Method for identifying travel intention and extracting attributes of freight vehicles based on trajectories

Abstract

The present invention discloses a method for identifying the travel intention and extracting attributes of freight vehicles based on trajectories, which mainly solves the technical problem of inaccurate discrimination of travel intention based on vehicle trajectories. The solution includes: 1) obtaining relevant data of freight vehicles and performing preprocessing; 2) extracting vehicle motion characteristics according to the vehicle trajectory characteristics after preprocessing; 3) using the kernel density method to extract the origin and destination of the vehicle, i.e., the OD points, according to the position and residence time of the vehicle; 4) fusing the POI point information within the buffer range of the OD points to determine the travel intention of the freight vehicle; 5) obtaining the multi-attribute information of the itinerary through technical means such as map matching according to the OD points of the freight itinerary, and obtaining the detailed attributes of the vehicle. The present invention can fully exploit the travel characteristics of freight vehicle GNSS trajectory data, identify the travel intention of the vehicle, and effectively improve the accuracy of vehicle travel trajectory information.

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and relates to the recognition and attribute extraction of motor vehicle travel intentions. Specifically, it is a method for recognizing and extracting the travel intentions and attributes of freight vehicles based on trajectories, which can be used to optimize logistics transportation and traffic management. Background Art

[0002] The recognition and attribute extraction of vehicle travel intentions is an important research area in the field of transportation. Since the travel behavior of freight vehicles has an important impact on economic operation and urban traffic planning, understanding and predicting the intentions and attributes of freight vehicles can provide valuable information for decision-makers, thereby optimizing logistics transportation and traffic management. It has broad and practical application value for improving logistics efficiency, optimizing traffic services, and achieving sustainable development, specifically reflected in:

[0003] 1. In terms of transportation scheduling optimization: By recognizing the travel intentions of freight vehicles, the arrival time of the vehicles and the destinations of the delivered goods can be predicted more accurately, thereby helping to optimize transportation scheduling. For example, the priority and route planning of the vehicles can be adjusted according to the travel intentions of the freight vehicles, reducing transportation time and costs.

[0004] 2. In terms of road condition monitoring and congestion management: By recognizing the travel intentions and attributes of freight vehicles, the driving status and route selection of the vehicles can be monitored in real time, thereby helping to predict and manage traffic congestion. For example, the signal timing of traffic lights can be adjusted according to the travel intentions of freight vehicles to optimize the traffic capacity of the road network.

[0005] 3. In terms of vehicle safety and environmental protection: Through attribute extraction, the status and characteristics of freight vehicles can be understood, thereby realizing the monitoring and management of vehicle safety and environmental protection. For example, attributes such as the fuel type and emissions of freight vehicles can be extracted to evaluate the impact of the vehicles on the environment, promoting green transportation and reducing pollution.

[0006] 4. In terms of logistics planning and decision support: By recognizing the travel intentions and attributes of freight vehicles, valuable information about logistics planning and scheduling can be provided to decision-makers. For example, according to the travel intentions of freight vehicles, future cargo flows and demands can be predicted, providing a reference basis for logistics planning and decision-making.

[0007] At the data level, the recognition of travel intentions is basic research data for understanding individual vehicle travel behavior and predicting travel demand. In early research, the acquisition of travel purpose information mainly relied on fieldwork or questionnaires, which had the problems of high cost and low accuracy. With the development of social technology, passive traffic data has gradually become popular in analysis applications such as geographic flow analysis. However, potential information such as travel intentions still cannot be directly obtained.

[0008] At the method level, as the meaning gradually becomes clear, the trends of inference methods, data, and elements are gradually changing and enriching. The research on vehicle travel intention has evolved from the early direct judgment based on rules to the gradual development of classification methods such as supervised and unsupervised methods.

[0009] Among them, the simple rule based on the threshold of aggregated survey data was the earliest method used, which had problems such as high cost and low accuracy; later, there was supervised learning which belongs to the non-aggregated method, and it was highly adaptable to individual complex trips, but the inference process was not suitable for wide application in travel behavior research; while the unsupervised learning method characterized by unlabeled data was more applicable, but the results obtained by the unsupervised learning method may not be well explained, and its method theory needs to be further matured, and methods considering more detailed travel characteristics and activity processes need to be proposed.

