Travel mode determination method, device and nonvolatile storage medium

Abstract

Embodiments of the present application disclose a travel mode determination method, device and nonvolatile storage medium, the travel mode determination method comprising: obtaining positioning data, the positioning data comprising at least travel data corresponding to a single trip, the travel data comprising a plurality of trajectory points; determining travel mode feature data of the single trip according to the travel data; determining a target model from a plurality of pre-trained travel mode recognition models according to the density of the trajectory points corresponding to the single trip; inputting the travel mode feature data corresponding to the single trip into the target model to recognize the travel mode, and obtaining the travel mode corresponding to the single trip. Through the present application, the problem of inaccurate travel mode determination result in the process of determining the travel mode of a user according to positioning data in the related art is solved.

Description

Technical Field

[0001] This invention relates to the field of urban travel mode analysis, and more particularly to a travel mode determination method, apparatus, and non-volatile storage medium. Background Technology

[0002] Traffic data is crucial for traffic planning and management departments to alleviate congestion and formulate policies, especially travel mode data, which is particularly important for understanding the current state of urban travel demand. Traditionally, travel mode data mainly comes from manual travel surveys, but these surveys suffer from problems such as long cycles, high costs, low frequency, and inconsistent questionnaire quality. Therefore, combining location data with travel mode analysis has become an important research direction.

[0003] In related technologies, when analyzing travel modes by combining location data, the trajectory of each user's trip varies greatly. The accuracy of travel mode identification by analyzing data is low, resulting in inaccurate travel mode determination results in related technologies when determining users' travel modes based on location data.

[0004] There is currently no effective solution to the above problems.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background art described herein. Therefore, the background art may contain information that would not be considered part of the prior art by those skilled in the art. Summary of the Invention

[0006] This invention provides a method, apparatus, and non-volatile storage medium for determining travel modes, thereby at least addressing the problem of inaccurate travel mode identification results in the process of determining a user's travel mode based on location data in related technologies.

[0007] According to a first aspect of the present invention, a method for determining a mode of transportation is provided, comprising: acquiring location data, the location data including at least one trip's travel data, the travel data including multiple trajectory points; determining travel mode feature data for a single trip based on the travel data; determining a target model from multiple pre-trained travel mode recognition models based on the density of the trajectory points corresponding to the single trip; and inputting the travel mode feature data corresponding to the single trip into the target model for travel mode recognition to obtain the travel mode corresponding to the single trip.

[0008] Furthermore, determining the target model from multiple pre-trained travel mode recognition models based on the density of trajectory points corresponding to a single trip includes: determining the trajectory point density classification of a single trip based on the density of trajectory points corresponding to a single trip; and selecting the model corresponding to the trajectory point density classification from multiple recognition models as the target model based on a preset correspondence.

[0009] Furthermore, the trajectory point density classification includes dense trajectories and sparse trajectories. The density of trajectory points corresponding to dense trajectories is higher than that corresponding to sparse trajectories. The target model corresponding to dense trajectories is a dense trajectory recognition model, and the target model corresponding to sparse trajectories is a sparse trajectory recognition model. The travel mode determination method also includes a dense trajectory recognition model training step, which includes: determining the travel modes corresponding to multiple dense trajectories; and inputting the travel mode feature data and the travel modes corresponding to multiple dense trajectories into the dense trajectory recognition model to train the dense trajectory recognition model.

[0010] Furthermore, the travel mode determination method also includes a sparse trajectory recognition model training step, which includes: downsampling the trajectory points of multiple dense trajectories to obtain multiple pseudo-sparse trajectories; inputting the travel mode feature data corresponding to the multiple pseudo-sparse trajectories and the travel modes of the multiple dense trajectories that correspond one-to-one with the multiple pseudo-sparse trajectories into the sparse trajectory recognition model to train the sparse trajectory recognition model.

[0011] Furthermore, downsampling the trajectory points of multiple dense trajectories to obtain multiple pseudo-sparse trajectories includes: obtaining the sampling time interval pattern from multiple sparse trajectories, the sampling time interval pattern representing the time interval between trajectory points at adjacent moments in the sparse trajectory; and downsampling the trajectory points of multiple dense trajectories according to the sampling time interval pattern to obtain multiple pseudo-sparse trajectories.

[0012] Further, downsampling the trajectory points of multiple dense trajectories to obtain multiple pseudo-sparse trajectories includes: calculating the time interval between trajectory points at each adjacent time in the multiple sparse trajectories; determining the distribution of transition probabilities based on the multiple time intervals, where the transition probability is the probability of any first interval value changing to any second interval value, and the first and second interval values ​​are the values ​​of two adjacent time intervals; determining multiple sampling intervals corresponding to each pseudo-sparse trajectory based at least on the distribution of transition probabilities; and sampling multiple trajectory points of each dense trajectory based on the multiple sampling intervals to determine each pseudo-sparse trajectory.

[0013] Furthermore, at least based on the distribution of transition probabilities, determining multiple sampling intervals corresponding to each pseudo-sparse trajectory includes: determining a first sampling interval, which is the time interval between any two adjacent sampling points; and determining the time interval with the highest transition probability as a second sampling interval based on the distribution of transition probabilities, which is the sampling interval located after and adjacent to the first sampling interval.

[0014] Furthermore, based on the density of trajectory points corresponding to a single trip, the trajectory point density classification of a single trip is determined as follows: if the density of trajectory points corresponding to a single trip is greater than or equal to a preset density threshold, the trajectory point density classification of a single trip is determined as a dense trajectory; if the density of trajectory points corresponding to a single trip is less than a preset density threshold, the trajectory point density classification of a single trip is determined as a sparse trajectory.

[0015] Furthermore, the method for determining travel modes also includes a step of determining a preset density threshold, which includes: determining one or more classification features, where the classification features belong to travel mode feature data; and determining the density value at which one or more classification features fail as the density of trajectory points corresponding to a single trip decreases, which is the preset density threshold.

[0016] Furthermore, the method for determining travel modes also includes: determining travel mode data based on the travel modes corresponding to each trip included in the location data; calculating the expansion coefficient based on the number of users involved in the location data and the number of urban population statistics; and expanding the travel mode data based on the expansion coefficient to obtain travel mode statistical data.

