Trajectory data processing method, device, equipment and storage medium

By mining vehicle trajectory data and road network data, extracting features and using machine learning models, the problem of long update cycle of traffic light data in intelligent traffic systems is solved, and more accurate identification and type determination of motor vehicle signal lights is achieved, which improves the Accuracy and user experience of transportation solutions.

CN114116854BActive Publication Date: 2024-11-08TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111502808.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-09
Publication Date
2024-11-08
Estimated Expiration
2041-12-09

AI Technical Summary

Technical Problem

In intelligent transportation systems, the update cycle of traffic light data is relatively long, which leads to deviations from the actual situation and the data used, making it difficult to timely and accurately identify the motor vehicle signal lights and their types.

Method used

By mining vehicle trajectory data and road network data, road condition characteristics and complexity characteristics are extracted, and machine learning models are used to train to determine the existence and type of motor vehicle signal lights.

Benefits of technology

It realizes more timely and accurate motor vehicle signal light recognition, saves labor costs for subsequent verification, and improves the accuracy and user experience of transportation solutions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114116854B_ABST
    Figure CN114116854B_ABST
Patent Text Reader

Abstract

The embodiments of the present disclosure provide a trajectory data processing method, device, equipment and computer-readable storage medium. The method provided by the embodiments of the present disclosure performs data mining based on road network data and vehicle travel trajectory data, extracts multiple representative features related to motor vehicle signal lights for training machine learning models, thereby realizing the determination of whether motor vehicle signal lights exist. In addition, when the method provided by the embodiments of the present disclosure determines the existence of motor vehicle signal lights, it further determines the type of the motor vehicle signal lights based on the road network data and trajectory data in a specific driving direction. The method of the present disclosure improves the accuracy of identifying motor vehicle signal lights, saves the manual operation cost for subsequent verification, and the traffic solution provided based on more accurate motor vehicle signal light identification significantly improves the user experience. The trajectory data processing method of the present disclosure can be applied to application fields such as maps and intelligent transportation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of data mining, and more specifically, to a trajectory data processing method, device, equipment and storage medium. Background Art

[0002] With the continuous development of urbanization, urban traffic problems are becoming more and more prominent. Intelligent transportation system is a new type of transportation system that aims to effectively solve the current traffic problems. Data mining uses a large amount of data generated by daily traffic behavior to extract useful information for traffic prediction. With the development of intelligent transportation, these data make data mining technology more and more widely used in the field of transportation. Now all kinds of transportation solution providers rely on data mining of traffic information.

[0003] At present, the main research direction of intelligent transportation systems is to control and guide traffic flow, which can be achieved by using traffic lights. Traffic lights are one of the most important elements on the road. Traffic solution providers can determine the traffic lights on the road based on the road network data they obtain. However, due to the limitation of the road network data collection cycle, the update cycle of traffic light data will also be longer, and the actual traffic lights may be adjusted frequently according to urban planning. Therefore, the traffic light data used by the traffic solution provider may deviate from the current actual situation.

[0004] Therefore, an efficient and accurate traffic light recognition method is needed. Summary of the invention

[0005] In order to solve the above problems, the present invention mines vehicle travel trajectory data and road network data, extracts multiple representative features for machine learning model training, and thus realizes the determination of the existence and type of traffic lights.

[0006] Embodiments of the present disclosure provide a trajectory data processing method, apparatus, device, and computer-readable storage medium.

[0007] An embodiment of the present disclosure provides a trajectory data processing method, comprising: acquiring trajectory data and road network data, the trajectory data comprising vehicle trajectory point data within a predetermined range of an intersection; extracting road condition features of the intersection from the trajectory data, the road condition features comprising trajectory point density features indicating a distribution pattern of vehicle trajectory points in a direction of travel toward the intersection; obtaining complexity features of the intersection based on the road network data; and determining whether there is a motor vehicle signal light at the intersection based on the road condition features and complexity features of the intersection.

[0008] An embodiment of the present disclosure provides a trajectory data processing device, comprising: a data acquisition module, configured to acquire trajectory data and road network data, the trajectory data including vehicle trajectory point data within a predetermined range of an intersection; an intersection feature extraction module, configured to extract road condition features of the intersection from the trajectory data, the road condition features including trajectory point density features indicating the distribution pattern of vehicle trajectory points in the direction of travel toward the intersection; a complexity feature extraction module, configured to obtain complexity features of the intersection based on the road network data; and a traffic light determination module, configured to determine whether there is a motor vehicle signal light at the intersection based on the road condition features and complexity features of the intersection.

[0009] An embodiment of the present disclosure provides a trajectory data processing device, comprising: one or more processors; and one or more memories, wherein a computer executable program is stored in the one or more memories, and when the computer executable program is executed by the processor, the trajectory data processing method described above is executed.

[0010] An embodiment of the present disclosure provides a computer-readable storage medium having computer-executable instructions stored thereon, and when the instructions are executed by a processor, the method for processing trajectory data as described above is implemented.

[0011] The embodiment of the present disclosure provides a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the trajectory data processing method according to the embodiment of the present disclosure.

[0012] Compared with the traditional method of using road images for identification, the method provided by the embodiment of the present disclosure uses trajectory data with a higher update frequency for data mining, has a shorter mining cycle, and can identify changes in motor vehicle signal lights more timely.

[0013] Compared with the existing motor vehicle signal light recognition method based on trajectory data for data mining, the method provided by the embodiment of the present disclosure can not only determine whether the motor vehicle signal light exists, but also further mine the type of the motor vehicle signal light when the motor vehicle signal light exists.

[0014] The method provided by the embodiment of the present disclosure performs data mining based on road network data and vehicle travel trajectory data, extracts multiple representative features related to motor vehicle signal lights for use in training machine learning models, thereby achieving the determination of whether a motor vehicle signal light exists. In addition, when the method provided by the embodiment of the present disclosure determines that a motor vehicle signal light exists, it further determines the type of the motor vehicle signal light based on the road network data and trajectory data in a specific travel direction. The method of the embodiment of the present disclosure improves the accuracy of identifying motor vehicle signal lights, saves the manual operation cost for subsequent verification, and the traffic solution provided based on more accurate motor vehicle signal light identification significantly improves the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some exemplary embodiments of the present disclosure, and a person of ordinary skill in the art can obtain other drawings based on these drawings without creative work.