[0010] The existing methods for identifying the travel intention and extracting attributes of freight vehicles usually have the following problems: 1) The commonly used vehicle travel intention has no pertinence for the scene discrimination of freight trips; freight industry vehicles have their own unique travel needs and route selection preferences, and conventional methods may not be able to meet the needs of this industry for feature extraction; 2) The resolution of the travel itinerary restoration is generally low, and it cannot finely restore the travel intention and specific attributes; conventional methods often lack spatio-temporal distribution characteristics of data such as travel time and mileage of freight vehicles, and the results in spatio-temporal analysis are insufficient, and it is impossible to timely and effectively grasp the recent industry travel situation in a certain area. Summary of the Invention

[0011] The purpose of the present invention is to propose a method for identifying the travel intention and extracting attributes of freight vehicles based on trajectories in view of the above deficiencies of the existing technologies, so as to solve the technical problem of inaccurate discrimination of travel intention based on vehicle trajectories. The present invention can effectively utilize the massive and easily obtainable GNSS trajectory data to identify the travel intention and extract attributes of freight vehicles.

[0012] The idea of implementing the present invention is as follows: First, analyze the information characteristics of freight vehicle GNSS trajectory data and perform preprocessing of trajectory point data; secondly, use publicly available POI point data to screen POI points in the freight industry; then, extract the motion characteristics of the preprocessed vehicle trajectories, and further extract their origin and destination using the kernel density method according to their positions and residence times; further, fuse the name information of POI points and the position information of the starting and ending points of each trip to construct a mapping relationship table between travel intention and related POI categories, obtain the semantic information of the vehicle at the origin and destination, and determine its travel intention; finally, according to the OD point information and road network data of the freight trip, obtain the multi-attribute information of the trip through technical means such as map matching.

[0013] The specific steps for the present invention to achieve the above object are as follows:

[0014] (1) Obtain the GNSS trajectory data information and regional road network data information of freight vehicles, and preprocess the GNSS trajectory data information to obtain the preprocessed GNSS trajectory of freight vehicles;

[0015] (2) Obtain all POI point data in the research area and preprocess it, and then screen it according to location, category, and keyword information, that is, according to the characteristics and requirements of the freight industry, use feature selection algorithms and information gain methods to extract freight-related features from the preprocessed POI point data, and eliminate information unrelated to freight;

[0016] (3) Extract the motion features of the preprocessed GNSS trajectory of freight vehicles;

[0017] (4) For the preprocessed GNSS trajectory of freight vehicles, use the kernel density method to extract its origin and destination according to the position and residence time of the vehicle in it, and obtain the itinerary data table of the freight vehicle's travel;

[0018] (5) Integrate the name information of POI points within the preset buffer range of the origin and destination, construct a mapping relationship table between travel intentions and their related POI categories, obtain the semantic information of the vehicle at the origin and destination, and determine its travel intention;

[0019] (6) According to the starting point of the freight itinerary and the regional road network data information, use the road network matching algorithm ST-Match to obtain the multi-attribute information of the itinerary, and obtain the detailed attribute information of the highway freight vehicle itinerary.

[0020] The present invention has the following advantages compared with the prior art:

[0021] First, for the attribute extraction of freight vehicles, the present invention adopts a method of extracting residence points based on kernel density, which can effectively improve the accuracy of freight vehicle behavior recognition;

[0022] Second, because the present invention performs targeted processing on POI data, retains POI points with high industry relevance, and improves the accuracy of matching between residence points and POI points, thus achieving the effect of accurately identifying travel intentions;

[0023] Third, because the present invention adopts an accurate road network matching algorithm, the trajectory restoration degree of vehicle travel is higher, thus achieving the purpose of accurately extracting vehicle travel attributes through trajectory data;

[0024] In summary, the present invention fully exploits the OD travel characteristics of highway freight vehicle trajectory data, combines the location, name, and category information of POI points, infers the travel intentions of highway freight vehicles, and combines map matching and other methods to calculate detailed attribute information of vehicle travel segments. This method provides an effective solution for information processing of massive trajectory data and identification of travel intentions of massive vehicles. The vehicle travel trajectory data required by the present invention is simple to obtain and has a large data scale. The algorithm is highly efficient and suitable for extracting travel information from large quantities of trajectory data. The generated road-level vehicle travel trajectory information is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Flowchart for the implementation of the present invention;

[0026] Figure 2 This is a schematic diagram of extracting the starting and ending points of vehicle trips based on the kernel density function of the present invention;

[0027] Figure 3 A schematic diagram of selecting POI points in the starting point and ending point buffers of the present invention;

