WO2020244220A1 - 一种交通融合分析预测方法、系统及电子设备 - Google Patents

一种交通融合分析预测方法、系统及电子设备 Download PDF

Info

Publication number
WO2020244220A1
WO2020244220A1 PCT/CN2019/130556 CN2019130556W WO2020244220A1 WO 2020244220 A1 WO2020244220 A1 WO 2020244220A1 CN 2019130556 W CN2019130556 W CN 2019130556W WO 2020244220 A1 WO2020244220 A1 WO 2020244220A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
vehicle
information
traffic
mobile phone
Prior art date
Application number
PCT/CN2019/130556
Other languages
English (en)
French (fr)
Inventor
虞鹏飞
胡金星
宋亦然
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Publication of WO2020244220A1 publication Critical patent/WO2020244220A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
    • G08G1/0116—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from roadside infrastructure, e.g. beacons
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
    • G08G1/012—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from other sources than vehicle or roadside beacons, e.g. mobile networks
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/065—Traffic control systems for road vehicles by counting the vehicles in a section of the road or in a parking area, i.e. comparing incoming count with outgoing count
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/02—Services making use of location information
    • H04W4/029—Location-based management or tracking services
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W8/00—Network data management
    • H04W8/18—Processing of user or subscriber data, e.g. subscribed services, user preferences or user profiles; Transfer of user or subscriber data
    • H04W8/183—Processing at user equipment or user record carrier