[0017] Furthermore, after expanding the travel mode data according to the expansion coefficient to obtain travel mode statistics, the travel mode determination method also includes: obtaining urban public transportation card swipe data; and calibrating the travel mode statistics based on the urban public transportation card swipe data.

[0018] According to a second aspect of the present invention, a travel mode determination apparatus is also provided, comprising: a first acquisition unit, configured to acquire positioning data, the positioning data including at least one travel data corresponding to a single trip, the travel data including multiple trajectory points; a first determination unit, configured to determine travel mode feature data of a single trip based on the travel data; a second determination unit, configured to determine a target model from multiple pre-trained travel mode recognition models based on the density of the trajectory points corresponding to the single trip; and a recognition unit, configured to input the travel mode feature data corresponding to the single trip into the target model for travel mode recognition, thereby obtaining the travel mode corresponding to the single trip.

[0019] According to a third aspect of the present invention, a method for determining traffic evaluation indicators is also provided, comprising: using the above-described method for determining travel modes to determine the travel modes of multiple trips included in the location data; and calculating traffic evaluation indicators based at least on the travel modes of the multiple trips, wherein the traffic evaluation indicators are used to characterize the traffic quality within a target area.

[0020] According to a fourth aspect of the present invention, a non-volatile storage medium is also provided, wherein the non-volatile storage medium includes a stored program, wherein, when the program is running, the device where the non-volatile storage medium is located executes the above-described travel mode determination method; or, when the program is running, the device where the non-volatile storage medium is located executes the above-described traffic evaluation index determination method.

[0021] The travel mode determination method of this invention, after acquiring location data, determines travel mode feature data based on travel data corresponding to a specific trip within the location data. This travel mode feature data is related to the user's travel mode, and different travel modes will result in different feature data. After obtaining the travel mode feature data, a target model is selected from multiple pre-trained travel mode recognition models based on the density of trajectory points corresponding to a single user trip. This target model is then used to identify the aforementioned travel mode feature data, thereby determining the travel mode used by the user for this trip. This travel mode determination method does not directly input the travel mode feature data into a uniform recognition model for travel mode identification. Instead, it selects different travel mode recognition models based on the different density of trajectory points. This targeted selection of appropriate recognition models based on different trajectory point densities effectively reduces or even avoids the adverse effects of varying trajectory point densities on travel mode recognition accuracy, improving the accuracy of travel mode recognition and solving the problem of inaccurate travel mode recognition results in related technologies when determining a user's travel mode based on location data. Attached Figure Description

[0022] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0023] Figure 1 This is a flowchart illustrating a method for determining a mode of transportation provided in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of a travel mode determination device provided in an embodiment of the present invention;

[0025] Figure 3This is a schematic diagram illustrating the use of the mode of travel determination method according to an embodiment of the present invention;

[0026] Figure 4 This is a schematic diagram illustrating the travel mode identification method according to an embodiment of the present invention. Detailed Implementation

[0027] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0028] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use, and processing of such data must comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse. The terms "first," "second," etc., in the specification, claims, and drawings of this invention are used to distinguish different objects, not to limit a specific order.

[0029] The applicant's research and analysis revealed that different travel trajectories contain varying density of trajectory points—some trajectories have densely packed points, while others have sparsely packed points. This density affects the calculation results of travel mode feature data, significantly impacting the accuracy of travel mode identification based on this data. For example, some travel mode feature data is highly sensitive to changes in trajectory point density (e.g., speed distribution at different percentiles during a trip); variations in trajectory point density within a certain range result in significant changes in the value, thus affecting the accuracy of travel mode identification. Conversely, some travel mode feature data is insensitive to changes in trajectory point density (e.g., average speed during a trip). To reduce or avoid the adverse effects of varying trajectory point density on travel mode identification results, this application proposes a travel mode determination method.

[0030] Figure 1 This is a method for determining the mode of transportation according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0031] Step S102: Obtain location data. The location data includes at least one trip's travel data, and the travel data includes multiple trajectory points.

[0032] Step S104: Based on the travel data, determine the travel mode characteristic data for a single trip;

[0033] Step S106: Determine the target model from multiple pre-trained travel mode recognition models based on the density of trajectory points corresponding to a single trip;

[0034] Step S108: Input the travel mode feature data corresponding to a single trip into the target model to identify the travel mode and obtain the travel mode corresponding to the single trip.

[0035] The travel mode determination method described above, after acquiring location data, determines travel mode feature data based on the travel data corresponding to a specific trip within the location data. This travel mode feature data is related to the user's travel mode, and different travel modes will correspond to different feature data. After obtaining the travel mode feature data, a target model is selected from multiple pre-trained travel mode recognition models based on the density of trajectory points corresponding to a single user trip. This target model is then used to identify the aforementioned travel mode feature data, thereby determining the user's travel mode for that trip. This method does not directly input the travel mode feature data into a uniform recognition model for travel mode identification. Instead, it selects different travel mode recognition models based on the varying density of trajectory points. This targeted selection of appropriate recognition models based on different trajectory point densities effectively reduces or even avoids the adverse effects of varying trajectory point densities on travel mode recognition accuracy, improving the accuracy of travel mode identification and solving the problem of inaccurate travel mode recognition results in related technologies when determining user travel modes based on location data.

[0036] It should be noted that the travel data for a single trip mentioned above includes multiple trajectory points, which belong to multiple times. There is no explicit correspondence between these times and trajectory points; one time may correspond to multiple trajectory points, and one trajectory point may correspond to multiple times. The travel mode identification model mentioned above can be flexibly selected according to the actual situation, such as AdaBoost, logistic regression, or random forest models. The travel mode feature data mentioned above can be various types of feature data, as long as it is related to the travel mode. For example, it can include travel time, travel distance, average travel speed, speed distribution at different percentiles, trajectory point interval time, average distance between the travel trajectory and the public transport network, average distance between the travel trajectory and the subway network, average distance between the travel trajectory and subway stations, etc.