[0028] Figure 4 A schematic diagram of map matching for a travel segment trajectory according to the present invention. DETAILED DESCRIPTION

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0030] Example 1: Refer to the attached Figure 1 The present invention proposes a method for identifying and extracting the travel intentions of freight vehicles based on trajectories, which specifically includes the following steps:

[0031] Step 1. Obtain the GNSS trajectory data and regional road network data of the freight vehicle and preprocess the GNSS trajectory data. This involves cleaning the GNSS trajectory data and removing outliers and abnormal points from the trajectory point data, ultimately obtaining the preprocessed GNSS trajectory of the freight vehicle. The GNSS trajectory data includes the freight vehicle ID, latitude and longitude coordinates, and sampling time information. The regional road network data is derived from OSM road network data. Outliers are points that deviate significantly from the preceding and following trajectory paths, while abnormal points are points that deviate significantly from the preceding and following trajectory points.

[0032] Step 2. Obtain all Point of Interest (POI) data within the research area and perform preprocessing. POIs are the so-called POI points. Then, filter them according to location, category, and keyword information. That is, based on the characteristics and requirements of the freight industry, use feature selection algorithms and information gain methods to extract freight-related features from the preprocessed POI point data and eliminate information unrelated to freight.

[0033] In this step, it is necessary to preprocess the collected POI point data, including data cleaning, removing duplicate data, handling missing values, etc., to ensure the quality and consistency of the data. According to the characteristics and requirements of the freight industry, use feature selection algorithms, information gain, etc. to extract freight-related features from the POI point data. These features include logistics centers, freight stations, warehouses, transportation companies, etc. In this way, features with a low correlation with the freight industry are excluded to improve the accuracy and efficiency of the model.

[0034] Step 3. Extract the motion features from the preprocessed GNSS trajectories of freight vehicles, including the following:

[0035] Jerk, that is, the rate of change of acceleration j i :

[0036]

[0037] The change in azimuth per unit length of the vehicle, that is, the rate of change of the direction angle bc i :

[0038] bc i = θ i+1 - θ i ,

[0039] The rate of change of azimuth per unit length of the vehicle, that is, the rate of change of the direction angle ba i :

[0040]

[0041] where, a i represents the acceleration value of the i-th trajectory point, a i+1 represents the acceleration value of the (i + 1)-th trajectory point, θ i represents the azimuth of the vehicle's travel at the i-th trajectory point, θ i+1 represents the azimuth of the vehicle's travel at the (i + 1)-th trajectory point, and ΔT is the time difference between the i-th trajectory point and the (i + 1)-th trajectory point.

[0042] Step 4. Refer to Figure 2, Schematic diagram of the present invention for extracting the starting point and ending point of vehicle trips based on the kernel density function. For the preprocessed GNSS trajectories of freight vehicles, according to the vehicle's position and residence time therein, the kernel density method is used to extract its origin and destination, and a travel data table of the freight vehicle's travel is obtained.

[0043] The above extraction of the vehicle's origin and destination is achieved by calculating the spatio-temporal characteristics, sampling frequency, instantaneous speed, and sampling interval time of the vehicle trajectory points, and further based on the time the vehicle stays at each position. The kernel density function is used to divide the preprocessed GNSS trajectories of freight vehicles into two parts: a stationary segment and a moving segment. The area with high kernel density is determined as the vehicle's residence point, and the starting point and ending point of each segment during the vehicle's travel are extracted by combining the spatio-temporal information of the moving segment. The travel data table includes the travel starting point, travel duration, travel trajectory line, travel time points, etc.

[0044] Step 5. Refer to Figure 3 , Schematic diagram of the present invention for selecting POI points in the buffer zones of the starting point and ending point. For the origin and destination information of each obtained travel segment, according to its coordinate position, that is, the starting point and ending point coordinates, a buffer zone with a certain range is preset. The setting of the buffer zone size is determined comprehensively according to the actual situation; the name information of POI points is extracted in this area, and the name information of POI points within the preset buffer zones of the origin and destination is fused to judge the attribute of this area, construct a mapping relationship table of travel intentions and their related POI categories, obtain the semantic information of the vehicle at the origin and destination, and determine its travel intention.

[0045] Step 6. According to the starting point of the freight travel and the regional road network data information, the road network matching algorithm ST-Match is used to obtain the multi-attribute information of the travel, and the detailed attribute information of the highway freight vehicle travel is obtained.