Definitions

  • This application belongs to the field of intelligent transportation technology, and particularly relates to a traffic fusion analysis and prediction method, system and electronic equipment.
  • the existing traffic data acquisition methods include fixed detection equipment data acquisition and GPS-based data acquisition.
  • fixed detection equipment includes loop coil detector, infrared detector, microwave detector, acoustic wave detector and video image detection equipment, etc. [Ding Youjin. Talking about the development and application of video detection in the field of highway information acquisition[J ]. China Highway Traffic Information Industry, 2004].
  • the toroidal coil detector In the process of data collection, although the toroidal coil detector has the advantages of high accuracy, low cost, and simple technology, it is difficult to install, difficult to maintain, and has a certain life span;
  • the infrared detector is mainly used for vehicle detection by receiving light reflection energy, so it is affected by radiation interference, defect size, and buried depth; microwave detectors record vehicle information by recording frequency changes caused by the mutual influence of microwave sources and moving vehicles. Due to the influence of the surrounding terrain conditions, the accuracy is poor; the performance of the acoustic wave detector is easily affected by the temperature, airflow and other environments, which leads to a decrease in the accuracy of the detection; these fixed-point devices have some common problems in addition to their own defects.
  • the vehicle data obtained at a fixed point is limited, and only vehicle information of a certain road section can be obtained. For some special scenarios, such as real-time tracking of the vehicle trajectory, it is even more powerless.
  • GPS data acquisition is usually composed of three parts: positioning part, communication part and monitoring platform.
  • Each vehicle obtains its current position, time and other information through a GPS receiver, and processes the information through a dedicated interface, and then communicates to The data center transmits, the monitoring platform processes and analyzes the data, and after matching through the GIS electronic map, the current position is displayed, which is more common in practical applications.
  • GPS to acquire vehicle information, although the acquired data is accurate, the coverage and acquisition methods are limited. Generally, it is acquired through volunteers carrying positioning equipment or floating vehicles, which has certain limitations.
  • remote sensing technology has also been widely used in the transportation field.
  • aviation equipment can perceive the reflection or radiation of long-distance targets in the sky through remote sensors.
  • remote sensing technology has mature applications in traffic infrastructure information extraction, traffic flow data collection, and traffic disaster environment detection, the existing remote sensing technology has mainly collected static traffic infrastructure information and lacks integration of dynamic traffic data. .
  • PJ Tseng and CC Hung proposed an urban traffic state estimation model based on taxi GPS data, which converts the positioning data into a flow relationship model between roads, and uses the local optimal solution of the linear complexity problem based on the greedy algorithm to accurately Reflect the city’s traffic status and prove the accuracy of the model through experiments [Tseng P J, Hung C, Chang T H, etal. Real-time urban traffic sensing with GPS equipped Probe Vehicles[C]. Interbational Conference on ITS Telecommunications .IEEE,2013:306-310].
  • Zhou Qing, Qin Kun and others proposed a method for detecting urban hot spots based on taxi trajectory points, using the theory of physics midfield to calculate the degree of association between urban areas and analyzing the spatial clustering mode of urban traffic flow And research[Zhou Qing, Qin Kun, Chen Yixiang, Li Zhixin. A method for detecting hot spots of taxi trajectory based on data field[J].Geography and Geo-Information Science,2016,32(06):51-56.]. N Caceres and JP Wideberg et al.
  • This application provides a traffic fusion analysis and prediction method, system, and electronic equipment, which aim to solve at least one of the above-mentioned technical problems in the prior art to a certain extent.
  • a traffic fusion analysis and prediction method includes the following steps:
  • Step a Use fixed-point electronic capture equipment combined with mobile intelligent information collection equipment to obtain vehicle data information and obtain mobile phone signaling data;
  • Step b Extract vehicle OD data and user OD data respectively according to the vehicle data information and mobile phone signaling data;
  • Step c Construct a network topology map based on the vehicle OD data and user OD data, and use a deep learning model based on the spatiotemporal graph convolution network to perform spatiotemporal convolution operations on the spatiotemporal correlation network topology to establish a traffic flow prediction model ;
  • Step d Perform traffic flow prediction and population distribution prediction through the traffic flow prediction model.
  • the technical solution adopted in the embodiment of the present application further includes: in the step a, the use of a fixed-point electronic capture device combined with a mobile intelligent information collection device to obtain vehicle data information is specifically: collected by a fixed-point electronic capture device deployed on the road
  • the license plate number, identification site, longitude, latitude, time, license plate picture data information of the passing vehicles, and the collected vehicle data information is transferred to the database table through the sensor device;
  • the mobile intelligent information collection device is smart glasses, and the Smart glasses capture pictures of vehicles and automatically trigger instructions to send pictures.
  • the vehicle pictures are recognized through the trained deep learning model, and the recognized license plate number, longitude, latitude, time, and license plate picture data are transferred to the database table.
  • the technical solution adopted in the embodiment of the application further includes: in the step b, extracting the vehicle OD data and the user OD data according to the vehicle data information and the mobile phone signaling data, respectively, also includes: comparing the vehicle data information and the mobile phone signal
  • the data is preprocessed; the preprocessing specifically includes: validating identification of vehicle data information and mobile phone signaling data in the database table, deleting worthless data; and dealing with incomplete, noisy, repetitive and inconsistent data
  • the vehicle data information and cell phone signaling data are cleaned.
  • the technical solution adopted in the embodiment of the application further includes: in the step b, extracting the vehicle OD data and the user OD data according to the vehicle data information and the mobile phone signaling data, respectively, also includes: according to the preprocessed vehicle data information and The mobile phone signaling data obtains vehicle trajectory data and user trajectory data respectively, and extracts vehicle OD data and user OD data of each road section according to the vehicle trajectory data and user trajectory data.
  • the vehicle trajectory data acquisition method is specifically: sorting the vehicle data information according to the license plate number and time field, extracting the trajectory data of each vehicle from it, and capturing the data according to the smart glasses
  • Vehicle trajectory data repairs vehicle trajectory data at sparse location points; filters the repaired vehicle trajectory data according to time period, important road sections, and key nodes to form complete vehicle trajectory data;
  • the user trajectory data acquisition method is specifically: according to the LAC and CI fields, look up the corresponding base station longitude and latitude coordinates in the base station information table as the approximate coordinates of the mobile phone signaling data collection; delete the repeated data between adjacent mobile phone signaling data, Delete the data where the ping-pong handover occurs, delete the data that drifts, and finally extract the user track data.
  • a traffic fusion analysis and prediction system including:
  • Vehicle data collection module used to use fixed-point electronic capture equipment combined with mobile intelligent information collection equipment to obtain vehicle data information;
  • Mobile phone data acquisition module used to acquire mobile phone signaling data
  • OD data extraction module used to extract vehicle OD data and user OD data respectively according to the vehicle data information and mobile phone signaling data;
  • Network topology map construction module used to construct a network topology map according to the vehicle OD data and user OD data;
  • Predictive model building module used to use a deep learning model based on spatiotemporal graph convolution network to perform spatiotemporal convolution operations on network topology graphs with spatiotemporal correlation to establish a traffic flow prediction model;
  • Traffic prediction module used for traffic flow prediction and population distribution prediction through the traffic flow prediction model.
  • the technical solution adopted in the embodiment of the present application further includes: the vehicle data collection module adopts fixed-point electronic capture equipment combined with mobile intelligent information collection equipment to obtain vehicle data information.
  • the vehicle data information is collected by fixed-point electronic capture equipment deployed on the road. License plate number, identification site, longitude, latitude, time, license plate picture data information, and transfer the collected vehicle data information to the database table through the sensor device;
  • the mobile smart information collection device is smart glasses, and the smart glasses are used to capture photos Vehicle pictures, and automatically trigger the instruction to send pictures.
  • the vehicle pictures are recognized through the trained deep learning model, and the recognized license plate number, longitude, latitude, time, and license plate picture data information are transferred to the database table.
  • the technical solution adopted in the embodiment of the application further includes a trajectory data acquisition module, which is used to preprocess the vehicle data information and mobile phone signaling data; the preprocessing specifically includes: Carry out validity identification of vehicle data information and mobile phone signaling data, delete worthless data; and perform data cleaning on incomplete, noisy, repetitive and inconsistent vehicle data information and mobile phone signaling data.
  • a trajectory data acquisition module which is used to preprocess the vehicle data information and mobile phone signaling data; the preprocessing specifically includes: Carry out validity identification of vehicle data information and mobile phone signaling data, delete worthless data; and perform data cleaning on incomplete, noisy, repetitive and inconsistent vehicle data information and mobile phone signaling data.
  • the technical solution adopted in the embodiment of the application further includes: the trajectory data acquisition module is also used to acquire vehicle trajectory data and user trajectory data according to the preprocessed vehicle data information and mobile phone signaling data; the OD data extraction module The vehicle trajectory data and user trajectory data respectively extract vehicle OD data and user OD data of each road section.
  • the vehicle trajectory data acquisition method is specifically: sorting the vehicle data information according to the license plate number and time field, extracting the trajectory data of each vehicle from it, and capturing the data according to the smart glasses
  • Vehicle trajectory data repairs vehicle trajectory data at sparse location points; filters the repaired vehicle trajectory data according to time period, important road sections, and key nodes to form complete vehicle trajectory data;
  • the user trajectory data acquisition method is specifically: according to the LAC and CI fields, look up the corresponding base station longitude and latitude coordinates in the base station information table as the approximate coordinates of the mobile phone signaling data collection; delete the repeated data between adjacent mobile phone signaling data, Delete the data where the ping-pong handover occurs, delete the data that drifts, and finally extract the user track data.
  • an electronic device including:
  • At least one processor At least one processor
  • a memory communicatively connected with the at least one processor; wherein,
  • the memory stores instructions that can be executed by the one processor, and the instructions are executed by the at least one processor, so that the at least one processor can perform the following operations of the foregoing traffic fusion analysis and prediction method:
  • Step a Use fixed-point electronic capture equipment combined with mobile intelligent information collection equipment to obtain vehicle data information and obtain mobile phone signaling data;
  • Step b Extract vehicle OD data and user OD data respectively according to the vehicle data information and mobile phone signaling data;
  • Step c Construct a network topology map based on the vehicle OD data and user OD data, and use a deep learning model based on the spatiotemporal graph convolution network to perform spatiotemporal convolution operations on the spatiotemporal correlation network topology to establish a traffic flow prediction model ;