[0037] Determining the travel mode characteristics of a single trip based on travel data includes: determining the travel trajectory corresponding to the single trip based on the travel data; and determining the travel mode characteristics of the single trip based on the travel trajectory. In other words, after obtaining the travel data for a single trip, the trajectory of the single trip can be determined based on the multiple trajectory points included, and then the travel mode characteristics related to the travel mode can be analyzed. For example, if the distribution of multiple trajectory points included in the travel data is relatively regular, corresponding to different trajectory points at different times, then the trajectory points corresponding to different times can be connected to form the travel trajectory corresponding to the user's single trip.

[0038] In some complex scenarios, the distribution of trajectory points in travel data can be quite complex. For example, if the data comes from different apps, due to the different positioning methods and times used by each data source, there may be several temporal and / or spatial anomalies, leading to positioning drift, positioning oscillation, etc. In this case, this embodiment will perform data cleaning on the travel data to improve the accuracy of travel trajectory determination, thereby removing abnormal trajectory points and increasing data quality and robustness. For example, preset rules or algorithms (such as the 3-sigma model) can be used to identify and clean drifting, oscillating, or repetitive trajectory points. After cleaning the data, in a preferred embodiment, the travel mode determination method further includes: clustering multiple trajectory points in the travel trajectory, for example, using the ST-DBSCAN (Spatial Temporal-Density-Based Spatial Clustering of Application with Noise) spatial clustering model to cluster the trajectory points, obtaining multiple clusters corresponding to multiple times; and determining the travel trajectory corresponding to the above-mentioned single trip based on multiple clusters. By clustering trajectory points, the complex distribution of trajectory points can be simplified into multiple clusters, each of which can be considered a stop point. After obtaining multiple clusters, the travel trajectory can be easily determined. Moreover, this processing method has strong robustness and is less affected by noise points and outliers. The 3-sigma model is an outlier filtering model. It assumes that a set of detection data contains only random errors, calculates the standard deviation, and determines an interval with a certain probability. Errors exceeding this interval are considered gross errors, not random errors, and data containing such errors should be discarded.

[0039] In one specific embodiment, to facilitate the selection of a target model for analyzing travel modes, determining the target model from multiple pre-trained travel mode recognition models based on the density of trajectory points corresponding to a single trip includes: determining the trajectory point density category to which the single trip belongs based on the density of trajectory points corresponding to the single trip; and selecting the model corresponding to the trajectory point density category from multiple recognition models as the target model according to a preset correspondence. In this embodiment, different trajectory point density categories are defined based on the different densities of trajectory points in a single trip, and each trajectory point density category has a corresponding travel mode recognition model. Thus, after determining the trajectory point density category to which the single trip belongs, the target model can be easily selected based on the aforementioned correspondence.

[0040] In one specific embodiment, the trajectory point density classification includes dense trajectories and sparse trajectories. The density of trajectory points corresponding to dense trajectories is higher than that corresponding to sparse trajectories. Here, density can refer to the density of all trajectory points in the travel trajectory or the density of a portion of the trajectory points. The target model for dense trajectories is a dense trajectory recognition model, and the target model for sparse trajectories is a sparse trajectory recognition model. The travel mode determination method also includes a dense trajectory recognition model training step, which includes: determining the travel modes corresponding to multiple dense trajectories; and inputting the travel mode feature data and the travel modes corresponding to multiple dense trajectories into the dense trajectory recognition model to train the dense trajectory recognition model. In this embodiment, trajectory point density is divided into two categories: dense trajectories and sparse trajectories. For dense trajectories, their travel modes can be determined more conveniently and accurately, which is then used to train the dense trajectory recognition model.

[0041] In practice, there are various methods to determine the travel modes corresponding to dense trajectories. For example, in one preferred method, some dense trajectories contain navigation source information. This information represents the navigation category corresponding to a user's single trip, such as driving, public transportation, cycling, walking, etc. Based on this navigation source information, the travel mode corresponding to the dense trajectories can be easily determined, enabling automatic tagging of travel modes. Another example is the use of manual methods based on experience to determine the travel modes corresponding to dense trajectories.

[0042] In the training process of a sparse trajectory recognition model, manual labeling of sparse trajectories based on experience is required to obtain training samples. This process is not only time-consuming and labor-intensive, but the accuracy of manual travel mode labeling is also poor, easily leading to low recognition accuracy of the final trained sparse trajectory recognition model. To solve this problem, the model training steps in this embodiment also include a sparse trajectory recognition model training step, which includes: downsampling the trajectory points of multiple dense trajectories to obtain multiple pseudo-sparse trajectories; and inputting the travel mode feature data corresponding to the multiple pseudo-sparse trajectories and the travel modes of the multiple dense trajectories that correspond one-to-one with the multiple pseudo-sparse trajectories into the sparse trajectory recognition model to train the sparse trajectory recognition model.

[0043] The travel mode determination method in this embodiment simulates multiple pseudo-sparse trajectories by downsampling the trajectory points of multiple dense trajectories. These simulated pseudo-sparse trajectories are then used to train a sparse trajectory recognition model. In this process, downsampling the trajectory points of dense trajectories artificially reduces the number of trajectory points, thus simulating the sparse trajectory scenario. The travel modes corresponding to the dense trajectories can then be directly used as the travel modes for the pseudo-sparse trajectories. Since the accuracy of travel modes corresponding to dense trajectories is high, this method results in higher accuracy for training the sparse trajectory recognition model. Furthermore, it eliminates the need for manual labeling of travel modes on the sparse trajectories, effectively simplifying the training process.

[0044] Since the training samples used in the sparse trajectory recognition model training process are pseudo-sparse trajectories simulated by downsampling dense trajectories, and are not real sparse trajectories, the simulation accuracy of the pseudo-sparse trajectories directly affects the recognition accuracy of the sparse trajectory recognition model. In this embodiment, in order to improve the simulation accuracy of the pseudo-sparse trajectories and make them closer to real sparse trajectories, the trajectory points of multiple dense trajectories are downsampled to obtain multiple pseudo-sparse trajectories. This includes: obtaining the sampling time interval pattern from multiple sparse trajectories, where the sampling time interval pattern represents the time interval between trajectory points at adjacent moments in the sparse trajectory; and downsampling the trajectory points of multiple dense trajectories according to the sampling time interval pattern to obtain multiple pseudo-sparse trajectories.

