CN-122024486-A - Service area vehicle track monitoring system and method based on data analysis

CN122024486ACN 122024486 ACN122024486 ACN 122024486ACN-122024486-A

Abstract

The invention relates to the field of vehicle monitoring, in particular to a service area vehicle track monitoring system and method based on data analysis, comprising the following steps: the system comprises a track restoration module, a path characteristic module, a demand analysis module, a vehicle prediction module and a service area management module, wherein the track restoration module is used for supplementing missing track points, the path characteristic module is used for outputting track characteristics of each vehicle, the demand analysis module is used for predicting the probability of the vehicle going to each destination, the vehicle prediction module is used for obtaining a full-connection diagram of the vehicle track, and the service area management module is used for predicting the flow of each road section in a rolling way.

Inventors

WANG YINGJIAN
ZHANG XIYA
WANG JIAN
WANG LIANG
CAO LEI
YANG ZHEN
SHEN JINGJIN
WANG TIANSHU

Assignees

江苏高速公路联网营运管理有限公司

Dates

Publication Date: 20260512
Application Date: 20260320

Claims (10)

1. A service area vehicle track monitoring method based on data analysis, which is characterized by comprising the following steps: s1, obtaining a vehicle visible track based on a vehicle GPS signal and a camera monitoring image, using a multilayer LSTM to encode each vehicle historical visible track, using a track error as a loss function to train a network model to obtain a track prediction model, using the track prediction model to predict a track point sequence, and supplementing a missing track; S2, calculating a speed distribution curve of each vehicle on each road section according to the track point sequence, classifying the vehicles by using embedded feature clusters of the track point sequence, establishing VGMM mixed clusters aiming at the statistical features of each type of vehicles, and outputting track features; S3, constructing a demand prediction model fused with Monte Carlo simulation according to the vehicle type and the vehicle track characteristics, predicting the demand of the vehicle for each resource, constructing a target prediction model according to the position of the supply point in the service area, inputting vehicle information in the model, and predicting the probability of the vehicle going to each destination; S4, modeling the vehicles and the destinations in the service area as graph nodes, generating tracks according to track characteristics of the vehicles, taking the probability of going to each destination as the connection probability among the nodes, generating edge connection among the nodes by adopting an edge generation strategy, and outputting a full connection graph of the track of the vehicles in the service area; S5, according to the full-connection graph, using a graph neural network learning node to represent, building an LSTM-GRU prediction model by using the trained graph neural network, predicting expected flow of each road section of the service area in each time interval in the future in a rolling mode, and carrying out road section maintenance and vehicle guiding according to the expected flow.
2. The method for monitoring the track of a service area vehicle based on data analysis according to claim 1, wherein the step S1 comprises the following steps: S11, accessing a vehicle GPS through a network protocol, collecting vehicle position coordinates, processing a camera image through a computer vision technology, acquiring a vehicle track based on a target detection and tracking algorithm, and performing data cleaning and track segmentation on the visible track; S12, training a multi-layer LSTM network model by taking the track error as a loss function, dynamically weighting the state weights of all time intervals through a attention mechanism, and generalizing data through a RMSprop optimizer to obtain a track prediction model, wherein the track prediction model takes a historical track sequence as input, predicted track points as output, and a complete track point sequence is output.
3. The method for monitoring the track of a service area vehicle based on data analysis according to claim 2, wherein the step S2 comprises the steps of: S21, calculating a speed distribution curve of each vehicle in a service area section based on time marks and change distances of track points, classifying the vehicles according to a driving path by using an unsupervised clustering algorithm, and outputting track characteristics of each vehicle, wherein the track characteristics comprise vehicle speed, acceleration, path length, angular speed, path complexity, curvature and stay probability; S22, establishing VGMM a hybrid clustering algorithm aiming at the statistical characteristics of the vehicle, and outputting the track characteristics of each vehicle, wherein the statistical characteristics comprise a running starting point, a running end point, a path inflection point, a running direction, a track length and a stay point; The step S3 comprises the following steps: S31, identifying the type of a vehicle running across a service area through vehicle information, and constructing a demand prediction model fused with Monte Carlo simulation through coupling calculation of the type, weight, speed, flow and energy consumption of the vehicle to simulate the resource consumption distribution of the vehicle; S32, establishing a target prediction model containing influences of factors of distance, congestion degree, resource allowance and waiting time according to positions of supply points in a service area, wherein the supply points comprise gas stations, charging piles, parking lots, restaurants and toilets, vehicle information is input into the model, and prediction probability of vehicles going to each destination is output.
4. The method for monitoring the track of a service area vehicle based on data analysis according to claim 3, wherein the step S4 comprises the steps of: S41, taking each vehicle in a service area as a class-one node, taking a replenishment point as a class-two node, generating an edge track according to track characteristics of the vehicle, taking the prediction probability of going to each destination as a node connection probability, projecting each class-one node to each class-two node by adopting an edge generation strategy, and generating a lane track among the nodes; S42, mapping the track characteristics into edge weights by using a learnable function, measuring relative angle factor vectors, and performing edge connection between nodes to construct a fully-connected network diagram in the service area.
5. The method for monitoring the vehicle track in the service area based on the data analysis of claim 4, wherein the step S5 comprises the following steps: s51, aggregating the node information of the full-connection network graph by using a graph convolution network, carrying out random masking on node characteristics when training is iterated each time, and minimizing cross entropy loss of destination prediction through a dynamic information discarding mechanism to obtain a trained graph neural network; S52, using LSTM to encode the historical track of each vehicle, carrying out multi-step regression according to the output of the graphic neural network, predicting the flow of each road section in the service area, uploading the time-division flow map to a visualization platform, and automatically carrying out lane diversion and maintenance early warning.