  • Step d Perform traffic flow prediction and population distribution prediction through the traffic flow prediction model.
  • the beneficial effects produced by the embodiments of the present application are: the traffic fusion analysis and prediction method, system and electronic equipment of the embodiments of the present application perform a combination of fixed-point electronic capture equipment and mobile intelligent information collection equipment on vehicles.
  • Tracking recognition not only retains the advantages of video tracking without missing cars, high-definition image recognition, and adapting to changing environments, but also makes up for the limitations of fixed point recognition in video tracking, and realizes the repair of sparse location point trajectory data.
  • a traffic fusion analysis prediction model is constructed.
  • the traffic flow prediction model is used to predict the future traffic volume and population distribution of each traffic district. If the prediction exceeds expectations, it can be advanced Carry out traffic control, provide decision support for traffic control and guidance in key areas, and support means for traffic monitoring and control during peak periods.
  • Figure 1 is a flowchart of a traffic fusion analysis and prediction method according to an embodiment of the present application
  • Figure 2 is a schematic diagram of the process of extracting user OD information through mobile phone signaling data
  • Figure 3 is a diagram of the temporal and spatial structure of traffic data
  • FIG. 4 is a schematic diagram of a deep learning model based on a spatiotemporal graph convolutional network according to an embodiment of the present application
  • Fig. 5 is a schematic structural diagram of a traffic fusion analysis and prediction system according to an embodiment of the present application.
  • FIG. 6 is a schematic diagram of the hardware device structure of the traffic fusion analysis and prediction method provided by an embodiment of the present application.
  • this application uses a multi-source video monitoring integration method to obtain vehicle information, combines traditional fixed-point video recognition technology with mobile smart wearable terminal technology, and implements vehicle intelligence through video license plate recognition combined with road network information Tracking and acquiring vehicle trajectory characteristics, acquiring regional population temporal and spatial distribution characteristics based on cell phone signaling data, combining information such as weather, seasons, weekends, holidays, etc., fusing vehicle data with user OD data, and constructing a deep learning model based on spatiotemporal graph convolutional network
  • the traffic fusion analysis prediction model combined with real-time road condition monitoring, can predict the future traffic flow trend of the road section, and provide decision support for vehicle traffic control and guidance.
  • FIG. 1 is a flowchart of a traffic fusion analysis and prediction method according to an embodiment of the present application.
  • the traffic fusion analysis and prediction method of the embodiment of the application includes the following steps:
  • Step 100 Use a fixed-point electronic capture device combined with a mobile intelligent information collection device to perform video intelligent tracking of the vehicle to obtain vehicle data information;
  • step 100 the vehicle data information such as the license plate number, identification site, longitude, latitude, time, and license plate picture of the passing vehicles are collected through the fixed point electronic capture equipment that has been deployed on the road, and the collected vehicle data information is transmitted through the sensor device.
  • the mobile smart information collection device is smart glasses, and smart glasses capture is flexible, and dynamically distributed based on information such as traffic flow, morning rush hour, road location, etc.; smart glasses capture the vehicle picture and automatically trigger the sending
  • the picture instructions are used to identify vehicle pictures through the trained deep learning model, and the vehicle data information such as the license plate number, longitude, latitude, time, and license plate picture obtained from the recognition is transferred to the database table through communication protocols such as WebSocket.
  • the mobile smart information collection device in the embodiments of the present application may also be other smart electronic devices other than smart glasses.
  • This application uses fixed-point electronic capture equipment combined with mobile intelligent information collection equipment to track and identify vehicles, which not only retains the advantages of video tracking without missing cars, high-definition image recognition, and adapting to changing environments, but also makes up for video tracking With the limitations of fixed point recognition, the repair of vehicle trajectory data at sparse location points can be achieved through mobile intelligent information collection equipment.
  • Step 200 Obtain cell phone signaling data
  • the mobile phone signaling data is data generated by a mobile terminal such as a mobile phone in contact with a mobile communication network.
  • the mobile phone signaling data includes data used to distinguish different users, time data of signaling data collection, area number data of signaling data collection location, type data of signaling data collection, and reason code data of signaling data collection, etc.
  • Step 300 After preprocessing the vehicle data information and mobile phone signaling data, obtain vehicle trajectory data and user trajectory data respectively;
  • step 300 the preprocessing is specifically as follows: first, the validity of the received vehicle data information and mobile phone signaling data is recognized, and the worthless data is deleted; then, for incomplete, noisy, repetitive, and inconsistent vehicles Data information and cell phone signaling data are cleaned.
  • the vehicle trajectory data acquisition method is specifically: sorting the vehicle data information according to the license plate number and time field, and extracting the trajectory data of each vehicle from it. Since fixed-point electronic capture equipment is deployed in major road sections and intersections, it is easy to cause location information to be lost during the random driving of vehicles. In order to build complete vehicle trajectory data, it is also necessary to compare sparse location points based on the vehicle trajectory data captured by smart glasses. The vehicle trajectory data is repaired. Finally, the repaired vehicle trajectory data is screened according to the time period, important road sections, and key nodes, and finally complete vehicle trajectory data is formed.
  • the specific method of obtaining user trajectory data is as follows: Since the mobile phone signaling data does not have direct latitude and longitude coordinates, it is necessary to find the corresponding base station latitude and longitude coordinates in the base station information table according to the LAC (location area) and CI (cell) fields as the mobile phone signaling data collection Approximate coordinates. There is a large error in mobile phone signaling data. In order to improve the accuracy of traffic prediction, delete the repeated data between adjacent mobile phone signaling data, delete the data that has ping-pong handover, delete the drifting data, and finally extract the user Track data.
  • LAC location area
  • CI cell
  • Step 400 Extract vehicle OD (traffic start and end point) data and user OD data of each road section according to the vehicle trajectory data and user trajectory data;
  • step 400 the vehicle OD data of the road section is extracted according to the trajectory tracked by the vehicle video intelligence.
  • the mobile phone signaling data cannot directly extract the user OD data. It is necessary to obtain the user's spatial position at a certain moment and the mobile position that changes with time through the mobile phone signaling data, using the LAC (location area) and CI (cell) fields of the signaling data, Find the longitude and latitude data of the corresponding base station in the base station information table as the approximate location of the current mobile phone signaling data collection.
  • LAC location area
  • CI cell
  • Figure 2 it is a schematic diagram of the process of extracting user OD information from mobile phone signaling data.
  • Figure 2(a) constructs a communication network composed of 36 location areas, and a small square grid represents the signal coverage of a location area.
  • the arrow indicates the change process of the user's travel location.
  • the change sequence of the location area can be obtained from the visitor location register ⁇ (L1,T1),(L2,T2)........(L16,T16),(L1,T17) ⁇ , according to the length of time the mobile phone stays in the location area and the user's moving speed in the location area to determine the starting and ending points.
  • the location area layout is converted to the traffic area layout of the road network as shown in Figure 2(b).
  • the home is located in the T1 area and the company is located in the T2 area.
  • the convenience store is located in the T3 area.
  • user OD information of "T1-T2, T2-T3, T3-T1" is obtained.
  • Step 500 Integrate and analyze the vehicle OD data, user OD data, traffic district, road section number, time, weather, traffic flow and other data, and construct a network topology map with road sections and traffic districts as network edges and nodes;
  • Step 600 Use a deep learning model based on a spatiotemporal graph convolution network to perform spatiotemporal convolution operations on a network topology map with spatiotemporal correlation to establish a traffic flow prediction model;
  • step 600 please refer to Figure 3, which is a time-space structure diagram of traffic data.
  • Each time slice is a spatial graph G, and the depth of nodes and edges represents the size of traffic flow and population distribution.
  • the traffic flow has a strong correlation in the space-time dimension. Therefore, this application adopts a deep learning model based on the space-time graph convolution network, and performs graph convolution in the space dimension and convolution operation in the time dimension. , Capture the temporal and spatial characteristics of traffic data, and establish a traffic flow prediction model.
  • FIG. 4 is a schematic diagram of a deep learning model based on a spatio-temporal graph convolutional network according to an embodiment of the present application.
  • the urban traffic flow has spatio-temporal correlation
  • the input value of the deep learning model based on spatio-temporal graph convolutional network is historical data associated with the predicted moment.
  • X 1 represents the hourly cycle time series segment
  • X 2 represents the daily cycle time series segment
  • X 3 represents the weekly cycle time series segment
  • GCN represents the convolution operation of the spatial dimension on the road network topology
  • Conv represents the time dimension Corresponding nodes do convolution operations in different time periods
  • FC means fully connected
  • y 1 , y 2 , and y 3 represent the traffic value predicted by the model
  • Fusion represents the fusion of the traffic forecast values of each input time period
  • y represents the merged traffic flow Predicted value
  • Loss represents the loss function
  • Y represents the actual traffic flow value.
  • Step 601 Select the hour, day, and weekly cycle time sequence fragments associated with the predicted time as input;
  • Step 602 Perform a graph convolution operation on the road network topology structure graph G of each time series segment, and the graph convolution operator is:
  • g ⁇ represents the convolution kernel
  • G represents the topological graph
  • the graph convolution adopts the method of spectrogram, so a graph is represented by its corresponding Laplacian matrix L.
  • L the Laplacian matrix
  • U the basis of Fourier
  • ⁇ is a diagonal matrix composed of L eigenvalues.
  • Step 603 Perform a convolution operation on the time dimension of each node to capture the time dimension feature, and the information of the node is updated by the adjacent time slice information of the node;
  • Step 604 After multiple layers of convolution of the time dimension and the space dimension, the result of the spatiotemporal convolution is made consistent with the predicted target dimension through a fully connected operation;
  • Step 605 Combine the output results of the hourly, daily, and weekly periods to obtain the final predicted value.
  • Step 700 Input relevant time and point traffic monitoring data, and perform traffic flow prediction and population distribution prediction of road sections and traffic districts based on the traffic flow prediction model.
  • FIG. 5 is a structural diagram of a traffic fusion analysis and prediction system according to an embodiment of the present application.
  • the traffic fusion analysis and prediction system of the embodiment of the application includes a vehicle data acquisition module, a mobile phone data acquisition module, a trajectory data acquisition module, an OD data extraction module, a network topology map building module, a prediction model building module, and a traffic prediction module.
  • Vehicle data collection module used to collect vehicle data information. Specifically, the vehicle data collection module includes:
  • Fixed-point electronic capture equipment used to collect vehicle data information such as license plate numbers, identification sites, longitude, latitude, time, license plate pictures of passing vehicles, and transfer the collected vehicle data information to the database table through the sensor device;