[0045] In other words, during the downsampling of trajectory points in a dense trajectory, the sampling time interval pattern is first obtained from multiple real sparse trajectories. Then, based on this pattern, the trajectory points of the dense trajectory are downsampled. In the process of downsampling trajectory points in a dense trajectory, pseudo-sparse trajectories simulated using some simple downsampling methods have poor fidelity (e.g., taking a sample from the trajectory points of the dense trajectory at fixed time intervals). In this embodiment, by obtaining the sampling time interval pattern from real sparse trajectories and then downsampling the dense trajectory according to this pattern, the generated pseudo-sparse trajectory can be made closer to the real sparse trajectory, thereby improving the recognition accuracy of the trained sparse trajectory recognition model.

[0046] As the name suggests, the sampling time interval pattern described above represents the time interval between adjacent trajectory points in a sparse trajectory, that is, how often a trajectory point appears in a real sparse trajectory. After obtaining this pattern, downsampling operations can be performed more effectively on dense trajectories, ensuring that the pseudo-sparse trajectory generated by downsampling is closer to the real sparse trajectory.

[0047] In one specific implementation, to facilitate downsampling and ensure the simulation effect of pseudo-sparse trajectories on real sparse trajectories, the sparse trajectory collection process is regarded as a first-order Markov process (a Markov process is a type of stochastic process that assumes the current state depends only on the previous state), meaning the next collection time depends only on the previous collection time. After determining the distribution pattern of the Markov process, if the time interval between two trajectory point collections is known, the next time interval can also be determined. Specifically, downsampling the trajectory points of multiple dense trajectories to obtain multiple pseudo-sparse trajectories includes: calculating the time interval between trajectory points at each adjacent time in the multiple sparse trajectories; determining the distribution of transition probabilities based on the multiple time intervals, where the transition probability is the probability of any first interval value changing to any second interval value, and the first and second interval values ​​are the values ​​of two adjacent time intervals; determining multiple sampling intervals corresponding to each pseudo-sparse trajectory based at least on the distribution of transition probabilities; and sampling multiple trajectory points of each dense trajectory based on the multiple sampling intervals to determine each pseudo-sparse trajectory.

[0048] In this embodiment, to determine the distribution of Markov processes during the acquisition of trajectory points in a sparse trajectory, the time intervals between trajectory points at adjacent moments in the sparse trajectory are first calculated. Then, based on these time intervals, the distribution of transition probabilities is determined through calculation. A transition probability is the probability that any first interval value changes to any second interval value, where the first and second interval values ​​are the values ​​of two adjacent time intervals. The first and second interval values ​​can be the same or different. For example, for ease of understanding, a transition probability P(1,2) = 0.3 represents the probability that the time interval between two adjacent sampling time intervals changes from 1 second to 2 seconds; a transition probability P(1,4) = 0.1 represents the probability that the time interval between two adjacent sampling time intervals changes from 1 second to 4 seconds; and a transition probability P(5,3) represents the probability that the time interval between two adjacent sampling time intervals changes from 5 seconds to 3 seconds. Knowing the distribution of transition probabilities, only the time interval between two trajectory points needs to be known to calculate the relationship between the interval length of the next sampling time interval and its probability of occurrence. For example, knowing that the time interval between the two currently sampled trajectory points is 1 second, then according to P(1,2) and P(1,4) above, the probability of the next sampling time interval being 2 seconds is 0.3, and the probability of the next sampling time interval being 4 seconds is 0.1. By downsampling the dense trajectory based on the distribution of the above transition probabilities, the sampling pattern of the real sparse trajectory can be easily and accurately fitted, thus making the generated pseudo-sparse trajectory closer to the real sparse trajectory.

[0049] In this embodiment, determining multiple sampling intervals corresponding to each pseudo-sparse trajectory based on the distribution of transition probabilities includes: determining a first sampling interval, which is the time interval between any two sampling points; and determining the time interval with the highest transition probability as a second sampling interval, which is the sampling interval following and adjacent to the first sampling interval. In other words, knowing the first sampling interval (first sampling time interval), in order to determine the next sampling time interval and perform downsampling, the time interval with the highest transition probability is determined as the next sampling time interval (i.e., the second sampling interval) based on the distribution of transition probabilities. For example, since P(1,2) = 0.3 is greater than P(1,4) = 0.1, the next sampling time interval (second sampling interval) is determined to be 2 seconds if the first sampling interval is 1 second. This probability-based method of determining the sampling time interval can better fit the sampling interval pattern of trajectory points in a real sparse trajectory, thereby better simulating the pseudo-sparse trajectory.

[0050] As described above, the travel mode determination method of this embodiment of the invention employs different recognition models for sparse and dense trajectories when identifying travel modes. The criteria for classifying dense and sparse trajectories can be flexibly set based on experience or actual needs. In an optional implementation, determining the trajectory point density classification of a single trip based on the density of trajectory points corresponding to that trip includes: classifying the travel trajectory as a dense trajectory if the density of trajectory points corresponding to that trip is greater than or equal to a preset density threshold; and classifying the travel trajectory as a sparse trajectory if the density of trajectory points corresponding to that trip is less than the preset density threshold. In this embodiment, by setting a preset density threshold, a travel trajectory is determined to be a dense trajectory if the density of trajectory points corresponding to that trip is greater than or equal to the preset density threshold; otherwise, it is determined to be a sparse trajectory. The density of trajectory points mentioned here can be the average density of trajectory points across the entire travel trajectory, or the density of trajectory points at a certain segment of the travel trajectory.

[0051] Building upon the aforementioned trajectory point density classification based on a preset density threshold, to make the trajectory point density classification more targeted and to ensure that the classified dense and sparse trajectories can be better identified by dense trajectory recognition models and sparse trajectory recognition models, this embodiment further includes a step of determining a preset density threshold. This step includes: determining one or more classification features, where the classification features belong to travel mode feature data; and determining the density value at which one or more classification features become ineffective as the trajectory point density corresponding to a single trip decreases, thus establishing the preset density threshold.