[0016] Figure 1A is a schematic diagram showing a scenario of processing a traffic solution request from a user terminal according to an embodiment of the present disclosure;

[0017] Figure 1B is a schematic diagram showing different return traffic schemes according to an embodiment of the present disclosure;

[0018] Figure 1C is a schematic diagram showing an example traffic scenario according to an embodiment of the present disclosure;

[0019] Figure 2 is a flow chart showing a trajectory data processing method according to an embodiment of the present disclosure;

[0020] Figure 3A is a trajectory point density distribution diagram showing an intersection with / without a motor vehicle signal light according to an embodiment of the present disclosure;

[0021] Figure 3B is a flow chart showing a method of extracting traffic characteristics of an intersection according to an embodiment of the present disclosure;

[0022] Figure 4A is a flow chart showing a method for extracting complexity features of an intersection according to an embodiment of the present disclosure;

[0023] Figure 4B is a schematic diagram showing an example topological structure between intersections according to an embodiment of the present disclosure;

[0024] Figure 5Ais a schematic flow chart showing the process of determining whether there is a direction indicator light at an intersection according to an embodiment of the present disclosure;

[0025] Figure 5B is a flow chart showing determining whether there is a direction indicator light at an intersection based on at least a portion of trajectory data according to an embodiment of the present disclosure;

[0026] Figure 5C is a diagram showing an example road and motor vehicle signal light arrangement according to an embodiment of the present disclosure;

[0027] Figure 6 is a schematic flow chart showing the recognition of a signal light of a motor vehicle according to an embodiment of the present disclosure;

[0028] Figure 7 is a schematic diagram showing a trajectory data processing device according to an embodiment of the present disclosure;

[0029] Figure 8 A schematic diagram of a trajectory data processing device according to an embodiment of the present disclosure is shown;

[0030] Fig. 9 A schematic diagram illustrating the architecture of an exemplary computing device according to an embodiment of the present disclosure; and

[0031] Fig.10 A schematic diagram of a storage medium according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present disclosure more obvious, the exemplary embodiments according to the present disclosure will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments of the present disclosure, and it should be understood that the present disclosure is not limited to the exemplary embodiments described here.

[0033] In this specification and the accompanying drawings, substantially the same or similar steps and elements are represented by the same or similar reference numerals, and repeated descriptions of these steps and elements will be omitted. At the same time, in the description of the present disclosure, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance or ranking.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present disclosure pertains. The terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0035] To facilitate description of the present disclosure, concepts related to the present disclosure are introduced below.

[0036] The trajectory data processing method disclosed in the present invention may be based on artificial intelligence (AI). Artificial intelligence is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. For example, for the trajectory data processing method based on artificial intelligence, it can determine the motor vehicle signal lights and their types on the road in a way similar to how humans identify various traffic signs and signal lights on the road with the naked eye. By studying the design principles and implementation methods of various intelligent machines, artificial intelligence enables the trajectory data processing method disclosed in the present invention to have the function of extracting features related to motor vehicle signal lights from the trajectory data and road network data of vehicle travel in real time and accurately, and determining the existence and type of motor vehicle signal lights based on these features.

[0037] The trajectory data processing method disclosed in the present invention may be based on an intelligent traffic system (ITS). An intelligent traffic system, also known as an intelligent transportation system, is an effective and comprehensive application of advanced science and technology (information technology, computer technology, data communication technology, sensor technology, electronic control technology, automatic control theory, operations research, artificial intelligence, etc.) to transportation, service control and vehicle manufacturing, strengthening the connection between vehicles, roads and users, thereby forming a comprehensive transportation system that ensures safety, improves efficiency, improves the environment and saves energy.

[0038] Optionally, the trajectory data processing method disclosed in the present invention can be based on data mining technology. Data mining technology uses large-scale data generated by daily traffic behavior to mine implicit patterns in large-scale data to provide powerful and flexible analysis and processing functions for large-scale data. As a data analysis and processing technology generated by applications and oriented to applications, data mining technology can quickly, effectively and deeply analyze massive traffic information and mine traffic patterns implicit in large amounts of traffic data. In the process of traffic data processing, data mining technology summarizes the association between data by analyzing the potential relationship between attributes, improves the depth of data and derives a lot of relational data, and then can judge the current state and make predictions, provide guidance for traffic management, and give full play to the value of data. For example, in the trajectory data processing method disclosed in the present invention, by analyzing the potential relationship between the trajectory data of vehicle travel and the road network data and the motor vehicle signal lights, multiple features that best represent this potential relationship are extracted to represent the association between these data and the motor vehicle signal lights, so as to give full play to the potential value of trajectory data and road network data.

[0039] Optionally, the training model adopted by the embodiments of the present disclosure may be a machine learning model based on artificial intelligence. Generally, machine learning models based on artificial intelligence can be divided into supervised learning, unsupervised learning and reinforcement learning based on the form of learning, wherein supervised learning is to learn a model from labeled training data in order to make predictions about unknown or future data, and the term "supervision" refers to the output signal or label required for a known sample. The present disclosure may adopt a supervised machine learning model, train the model based on samples with labels (presence / absence), and then use the trained model to predict the possibility of the presence of a motor vehicle signal light at a new intersection. Supervised machine learning models may include algorithm models such as extreme gradient boosting (eXtreme Gradient Boosting, XGBoost), which is a tree model algorithm derived from ensemble learning, and can be regarded as an integration of different single decision trees (Decision Tree, DT), transforming the input features through enhanced DT, and splicing the results of all trees to obtain new combined features, which has been proven to be able to achieve very powerful feature conversion. Therefore, optionally, the embodiments of the present disclosure below may adopt the above-mentioned XGBoost model for training, but it should be understood that other machine learning models that can achieve similar technical effects may also be applicable to the method of the present disclosure, and the XGBoost model is only used as an example rather than a limitation in the present disclosure.

[0040] In summary, the solutions provided by the embodiments of the present disclosure involve technologies such as artificial intelligence and data mining. The embodiments of the present disclosure will be further described below in conjunction with the accompanying drawings.

[0041] Figure 1A 1 is a schematic diagram illustrating a scenario 100 of processing a traffic solution request from a user terminal according to an embodiment of the present disclosure. Figure 1B is a schematic diagram showing different return traffic schemes according to an embodiment of the present disclosure. Figure 1C is a schematic diagram showing an example traffic scenario according to an embodiment of the present disclosure.

[0042] Currently, there are applications that generate multiple transportation solutions for the starting point and the end point input by the user (for example, various applications represented by Tencent Maps, etc.), and the user can initiate a transportation solution request in these applications installed on his user terminal. Then the user terminal transmits its input data, such as the location information of the starting point and the end point, to the server of the application through the network.

[0043] Optionally, the user terminal may specifically include a smart phone, a tablet computer, a laptop, a vehicle-mounted terminal, a wearable device, and the like. The user terminal may also be a client that installs a browser or various applications (including system applications and third-party applications). The network may be an Internet of Things based on the Internet and / or telecommunications network, which may be a wired network or a wireless network, for example, it may be an electronic network that can realize information exchange functions such as a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a cellular data communication network, etc. The server may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0044] like Figure 1A As shown, the server can determine multiple possible traffic plans online based on the received input data, and determine the best traffic plan as a recommended plan for the user, so that the user can easily obtain the best plan and freely choose other traffic plans as needed. Subsequently, the server can return the determined multiple traffic plans to the user terminal through the network to display them to the user.