  • Mobile intelligent information collection equipment used to capture vehicle pictures, and automatically trigger instructions to send pictures, recognize vehicle pictures through the trained deep learning model, and identify the license plate number, longitude, latitude, time, license plate picture, etc.
  • Vehicle data information is transferred to the database table through communication protocols such as WebSocket.
  • the mobile smart information collection device is smart glasses, and the smart glasses have flexibility to capture photos, and dynamically distribute with reference to information such as traffic flow, morning rush hour, road location, etc.; it is understandable that the mobile in the embodiment of this application
  • the smart information collection device may also be other smart electronic devices other than smart glasses.
  • This application uses fixed-point electronic capture equipment combined with mobile intelligent information collection equipment to track and identify vehicles, which not only retains the advantages of video tracking without missing cars, high-definition image recognition, and adapting to changing environments, but also makes up for video tracking With the limitations of fixed point recognition, the repair of vehicle trajectory data at sparse location points can be achieved through mobile intelligent information collection equipment.
  • Mobile phone data acquisition module used to acquire mobile phone signaling data; mobile phone signaling data is data generated by mobile terminals such as mobile phones contacting the mobile communication network.
  • the mobile phone signaling data includes data used to distinguish different users, time data of signaling data collection, area number data of signaling data collection location, type data of signaling data collection, and reason code data of signaling data collection, etc.
  • Trajectory data acquisition module used to obtain vehicle trajectory data and user trajectory data after preprocessing the vehicle data information and cell phone signaling data; specifically, the trajectory data acquisition module includes:
  • Preprocessing unit used to identify the validity of the received vehicle data information and mobile phone signaling data, and delete worthless data; then, for incomplete, noisy, repeated and inconsistent vehicle data information and mobile phone signaling Let the data undergo data cleaning.
  • Vehicle trajectory data acquisition unit used to sort the vehicle data information according to the license plate number and time field, and extract the trajectory data of each vehicle from it. Since fixed-point electronic capture equipment is deployed in major road sections and intersections, it is easy to cause location information to be lost during the random driving of vehicles. In order to build complete vehicle trajectory data, it is also necessary to compare sparse location points based on the vehicle trajectory data captured by smart glasses. The vehicle trajectory data is repaired. Finally, the repaired vehicle trajectory data is filtered according to the time period, important road sections, and key nodes, and finally complete vehicle trajectory data is formed.
  • User trajectory data acquisition unit used to find the corresponding base station longitude and latitude coordinates in the base station information table according to the LAC (location area) and CI (cell) fields as the approximate coordinates for mobile phone signaling data collection.
  • LAC location area
  • CI cell
  • OD data extraction module used to extract vehicle OD (traffic start and end) data and user OD data of each road section according to vehicle trajectory data and user trajectory data; specifically, the OD data extraction module includes:
  • Vehicle OD data extraction unit used to extract vehicle OD data of road sections according to the trajectory of vehicle video intelligent tracking
  • User OD data extraction unit Since the user's spatial location at a certain moment and the mobile location that changes over time are obtained through mobile phone signaling data, the LAC (location area) and CI (cell) fields of the signaling data are used to search in the base station information table The longitude and latitude data to the corresponding base station is used as the approximate location when the current search mobile signaling data is collected. Based on the road network, traffic cell division and other data, the network topology of each OD area is constructed, and it is obtained according to the user's stay time at the base station and the base station topology network User travel trajectory, and extract user OD data. Specifically, as shown in Figure 2, it is a schematic diagram of the process of extracting user OD information from mobile phone signaling data.
  • Figure 2(a) constructs a communication network composed of 36 location areas, and a small square grid represents the signal coverage of a location area.
  • the arrow indicates the change process of the user's travel location.
  • the change sequence of the location area can be obtained from the visitor location register ⁇ (L1,T1),(L2,T2)........(L16,T16),(L1,T17) ⁇ , according to the length of time the mobile phone stays in the location area and the user's moving speed in the location area to determine the starting and ending points.
  • the location area layout is converted to the traffic area layout of the road network as shown in Figure 2(b).
  • the home is located in the T1 area and the company is located in the T2 area.
  • the convenience store is located in the T3 area.
  • Network topology diagram building module It is used for integrated analysis of vehicle OD data, user OD data, traffic district, road section number, time, weather, traffic flow and other data, and uses road sections and traffic districts as network diagram edges and nodes to construct network topology Figure;
  • Prediction model building module used to use the deep learning model based on the spatio-temporal graph convolution network to perform spatio-temporal convolution operations on the spatio-temporal correlation network topology map to establish a traffic flow prediction model; among them, please refer to Figure 3, which is traffic Data spatio-temporal structure diagram, each time slice is a spatial graph G, the depth of nodes and edges represents the size of vehicle flow and population distribution. It can be seen from Figure 3 that the traffic flow has a strong correlation in the space-time dimension. Therefore, this application adopts a deep learning model based on the space-time graph convolution network, and performs graph convolution in the space dimension and convolution operation in the time dimension. , Capture the temporal and spatial characteristics of traffic data, and establish a traffic flow prediction model.
  • FIG. 4 is a schematic diagram of a deep learning model based on a spatio-temporal graph convolutional network according to an embodiment of the present application.
  • the urban traffic flow has spatio-temporal correlation
  • the input value of the deep learning model based on spatio-temporal graph convolutional network is historical data associated with the predicted moment.
  • X 1 represents the hourly cycle time series segment
  • X 2 represents the daily cycle time series segment
  • X 3 represents the weekly cycle time series segment
  • GCN represents the convolution operation of the spatial dimension on the road network topology
  • Conv represents the time dimension Corresponding nodes do convolution operations in different time periods
  • FC means fully connected
  • y 1 , y 2 , and y 3 represent the traffic value predicted by the model
  • Fusion represents the fusion of the traffic forecast values of each input time period
  • y represents the merged traffic flow Predicted value
  • Loss represents the loss function
  • Y represents the actual traffic flow value.
  • g ⁇ represents the convolution kernel
  • G represents the topological graph
  • the graph convolution adopts the method of spectrogram, so a graph is represented by its corresponding Laplacian matrix L.
  • L the Laplacian matrix
  • U the basis of Fourier
  • ⁇ is a diagonal matrix composed of L eigenvalues.
  • Traffic prediction module used to input relevant time and point traffic monitoring data, and predict traffic flow and population distribution of road sections and traffic districts based on the traffic flow prediction model.
  • FIG. 6 is a schematic diagram of the hardware device structure of the traffic fusion analysis and prediction method provided by an embodiment of the present application.
  • the device includes one or more processors and memory. Taking a processor as an example, the device may also include: an input system and an output system.
  • the processor, the memory, the input system, and the output system may be connected by a bus or other methods.
  • the connection by a bus is taken as an example.
  • the memory can be used to store non-transitory software programs, non-transitory computer executable programs, and modules.
  • the processor executes various functional applications and data processing of the electronic device by running non-transitory software programs, instructions, and modules stored in the memory, that is, realizing the processing methods of the foregoing method embodiments.
  • the memory may include a program storage area and a data storage area, where the program storage area can store an operating system and an application program required by at least one function; the data storage area can store data and the like.
  • the memory may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid state storage devices.
  • the storage may optionally include storage remotely arranged with respect to the processor, and these remote storages may be connected to the processing system through a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.
  • the input system can receive input digital or character information, and generate signal input.
  • the output system may include display devices such as a display screen.
  • the one or more modules are stored in the memory, and when executed by the one or more processors, the following operations of any of the foregoing method embodiments are performed:
  • Step a Use fixed-point electronic capture equipment combined with mobile intelligent information collection equipment to obtain vehicle data information and obtain mobile phone signaling data;
  • Step b Extract vehicle OD data and user OD data respectively according to the vehicle data information and mobile phone signaling data;
  • Step c Construct a network topology map based on the vehicle OD data and user OD data, and use a deep learning model based on the spatiotemporal graph convolution network to perform spatiotemporal convolution operations on the spatiotemporal correlation network topology to establish a traffic flow prediction model ;
  • Step d Perform traffic flow prediction and population distribution prediction through the traffic flow prediction model.
  • the embodiments of the present application provide a non-transitory (non-volatile) computer storage medium, the computer storage medium stores computer executable instructions, and the computer executable instructions can perform the following operations:
  • Step a Use fixed-point electronic capture equipment combined with mobile intelligent information collection equipment to obtain vehicle data information and obtain mobile phone signaling data;
  • Step b Extract vehicle OD data and user OD data respectively according to the vehicle data information and mobile phone signaling data;
  • Step c Construct a network topology map based on the vehicle OD data and user OD data, and use a deep learning model based on the spatiotemporal graph convolution network to perform spatiotemporal convolution operations on the spatiotemporal correlation network topology to establish a traffic flow prediction model ;
  • Step d Perform traffic flow prediction and population distribution prediction through the traffic flow prediction model.
  • the embodiment of the present application provides a computer program product, the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, when the program instructions are executed by a computer To make the computer do the following:
  • Step a Use fixed-point electronic capture equipment combined with mobile intelligent information collection equipment to obtain vehicle data information and obtain mobile phone signaling data;
  • Step b Extract vehicle OD data and user OD data respectively according to the vehicle data information and mobile phone signaling data;
  • Step c Construct a network topology map based on the vehicle OD data and user OD data, and use a deep learning model based on the spatiotemporal graph convolution network to perform spatiotemporal convolution operations on the spatiotemporal correlation network topology to establish a traffic flow prediction model ;
  • Step d Perform traffic flow prediction and population distribution prediction through the traffic flow prediction model.
  • the traffic fusion analysis and prediction method, system and electronic equipment of the embodiments of the present application track and recognize vehicles by combining fixed-point electronic capture equipment and mobile intelligent information collection equipment, which not only retains video tracking without missing cars, and high-definition image recognition , Adapt to changing environments and other advantages, but also make up for the limitations of video tracking fixed point recognition, and realize the repair of sparse location point track data.
  • the deep learning model based on the spatio-temporal graph convolutional network is used to construct a traffic fusion analysis prediction model, and the traffic flow prediction model is used to predict the future traffic volume and population distribution of each traffic district. If the forecast exceeds expectations, traffic control can be carried out in advance to provide decision support for traffic control and guidance in key areas, and support means for traffic monitoring and control during peak hours.