[0052] The classification features, which belong to the aforementioned travel mode feature data, can be various types of features, such as travel time, travel distance, average travel speed, speed distribution at different percentiles, trajectory point interval time, average distance between the travel trajectory and the public transport network, average distance between the travel trajectory and the subway network, average distance between the travel trajectory and subway stations, etc. The determination of the preset density threshold is combined with the aforementioned classification features. As the density of trajectory points corresponding to the travel trajectory decreases, when one or more classification features become ineffective, the corresponding density value is determined to be the aforementioned preset density threshold. The criteria for determining the ineffectiveness of the aforementioned classification features can be flexibly determined according to the actual situation. This method can link trajectory point density with travel mode feature data, identifying which travel mode features are more sensitive to changes in trajectory point density, thereby classifying travel trajectory point density and using differentiated travel mode identification models for recognition. This allows travel trajectories belonging to different density classifications to be better identified by the corresponding travel mode identification models, thus improving the accuracy of travel mode identification. For example, in one embodiment, only the speed distribution at different percentiles during the trip is used as a classification feature. When the trajectory point density is high, this speed distribution is more accurate; as the trajectory point density decreases, the accuracy of this speed distribution becomes increasingly worse. When a critical condition is determined and defined as the failure of the classification feature, the critical density value can be determined as a preset density value. In this way, the determination of the preset density threshold is correlated with the travel mode feature of speed distribution at different percentiles, thereby improving the accuracy of trajectory point density classification and subsequent identification.

[0053] The above-described method for determining travel modes can identify the travel modes used for each trip included in the location data after travel mode identification. However, the population counted may not represent the entire population of the city or a specific area within the city, thus introducing sampling bias. To address this issue, the travel mode determination method in this embodiment further includes: determining travel mode data based on the travel modes corresponding to each trip included in the location data; calculating an expansion coefficient based on the number of users involved in the location data and the city's population statistics; and expanding the travel mode data according to the expansion coefficient to obtain travel mode statistical data.

[0054] By combining the number of users involved in location data with the city's population statistics, an expansion coefficient can be calculated, thereby expanding the travel mode data. The city's population statistics can be the total population of the city or the population of a specific area (a street or region). The expanded travel mode statistics provide a more complete picture of citizens' travel patterns, assisting urban traffic planning and management departments in congestion mitigation and policy development. It can also serve various application platforms such as the Transportation Operations Coordination Center (TOCC) and the Traffic Management Bureau's "Traffic Brain," helping urban traffic managers better monitor and analyze traffic conditions.

[0055] After expanding the travel mode data to obtain travel mode statistics, these statistics may still deviate from actual conditions because they are derived from sampling and expansion. Therefore, the method for determining travel modes after expanding the travel mode data using expansion coefficients also includes: acquiring urban public transportation card swipe data; and calibrating the travel mode statistics based on this data.

[0056] In this embodiment, the statistical data on the mode of travel will be calibrated based on urban public transportation card swiping data (such as bus card swiping data, rail transit card swiping data, etc.) so that the final calibrated data can more accurately and comprehensively reflect the urban transportation mode of travel.

[0057] As mentioned above, travel mode characteristic data can be various types of characteristic data, as long as they are related to the travel mode. For example, it can include travel time, travel distance, average travel speed, speed distribution at different percentiles, trajectory point interval time, average distance between the travel trajectory and the public transport network, average distance between the travel trajectory and the subway network, average distance between the travel trajectory and subway stations, etc. In this embodiment, travel mode characteristic data includes motion characteristic data and / or geographic characteristic data. Motion characteristic data is related to the user's motion state during travel, such as travel time, travel distance, average travel speed, speed distribution at different percentiles, trajectory point interval time, etc. Geographic characteristic data is related to the geographical locations the user passes through during travel, such as the average distance between the travel trajectory and the public transport network, the average distance between the travel trajectory and the subway network, the average distance between the travel trajectory and subway stations, the distance between the travel trajectory and a point of interest, etc.

[0058] In a preferred embodiment, the location data is location-based service data. Location-based service data, or LBS data, utilizes various types of positioning technologies to obtain the current location of the positioning device and provides information resources and basic services to the positioning device via the mobile internet. LBS data can integrate data from multiple apps using location SDKs (Software Development Kits, a collection of development tools used by software engineers to build application software for specific software packages, frameworks, hardware platforms, operating systems, etc.) to reconstruct a user's activity trajectory. Compared with positioning methods using mobile phone signaling, it has the following advantages: High positioning accuracy: The location SDK uses a hybrid positioning mode of GPS + base station (including 5G base station) + WIFI, achieving meter-level accuracy in most cases; Good data continuity: LBS data integrates data from multiple apps using location SDKs to reconstruct a user's activity trajectory, ensuring accurate identification of important travel information; High data mining potential: The high positioning accuracy of LBS data, combined with rich internet map data, such as public transportation networks, road networks, and Points of Interest (POI) data, can mine important information such as transportation modes and travel purposes, ensuring accurate identification of important travel information.