[0045] An interface that displays multiple transportation options to the user could be something like Figure 1B The user can enter the starting and ending locations in the address boxes at the top of interface (a) or (b) (for example, Figure 1BIn the example, the starting point is the user's current location (shown as "My Location"), and the end point is Subway Station A) to request a transportation plan. The user can also freely select a variety of travel modes according to actual needs, including but not limited to taxi, driving, public transportation, walking or cycling as shown in interfaces (a) and (b). According to the user's choice of travel mode, the server can determine different transportation plans for the user. For example, when the user chooses the walking mode, the transportation plan may include multiple small roads that motor vehicles cannot pass. Since the purpose of the present disclosure is to determine the existence and type of motor vehicle signal lights, the following description is based on the driving mode.

[0046] In interfaces (a) and (b), three traffic plans (Traffic Plan 1, Traffic Plan 2, and Traffic Plan 3) returned from the server are shown, where each traffic plan includes the corresponding travel time, travel distance, and the number of motor vehicle traffic lights passed. Interfaces (a) and (b) respectively use Traffic Plan 1 and Traffic Plan 3 as examples to show their respective travel routes, where the positioning symbol represents the user's current location, i.e., the starting point, and the mark "end" represents the destination location, i.e., the end point. The black bold curve between the starting point and the end point represents the travel route corresponding to the current traffic plan.

[0047] As shown in interfaces (a) and (b), Traffic Plan 1 is the optimal traffic plan (i.e., the recommended plan), with a travel time of 23 minutes, a travel distance of 12 kilometers, and 11 motor vehicle traffic lights. Traffic Plan 3 has a shorter travel distance (11.5 kilometers) but more motor vehicle traffic lights (15) than Traffic Plan 1, and its estimated travel time is 24 minutes. This may be because the travel distances of Traffic Plan 1 and Traffic Plan 3 are not much different, but Traffic Plan 3 needs to pass more motor vehicle traffic lights, resulting in an increase in the required travel time. Traffic Plan 2 has the same travel distance as Traffic Plan 1 (12 kilometers) but more motor vehicle traffic lights (20) than Traffic Plan 3, and its travel time may be longer (26 minutes).

[0048] As mentioned above, the number of motor vehicle traffic lights on a vehicle's route may directly affect the vehicle's travel time and the choice of traffic plans. Therefore, accurate identification of motor vehicle traffic lights on the road is of great significance to traffic plan selection and user experience.

[0049] In the existing motor vehicle signal light recognition technology, road images are usually used for recognition. This technology uses road images as input. However, since the available road images are usually updated for a long time (generally, they are updated once every few days or even months), and the actual motor vehicle signal lights may often be adjusted according to urban planning, the input road image may no longer be the same as the road conditions when the user initiates the traffic plan request. Therefore, in this case, the motor vehicle signal light recognition technology based on road images may provide erroneous traffic plans, resulting in errors and even dangers during vehicle driving. In addition, because the motor vehicle signal light recognized by the recognition technology based on the road image is a point, it is impossible to determine the specific road direction on which the motor vehicle signal light acts. For example, Figure 1C As shown, for the intersection connecting five roads A, B, C, D and E, vehicles traveling in the direction of the arrow on road E are not controlled by the motor vehicle signal lights at the intersection, but the recognition method based on road images can only identify whether there are motor vehicle signal lights at the intersection, but cannot identify which roads the motor vehicle signal lights specifically control.

[0050] Therefore, there is also a technology that uses data mining on the vehicle driving trajectory data on the road to optimize the recognition of motor vehicle signal lights. However, the above two types of motor vehicle signal light recognition methods can only determine whether there is a motor vehicle signal light at the intersection, but cannot determine the specific driving direction of the motor vehicle signal light, that is, they cannot determine the type of the motor vehicle signal light. However, in the actual vehicle driving process, it is very necessary to determine the type of motor vehicle signal light, because although the right turn of vehicles at most intersections is usually not controlled by motor vehicle signal lights, if there is a signal light that controls the right turn at the intersection, the above recognition method may have errors when providing path planning and navigation to the user, resulting in the driver not actually driving according to the true optimal traffic plan, and may also cause the driver to violate traffic rules or even get into danger due to ignoring the motor vehicle signal light that controls a specific driving direction.

[0051] Based on this, the present disclosure provides a trajectory data processing method, which extracts multiple representative features for machine learning model training by performing data mining on both vehicle trajectory data and road network data, thereby realizing the determination of the existence and type of traffic lights.

[0052] Compared with the traditional method of using road images for identification, the method provided by the embodiment of the present disclosure uses trajectory data with a higher update frequency for data mining, has a shorter mining cycle, and can identify changes in motor vehicle signal lights more timely.

[0053] Compared with the existing motor vehicle signal light recognition method based on trajectory data for data mining, the method provided by the embodiment of the present disclosure can not only determine whether the motor vehicle signal light exists, but also further mine the type of the motor vehicle signal light when the motor vehicle signal light exists.

[0054] The method provided by the embodiment of the present disclosure performs data mining based on road network data and vehicle travel trajectory data, extracts multiple representative features related to motor vehicle signal lights for use in training machine learning models, thereby achieving the determination of whether a motor vehicle signal light exists. In addition, when the method provided by the embodiment of the present disclosure determines that a motor vehicle signal light exists, it further determines the type of the motor vehicle signal light based on the road network data and trajectory data in a specific travel direction. The method of the embodiment of the present disclosure improves the accuracy of identifying motor vehicle signal lights, saves the manual operation cost for subsequent verification, and the traffic solution provided based on more accurate motor vehicle signal light identification significantly improves the user experience.

[0055] Figure 2 is a flow chart showing a trajectory data processing method 200 according to an embodiment of the present disclosure.

[0056] In step 201, trajectory data and road network data may be acquired, and the trajectory data may include vehicle trajectory point data within a predetermined range of an intersection.

[0057] According to the embodiments of the present disclosure, traffic data can be divided into two categories: static traffic data and dynamic traffic data. Among them, static traffic data can be data that is no longer updated or is updated very slowly after the setting is completed, that is, the road network data in the present disclosure, which can mainly include, for example, basic geographic information (for example, urban road network, intersection layout, etc.), management entity information (for example, electronic eye layout, speed limit signs, pedestrian crossing line settings, etc.), traffic management information (regional boundaries, safety ranges, distribution of traffic police, posts, law enforcement stations, etc.), traffic passenger information (for example, passenger lines, bus station information, line radiation maps, transportation hubs, etc.), basic spatial data of urban traffic, and basic road information (for example, road grade, length, toll information, etc.). Among them, when determining whether there is a motor vehicle signal light at an intersection, the above-mentioned road network data can be used as a reference. For example, at intersections with more complex urban roads (for example, Figure 1C For example, the intersection connecting five roads A, B, C, D and E shown in the figure, or the intersection with stricter traffic management (for example, the intersection with more management entities or the intersection connected by roads with more management entities), it may be more inclined to set up motor vehicle traffic lights. The following describes the complexity of the intersection by taking the above two points (basic geographic information and management entity information) as examples of the adopted road network data rather than limitations.