Abstract

一种交通融合分析预测方法、系统及电子设备，通过固定点电子抓拍设备和移动智能信息采集设备相结合的方式对车辆进行追踪识别，弥补了视频追踪固定点识别的局限性，为重点区域交通控制与诱导提供决策支持。方法包括：步骤a：采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息（100），并获取手机信令数据（200）；步骤b：根据车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据（300,400）；步骤c：根据车辆OD数据和用户OD数据构建网络拓扑图（500），并采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型（600）；步骤d：通过交通流预测模型进行交通流预测和人口分布预测（700）。

Description

一种交通融合分析预测方法、系统及电子设备 技术领域
本申请属于智能交通技术领域，特别涉及一种交通融合分析预测方法、系统及电子设备。
背景技术
随着城市经济的快速发展，对车辆出行需求日渐增加，城市及区域路网中机动车流量逐年提高。机动化水平的不断提高，不但增大了道路交通的压力，同时也对交通管理带来了严峻的考验，行车难、监管难、停车难等难题是城市交通亟需解决的问题之一。
交通数据作为交通信息服务的基础，对解决交通问题具有重要的价值意义。现有的交通数据获取方式有固定式检测设备数据采集、基于GPS数据采集等方式。其中，固定式检测设备有环形线圈检测器、红外线检测器、微波检测器、声波检测器和视频图像检测设备等[丁有进.浅谈视频检测在高速公路信息采集领域的发展和应用[J].中国公路交通信息产业,2004]，在数据采集过程中，采用环形线圈检测器虽然具有精度高、成本低、技术简单等优点，但是安装难度大，维护困难，使用寿命也有一定的期限；采用红外检测器主要通过接收光反射能量进行车辆检测，所以受辐射干扰、缺陷大小、埋藏深度的影响；微波检测器通过记录微波源和运动车辆相互影响导致的频率变化记录车辆信息，该检测器由于受到周围地形条件的影响，精度较差；声波检测器其性能容易受到温度、气流等环境的影响导致检测的准确率降低；这些固定点设备除了自身存在的一些缺陷外，还存在共同的问题是固定点获取的车辆数据有限，只能获取某一路段的车辆信息，对于某些特殊情景，如需要实时追踪车辆行驶轨迹，更显的无能为力。
GPS数据采集通常由定位部分、通信部分和监控平台三部分组成，各个车辆通过配备GPS接收机获取自己当前的位置、时间等信息，并将这些信息经过专用接口加以处理，再通过无线数据通信向数据中心传递，监控平台将数据加以处理分析，并通过GIS电子地图相匹配后，显示当前位置，在实际应用中较为普遍。采用基于GPS获取车辆信息，虽然获取的数据精确，但是覆盖范围以及获取方式有限，一般通过志愿者携带定位设备或者浮动车采集获得，具有一定的局限性。
目前遥感技术在交通领域也得到了广泛的应用，随着飞机、航天器、无人机等航空设备在航空摄影中的应用，航空设备通过搭载的遥感器能感知到高空中远距离目标反射或辐射出的磁信息，对路面上行驶的车辆进行探测和识别[殷林.高分卫星遥感技术在交通运输领域的研究探索[J].数字通信世界,2018]。遥感技术虽然在交通基础设施信息提取、交通流数据采集、交通灾害环境检测等方面已有成熟的应用，但是现有的遥感技术已采集静态交通基础设施信息为主，缺乏对动态交通数据的整合。
城市交通领域的研究范围十分广泛，不少学者在城市交通领域做出了自己的分析和研究。PJ Tseng和CC Hung提出了基于出租车GPS数据的城市交通状态估计模型，将定位数据转换为道路之间的流量关系模型，并通过基于贪心算法的线性复杂度问题的局部最优解，准确的反应城市的交通状态，并通过实验证明了模型的准确性[Tseng P J,Hung C C,Chang T H,etal.Real-time urban traffic sensing with GPS equipped Probe Vehicles[C].Interbational Conference on ITS Telecommunications.IEEE,2013:306-310]。周勍、秦昆等人提出了一种基于出租车轨迹点的城市热点区域探测方法，运用物理学中场的理论计算城市区域间的关联程度，并对城市交通流量的空间聚类模式进行分析和研究[周勍,秦昆, 陈一祥,李志鑫.基于数据场的出租车轨迹热点区域探测方法[J].地理与地理信息科学,2016,32(06):51-56.]。N Caceres和JP Wideberg等提出了一种基于手机信令数据对城市区域之间进行OD流量进行估计，获取城市某一时间点交通状况用于分析和研究城市交通情况[Caceres N,Wideberg J P,Benitez F G.Deriving origin destination data from a mobile phone network[J].Intelligent Transport Systems Iet,2007,1(1):15-26]。然而，现有技术都只使用了单一的交通数据进行研究，而忽略了那些会对交通产生影响的非交通数据。
发明内容
本申请提供了一种交通融合分析预测方法、系统及电子设备，旨在至少在一定程度上解决现有技术中的上述技术问题之一。
为了解决上述问题，本申请提供了如下技术方案：
一种交通融合分析预测方法，包括以下步骤：
步骤a：采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息，并获取手机信令数据；
步骤b：根据所述车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据；
步骤c：根据所述车辆OD数据和用户OD数据构建网络拓扑图，并采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
步骤d：通过所述交通流预测模型进行交通流预测和人口分布预测。
本申请实施例采取的技术方案还包括：在所述步骤a中，所述采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息具体为：通过道路上部署的固定点电子抓拍设备采集来往车辆的车牌号、识别站点、经度、纬度、 时间、车牌图片数据信息，并通过传感器设备将采集的车辆数据信息传入数据库表中；所述移动智能信息采集设备为智能眼镜，通过所述智能眼镜抓拍车辆图片，并自动触发发送图片的指令，通过训练好的深度学习模型对车辆图片进行识别，将识别得到的车牌号、经度、纬度、时间、车牌图片数据信息传入数据库表中。
本申请实施例采取的技术方案还包括：在所述步骤b中，所述根据车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据还包括：对所述车辆数据信息和手机信令数据进行预处理；所述预处理具体为：对数据库表中的车辆数据信息和手机信令数据进行有效性识别，删除无价值的数据；并对不完整的、含噪声的、重复以及不一致的车辆数据信息和手机信令数据进行数据清洗。
本申请实施例采取的技术方案还包括：在所述步骤b中，所述根据车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据还包括：根据预处理后的车辆数据信息和手机信令数据分别获取车辆轨迹数据和用户轨迹数据，并根据所述车辆轨迹数据和用户轨迹数据分别提取各路段的车辆OD数据和用户OD数据。
本申请实施例采取的技术方案还包括：所述车辆轨迹数据获取方式具体为：将车辆数据信息根据车牌号与时间字段进行排序，从中提取每辆车的轨迹数据，并根据智能眼镜抓拍到的车辆轨迹数据对稀疏位置点的车辆轨迹数据进行修补；根据时段、重要路段、关键节点对修补后的车辆轨迹数据进行筛选，形成完整的车辆轨迹数据；
所述用户轨迹数据获取方式具体为：根据LAC与CI字段，在基站信息表中查找对应的基站经纬度坐标作为手机信令数据采集的近似坐标；对手机信令数 据相邻间重复数据进行删除、对发生乒乓切换的数据进行删除、对漂移的数据进行删除，最终提取出用户轨迹数据。
本申请实施例采取的另一技术方案为：一种交通融合分析预测系统，包括：
车辆数据采集模块：用于采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息；
手机数据获取模块：用于获取手机信令数据；
OD数据提取模块：用于根据所述车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据；
网络拓扑图构建模块：用于根据所述车辆OD数据和用户OD数据构建网络拓扑图；
预测模型构建模块：用于采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
交通预测模块：用于通过所述交通流预测模型进行交通流预测和人口分布预测。
本申请实施例采取的技术方案还包括：所述车辆数据采集模块采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息具体为：通过道路上部署的固定点电子抓拍设备采集来往车辆的车牌号、识别站点、经度、纬度、时间、车牌图片数据信息，并通过传感器设备将采集的车辆数据信息传入数据库表中；所述移动智能信息采集设备为智能眼镜，通过所述智能眼镜抓拍车辆图片，并自动触发发送图片的指令，通过训练好的深度学习模型对车辆图片进行识别，将识别得到的车牌号、经度、纬度、时间、车牌图片数据信息传入数据库表中。
本申请实施例采取的技术方案还包括轨迹数据获取模块，所述轨迹数据获取模块用于对所述车辆数据信息和手机信令数据进行预处理；所述预处理具体为：对数据库表中的车辆数据信息和手机信令数据进行有效性识别，删除无价值的数据；并对不完整的、含噪声的、重复以及不一致的车辆数据信息和手机信令数据进行数据清洗。
本申请实施例采取的技术方案还包括：所述轨迹数据获取模块还用于根据预处理后的车辆数据信息和手机信令数据分别获取车辆轨迹数据和用户轨迹数据；所述OD数据提取模块根据所述车辆轨迹数据和用户轨迹数据分别提取各路段的车辆OD数据和用户OD数据。
本申请实施例采取的技术方案还包括：所述车辆轨迹数据获取方式具体为：将车辆数据信息根据车牌号与时间字段进行排序，从中提取每辆车的轨迹数据，并根据智能眼镜抓拍到的车辆轨迹数据对稀疏位置点的车辆轨迹数据进行修补；根据时段、重要路段、关键节点对修补后的车辆轨迹数据进行筛选，形成完整的车辆轨迹数据；
所述用户轨迹数据获取方式具体为：根据LAC与CI字段，在基站信息表中查找对应的基站经纬度坐标作为手机信令数据采集的近似坐标；对手机信令数据相邻间重复数据进行删除、对发生乒乓切换的数据进行删除、对漂移的数据进行删除，最终提取出用户轨迹数据。
本申请实施例采取的又一技术方案为：一种电子设备，包括：
至少一个处理器；以及
与所述至少一个处理器通信连接的存储器；其中，
所述存储器存储有可被所述一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行上述的交通融合分析预测 方法的以下操作：
步骤a：采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息，并获取手机信令数据；
步骤b：根据所述车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据；
步骤c：根据所述车辆OD数据和用户OD数据构建网络拓扑图，并采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
步骤d：通过所述交通流预测模型进行交通流预测和人口分布预测。
相对于现有技术，本申请实施例产生的有益效果在于：本申请实施例的交通融合分析预测方法、系统及电子设备通过固定点电子抓拍设备和移动智能信息采集设备相结合的方式对车辆进行追踪识别，既保留了视频追踪无漏车、高清晰图像识别、适应多变环境等优势，也弥补了视频追踪固定点识别带来的局限性，实现稀疏位置点轨迹数据的修补。通过对车辆视频智能追踪数据与用户OD数据进行多源数据融合分析，构建交通融合分析预测模型，通过交通流预测模型进行各交通小区未来车流量及人口分布的预测，如预测超出预期则可提前进行交通调控，为重点区域交通控制与诱导提供决策支持，为高峰期的交通监控调控支撑手段。
附图说明
图1是本申请实施例的交通融合分析预测方法的流程图；
图2为通过手机信令数据提取用户OD信息的过程示意图；
图3是交通数据时空结构图；
图4是本申请实施例的基于时空图卷积网络的深度学习模型示意图；
图5是本申请实施例的交通融合分析预测系统的结构示意图；
图6是本申请实施例提供的交通融合分析预测方法的硬件设备结构示意图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。
针对现有技术存在的不足，本申请采用多源视频监测集成方式进行车辆信息获取，将传统固定点位视频识别技术与移动智能穿戴终端技术相结合，通过视频车牌识别结合路网信息进行车辆智能追踪获取车辆轨迹特征，基于手机信令数据获取区域人口时空分布特征，结合气象、时节、周末、节假日等信息，将车辆数据与用户OD数据相融合，基于时空图卷积网络的深度学习模型构建交通融合分析预测模型，结合实时路况监测即可预测路段未来交通流趋势，为车辆交通控制与诱导提供决策支持。
具体地，请参阅图1，是本申请实施例的交通融合分析预测方法的流程图。本申请实施例的交通融合分析预测方法包括以下步骤：
步骤100：采用固定点电子抓拍设备结合移动智能信息采集设备对车辆进行视频智能追踪，获取车辆数据信息；
步骤100中，通过道路上已经部署好的固定点电子抓拍设备采集来往车辆的车牌号、识别站点、经度、纬度、时间、车牌图片等车辆数据信息，并通过传感器设备将采集的车辆数据信息传入数据库表中；移动智能信息采集设备为智能眼镜，智能眼镜抓拍具有灵活性，以车流量、早高峰时间点、道路位置等 信息参考进行动态分布；智能眼镜抓拍到车辆图片后，自动触发发送图片的指令，通过训练好的深度学习模型对车辆图片进行识别，并将识别得到的车牌号、经度、纬度、时间、车牌图片等车辆数据信息通过WebSocket等通信协议传入数据库表中。可以理解，本申请实施例中的移动智能信息采集设备也可以是智能眼镜以外的其他智能电子设备。本申请通过固定点电子抓拍设备结合移动智能信息采集设备相结合的方式对车辆进行追踪识别，既保留了视频追踪无漏车、高清晰图像识别、适应多变环境等优势，也弥补了视频追踪固定点识别带来的局限性，通过移动智能信息采集设备可实现稀疏位置点车辆轨迹数据的修补。
步骤200：获取手机信令数据；
步骤200中，手机信令数据是手机等移动终端与移动通信网络联系产生的数据。手机信令数据包括用于区分不同用户的数据、信令数据采集的时间数据、信令数据采集位置区域编号数据、信令数据采集的类型数据、信令数据采集的原因编码数据等。
步骤300：对车辆数据信息和手机信令数据进行预处理后，分别获取车辆轨迹数据和用户轨迹数据；
步骤300中，预处理具体为：首先，对接收到的车辆数据信息和手机信令数据进行有效性识别，删除无价值的数据；然后，对于不完整的、含噪声的、重复以及不一致的车辆数据信息和手机信令数据进行数据清洗。
车辆轨迹数据获取方式具体为：将车辆数据信息根据车牌号与时间字段进行排序，从中提取每辆车的轨迹数据。由于固定点电子抓拍设备部署于主要的路段、路口，在车辆随机行驶过程中，容易造成位置信息丢失，为了构建完整的车辆轨迹数据，还需要根据智能眼镜抓拍到的车辆轨迹数据对稀疏位置点的车辆轨迹数据进行修补。最后，根据时段、重要路段、关键节点对修补后的车 辆轨迹数据进行筛选，并最终形成完整的车辆轨迹数据。