[0059] Figure 3 This is a schematic diagram illustrating the use of an embodiment of the travel mode determination method of the present invention, as shown below. Figure 3As shown, in this embodiment, the basic data relied upon for determining travel modes using this method includes three parts: 1. LBS trajectory data; 2. Road network and public transport network data; 3. Population and other statistical data. Based on the above data, step 1, data preprocessing and travel identification, is performed first: abnormal data in the LBS trajectory data is cleaned as described above, and the origin and destination points of a single trip are identified to obtain the travel trajectory of that single trip. Then, step 2, travel mode feature calculation, is performed: travel mode-related features (i.e., the travel mode feature data mentioned above) are calculated. As mentioned above, travel mode feature data can be various types of feature data, as long as they are related to the travel mode. For example, it can include travel time, travel distance, average travel speed, speed distribution at different percentiles, trajectory point interval time, average distance between the travel trajectory and the public transport network, average distance between the travel trajectory and the public transport network, average distance between the travel trajectory and the public transport network, average distance between the travel trajectory and the public transport station, etc. During the travel mode identification process, different target models are selected for identification based on the density of trajectory points of a single trip. Step 3, travel density classification, is performed: trajectory point density is classified based on the density of trajectory points of a single trip. Steps 4.1 and 4.2 demonstrate the methods for acquiring training data during the training process of different travel mode recognition models. In this embodiment, trajectory point density classification includes dense trajectories and sparse trajectories. Correspondingly, the travel mode recognition models include dense trajectory recognition models and sparse trajectory recognition models. For dense trajectories, as shown in steps 4.1 and 5.1, for some travel trajectories in LBS positioning data, there is navigation source information. The navigation source information represents the navigation category corresponding to a user's single trip, such as driving, public transportation, cycling, walking, etc. Based on the navigation source information, the travel mode corresponding to the dense trajectory can be easily determined, realizing automatic labeling of the travel mode. Then, based on the travel mode characteristics of a single trip corresponding to the dense trajectory and the above-mentioned labeled tags, the dense trajectory model (i.e., the dense trajectory recognition model) can be trained. After the dense trajectory recognition model is trained, the travel mode feature data corresponding to the dense trajectory to be identified is input into the model, and the corresponding travel mode can be identified. In the training process of sparse trajectory recognition models, if sparse trajectories are manually labeled based on experience and then used for model training in order to obtain training samples, the entire process is not only time-consuming and labor-intensive, but the accuracy of manual labeling of travel modes is also poor, which can easily lead to low recognition accuracy of the final trained sparse trajectory recognition model. To solve this problem, as shown in steps 4.2 and 5.2, in this embodiment, the labeled dense trajectories are downsampled to make them similar to the trajectory point density of the real sparse trajectories, thus obtaining pseudo-sparse trajectories with travel mode labels.By employing this pseudo-sparse trajectory to determine travel mode feature data, and training a sparse trajectory recognition model based on this feature data and travel mode labels, the model can be trained by inputting the travel mode feature data corresponding to the sparse trajectory to be identified. After identifying the travel modes corresponding to multiple trips based on LBS trajectory data, travel mode data is obtained. Then, step 6 is used to expand and calibrate this data, using population data for data expansion and calibrating the data based on different travel mode statistics. Specifically, after identifying travel modes using the above-described travel mode determination method, the travel modes used for each trip included in the location data can be determined. However, the population counted may not represent the entire population of the city or a certain area of ​​the city, thus introducing sampling bias. By calculating the expansion coefficient based on the number of users involved in the LBS trajectory data and the city's population statistics, and expanding the travel mode data, the obtained travel mode statistics can better reflect the travel mode situation of all or part of the city as a whole, reducing or eliminating data errors caused by sampling bias. After expanding the sample size of travel mode data to obtain travel mode statistics, these statistics may still contain some discrepancies compared to actual conditions because they are derived from sampling and expansion. Calibration using statistics from different travel modes (such as bus card swipe statistics, rail transit card swipe statistics, etc.) can make the data more accurately reflect the distribution of various travel modes.

[0060] Figure 4 This diagram illustrates a method for determining travel modes according to an embodiment of the present invention. For any single trip's trajectory, it is classified based on the density of its trajectory points. In this embodiment, the travel trajectory is classified into dense trajectories (Level I in the diagram) and sparse trajectories (Level II in the diagram) based on the density of the trajectory points. As mentioned above, the density used for trajectory point density classification can be the density of all trajectory points in the travel trajectory or the density of only some trajectory points. In this embodiment, the presence or absence of dense segments in the travel trajectory distinguishes it into dense trajectories (Level I) and sparse trajectories. For sparse trajectories (Level II), they are further classified based on whether they include public transport features. That is, some sparse trajectories can identify public transport features (the matching with bus routes or bus stops meets certain requirements), while some sparse trajectories cannot identify public transport features. For dense trajectories, their... Figure 4In this embodiment, non-navigation trajectories are named locally dense. For sparse trajectories, if public transport features (a type of travel mode feature data) can be identified, they are named sparse non-navigation trajectories; if public transport features cannot be identified, they are named extremely sparse non-navigation trajectories. Different travel mode identification models are used for dense trajectories and sparse trajectories. Specifically, for dense trajectories (Level I), this embodiment uses a dense trajectory identification model to identify the travel mode of each dense segment within the dense trajectory (i.e.,...). Figure 4 The process involves point-by-point identification, followed by a voting-based approach to determine the final mode of transportation, or determining the final mode of transportation based on the weight of each mode. For sparse trajectories (Level II), a sparse trajectory recognition model is used for mode of transportation identification. It's easy to understand that since the mode of transportation feature data that can be used during model training is diverse, the trained sparse trajectory recognition model can also be different depending on the mode of transportation features used in the training. Therefore, based on the judgment result of whether public transportation features can be identified in the sparse trajectory, the sparse trajectory will be distributed to the corresponding sparse trajectory recognition model for mode of transportation identification (i.e.,...). Figure 4 (Overall identification in the process).

[0061] Secondly, such as Figure 2 As shown, embodiments of the present invention also provide a travel mode determination device, comprising: a first acquisition unit for acquiring location data, the location data including at least one travel data corresponding to a single trip, the travel data including multiple trajectory points; a first determination unit for determining travel mode feature data of a single trip based on the travel data; a second determination unit for determining a target model from multiple pre-trained travel mode recognition models based on the density of trajectory points corresponding to a single trip; and a recognition unit for inputting the travel mode feature data corresponding to a single trip into the target model for travel mode recognition, thereby obtaining the travel mode corresponding to the single trip.

[0062] In the travel mode determination device using the above configuration, after acquiring location data, the first acquisition unit determines travel mode feature data based on the travel data corresponding to a specific trip within the location data. This travel mode feature data is related to the user's travel mode, and different travel modes will result in different feature data. After obtaining the travel mode feature data, the second determination unit selects a target model from multiple pre-trained travel mode recognition models based on the density of trajectory points corresponding to the user's single trip. The recognition unit then uses this target model to identify the aforementioned travel mode feature data, thereby determining the user's travel mode for this trip. This travel mode determination device does not directly input the travel mode feature data into a uniform recognition model for travel mode identification. Instead, it selects different travel mode recognition models based on the different density of trajectory points. This targeted selection of appropriate recognition models based on different trajectory point densities effectively reduces or even avoids the adverse effects of varying trajectory point densities on travel mode recognition accuracy, improving the accuracy of travel mode recognition and solving the problem of inaccurate travel mode recognition results in related technologies when determining the user's travel mode based on location data.