[0046] Refer to Figure 4 , Schematic diagram of map matching for the travel segment trajectory of the present invention; the above use of the road network matching algorithm ST-Match to obtain the multi-attribute information of the travel is based on the regional road network data information, and information is obtained through map matching. The attributes of the freight vehicle travel segments corresponding to each travel data table are corrected and supplemented, and a refined travel trajectory table is obtained by fusing the road data table. The finally obtained detailed attribute information of the highway freight vehicle travel includes travel duration, travel mileage, travel average speed, movement characteristics, and multi-attribute information; among them, the travel mileage is the navigable road distance rather than the Euclidean distance.

[0047] Embodiment 2: The overall implementation steps of this embodiment are the same as those of Embodiment 1. Now, the following further description is made for the kernel density function in its Step 4:

[0048] The definition of the kernel function is as follows:

[0049]

[0050] Among them, α and β represent the horizontal and vertical coordinate parameters of the spatial position, and γ represents the time parameter.

[0051] In this invention, a time variable is introduced in this step, and the density estimation value of the number of trajectory points in each spatio-temporal unit is expressed as:

[0052]

[0053] where K s (·) and K t (·) respectively represent the Gaussian kernel functions in the spatial dimension and the time dimension, n is the number of trajectory points in the trajectory, h1 and h2 are the spatial bandwidth and the time bandwidth respectively, x and y are the horizontal and vertical coordinates of the spatial position of the spatio-temporal unit respectively, and t represents time; x i , y i are respectively the horizontal and vertical coordinates of the position of the i-th trajectory point, and t i represents the time when the record of the i-th trajectory point occurs.

[0054] Embodiment 3: The overall implementation steps of this embodiment are the same as those of Embodiment 1. Now, the multi-attribute information obtained by using the road network matching algorithm ST-Match according to the starting point of the freight itinerary and the regional road network data information in step 6 of this embodiment is further described as follows:

[0055] In this embodiment, the multi-attribute information of the itinerary obtained based on the OD points of the freight itinerary, that is, the starting place and the destination, through technical means such as map matching, specifically includes the following information:

[0056] Total driving distance (d): reflecting the cumulative value of the Euclidean distance traveled by the vehicle;

[0057] Total driving time (t): reflecting the cumulative driving duration of the vehicle;

[0058] Average speed (v): reflecting the average speed of the vehicle;

[0059] Turning radius (rg): reflecting the spatial range of the vehicle's activity position;

[0060] Number of stops (sn): reflecting the number of times the vehicle stops;

[0061] Residence time (st): reflecting the time the vehicle stays;

[0062] Average sampling interval (inr): reflecting the average sampling time interval of the trajectory.

[0063] Among them, the map matching for calculating the total driving distance uses the road network matching algorithm ST-Match method, which is based on the Hidden Markov Model (HMM). By comprehensively considering the time and space characteristics of the trajectory data, as well as the topological structure and road attributes of the road network, it realizes the accurate matching of the trajectory data points to the road network. In order to achieve a better matching effect, it is necessary to preprocess the road network data, including extracting information such as the topological structure, nodes, and road attributes of the road network, as well as topological construction and time alignment, etc.

[0064] The present invention fully exploits the OD travel characteristics of the trajectory data of highway freight vehicles, combines the location, name, and category information of POI points, infers the travel intention of highway freight vehicles, and calculates the detailed attribute information of the vehicle travel section by combining methods such as map matching. This method provides a solution for the information processing of massive trajectory data and the identification of the travel intentions of massive vehicles.

[0065] The vehicle travel trajectory data required by the present invention is obtained in a simple manner, has a large data scale, and the adopted algorithm is efficient, which is suitable for extracting travel information from a large number of trajectory data; the generated vehicle travel trajectory information at the road level is more accurate.

[0066] The parts not described in detail in the present invention belong to the common general knowledge of those skilled in the art.

[0067] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Obviously, for those skilled in the art, after understanding the content and principle of the present invention, various modifications and changes in form and details may be made without departing from the principle and structure of the present invention. However, these modifications and changes based on the idea of the present invention are still within the protection scope of the claims of the present invention.