用户轨迹数据获取方式具体为：由于手机信令数据没有直接的经纬度坐标，需要根据LAC(位置区)与CI(小区)字段，在基站信息表中查找对应的基站经纬度坐标作为手机信令数据采集的近似坐标。手机信令数据存在较大的误差，为了提高流量预测的精度，对手机信令数据相邻间重复数据进行删除、对发生乒乓切换的数据进行删除、对漂移的数据进行删除，最终提取出用户轨迹数据。
步骤400：根据车辆轨迹数据和用户轨迹数据分别提取各路段的车辆OD(交通起止点)数据和用户OD数据；
步骤400中，根据车辆视频智能追踪的轨迹，提取路段的车辆OD数据。手机信令数据并不能直接提取用户OD数据，需要通过手机信令数据获取用户某时刻的空间位置和随时间变化的移动位置，利用信令数据的LAC(位置区)和CI(小区)字段，在基站信息表中查找到对应基站的经纬度数据作为当前搜手机信令数据采集时的近似位置，基于道路网络、交通小区划分等数据，构建各OD区域的网络拓扑图，并根据用户在基站停留时间和基站拓扑网络获取用户出行轨迹，并提取到用户OD数据。具体如图2所示，为通过手机信令数据提取用户OD信息的过程示意图。其中，图2(a)构建了一个有36个位置区组成的通信网络，一个方形小格表示一个位置区的信号覆盖范围。箭头表示的是用户的出行位置变化过程，从来访用户位置寄存器可获得位置区的变化序列{(L1,T1),(L2,T2)……..(L16,T16),(L1,T17)},根据手机在位置区域内停留的时间长短和用户在位置区域内的移动速度来判断出现的起讫点。通过建立通信网络中的位置区域和路网中交通区域之间的对应关系，将位置区布局转换为路网的交通区域布局如图2(b)所示，家位于T1区域、公司位于T2区域、便 利店位于T3区域。通过位置变化分析如图2(c)所示，获得“T1-T2、T2-T3、T3-T1、”的用户OD信息。
步骤500：将车辆OD数据、用户OD数据以及交通小区、路段编号、时间、天气、车流量等数据进行集成分析，以路段、交通小区为网络图边和节点，构建网络拓扑图；
步骤600：采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
步骤600中，请一并参阅图3，是交通数据时空结构图，每个时间片为一个空间图G，节点和边的深浅代表了车流量、人口分布的大小。从图3中可以看出交通流在时空维度上有很强的相关性，因此，本申请采用基于时空图卷积网络的深度学习模型，在空间维做图卷积、时间维做卷积操作，捕获交通数据的时空特性，建立交通流预测模型。
请参阅图4，是本申请实施例的基于时空图卷积网络的深度学习模型示意图。城市交通流存在时空相关性，基于时空图卷积网络的深度学习模型输入值是与预测时刻相关联的历史数据。图4中，X 1表示小时周期时间序列片段、X 2表示日周期时间序列片段、X 3表示周周期时间序列片段；GCN表示空间维对路网拓扑结构做图卷积操作；Conv表示时间维对应节点在不同时间段做卷积操作；FC表示全连接；y 1、y 2、y 3表示模型预测的流量值；Fusion表示将各个输入时间段的流量预测值融合；y表示融合后的车流预测值；Loss表示损失函数；Y表示实际的车流量值。
本申请实施例中，时空卷积操作具体步骤如下：
步骤601：选取与预测时刻相关联的小时、日、周周期时间序列片段作为输入；
步骤602：对每个时间序列片段的路网拓扑结构图G做图卷积操作，图卷积算子为：
g θ×G X＝g θ(L)X＝g θ(UΛU T)X＝Ug θ(Λ)U TX   (1)
公式(1)中，g θ表示卷积核，G表示拓扑图，图卷积采用谱图的方法，所以一个图用其对应的拉普拉斯矩阵L来表示，通过分析拉普拉斯矩阵及其特征值就可以得到图结构的性质。对拉普拉斯矩阵进行特征分解L＝UΛU T，U是傅里叶的基，Λ是L特征值组成的对角矩阵。
步骤603：对每个节点的时间维做卷积操作，捕获时间维特征，节点的信息被该节点相邻时间片信息更新；
步骤604：经过多层时间维与空间维的卷积后，再通过全连接操作使时空卷积的结果与预测目标维数一致；
步骤605：将小时、日、周周期的输出结果进行融合，得出最终的预测值。
步骤700：输入相关时间、点位交通监测数据，基于交通流预测模型进行路段及交通小区的交通流预测和人口分布预测。
请参阅图5，是本申请实施例的交通融合分析预测系统的结构图。本申请实施例的交通融合分析预测系统包括车辆数据采集模块、手机数据获取模块、轨迹数据获取模块、OD数据提取模块、网络拓扑图构建模块、预测模型构建模块和交通预测模块。
车辆数据采集模块：用于采集车辆数据信息，具体的，车辆数据采集模块包括：
固定点电子抓拍设备：用于采集来往车辆的车牌号、识别站点、经度、纬度、时间、车牌图片等车辆数据信息，并通过传感器设备将采集的车辆数据信息传入数据库表中；
移动智能信息采集设备：用于抓拍车辆图片，并自动触发发送图片的指令，通过训练好的深度学习模型对车辆图片进行识别，并将识别得到的车牌号、经度、纬度、时间、车牌图片等车辆数据信息通过WebSocket等通信协议传入数据库表中。本申请实施例中，移动智能信息采集设备为智能眼镜，智能眼镜抓拍具有灵活性，以车流量、早高峰时间点、道路位置等信息参考进行动态分布；可以理解，本申请实施例中的移动智能信息采集设备也可以是智能眼镜以外的其他智能电子设备。本申请通过固定点电子抓拍设备结合移动智能信息采集设备相结合的方式对车辆进行追踪识别，既保留了视频追踪无漏车、高清晰图像识别、适应多变环境等优势，也弥补了视频追踪固定点识别带来的局限性，通过移动智能信息采集设备可实现稀疏位置点车辆轨迹数据的修补。
手机数据获取模块：用于获取手机信令数据；手机信令数据是手机等移动终端与移动通信网络联系产生的数据。手机信令数据包括用于区分不同用户的数据、信令数据采集的时间数据、信令数据采集位置区域编号数据、信令数据采集的类型数据、信令数据采集的原因编码数据等。
轨迹数据获取模块：用于对车辆数据信息和手机信令数据进行预处理后，分别获取车辆轨迹数据和用户轨迹数据；具体的，轨迹数据获取模块包括：
预处理单元：用于对接收到的车辆数据信息和手机信令数据进行有效性识别，删除无价值的数据；然后，对于不完整的、含噪声的、重复以及不一致的车辆数据信息和手机信令数据进行数据清洗。
车辆轨迹数据获取单元：用于将车辆数据信息根据车牌号与时间字段进行排序，从中提取每辆车的轨迹数据。由于固定点电子抓拍设备部署于主要的路段、路口，在车辆随机行驶过程中，容易造成位置信息丢失，为了构建完整的车辆轨迹数据，还需要根据智能眼镜抓拍到的车辆轨迹数据对稀疏位置点的车 辆轨迹数据进行修补。最后，根据时段、重要路段、关键节点对修补后的车辆轨迹数据进行筛选，并最终形成完整的车辆轨迹数据。
用户轨迹数据获取单元：用于根据LAC(位置区)与CI(小区)字段，在基站信息表中查找对应的基站经纬度坐标作为手机信令数据采集的近似坐标。手机信令数据存在较大的误差，为了提高流量预测的精度，对手机信令数据相邻间重复数据进行删除、对发生乒乓切换的数据进行删除、对漂移的数据进行删除，最终提取出用户轨迹数据。
OD数据提取模块：用于根据车辆轨迹数据和用户轨迹数据分别提取各路段的车辆OD(交通起止点)数据和用户OD数据；具体的，OD数据提取模块包括：
车辆OD数据提取单元：用于根据车辆视频智能追踪的轨迹，提取路段的车辆OD数据；
用户OD数据提取单元：由于通过手机信令数据获取用户某时刻的空间位置和随时间变化的移动位置，利用信令数据的LAC(位置区)和CI(小区)字段，在基站信息表中查找到对应基站的经纬度数据作为当前搜手机信令数据采集时的近似位置，基于道路网络、交通小区划分等数据，构建各OD区域的网络拓扑图，并根据用户在基站停留时间和基站拓扑网络获取用户出行轨迹，并提取到用户OD数据。具体如图2所示，为通过手机信令数据提取用户OD信息的过程示意图。其中，图2(a)构建了一个有36个位置区组成的通信网络，一个方形小格表示一个位置区的信号覆盖范围。箭头表示的是用户的出行位置变化过程，从来访用户位置寄存器可获得位置区的变化序列{(L1,T1),(L2,T2)……..(L16,T16),(L1,T17)},根据手机在位置区域内停留的时间长短和用户在位置区域内的移动速度来判断出现的起讫点。通过建立通信网络中的位 置区域和路网中交通区域之间的对应关系，将位置区布局转换为路网的交通区域布局如图2(b)所示，家位于T1区域、公司位于T2区域、便利店位于T3区域。通过位置变化分析如图2(c)所示，获得“T1-T2、T2-T3、T3-T1、”的用户OD信息。
网络拓扑图构建模块：用于将车辆OD数据、用户OD数据以及交通小区、路段编号、时间、天气、车流量等数据进行集成分析，以路段、交通小区为网络图边和节点，构建网络拓扑图；
预测模型构建模块：用于采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；其中，请一并参阅图3，是交通数据时空结构图，每个时间片为一个空间图G，节点和边的深浅代表了车流量、人口分布的大小。从图3中可以看出交通流在时空维度上有很强的相关性，因此，本申请采用基于时空图卷积网络的深度学习模型，在空间维做图卷积、时间维做卷积操作，捕获交通数据的时空特性，建立交通流预测模型。
请参阅图4，是本申请实施例的基于时空图卷积网络的深度学习模型示意图。城市交通流存在时空相关性，基于时空图卷积网络的深度学习模型输入值是与预测时刻相关联的历史数据。图4中，X 1表示小时周期时间序列片段、X 2表示日周期时间序列片段、X 3表示周周期时间序列片段；GCN表示空间维对路网拓扑结构做图卷积操作；Conv表示时间维对应节点在不同时间段做卷积操作；FC表示全连接；y 1、y 2、y 3表示模型预测的流量值；Fusion表示将各个输入时间段的流量预测值融合；y表示融合后的车流预测值；Loss表示损失函数；Y表示实际的车流量值。
本申请实施例中，时空卷积操作具体过程如下：
1：选取与预测时刻相关联的小时、日、周周期时间序列片段作为输入；
2：对每个时间序列片段的路网拓扑结构图G做图卷积操作，图卷积算子为：
g θ×G X＝g θ(L)X＝g θ(UΛU T)X＝Ug θ(Λ)U TX  (1)
公式(1)中，g θ表示卷积核，G表示拓扑图，图卷积采用谱图的方法，所以一个图用其对应的拉普拉斯矩阵L来表示，通过分析拉普拉斯矩阵及其特征值就可以得到图结构的性质。对拉普拉斯矩阵进行特征分解L＝UΛU T，U是傅里叶的基，Λ是L特征值组成的对角矩阵。
3：对每个节点的时间维做卷积操作，捕获时间维特征，节点的信息被该节点相邻时间片信息更新；
4：经过多层时间维与空间维的卷积后，再通过全连接操作使时空卷积的结果与预测目标维数一致；
5：将小时、日、周周期的输出结果进行融合，得出最终的预测值。
交通预测模块：用于输入相关时间、点位交通监测数据，基于交通流预测模型进行路段及交通小区的交通流预测和人口分布预测。
图6是本申请实施例提供的交通融合分析预测方法的硬件设备结构示意图。如图6所示，该设备包括一个或多个处理器以及存储器。以一个处理器为例，该设备还可以包括：输入系统和输出系统。
处理器、存储器、输入系统和输出系统可以通过总线或者其他方式连接，图6中以通过总线连接为例。
存储器作为一种非暂态计算机可读存储介质，可用于存储非暂态软件程序、非暂态计算机可执行程序以及模块。处理器通过运行存储在存储器中的非暂态软件程序、指令以及模块，从而执行电子设备的各种功能应用以及数据处理， 即实现上述方法实施例的处理方法。
存储器可以包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需要的应用程序；存储数据区可存储数据等。此外，存储器可以包括高速随机存取存储器，还可以包括非暂态存储器，例如至少一个磁盘存储器件、闪存器件、或其他非暂态固态存储器件。在一些实施例中，存储器可选包括相对于处理器远程设置的存储器，这些远程存储器可以通过网络连接至处理系统。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
输入系统可接收输入的数字或字符信息，以及产生信号输入。输出系统可包括显示屏等显示设备。
所述一个或者多个模块存储在所述存储器中，当被所述一个或者多个处理器执行时，执行上述任一方法实施例的以下操作：
步骤a：采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息，并获取手机信令数据；
步骤b：根据所述车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据；
步骤c：根据所述车辆OD数据和用户OD数据构建网络拓扑图，并采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
步骤d：通过所述交通流预测模型进行交通流预测和人口分布预测。
上述产品可执行本申请实施例所提供的方法，具备执行方法相应的功能模块和有益效果。未在本实施例中详尽描述的技术细节，可参见本申请实施例提供的方法。
本申请实施例提供了一种非暂态(非易失性)计算机存储介质，所述计算机存储介质存储有计算机可执行指令，该计算机可执行指令可执行以下操作：
步骤a：采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息，并获取手机信令数据；
步骤b：根据所述车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据；
步骤c：根据所述车辆OD数据和用户OD数据构建网络拓扑图，并采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
步骤d：通过所述交通流预测模型进行交通流预测和人口分布预测。
本申请实施例提供了一种计算机程序产品，所述计算机程序产品包括存储在非暂态计算机可读存储介质上的计算机程序，所述计算机程序包括程序指令，当所述程序指令被计算机执行时，使所述计算机执行以下操作：
步骤a：采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息，并获取手机信令数据；
步骤b：根据所述车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据；
步骤c：根据所述车辆OD数据和用户OD数据构建网络拓扑图，并采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
步骤d：通过所述交通流预测模型进行交通流预测和人口分布预测。
本申请实施例的交通融合分析预测方法、系统及电子设备通过固定点电子抓拍设备和移动智能信息采集设备相结合的方式对车辆进行追踪识别，既保留 了视频追踪无漏车、高清晰图像识别、适应多变环境等优势，也弥补了视频追踪固定点识别带来的局限性，实现稀疏位置点轨迹数据的修补。通过对车辆视频智能追踪数据与用户OD数据进行融合分析，基于时空图卷积网络的深度学习模型构建交通融合分析预测模型，通过交通流预测模型进行各交通小区未来车流量及人口分布的预测，如预测超出预期则可提前进行交通调控，为重点区域交通控制与诱导提供决策支持，为高峰期的交通监控调控支撑手段。
以上所述仅是本发明的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本发明的保护范围。