[0063] In one specific embodiment, the first determining unit includes a first determining module and a second determining module: the first determining module is used to determine the travel trajectory corresponding to a single trip based on travel data; the second determining module is used to determine the travel mode characteristic data of a single trip based on the travel trajectory.

[0064] The second determining unit includes: a third determining module, used to determine the trajectory point density classification of a single trip based on the density of trajectory points corresponding to the single trip; and a fourth determining module, used to select the model corresponding to the trajectory point density classification from multiple recognition models as the target model based on a preset correspondence.

[0065] The trajectory point density classification includes dense trajectories and sparse trajectories. The density of trajectory points corresponding to dense trajectories is higher than that corresponding to sparse trajectories. The target model corresponding to dense trajectories is a dense trajectory recognition model, and the target model corresponding to sparse trajectories is a sparse trajectory recognition model. The travel mode determination device also includes a first training unit, which is used to execute the training steps of the dense trajectory recognition model. The first training unit includes: a fifth determination module, used to determine the travel modes corresponding to multiple dense trajectories; and a first training module, used to input the travel mode feature data and the travel modes corresponding to multiple dense trajectories into the dense trajectory recognition model to train the dense trajectory recognition model.

[0066] The travel mode determination device also includes a second training unit, which is used to execute the sparse trajectory recognition model training steps. The second training unit includes: a downsampling module, which is used to downsample the trajectory points of multiple dense trajectories to obtain multiple pseudo-sparse trajectories; and a second training module, which is used to input the travel mode feature data corresponding to the multiple pseudo-sparse trajectories and the travel modes of the multiple dense trajectories that correspond one-to-one with the multiple pseudo-sparse trajectories into the sparse trajectory recognition model to train the sparse trajectory recognition model.

[0067] The downsampling module includes: an acquisition submodule, used to acquire the sampling time interval pattern from multiple sparse trajectories, the sampling time interval pattern representing the time interval between adjacent trajectory points in the sparse trajectory; and a first sampling submodule, used to downsample the trajectory points of multiple dense trajectories according to the sampling time interval pattern to obtain multiple pseudo-sparse trajectories.

[0068] The downsampling module includes: a calculation submodule for calculating the time interval between trajectory points at adjacent time points in multiple sparse trajectories; a first determination submodule for determining the distribution of transition probabilities based on multiple time intervals, where the transition probability is the probability of any first interval value changing to any second interval value, and the first and second interval values ​​are the values ​​of two adjacent time intervals; a second determination submodule for determining multiple sampling intervals corresponding to each pseudo-sparse trajectory based at least on the distribution of transition probabilities; and a second sampling submodule for sampling multiple trajectory points of each dense trajectory based on multiple sampling intervals to determine each pseudo-sparse trajectory.

[0069] The second determining submodule is used to: determine the first sampling interval, which is the time interval between any two adjacent sampling points; and determine the time interval with the highest transition probability as the second sampling interval based on the distribution of the transition probability, which is the sampling interval that is located after the first sampling interval and adjacent to the first sampling interval.

[0070] The third determining module is used to: classify the trajectory point density of a trip as a dense trajectory when the density of trajectory points corresponding to a single trip is greater than or equal to a preset density threshold; and classify the trajectory point density of a trip as a sparse trajectory when the density of trajectory points corresponding to a single trip is less than a preset density threshold.

[0071] The travel mode determination device also includes a threshold determination unit, which is used to perform a preset density threshold determination step. The preset density threshold determination step includes: determining one or more classification features, which belong to travel mode feature data; and determining the density value of one or more classification features that fail as the density of trajectory points corresponding to a single trip decreases, which is the preset density threshold.

[0072] The travel mode determination device also includes a sampling unit, which is used to determine travel mode data based on the travel modes corresponding to each trip included in the location data; calculate the sampling coefficient based on the number of users involved in the location data and the number of urban population statistics; and expand the travel mode data based on the sampling coefficient to obtain travel mode statistical data.

[0073] After expanding the travel mode data according to the expansion coefficient to obtain travel mode statistics, the travel mode determination device also includes a calibration unit, which is used to: acquire urban public transportation card swipe data; and calibrate the travel mode statistics based on the urban public transportation card swipe data.

[0074] In addition, this embodiment of the invention also provides a method for determining traffic evaluation indicators, including: using the above-described method for determining travel modes to determine the travel modes of multiple trips included in the location data; and calculating traffic evaluation indicators based at least on the travel modes of the multiple trips, wherein the traffic evaluation indicators are used to characterize the traffic quality within the target area.

[0075] Furthermore, this embodiment of the invention also provides a non-volatile storage medium, wherein the non-volatile storage medium includes a stored program, wherein, when the program is running, it controls the device where the non-volatile storage medium is located to execute the above-mentioned travel mode determination method; or, when the program is running, it controls the device where the non-volatile storage medium is located to execute the above-mentioned traffic evaluation index determination method.

[0076] The sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Moreover, the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowcharts, in some cases the steps shown or described may be performed in a different order than that shown here.

[0077] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0078] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.

[0079] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0080] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0081] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0082] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining a mode of transportation, comprising: Acquire location data, which includes at least one trip's travel data, and the travel data includes multiple trajectory points; Based on the travel data, determine the travel mode characteristics of a single trip; Based on the density of trajectory points corresponding to a single trip, a target model is determined from multiple pre-trained travel mode recognition models; The travel mode feature data corresponding to the single trip is input into the target model for travel mode identification to obtain the travel mode corresponding to the single trip; Among them, determining the target model from multiple pre-trained travel mode recognition models based on the density of trajectory points corresponding to the single trip includes: Based on the density of the trajectory points corresponding to a single trip, the trajectory point density classification of the single trip is determined. The trajectory point density classification includes dense trajectories and sparse trajectories. The density of the trajectory points corresponding to the dense trajectory is higher than the density of the trajectory points corresponding to the sparse trajectory. According to a preset correspondence, a model corresponding to the trajectory point density classification is selected from multiple recognition models as the target model. The target model corresponding to the dense trajectory is a dense trajectory recognition model, and the target model corresponding to the sparse trajectory is a sparse trajectory recognition model.