[0058] According to the embodiments of the present disclosure, dynamic traffic data may include data obtained by remote sensing means such as satellite remote sensing, aerial photogrammetry, low-altitude UAV emergency platform, ground measurement vehicle, ground video, etc., and traffic data collected by sensor devices and mobile terminals such as video, mobile phone, bus card, ground sensor coil, etc. in the ground intelligent transportation system. Among them, the trajectory data used in the motor vehicle signal light recognition method of the present disclosure can be obtained from the collected traffic data, including vehicle trajectory data, vehicle real-time position data, vehicle trajectory point data within a predetermined range of the intersection, etc.

[0059] The driving trajectory of a vehicle consists of a continuous sequence of trajectory points, and the trajectory points are points collected for positioning the vehicle during its driving process (here it is assumed that the collection of the vehicle's driving trajectory points has a certain periodicity, such as collecting once per second or once every five seconds), which contains the longitude, latitude, time and other information of the positioning. Therefore, after matching the collected trajectory points with the road network, it is possible to know which road the trajectory point is on. Since when a vehicle is driving on a road, if it is controlled by a traffic light at an intersection with a motor vehicle signal light, it needs to stop when a red light appears, and can only pass when the light is green, the trajectory points on the driving trajectory of the vehicle usually present a non-uniform distribution. Because at an intersection with a motor vehicle signal light, when a vehicle is waiting for a green light, the number of trajectory points in the waiting area will gradually increase over time, and after the red light changes to a green light, the number of trajectory points in the waiting area will gradually decrease. Therefore, at intersections with motor vehicle traffic lights, the density distribution of trajectory points may have a certain periodicity as the motor vehicle traffic lights change, while at intersections without motor vehicle traffic lights, the distribution of trajectory points is usually more uniform and there is no obvious periodicity.

[0060] Optionally, the predetermined range of the intersection can be a road area within a predetermined distance range from the intersection on the road in the direction of travel to the intersection. The predetermined distance range can be selected based on the periodicity of the above-mentioned trajectory point density distribution to better present the relationship between the trajectory point density and the distance from the intersection on the road in the direction of travel to the intersection, that is, the periodicity of the vehicle trajectory point density distribution within the predetermined range of the intersection.

[0061] Figure 3A is a diagram showing the density distribution of trajectory points at an intersection with / without a motor vehicle signal light according to an embodiment of the present disclosure.

[0062] Optionally, for intersections with motor vehicle traffic lights, vehicles will wait in the parking area (i.e., the waiting area) during the red light period, while for intersections without motor vehicle traffic lights, vehicles do not need to wait and can pass through the intersection directly. Therefore, this difference can be used to determine whether there is a motor vehicle traffic light at the intersection by comparing the trajectory point density distribution in the waiting area.

[0063] Therefore, in step 202, the road condition characteristics of the intersection can be extracted from the trajectory data, and the road condition characteristics include trajectory point density characteristics, which indicate the distribution pattern of vehicle trajectory points in the direction of traveling toward the intersection.

[0064] like Figure 3A As shown in FIG. 1 , at an intersection with a motor vehicle signal light, a large number of trajectory points may be gathered on the road in the direction of the intersection due to the influence of the motor vehicle signal light, and these trajectory points usually show a periodicity of peaks and troughs, which is usually related to the period of the motor vehicle signal light change. Figure 3A In (a), obvious peaks appear at 0m, 30m and 50m away from the intersection with the motor vehicle signal light. The appearance of the peak may be due to the switching of the motor vehicle signal light. For example, the motor vehicle signal light switches from red to green at the peak, so the speed of the vehicle that was originally in a parked or slow-moving state gradually increases, and the trajectory point density is thus reduced.

[0065] The appearance of a trough after a peak may be due to the switching of the motor vehicle signal light from green to red, which causes the speed of the vehicle after the trough to gradually decrease, and the density of the trajectory points increases. However, since the congestion level is usually lower at places farther from the intersection, the increased density of the trajectory points after the trough is usually less than the density of the trajectory points of the previous peak. Therefore, the distance between the two peaks can be understood as the distance traveled by the vehicle in a round of changes in the motor vehicle signal light (including switching from red to green, and then from green to red), which can reflect the degree of traffic flow at the intersection.

[0066] like Figure 3A As shown in (b), at an intersection without a motor vehicle traffic light, although the trajectory point density decreases with the distance from the intersection, the trajectory points are distributed evenly and the trajectory point density value is small, without obvious peaks and troughs.

[0067] Therefore, the intersection with the motor vehicle signal light can be determined based on the above-mentioned vehicle trajectory point distribution law in the direction of travel to the intersection, and the vehicle trajectory point distribution law can be indicated by the trajectory point density feature. Optionally, the trajectory point density feature can be determined based on the statistical information of the vehicle trajectory point density within a predetermined range of the intersection in the trajectory data.

[0068] Optionally, the trajectory point density feature extracted by the embodiment of the present disclosure may include, for example, at least one of the ratio of the corresponding trajectory point density at adjacent peaks and troughs in the above trajectory point density distribution, the ratio of the absolute distances between adjacent peaks, the ratio of the trajectory point density of adjacent peaks, etc. It should be understood that in addition to the above trajectory point density features, other trajectory point density features that can reflect the distribution law of the trajectory point density at the intersection with the motor vehicle signal light can also be used in the trajectory data processing method of the present disclosure, and the above several trajectory point density features are only used as examples and not as limitations.

[0069] Figure 3B Detailed description is a flow chart showing the method of extracting traffic characteristics of an intersection according to an embodiment of the present disclosure.

[0070] According to an embodiment of the present disclosure, step 202 may include: Figure 3B Steps shown.

[0071] In step 2021, the distribution of trajectory points in the direction of travel toward the intersection may be determined based on the trajectory data.

[0072] As mentioned above, you can Figure 3A As shown, the distribution of trajectory points within a predetermined range of an intersection is determined based on trajectory data. The distribution of trajectory points serves as a basis for extracting important features in the subsequent part and can reflect information related to the characteristics of an intersection with a motor vehicle signal light.

[0073] According to an embodiment of the present disclosure, the trajectory data may be vehicle trajectory point data acquired at various sampling time points within a predetermined time range.

[0074] According to an embodiment of the present disclosure, step 2021 may include: adding the vehicle trajectory point data acquired at each sampling time point within a predetermined time range to obtain an accumulated result of the trajectory point data within the predetermined time range; based on the accumulated result of the trajectory point data, extracting the accumulated result of the trajectory point data in the direction of travel toward the intersection; and determining the trajectory point distribution based on the accumulated result of the trajectory point data in the direction of travel toward the intersection.