6. The service area vehicle track monitoring system based on data analysis is characterized by comprising a track restoration module, a path characteristic module, a demand analysis module, a vehicle prediction module and a service area management module; The track recovery module is used for obtaining a vehicle visible track through a target detection and tracking algorithm based on a vehicle GPS signal and a camera monitoring image in a service area range, performing data cleaning and track segmentation on the visible track, using a multilayer LSTM to encode a historical visible track of each vehicle, training a network model by taking a track error as a loss function to obtain a track prediction model, taking a historical track sequence as an input, taking predicted track points as an output, performing sequence prediction by using the track prediction model, and supplementing missing track points; The path characteristic module is used for calculating a speed distribution curve of each vehicle in a service area road section based on time marks and change distances of track points, classifying the vehicles by using embedded characteristic clusters of track point sequences, establishing VGMM hybrid clustering algorithm aiming at statistical characteristics of each type of vehicles, and outputting track characteristics of each vehicle, wherein the statistical characteristics comprise a running starting point, a running end point, a path inflection point, a running direction, a track length and a stay point; the demand analysis module is used for constructing a demand prediction model fused with Monte Carlo simulation for vehicles in each service area according to the vehicle type and the vehicle track characteristics, predicting the demand of the vehicles for each resource, and establishing a target prediction model containing the influences of the factors of distance, crowding degree, resource allowance and waiting time according to the position of a supplement point in the service area, wherein the supplement point comprises a gas station, a charging pile, a parking lot, a restaurant and a toilet, inputting vehicle information in the model, and outputting the prediction probability of the vehicles going to each destination; The vehicle prediction module is used for taking each vehicle in the service area as one type of node, taking the replenishment points as two types of nodes, generating an edge track according to track characteristics of the vehicles, taking the prediction probability of going to each destination as node connection probability, projecting each type of node to each two types of nodes by adopting an edge generation strategy, generating a lane track among the nodes, and generating an edge connection by projecting the track into a characteristic expression vector, measuring relative angle factors of the vehicles, so as to obtain a full connection diagram of the vehicle track of the service area; The service area management module is used for learning node representation by using a graph neural network according to a full-connection graph, carrying out random masking on node characteristics when training is iterated each time, minimizing cross entropy loss of destination prediction by a dynamic information discarding mechanism, building an LSTM-GRU prediction model by using the trained graph neural network, rolling and predicting expected flow of each road section of a service area in each time interval in the future, carrying out road section maintenance and vehicle guidance according to the expected flow, and early warning congestion.
7. The service area vehicle track monitoring system based on data analysis according to claim 6, wherein the track restoration module comprises an information acquisition unit and a track coding unit; The information acquisition unit is used for accessing a vehicle GPS through a network protocol, collecting vehicle position coordinates, processing a camera image through a computer vision technology and obtaining a vehicle track; The track coding unit is used for dynamically weighting the state weights of all time intervals through an attention mechanism, generalizing data through a RMSprop optimizer and training a track prediction model; the path characteristic module comprises a speed distribution unit and a characteristic classification unit; the speed distribution unit is used for calculating the speed from the vehicle track and drawing a speed distribution curve for each road section of the service area; the feature classification unit is used for classifying the vehicles according to the driving paths by using an unsupervised clustering algorithm and outputting track features of each vehicle, wherein the track features comprise vehicle speed, acceleration, path length, angular speed, path complexity, curvature and stay probability.
8. The system for monitoring the track of the vehicle in the service area based on data analysis according to claim 7, wherein the demand analysis module comprises a resource consumption unit, a model prediction unit and a target replenishment unit; The resource consumption unit is used for identifying the type of the vehicle according to the vehicle information and simulating the distribution of the automobile resource consumption according to the coupling calculation of the type, weight, speed, flow and energy consumption of the vehicle running across the service area; The model prediction unit is used for simulating the resource consumption of the vehicle according to the energy consumption model and the stay decision, aggregating the simulation results and determining the demand intensity of the vehicle for replenishment; The target replenishment unit is used for calculating the probability of the vehicle going to each replenishment point by using a plurality of logistic regression according to the state of each replenishment point.
9. The service area vehicle track monitoring system based on data analysis of claim 8, wherein the vehicle prediction module comprises a vector measurement unit and a network generation unit; the vector measurement unit is used for modeling vehicles and destinations in the service area as graph nodes, measuring relative angle factor vectors and performing edge connection between the nodes; The network generation unit is used for mapping the track characteristics into edge weights by using a learnable function and constructing a fully-connected network diagram in the service area.
10. The service area vehicle track monitoring system based on data analysis of claim 9, wherein the service area management module comprises a track connection unit, a flow prediction unit and a service area response unit; the track connection unit is used for aggregating the node information of the full-connection network graph by using the graph convolution network, updating the attention coefficient of the node and minimizing the predicted cross entropy loss; The flow prediction unit is used for encoding the historical track of each vehicle by using the LSTM, carrying out multi-step regression according to the output of the graphic neural network, and predicting the flow of each road section in the service area; the service area response unit is used for uploading the time-division flow chart to the visualization platform and automatically carrying out lane diversion and maintenance early warning.