Claims (11)

  1. 一种交通融合分析预测方法，其特征在于，包括以下步骤：
    步骤a：采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息，并获取手机信令数据；
    步骤b：根据所述车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据；
    步骤c：根据所述车辆OD数据和用户OD数据构建网络拓扑图，并采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
    步骤d：通过所述交通流预测模型进行交通流预测和人口分布预测。
  2. 根据权利要求1所述的交通融合分析预测方法，其特征在于，在所述步骤a中，所述采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息具体为：通过道路上部署的固定点电子抓拍设备采集来往车辆的车牌号、识别站点、经度、纬度、时间、车牌图片数据信息，并通过传感器设备将采集的车辆数据信息传入数据库表中；所述移动智能信息采集设备为智能眼镜，通过所述智能眼镜抓拍车辆图片，并自动触发发送图片的指令，通过训练好的深度学习模型对车辆图片进行识别，将识别得到的车牌号、经度、纬度、时间、车牌图片数据信息传入数据库表中。
  3. 根据权利要求2所述的交通融合分析预测方法，其特征在于，在所述步骤b中，所述根据车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据还包括：对所述车辆数据信息和手机信令数据进行预处理；所述预处理具体为：对数据库表中的车辆数据信息和手机信令数据进行有效性识别，删除无 价值的数据；并对不完整的、含噪声的、重复以及不一致的车辆数据信息和手机信令数据进行数据清洗。
  4. 根据权利要求3所述的交通融合分析预测方法，其特征在于，在所述步骤b中，所述根据车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据还包括：根据预处理后的车辆数据信息和手机信令数据分别获取车辆轨迹数据和用户轨迹数据，并根据所述车辆轨迹数据和用户轨迹数据分别提取各路段的车辆OD数据和用户OD数据。
  5. 根据权利要求4所述的交通融合分析预测方法，其特征在于，所述车辆轨迹数据获取方式具体为：将车辆数据信息根据车牌号与时间字段进行排序，从中提取每辆车的轨迹数据，并根据智能眼镜抓拍到的车辆轨迹数据对稀疏位置点的车辆轨迹数据进行修补；根据时段、重要路段、关键节点对修补后的车辆轨迹数据进行筛选，形成完整的车辆轨迹数据；
    所述用户轨迹数据获取方式具体为：根据LAC与CI字段，在基站信息表中查找对应的基站经纬度坐标作为手机信令数据采集的近似坐标；对手机信令数据相邻间重复数据进行删除、对发生乒乓切换的数据进行删除、对漂移的数据进行删除，最终提取出用户轨迹数据。
  6. 一种交通融合分析预测系统，其特征在于，包括：
    车辆数据采集模块：用于采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息；
    手机数据获取模块：用于获取手机信令数据；
    OD数据提取模块：用于根据所述车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据；
    网络拓扑图构建模块：用于根据所述车辆OD数据和用户OD数据构建网络 拓扑图；
    预测模型构建模块：用于采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
    交通预测模块：用于通过所述交通流预测模型进行交通流预测和人口分布预测。
  7. 根据权利要求6所述的交通融合分析预测系统，其特征在于，所述车辆数据采集模块采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息具体为：通过道路上部署的固定点电子抓拍设备采集来往车辆的车牌号、识别站点、经度、纬度、时间、车牌图片数据信息，并通过传感器设备将采集的车辆数据信息传入数据库表中；所述移动智能信息采集设备为智能眼镜，通过所述智能眼镜抓拍车辆图片，并自动触发发送图片的指令，通过训练好的深度学习模型对车辆图片进行识别，将识别得到的车牌号、经度、纬度、时间、车牌图片数据信息传入数据库表中。
  8. 根据权利要求7所述的交通融合分析预测系统，其特征在于，还包括轨迹数据获取模块，所述轨迹数据获取模块用于对所述车辆数据信息和手机信令数据进行预处理；所述预处理具体为：对数据库表中的车辆数据信息和手机信令数据进行有效性识别，删除无价值的数据；并对不完整的、含噪声的、重复以及不一致的车辆数据信息和手机信令数据进行数据清洗。
  9. 根据权利要求8所述的交通融合分析预测系统，其特征在于，所述轨迹数据获取模块还用于根据预处理后的车辆数据信息和手机信令数据分别获取车辆轨迹数据和用户轨迹数据；所述OD数据提取模块根据所述车辆轨迹数据和用户轨迹数据分别提取各路段的车辆OD数据和用户OD数据。
  10. 根据权利要求9所述的交通融合分析预测系统，其特征在于，所述车 辆轨迹数据获取方式具体为：将车辆数据信息根据车牌号与时间字段进行排序，从中提取每辆车的轨迹数据，并根据智能眼镜抓拍到的车辆轨迹数据对稀疏位置点的车辆轨迹数据进行修补；根据时段、重要路段、关键节点对修补后的车辆轨迹数据进行筛选，形成完整的车辆轨迹数据；
    所述用户轨迹数据获取方式具体为：根据LAC与CI字段，在基站信息表中查找对应的基站经纬度坐标作为手机信令数据采集的近似坐标；对手机信令数据相邻间重复数据进行删除、对发生乒乓切换的数据进行删除、对漂移的数据进行删除，最终提取出用户轨迹数据。
  11. 一种电子设备，包括：
    至少一个处理器；以及
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行上述1至5任一项所述的交通融合分析预测方法的以下操作：
    步骤a：采用固定点电子抓拍设备结合移动智能信息采集设备获取车辆数据信息，并获取手机信令数据；
    步骤b：根据所述车辆数据信息和手机信令数据分别提取车辆OD数据和用户OD数据；
    步骤c：根据所述车辆OD数据和用户OD数据构建网络拓扑图，并采用基于时空图卷积网络的深度学习模型对具有时空相关性的网络拓扑图进行时空卷积操作，建立交通流预测模型；
    步骤d：通过所述交通流预测模型进行交通流预测和人口分布预测。
PCT/CN2019/130556 2019-06-05 2019-12-31 一种交通融合分析预测方法、系统及电子设备 WO2020244220A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910485586.8 2019-06-05
CN201910485586.8A CN110276947B (zh) 2019-06-05 2019-06-05 一种交通融合分析预测方法、系统及电子设备

Publications (1)

Publication Number Publication Date
WO2020244220A1 true WO2020244220A1 (zh) 2020-12-10

Family

ID=67960603

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/130556 WO2020244220A1 (zh) 2019-06-05 2019-12-31 一种交通融合分析预测方法、系统及电子设备

Country Status (2)

Country Link
CN (1) CN110276947B (zh)
WO (1) WO2020244220A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113570859A (zh) * 2021-07-23 2021-10-29 江南大学 一种基于异步时空膨胀图卷积网络的交通流量预测方法
CN114355954A (zh) * 2022-03-21 2022-04-15 北京理工大学 一种无人履带车辆转向过程的跟踪控制方法和系统
CN115223402A (zh) * 2022-06-29 2022-10-21 北京航空航天大学 一种基于时空图卷积网络的空域扇区复杂度预测方法