[0075] Optionally, based on the accumulated result of the trajectory point data within the predetermined time range, the number of trajectory points at various locations in the direction of travel to the intersection within the predetermined time range may be counted. Adding the vehicle trajectory point data acquired at each sampling time point within the predetermined time range may refer to superimposing the vehicle trajectory point data acquired within the predetermined time range on the same coordinate plane, thereby obtaining the accumulated result of the vehicle trajectory point data within the predetermined time range, for example Figure 3A The trajectory point density distribution diagram is shown.

[0076] In step 2022, the trajectory point density characteristics and the vehicle kinematic characteristics within the predetermined range of the intersection may be determined based on the trajectory point distribution, and the vehicle kinematic characteristics indicate the motion state of the vehicle passing through the intersection.

[0077] Alternatively, you can refer to Figure 3A The trajectory point density feature is determined based on the statistical information of the trajectory point distribution.

[0078] Optionally, the vehicle kinematic features within the predetermined range of the intersection may include the motion information of the vehicle within the predetermined range of the intersection. For example, the vehicle kinematic features may include at least one of the statistical information of the average time for the vehicle to pass through the intersection (such as the mean and quantile) and the statistical information of the average speed of the vehicle passing through the intersection (such as the mean and quantile). The average time / average speed of the vehicle passing through the intersection here may be the average time / average speed of the vehicle within the predetermined range of the intersection. It should be understood that in addition to the above-mentioned vehicle kinematic features, other vehicle kinematic features that can reflect the motion state of the vehicle passing through the intersection can also be used in the trajectory data processing method of the present disclosure. The above-mentioned vehicle kinematic features are only used as examples and not as limitations.

[0079] In step 2023, basic attribute characteristics of the intersection may be determined based on the trajectory point distribution and the road network data, where the basic attribute characteristics indicate a relationship between lanes at the intersection and the number of trajectories passing through the intersection.

[0080] According to an embodiment of the present disclosure, step 2023 may include: determining the number of lanes at the intersection based on the road network data; determining the number of trajectories at the intersection and the number of trajectories on each lane at the intersection based on the trajectory point distribution and the number of lanes at the intersection; and determining basic attribute characteristics of the intersection based on the number of trajectories at the intersection and the number of trajectories on each lane at the intersection.

[0081] Optionally, the number of trajectories at the intersection, ie, the number of vehicles passing through the intersection, may be determined based on the trajectory point distribution, the trajectory being composed of a corresponding plurality of vehicle trajectory points.

[0082] According to an embodiment of the present disclosure, the basic attribute feature may reflect at least one of the vehicle congestion level at the intersection, the vehicle congestion level of each lane at the intersection, etc. For example, the vehicle congestion level of each lane at the intersection may be the basic usage of each lane at the intersection.

[0083] For example, the basic attribute feature may include at least a portion of the total number of trajectories passing through the intersection, the total number of lanes at the intersection, and the average number of trajectories on each lane (i.e., the total number of trajectories passing through the intersection / the total number of lanes at the intersection), etc. Therefore, the basic attribute feature of the intersection can reflect the frequency of use of each lane at the intersection, and the frequency of use is usually associated with the importance of the road. Generally, intersections with roads with higher usage frequencies are more likely to be equipped with motor vehicle signal lights. Therefore, the basic attribute feature of the intersection can be used as another reference for determining motor vehicle signal lights at intersections in the present disclosure.

[0084] It should be understood that in addition to the above-mentioned basic attribute characteristics, other basic attribute characteristics (for example, traffic flow, etc.) that can reflect the usage of each lane at the intersection can also be used in the trajectory data processing method disclosed in the present invention. The above-mentioned basic attribute characteristics are only used as examples and not limitations.

[0085] According to an embodiment of the present disclosure, the road condition characteristics of the intersection may further include the vehicle kinematic characteristics and the basic attribute characteristics.

[0086] As described above, the traffic condition characteristics of the intersection may include the track point density characteristics, vehicle kinematic characteristics and basic attribute characteristics of the intersection, which may reflect the degree of traffic flow at the intersection. In addition to these traffic condition characteristics, other characteristics that may represent the traffic characteristics of an intersection with a motor vehicle signal light may also be included in the intersection characteristics of the intersection, and the above intersection characteristics are only used as examples and not as limitations.

[0087] In addition, in the trajectory data collection, the number of trajectories collected for some intersections may be insufficient or even no trajectories may be collected, resulting in the obtained road condition features not being significant. In this case, the trajectory data processing method disclosed in the present invention can use the complexity of the intersection to predict whether there is a motor vehicle signal light at the intersection.

[0088] In step 203, the complexity characteristics of the intersection may be obtained based on the road network data.

[0089] Figure 4A is a flow chart showing the extraction of complexity features of an intersection according to an embodiment of the present disclosure.

[0090] According to an embodiment of the present disclosure, step 203 may include: Figure 4A Steps shown.

[0091] In step 2031, the topological structure and special attributes of the intersection can be determined based on the road network data, wherein the topological structure includes at least one other intersection connected to the intersection and the road between the intersection and the intersection, and the special attributes are related to specific physical entities at the intersection, and the specific physical entities include at least one of a crosswalk, an electronic eye and a speed limit sign.

[0092] According to the embodiments of the present disclosure, since intersections usually have a relatively complex relationship structure, they can be connected with surrounding intersections to form their own topological structure. The topological structure of the intersection can be determined based on the basic geographic information in the previously described road network data (for example, urban road network, intersection layout, etc.). Figure 4B is a schematic diagram showing an example topological structure between intersections according to an embodiment of the present disclosure.

[0093] like Figure 4B As shown in (a) and (b) in , (a) is an actual map showing the connection relationship of each intersection. Taking the prediction of the motor vehicle signal light at intersection A as an example, (b) is a road network topology structure centered on intersection A constructed based on the map. Intersection A can be connected to intersections B, C and D through roads. Generally, the length of the road between intersections can reflect the complexity of the intersection to a certain extent. For example, if an intersection is far away from the surrounding intersections and the number of trajectories passing through the intersection is small (the traffic volume at the intersection is small), the possibility of setting a motor vehicle signal light at the intersection is low. Generally, if the relationship characteristics of the intersection are more complex, its traffic situation will also be more complex. Therefore, in order to maintain the traffic order at the intersection, it may be more inclined to set a motor vehicle signal light at the intersection.

[0094] In addition, the various attribute information at the intersection can also reflect the complexity of determining the intersection, such as the management entity information in the road network data described previously (e.g., electronic eye layout, speed limit signs, pedestrian crossing line settings, etc.). Figure 4B The traffic management is usually stricter (for example, more management entities are set up at the intersection or on the road connected to the intersection), and motor vehicle traffic lights may be more likely to be set up at these intersections. Therefore, the attribute information of the intersection can be used as a reference for determining the complexity of the intersection.

