unit, a traffic prediction unit, and a service area response unit; The trajectory connection unit is used to aggregate graph node information of a fully connected network using a graph convolutional network, update the attention coefficients of the nodes, and minimize the predicted cross-entropy loss. The traffic prediction unit is used to encode the historical trajectory of each vehicle using LSTM, and perform multi-step regression based on the output of the graph neural network to predict the traffic flow of each road segment within the service area. The service area response unit is used to upload time-segmented traffic flow maps to the visualization platform and automatically perform lane diversion and maintenance warnings.