Claims

1. A method for identifying the travel intention and extracting attributes of freight vehicles based on trajectories, characterized in that, include: (1) Obtaining GNSS trajectory data information of the freight vehicle and regional road network data information, and preprocessing the GNSS trajectory data information to obtain the preprocessed GNSS trajectory of the freight vehicle; (2) Obtain all POI point data in the study area and preprocess them, then filter them according to location, category, and keyword information. That is, based on the characteristics and needs of the freight industry, use feature selection algorithms and information gain methods to extract freight-related features from the preprocessed POI point data and eliminate information that is not related to freight; (3) Extract motion features from the pre-processed GNSS trajectory of freight vehicles; (4) Based on the pre-processed GNSS trajectory of the freight vehicle and its location and residence time, the kernel density method is used to extract its origin and destination, and the travel data table of the freight vehicle is obtained; (5) Integrate the name information of POI points within the preset buffer zone of the origin and destination, build a mapping relationship table between travel intentions and their related POI categories, obtain the semantic information of the vehicle at the origin and destination, and determine its travel intention; (6) Based on the starting point of the freight trip and the regional road network data information, the road network matching algorithm ST-Match is used to obtain the multi-attribute information of the trip and obtain the detailed attribute information of the highway freight vehicle trip.

2. The method according to claim 1, wherein: The GNSS trajectory data information in step (1) includes the freight vehicle ID, longitude and latitude coordinates, and sampling time information.

3. The method according to claim 1, wherein: The preprocessing in step (1) is to clean the GNSS trajectory data of the freight vehicle, remove outliers and speeding abnormal points from the trajectory point data, and finally obtain the preprocessed GNSS trajectory of the freight vehicle.

4. The method according to claim 1, characterized in that: Step (2) pre-processes the collected POI data, including data cleaning, removing duplicate data, and processing missing values; the freight-related features include logistics centers, freight stations, warehouses, and transportation companies.

5. The method according to claim 1, wherein: The motion characteristics of step (3) include the following: Jerk, that is, the rate of change of acceleration j i : The change amount of the vehicle's azimuth per unit length, that is, the transformation rate bc of the direction angle i : bc i = θ i+1 -θ i , The change rate of the vehicle's azimuth per unit length, i.e., the change rate ba of the direction angle i : Among them, a i represents the acceleration value of the i-th trajectory point, and a i+1 represents the acceleration value of the (i + 1)-th trajectory point. θ i represents the azimuth angle of the vehicle's travel at the i-th trajectory point, and θ i+1 represents the azimuth angle of the vehicle's travel at the (i + 1)-th trajectory point. Δt is the time difference between the i-th trajectory point and the (i + 1)-th trajectory point.

6. The method according to claim 1, wherein: The vehicle origin and destination are extracted in step (4) by calculating the spatiotemporal characteristics, sampling frequency, instantaneous speed and sampling interval of the vehicle trajectory points. Based on the time the vehicle stays at each location, the pre-processed freight vehicle GNSS trajectory is divided into two parts: a stationary segment and a moving segment using a kernel density function. The area with high kernel density is determined as the vehicle's residence point. The starting point and end point of each segment during the vehicle's travel are extracted in combination with the spatiotemporal information of the moving segment.

7. The method according to claim 6, wherein: The kernel density function, in which a time variable needs to be added, and the density estimate value of the number of trajectory points in each spatio-temporal unit is expressed as: Among them, K s (·) and K t (·) respectively represent the Gaussian kernel functions in the spatial dimension and the temporal dimension. n is the number of trajectory points in the trajectory. h1 and h2 are the spatial bandwidth and the temporal bandwidth respectively. x and y are the abscissa and ordinate of the spatial position of the spatio-temporal unit respectively. t represents time; x i , y i are respectively the abscissa and ordinate of the position of the i-th trajectory point, and t i represents the time when the record of the i-th trajectory point occurs; the use of the road network matching algorithm ST-Match to obtain the multi-attribute information of the itinerary is based on the regional road network data information, obtains information through map matching, corrects and supplements the attributes of the freight vehicle itinerary segments corresponding to each itinerary data table, and fuses the road data tables to obtain a refined itinerary trajectory table.

8. The method according to claim 1, wherein: The travel data table in step (4) includes the travel starting point, travel time, travel trajectory and travel time point.

9. The method according to claim 1, wherein: The use of the road network matching algorithm ST-Match in step (6) to obtain the multi-attribute information of the trip is based on the regional road network data information, and the information is obtained through map matching. The attributes of the freight vehicle trip segment corresponding to each trip data table are corrected and supplemented, and the road data table is integrated to obtain a refined trip trajectory table.

10. The method according to claim 1, characterized in that: The detailed attribute information of the highway freight vehicle itinerary described in step (6) includes driving duration, driving mileage, average driving speed, motion characteristics, and multi-attribute information; among them, the driving mileage is the navigable road distance.

Images