2. The method for determining travel mode according to claim 1, wherein, The method for determining travel modes also includes a dense trajectory recognition model training step, which includes: Determine the travel modes corresponding to multiple of the aforementioned dense trajectories; The travel mode feature data corresponding to multiple dense trajectories and the travel modes corresponding to multiple dense trajectories are input into the dense trajectory recognition model to train the dense trajectory recognition model.

3. The method for determining travel mode according to claim 2, wherein, The method for determining travel modes further includes a sparse trajectory recognition model training step, which includes: Downsampling of the trajectory points of multiple dense trajectories yields multiple pseudo-sparse trajectories; The travel mode feature data corresponding to the multiple pseudo-sparse trajectories and the travel modes of the multiple dense trajectories that correspond one-to-one with the multiple pseudo-sparse trajectories are input into the sparse trajectory recognition model to train the sparse trajectory recognition model.

4. The method for determining travel mode according to claim 3, wherein, Downsampling the trajectory points of multiple dense trajectories yields multiple pseudo-sparse trajectories, including: The sampling time interval pattern is obtained from multiple sparse trajectories, and the sampling time interval pattern represents the time interval between trajectory points at adjacent times in the sparse trajectory; Based on the sampling time interval pattern, the trajectory points of multiple dense trajectories are downsampled to obtain multiple pseudo-sparse trajectories.

5. The method for determining travel mode according to claim 3, wherein, Downsampling the trajectory points of multiple dense trajectories yields multiple pseudo-sparse trajectories, including: Calculate the time interval between the trajectory points at each adjacent time in the multiple sparse trajectories; The distribution of the transition probability is determined based on the multiple time intervals, wherein the transition probability is the probability of any first interval value changing to any second interval value, and the first interval value and the second interval value are the values ​​of two adjacent time intervals; Based at least on the distribution of the transition probabilities, determine multiple sampling intervals corresponding to each of the pseudo-sparse trajectories; Based on the multiple sampling intervals, multiple trajectory points of each of the dense trajectories are sampled to determine each of the pseudo-sparse trajectories.

6. The method for determining travel mode according to claim 5, wherein, Based at least on the distribution of the transition probabilities, determining the multiple sampling intervals corresponding to each of the pseudo-sparse trajectories includes: Determine a first sampling interval, which is the time interval between any two adjacent sampling points; Based on the distribution of the transition probabilities, the time interval with the highest transition probability is determined as the second sampling interval, which is the sampling interval that is located after the first sampling interval and adjacent to the first sampling interval.

7. The method for determining travel mode according to claim 1, wherein, The determination of the trajectory point density classification of a single trip based on the density of the trajectory points corresponding to that single trip includes: if the density of the trajectory points corresponding to that single trip is greater than or equal to a preset density threshold, the trajectory point density classification of the single trip is determined to be a dense trajectory; if the density of the trajectory points corresponding to that single trip is less than the preset density threshold, the trajectory point density classification of the single trip is determined to be a sparse trajectory.

8. The method for determining travel mode according to claim 7, wherein, The method for determining travel modes further includes a step of determining the preset density threshold, which includes: Determine one or more classification features, wherein the classification features belong to the travel mode feature data; As the density of trajectory points corresponding to a single trip decreases, the density value at which one or more of the classification features fail is determined to be the preset density threshold.

9. The method for determining travel mode according to claim 1, wherein, The method for determining the mode of travel also includes: Based on the travel modes corresponding to each trip included in the location data, determine the travel mode data; Calculate the expansion coefficient based on the number of users and the number of urban population statistics involved in the location data; Based on the expansion coefficient, the travel mode data is expanded to obtain travel mode statistics.

10. The method for determining travel mode according to claim 9, wherein, After expanding the travel mode data according to the expansion coefficient to obtain travel mode statistics, the travel mode determination method further includes: Obtain urban public transportation card swipe data; The travel mode statistics are calibrated based on the city's public transportation card swiping data.

11. A mode of transportation determination device, comprising: The first acquisition unit is used to acquire location data, the location data including at least one trip data, the trip data including multiple trajectory points; The first determining unit is used to determine the travel mode characteristic data of a single trip based on the travel data; The second determining unit is used to determine the target model from multiple pre-trained travel mode recognition models based on the density of trajectory points corresponding to the single trip. The identification unit is used to input the travel mode feature data corresponding to the single trip into the target model to identify the travel mode and obtain the travel mode corresponding to the single trip. The second determining unit includes: The third determining module is used to determine the trajectory point density classification of a single trip based on the density of trajectory points corresponding to the single trip. The trajectory point density classification includes dense trajectories and sparse trajectories. The density of trajectory points corresponding to the dense trajectory is higher than the density of trajectory points corresponding to the sparse trajectory. The fourth determining module is used to select the model corresponding to the trajectory point density classification from multiple recognition models as the target model according to a preset correspondence relationship. The target model corresponding to the dense trajectory is a dense trajectory recognition model, and the target model corresponding to the sparse trajectory is a sparse trajectory recognition model.

12. A method for determining traffic evaluation indicators, comprising: Using the travel mode determination method according to any one of claims 1 to 10, the travel modes of multiple trips included in the location data are determined; Based at least on the travel modes of the multiple trips, a traffic evaluation index is calculated, which is used to characterize the traffic quality within the target area.

13. A non-volatile storage medium, wherein, The non-volatile storage medium includes a stored program, wherein, when the program is running, it controls the device containing the non-volatile storage medium to execute the travel mode determination method according to any one of claims 1 to 10; or, when the program is running, it controls the device containing the non-volatile storage medium to execute the traffic evaluation index determination method according to claim 12.

Images