Families Citing this family (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110276947B (zh) * 2019-06-05 2021-03-23 中国科学院深圳先进技术研究院 一种交通融合分析预测方法、系统及电子设备
CN110827540B (zh) * 2019-11-04 2021-03-12 黄传明 一种多模态数据融合的机动车移动模式识别方法及系统
CN110825604B (zh) * 2019-11-05 2023-06-30 北京博睿宏远数据科技股份有限公司 一种应用的用户轨迹和性能监控方法、装置、设备及介质
CN110909942B (zh) * 2019-11-27 2022-07-19 第四范式（北京）技术有限公司 训练模型的方法及系统和预测序列数据的方法及系统
CN112884190B (zh) * 2019-11-29 2023-11-03 杭州海康威视数字技术股份有限公司 一种流量预测方法及装置
CN111091708B (zh) * 2019-12-13 2020-11-03 中国科学院深圳先进技术研究院 车辆轨迹预测方法及装置
CN112991804B (zh) * 2019-12-18 2022-06-07 浙江大华技术股份有限公司 停留区域确定方法以及相关装置
CN111540198B (zh) * 2020-04-17 2021-07-27 浙江工业大学 基于有向图卷积神经网络的城市交通态势识别方法
CN112965466B (zh) 2021-02-18 2022-08-30 北京百度网讯科技有限公司 自动驾驶系统的还原测试方法、装置、设备及程序产品
CN113256968B (zh) * 2021-04-30 2023-02-17 山东金宇信息科技集团有限公司 一种基于手机活动数据的交通状态预测方法、设备及介质
CN113515581A (zh) * 2021-06-29 2021-10-19 湖北智凌数码科技有限公司 一种地名地址信息管理系统
CN113435356B (zh) * 2021-06-30 2023-02-28 吉林大学 一种克服观察噪声与感知不确定性的轨迹预测方法
CN113744525A (zh) * 2021-08-17 2021-12-03 东南大学 一种基于特征提取与深度学习的交通分布预测方法
CN114299727B (zh) * 2021-12-28 2022-12-09 杭州滨电信息技术有限公司 一种基于物联网和边缘计算的交通流预测系统及云平台
CN114416710B (zh) * 2021-12-29 2023-04-07 苏州大学 一种快速路车辆od位置提取方法及系统
CN114860976B (zh) * 2022-04-29 2023-05-05 长沙公交智慧大数据科技有限公司 一种基于大数据的图像数据查询方法和系统
CN116778292B (zh) * 2023-08-18 2023-11-28 深圳前海中电慧安科技有限公司 多模态车辆时空轨迹的融合方法、装置、设备及存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006085602A (ja) * 2004-09-17 2006-03-30 Gosei:Kk 交通解析システム
CN101694706A (zh) * 2009-09-28 2010-04-14 深圳先进技术研究院 基于多源数据融合的人口时空动态出行特征建模方法
CN104715601A (zh) * 2013-12-13 2015-06-17 吴建平 智能交通眼镜及基于该眼镜的工作方法
CN107040894A (zh) * 2017-04-21 2017-08-11 杭州市综合交通研究中心 一种基于手机信令数据的居民出行od获取方法
CN207164998U (zh) * 2017-09-01 2018-03-30 浙江志诚软件有限公司 一种移动式道路停车数据采集设备
CN108376472A (zh) * 2018-04-24 2018-08-07 浙江方大智控科技有限公司 基于智能交通灯od信息分析的路段管理方法及系统
CN110276947A (zh) * 2019-06-05 2019-09-24 中国科学院深圳先进技术研究院 一种交通融合分析预测方法、系统及电子设备

Family Cites Families (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101976505A (zh) * 2010-10-25 2011-02-16 中国科学院深圳先进技术研究院 交通评价方法及系统
CN104200667B (zh) * 2014-09-19 2016-07-27 上海美慧软件有限公司 一种基于手机信号数据的交通拥堵分级检测方法
US9754485B2 (en) * 2015-06-16 2017-09-05 DataSpark, PTE. LTD. Traffic prediction and real time analysis system
CN105702041A (zh) * 2016-04-21 2016-06-22 东南大学 基于神经网络的高速公路多源数据融合状态估计系统及方法
CN106571032B (zh) * 2016-11-01 2019-04-30 浙江大学 一种利用手机信令大数据和动态交通分配的od标定方法
CN106878952B (zh) * 2017-03-20 2020-07-10 迪爱斯信息技术股份有限公司 区域人员数量的预测方法及装置
CN106875686B (zh) * 2017-04-16 2020-05-08 北京工业大学 一种基于信令和浮动车数据的小汽车od提取方法
CN107480784A (zh) * 2017-06-28 2017-12-15 青岛科技大学 一种基于深度学习的手机信令数据行人交通轨迹预测方法
CN107134142B (zh) * 2017-07-10 2018-06-12 中南大学 一种基于多源数据融合的城市道路流量预测方法
CN109842848A (zh) * 2017-09-22 2019-06-04 江苏智谋科技有限公司 一种基于手机信令的区域人流量预测平台
CN107945509A (zh) * 2017-11-14 2018-04-20 武汉大学 一种道路路况图像导航方法及系统
CN108198416A (zh) * 2017-12-28 2018-06-22 金交恒通有限公司 一种手机信令与路网大数据的融合方法及其应用与系统
CN108711286B (zh) * 2018-05-29 2021-06-08 重庆市交通规划研究院 一种基于多源车联网和手机信令的交通量分配方法及系统
CN108765949A (zh) * 2018-06-06 2018-11-06 上海城市交通设计院有限公司 基于汽车电子标识技术的智能交通系统
CN109165779B (zh) * 2018-08-12 2022-04-08 北京清华同衡规划设计研究院有限公司 一种基于多源大数据与长短期记忆神经网络模型的人口数量预测方法
CN109448361B (zh) * 2018-09-18 2021-10-19 云南大学 居民交通出行流量预测系统及其预测方法
CN109637125A (zh) * 2018-11-30 2019-04-16 创发科技有限责任公司 智能路况监测系统、装置、方法和计算机可读存储介质
CN109544932B (zh) * 2018-12-19 2021-03-19 东南大学 一种基于出租车gps数据与卡口数据融合的城市路网流量估计方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006085602A (ja) * 2004-09-17 2006-03-30 Gosei:Kk 交通解析システム
CN101694706A (zh) * 2009-09-28 2010-04-14 深圳先进技术研究院 基于多源数据融合的人口时空动态出行特征建模方法
CN104715601A (zh) * 2013-12-13 2015-06-17 吴建平 智能交通眼镜及基于该眼镜的工作方法
CN107040894A (zh) * 2017-04-21 2017-08-11 杭州市综合交通研究中心 一种基于手机信令数据的居民出行od获取方法
CN207164998U (zh) * 2017-09-01 2018-03-30 浙江志诚软件有限公司 一种移动式道路停车数据采集设备
CN108376472A (zh) * 2018-04-24 2018-08-07 浙江方大智控科技有限公司 基于智能交通灯od信息分析的路段管理方法及系统
CN110276947A (zh) * 2019-06-05 2019-09-24 中国科学院深圳先进技术研究院 一种交通融合分析预测方法、系统及电子设备

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113570859A (zh) * 2021-07-23 2021-10-29 江南大学 一种基于异步时空膨胀图卷积网络的交通流量预测方法
CN113570859B (zh) * 2021-07-23 2022-07-22 江南大学 一种基于异步时空膨胀图卷积网络的交通流量预测方法
CN114355954A (zh) * 2022-03-21 2022-04-15 北京理工大学 一种无人履带车辆转向过程的跟踪控制方法和系统
CN115223402A (zh) * 2022-06-29 2022-10-21 北京航空航天大学 一种基于时空图卷积网络的空域扇区复杂度预测方法
CN115223402B (zh) * 2022-06-29 2023-05-26 北京航空航天大学 一种基于时空图卷积网络的空域扇区复杂度预测方法

Also Published As

Publication number Publication date
CN110276947A (zh) 2019-09-24
CN110276947B (zh) 2021-03-23

Similar Documents

Publication Publication Date Title
WO2020244220A1 (zh) 一种交通融合分析预测方法、系统及电子设备
Liu et al. A tailored machine learning approach for urban transport network flow estimation
KR20200121274A (ko) 전자 지도를 업데이트하기 위한 방법, 장치 및 컴퓨터 판독 가능한 저장 매체
CN107463940A (zh) 基于手机数据的车辆类型识别方法和设备
WO2021082464A1 (zh) 预测车辆的目的地的方法和装置
CN102542789A (zh) 行车路径重建方法、系统及计算机程序产品
CN105206057A (zh) 基于浮动车居民出行热点区域的检测方法及系统
CN101404119A (zh) 利用遥感影像探测和计数城市道路车辆的方法
CN107516417A (zh) 一种挖掘时空关联关系的实时高速公路流量估计方法
US20210256848A1 (en) Moving body tracking system, moving body tracking method, and program
CN105096590A (zh) 交通信息生成方法和交通信息生成设备
CN114328780A (zh) 基于六角格的智慧城市地理信息更新方法、设备及介质
Minnikhanov et al. Detection of traffic anomalies for a safety system of smart city
CN116778292A (zh) 多模态车辆时空轨迹的融合方法、装置、设备及存储介质
CN114419897A (zh) 一种基于v2x技术的城市交通cim系统及其展示方法
Zhou et al. Method for judging parking status based on yolov2 target detection algorithm
Kumar et al. Open-air Off-street Vehicle Parking Management System Using Deep Neural Networks: A Case Study
CN105761538A (zh) 基于视频识别的辅助报站方法、系统及车载终端
Yao et al. Trip end identification based on spatial-temporal clustering algorithm using smartphone positioning data
Telles et al. SParkSys: a framework for smart parking systems
Hu et al. A novel method for the detection of road intersections and traffic rules using big floating car data
Ramesh et al. Real-time vehicular traffic analysis using big data processing and IoT based devices for future policy predictions in smart transportation
CN114120631B (zh) 构建动态高精度地图的方法、装置及交通云控平台
KR102398493B1 (ko) 다중 cctv 비디오를 이용한 도시 교통 네트워크 모델링 방법 및 장치
CN114140289A (zh) 一种基于卫星遥感技术的智慧城市规划系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19931660

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19931660

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 19931660

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 13/06/2022)

122 Ep: pct application non-entry in european phase

Ref document number: 19931660

Country of ref document: EP

Kind code of ref document: A1