[0095] Therefore, as described above, the trajectory data processing method of the present disclosure extracts features that can represent the complexity of intersections (ie, intersection complexity features) based on road network data as another important feature for predicting whether there are motor vehicle traffic lights at intersections.

[0096] In step 2032, the complexity characteristics of the intersection can be obtained based on the topological structure and special attributes of the intersection through a pre-trained graph embedding model.

[0097] Optionally, the trajectory data processing method of the present disclosure can obtain a graph embedding model that can output multidimensional features representing the complexity of the intersection based on the input graph information through graph embedding deep network learning. The essence of the method of obtaining a graph embedding model based on graph embedding deep network learning is to represent the entire graph with a multidimensional vector, which is also the purpose of the present disclosure to adopt this method (i.e., to obtain complexity features based on the road network topology and node attribute information). Therefore, it should be understood that other methods that can achieve the same purpose are also applicable to the present disclosure. The method of obtaining a graph embedding model based on graph embedding deep network learning is only used as an example and not a limitation.

[0098] According to the embodiment of the present disclosure, taking the prediction of the traffic lights of motor vehicles at intersection A as an example, the following can be constructed: Figure 4B In the road network topology structure with intersection A as the center as shown in (b), the three intersections B, C, and D are connected to intersection A through corresponding roads respectively.

[0099] According to an embodiment of the present disclosure, the four intersections A, B, C, and D can be used as points in the graph network, and the length of the road between the intersections can be used as the weight of the edge between the points. A matrix containing the connection relationship of the four intersections A, B, C, and D and the length of the road between the intersections is constructed as the input of the graph embedding model used in the present disclosure. In addition, each point can also have its own attribute information, including but not limited to whether there is a crosswalk, whether there is an electronic eye, whether there is a speed limit sign, etc. These attribute information can also be input into the graph embedding model in the form of a matrix, so that the graph embedding model obtained by training can output better complexity features.

[0100] It should be understood that, in addition to being used when the above-obtained road condition features are not significant, the complexity features of the intersection can also be used together with the road condition features of the intersection to more accurately determine whether there are motor vehicle traffic lights at the intersection.

[0101] In step 204, it may be determined whether there is a traffic light for a motor vehicle at the intersection based on the road condition characteristics and complexity characteristics of the intersection.

[0102] According to an embodiment of the present disclosure, determining whether there is a motor vehicle signal light at the intersection based on the traffic conditions and complexity characteristics of the intersection in step 204 may include determining whether there is a motor vehicle signal light at the intersection based on the traffic conditions and complexity characteristics of the intersection by a pre-trained machine learning model.

[0103] According to an embodiment of the present disclosure, the machine learning model can be obtained by pre-training based on a sample set of trajectory data and road network data. The machine learning model takes the road condition characteristics and complexity characteristics extracted from the sample set of trajectory data and road network data as input, and uses the determination of whether there are motor vehicle traffic lights as output.

[0104] According to an embodiment of the present disclosure, the machine learning model may be an extreme gradient boosting model, wherein the extreme gradient boosting model takes the road condition characteristics and complexity characteristics of the intersection as input, constructs a new combined feature based on the road condition characteristics and complexity characteristics of the intersection, and outputs a determination of whether there is a motor vehicle traffic light at the intersection based on the constructed combined feature and the road condition characteristics and complexity characteristics of the intersection.

[0105] Optionally, the extracted road condition features and complexity features can be used as input samples of the XGBoost model, and whether there are motor vehicle traffic lights at the intersection can be used as the sample label of the input sample and passed into the XGBoost model for training.

[0106] For example, the XGBoost model can be used to discover and construct new effective combined features based on existing original features, so as to determine whether there is a motor vehicle traffic light at the intersection based on these combined features and the original features.

[0107] When training a machine learning model for determining whether there is a motor vehicle signal light at an intersection, the positive and negative sample ratios of the sample sets of the actual collected trajectory data and road network data are usually unbalanced, that is, the number of samples at intersections with motor vehicle signal lights is far less than the number of samples at intersections without motor vehicle signal lights. However, considering that the imbalance of positive and negative samples may cause overfitting of samples with a large ratio, the prediction results tend to be biased towards the classification with a large number of samples. Therefore, the method disclosed in the present invention can select an XGBoost model that performs well in the case of a serious imbalance in the positive and negative sample ratio, which is very suitable for processing such unbalanced data. Other machine learning models that can achieve the same effect can also be applied to the method disclosed in the present invention. The XGBoost model is used here only as an example and not as a limitation. In addition, the undersampling method can be used in model training to reduce the difference in the number of positive and negative samples.

[0108] Therefore, by inputting the newly obtained traffic condition characteristics and complexity characteristics of the intersection into the trained XGBoost model, a determination of whether there is a motor vehicle traffic light at the intersection can be output in real time.

[0109] As previously mentioned, in addition to determining whether there is a motor vehicle signal light at an intersection, the motor vehicle signal light recognition method of the embodiment of the present disclosure can further utilize the lane information about the intersection in the road network data and the periodicity of the distribution of trajectory points along each driving direction to determine the arrow information of the motor vehicle signal light at the intersection (i.e., whether the motor vehicle signal light is a direction indicator light controlling a specific driving direction). For example, if there is a right turn lane at the intersection and the distribution of the right turn trajectory points is periodic, then there is a high probability that there is a direction indicator light controlling the right turn driving direction at this intersection.

[0110] Therefore, according to an embodiment of the present disclosure, in addition to the above steps 201-204, the trajectory data processing method 200 may also include step 205, that is, when it is determined that there is a motor vehicle traffic light at the intersection, it can be determined whether there is a direction indicator signal light at the intersection based on the road network data and at least a part of the trajectory data.

[0111] According to an embodiment of the present disclosure, at least a portion of the trajectory data is associated with at least one trajectory along a specific trajectory direction at the intersection, and the direction indicator light corresponds to the specific trajectory direction. The specific trajectory direction may include at least one of a right turn, a left turn, and a U-turn. The specific trajectory direction is the above-mentioned specific driving direction, that is, at the intersection, the right turn, left turn, or U-turn of the vehicle may need to be carried out according to the instructions of the traffic light.

[0112] Figure 5A is a schematic flowchart showing the process of determining whether there is a direction indicator light at an intersection according to an embodiment of the present disclosure. Figure 5B The present invention is a flowchart showing a method of determining whether there is a direction indicator light at an intersection based on at least a portion of trajectory data according to an embodiment of the present disclosure. Figure 5C is a diagram showing an example road and motor vehicle signal light arrangement according to an embodiment of the present disclosure.

[0113] According to an embodiment of the present disclosure, the above-mentioned determination of whether there is a direction indicator light at the intersection based on at least a portion of the road network data and the trajectory data may include: Figure 5A Steps shown.

[0114] In step 2051, it can be determined whether there is a lane corresponding to the specific trajectory direction at the intersection based on the road network data.