Description

Service area vehicle track monitoring system and method based on data analysis Technical Field The invention relates to the field of vehicle monitoring, in particular to a service area vehicle track monitoring system and method based on data analysis. Background The service area is an important node in the expressway network, plays important functions of safety management, material supply, business service and the like, grasps the traffic state of the service area, can prevent local congestion, and better performs service supply and traffic guarantee. Because the stay time of vehicles in service is short, the traffic flow is frequent, the resource demand is high, the vehicle behavior needs to be predicted in order to conduct traffic management in advance in peak period, and the facility layout of a service area can be improved by researching the vehicle driving-in rate, behavior characteristics and driving tracks, so that the method has great significance for improving the service capacity of the service area. Most of the service areas have fewer entrances and exits, the internal roads are complex, and parking lots are more in vehicle searching, vehicle charging, purchasing and pedestrians. The mobility is large, the problems of difficult road searching, long waiting time, outstanding contradiction between supply and demand and the like are easy to occur, the traffic accident rate is high, and the vehicle guiding and behavior prediction are needed to coordinate the running of the vehicle. However, the monitoring granularity, coverage rate and prediction capability of the common satellite monitoring system are greatly limited due to the influence of the capability and deployment position of the sensing equipment, and the dynamic driving environment of the vehicle is difficult to describe, so that large data errors can be generated. In addition, in the process of regional traffic flow management, short-term prediction is needed for the track of each vehicle, the future traffic flow of each region is determined, the existing track prediction method based on the GNN has the problems of long training time, high calculation complexity, poor dynamic modeling capability and the like, the space dependence is difficult to express, the vehicle operation in the service area is driven by demand factors, and the stability of track prediction is limited. Disclosure of Invention The invention aims to provide a service area vehicle track monitoring system and method based on data analysis, so as to solve the problems in the background technology. In order to solve the technical problems, the invention provides a service area vehicle track monitoring system based on data analysis, which comprises a track restoration module, a path characteristic module, a demand analysis module, a vehicle prediction module and a service area management module; The track recovery module is used for obtaining a vehicle visible track through a target detection and tracking algorithm based on a vehicle GPS signal and a camera monitoring image in a service area range, performing data cleaning and track segmentation on the visible track, using a multilayer LSTM to encode a historical visible track of each vehicle, training a network model by taking a track error as a loss function to obtain a track prediction model, taking a historical track sequence as an input, taking predicted track points as an output, performing sequence prediction by using the track prediction model, and supplementing missing track points; The path characteristic module is used for calculating a speed distribution curve of each vehicle in a service area road section based on time marks and change distances of track points, classifying the vehicles by using embedded characteristic clusters of track point sequences, establishing VGMM hybrid clustering algorithm aiming at statistical characteristics of each type of vehicles, and outputting track characteristics of each vehicle, wherein the statistical characteristics comprise a running starting point, a running end point, a path inflection point, a running direction, a track length and a stay point; the demand analysis module is used for constructing a demand prediction model fused with Monte Carlo simulation for vehicles in each service area according to the vehicle type and the vehicle track characteristics, predicting the demand of the vehicles for each resource, and establishing a target prediction model containing the influences of the factors of distance, crowding degree, resource allowance and waiting time according to the position of a supplement point in the service area, wherein the supplement point comprises a gas station, a charging pile, a parking lot, a restaurant and a toilet, inputting vehicle information in the model, and outputting the prediction probability of the vehicles going to each destination; The vehicle prediction module is used for taking each vehicle in the service area as one type of node, taking t