[0115] Optionally, when it is determined that there is a motor vehicle traffic light at the intersection, it can be determined first whether there is a lane corresponding to a specific trajectory direction at the intersection. If so, it can be further determined whether there is a direction indicator signal light for controlling vehicles traveling along the specific trajectory direction on the lane. In any case, if there is no such lane, it can be considered that the intersection is not equipped with a direction indicator signal light corresponding to such a lane.

[0116] In step 2052, when it is determined that there is a lane corresponding to the specific trajectory direction, it is determined whether there is a direction indicator light at the intersection based on the trajectory data associated with the at least one trajectory in the trajectory data and the complexity characteristics of the intersection.

[0117] Optionally, the operation in step 2052 can be understood as an operation similar to steps 202-204, except that the operation in step 2052 is for controlling a direction indicator light in a specific track direction, and therefore only the data portion of the track data associated with at least one track at the intersection along the specific track direction is used. Figure 5C As shown, if you want to determine whether there is a direction indicator light controlling the right turn direction at intersection A, you can use the data portion of the trajectory data related to the trajectory of turning right at the intersection, that is, the trajectory data within a predetermined range of the intersection from intersection A to intersection B, to obtain the road condition characteristics of intersection A associated with the right turn direction.

[0118] According to an embodiment of the present disclosure, the above step 2052 may include: Figure 5B Steps shown.

[0119] In step 20521, based on the trajectory data associated with the at least one trajectory in the trajectory data, the distribution of vehicle trajectory points associated with the specific trajectory direction in the direction of travel toward the intersection can be determined.

[0120] Optionally, based on data in the trajectory data associated with at least one trajectory along the specific trajectory direction at the intersection, the trajectory point density distribution along the specific trajectory direction within the predetermined range of the intersection can be calculated, which reflects the distribution pattern of vehicle trajectory points along the specific trajectory direction within the predetermined range of the intersection.

[0121] In step 20522, trajectory point density characteristics and vehicle kinematic characteristics associated with the at least one trajectory at the intersection may be determined based on the trajectory point distribution, wherein the vehicle kinematic characteristics indicate kinematic characteristics of a vehicle passing through the intersection along the specific trajectory direction.

[0122] Alternatively, you can refer to Figure 3AThe statistical information based on the distribution of trajectory points determines a trajectory point density feature associated with at least one trajectory at the intersection.

[0123] Optionally, the kinematic characteristics of vehicles within a predetermined range of the intersection may include movement information of the vehicle along a specific trajectory direction within the predetermined range of the intersection, such as statistical information (such as mean and quantile) of the average time it takes for the vehicle to pass through the intersection along the specific trajectory direction and statistical information (such as mean and quantile) of the average speed of the vehicle passing through the intersection along the specific trajectory direction.

[0124] In step 20523, basic attribute characteristics of the intersection can be determined based on the trajectory point distribution and the road network data, and the basic attribute characteristics indicate the relationship between the lanes at the intersection along the specific trajectory direction and the number of trajectories passing through the intersection along the specific trajectory direction, wherein the road condition characteristics of the intersection also include the vehicle kinematic characteristics and the basic attribute characteristics.

[0125] Optionally, basic attribute characteristics of an intersection associated with a specific trajectory direction may be determined based on the trajectory point distribution along the specific trajectory direction and associated road network data, and the basic attribute characteristics may reflect basic usage of lanes associated with the specific trajectory direction at the intersection.

[0126] For example, the basic attribute features may include the total number of trajectories passing through the intersection along a specific trajectory direction, the total number of lanes associated with the specific trajectory direction at the intersection, and the average number of trajectories on the lanes associated with the specific trajectory direction (i.e., the total number of trajectories passing through the intersection / the total number of lanes at the intersection), etc.

[0127] In step 20524, based on the road condition characteristics and complexity characteristics of the intersection associated with the specific trajectory direction, a pre-trained machine learning model can be used to determine whether there is a direction indicator light at the intersection.

[0128] According to an embodiment of the present disclosure, the machine learning model may be an extreme gradient boosting model, wherein the extreme gradient boosting model takes the road condition characteristics and complexity characteristics of the intersection associated with the specific trajectory direction as input, and takes the determination of whether there is a direction indicator light at the intersection as output.

[0129] Optionally, the extracted road condition features and complexity features associated with a specific trajectory direction can be used as input samples of the XGBoost model, and whether there is a corresponding direction indicator light at the intersection can be used as a sample label of the input sample and passed into the XGBoost model for training.

[0130] Therefore, by inputting the newly obtained road condition characteristics and complexity characteristics associated with a specific trajectory direction at the intersection into the trained XGBoost model, a determination of whether there is a corresponding direction indicator light at the intersection can be output in real time. Optionally, one or more direction indicator lights that may exist at the intersection and are used to control different trajectory directions can be determined, thereby forming a complete identification of the traffic lights at the intersection.

[0131] Figure 6 is a schematic flowchart illustrating the recognition of a signal light of a motor vehicle according to an embodiment of the present disclosure.

[0132] like Figure 6 As shown, the trajectory data processing method according to the embodiment of the present disclosure mainly utilizes trajectory data and road network data.

[0133] For the original trajectory data, since the collection of trajectory data may be affected by external factors such as the network and equipment, as well as human factors, there may be some abnormal trajectories in the actual collected trajectory data. Therefore, the abnormal trajectory data needs to be screened before feature extraction.

[0134] Optionally, the operation of filtering abnormal trajectory data may include:

[0135] 1) Filter out too short trajectories. Generally, if the number of trajectory points on a trajectory is less than 3, it can be considered that the trajectory is too short and can be filtered directly.

[0136] 2) Filter out trajectories that deviate too far from the road. If a trajectory deviates too far from the real road network, then this trajectory may be artificially forged or caused by an abnormality in the device that collects the trajectory, so it can be filtered out directly.

[0137] 3) Filter trajectories with many trajectory points that cannot be matched with the road network. If a trajectory has many trajectory points that cannot be matched with the road network, then this trajectory may be artificially forged or caused by an abnormality in the device that collects the trajectory, so it can be directly filtered.

[0138] Therefore, for the preprocessed trajectory data, feature extraction can be performed to obtain the traffic data of the intersection, such as Figure 6 shown.

[0139] For the road network data, data related to the road network topology structure centered on the intersection to be determined can be input into the trained graph embedding model in matrix form to extract the complexity characteristics of the intersection.

[0140] Therefore, by inputting the extracted traffic characteristics and complexity characteristics of the intersection into the trained XGBoost model, it is possible to output in real time whether there is a motor vehicle traffic light at the intersection and then output a determination of whether there is a direction indicator light at the intersection.

[0141] Figure 7 is a schematic diagram showing a trajectory data processing device 700 according to an embodiment of the present disclosure.

[0142] The trajectory data processing device 700 may include a data acquisition module 701 , a road intersection feature extraction module 702 , a complexity feature extraction module 703 and a traffic light determination module 704 .

[0143] According to an embodiment of the present disclosure, the data acquisition module 701 may be configured to acquire trajectory data and road network data, wherein the trajectory data includes vehicle trajectory point data within a predetermined range of an intersection.

[0144] Optionally, the predetermined range of the intersection can be a road area within a predetermined distance range from the intersection on the road in the direction of travel to the intersection. The predetermined distance range can be selected based on the periodicity of the trajectory point density distribution to better present the relationship between the trajectory point density and the distance from the intersection on the road in the direction of travel to the intersection, that is, the periodicity of the vehicle trajectory point density distribution within the predetermined range of the intersection.

[0145] According to an embodiment of the present disclosure, the intersection feature extraction module 702 may be configured to extract the traffic characteristics of the intersection from the trajectory data, wherein the traffic characteristics include trajectory point density characteristics indicating the distribution pattern of vehicle trajectory points in the direction of travel toward the intersection.

[0146] Optionally, the intersection feature extraction module 702 extracts the road condition features of the intersection from the trajectory data, which may include the following: Figure 3B The operation described.

[0147] For example, the trajectory point density features extracted by the embodiments of the present disclosure may include, for example, the ratio of the corresponding trajectory point densities at adjacent peaks and troughs in the above-mentioned trajectory point density distribution, the ratio of the absolute distances between adjacent peaks, the ratio of the trajectory point densities of adjacent peaks, etc.

[0148] According to an embodiment of the present disclosure, the complexity feature extraction module 703 may be configured to obtain the complexity features of the intersection based on the road network data.

[0149] Since intersections usually have a relatively complex relationship structure, they can be connected with surrounding intersections to form their own topological structure. The topological structure of the intersection can be determined based on the basic geographic information in the previously described road network data (e.g., urban road network, intersection layout, etc.).

[0150] In addition, the various attribute information at the intersection can also reflect the complexity of determining the intersection, such as the management entity information in the road network data described previously (e.g., electronic eye layout, speed limit signs, pedestrian crossing line settings, etc.). Figure 4B The traffic management is usually stricter (for example, more management entities are set up at the intersection or on the road connected to the intersection), and motor vehicle traffic lights may be more likely to be set up at these intersections. Therefore, the attribute information of the intersection can be used as a reference for determining the complexity of the intersection.

[0151] Optionally, the complexity feature extraction module 703 obtains the complexity feature of the intersection based on the road network data, which may include the following: Figure 4A The operation described.

[0152] According to an embodiment of the present disclosure, the traffic light determination module 704 may be configured to determine whether there is a traffic light for a motor vehicle at the intersection based on the road condition characteristics and complexity characteristics of the intersection.

[0153] Optionally, the traffic light determination module 704 determines whether there is a motor vehicle traffic light at the intersection based on the traffic characteristics and complexity characteristics of the intersection, which may include determining whether there is a motor vehicle traffic light at the intersection based on the traffic characteristics and complexity characteristics of the intersection through a pre-trained extreme gradient boosting model.

[0154] For example, the extracted road condition features and complexity features can be used as input samples of the XGBoost model, and whether there is a motor vehicle signal light at the intersection can be used as the sample label of the input sample and passed into the XGBoost model for training. By inputting the newly obtained road condition features and complexity features of the intersection into the trained XGBoost model, the determination of whether there is a motor vehicle signal light at the intersection can be output in real time.

[0155] According to an embodiment of the present disclosure, in addition to the above-mentioned modules, the trajectory data processing device 700 may also include a signal light type determination module 705, which may be configured to determine whether there is a direction indication signal light at the intersection based on the road network data and at least a part of the trajectory data when it is determined that there is a motor vehicle signal light at the intersection.

[0156] Optionally, the traffic light type determination module 705 extracts the traffic condition characteristics of the intersection from the trajectory data, including the following: Figure 5A and 5B The operation described.

[0157] The traffic light type determination module 705 can further utilize the lane information about the intersection in the road network data and the periodicity of the distribution of trajectory points along each driving direction to achieve the purpose of determining the type of motor vehicle traffic lights at the intersection of the present disclosure.

[0158] According to yet another aspect of the present disclosure, a trajectory data processing device is provided. Figure 8 A schematic diagram of a trajectory data processing device 2000 according to an embodiment of the present disclosure is shown.

[0159] like Figure 8 As shown, the trajectory data processing device 2000 may include one or more processors 2010 and one or more memories 2020. The memory 2020 stores a computer readable code, and when the computer readable code is executed by the one or more processors 2010, the trajectory data processing method described above may be executed.

[0160] The processor in the embodiments of the present disclosure may be an integrated circuit chip having signal processing capabilities. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application may be implemented or executed. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc., and may be an X86 architecture or an ARM architecture.

[0161] In general, various example embodiments of the present disclosure may be implemented in hardware or dedicated circuits, software, firmware, logic, or any combination thereof. Certain aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device. When various aspects of the embodiments of the present disclosure are illustrated or described as block diagrams, flow charts, or using some other graphical representation, it will be understood that the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general purpose hardware or controllers or other computing devices, or some combination thereof as non-limiting examples.

[0162] For example, the method or device according to the embodiment of the present disclosure may also be implemented by Fig. 9 The architecture of the computing device 3000 shown in FIG. Fig. 9As shown, the computing device 3000 may include a bus 3010, one or more CPUs 3020, a read-only memory (ROM) 3030, a random access memory (RAM) 3040, a communication port 3050 connected to a network, an input / output component 3060, a hard disk 3070, etc. The storage device in the computing device 3000, such as the ROM 3030 or the hard disk 3070, may store various data or files used for processing and / or communication of the trajectory data processing method provided by the present disclosure and program instructions executed by the CPU. The computing device 3000 may also include a user interface 3080. Of course, Figure 8 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Fig. 9 One or more components of a computing device are shown.

[0163] According to yet another aspect of the present disclosure, a computer-readable storage medium is provided. Fig.10 A schematic diagram 4000 of a storage medium according to the present disclosure is shown.

[0164] like Fig.10 As shown, the computer storage medium 4020 stores computer readable instructions 4010. When the computer readable instructions 4010 are executed by the processor, the trajectory data processing method according to the embodiment of the present disclosure described with reference to the above figures can be executed. The computer readable storage medium in the embodiment of the present disclosure can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. The non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous connection dynamic random access memory (SLDRAM) and direct memory bus random access memory (DRRAM). It should be noted that memory of the methods described herein is intended to include, but is not limited to, these and any other suitable types of memory. It should be noted that memory of the methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0165] The embodiments of the present disclosure also provide a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. The processor of the computer device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device performs the trajectory data processing method according to the embodiments of the present disclosure.

[0166] Embodiments of the present disclosure provide a trajectory data processing method, apparatus, device, and computer-readable storage medium.

[0167] Compared with the traditional method of using road images for identification, the method provided by the embodiment of the present disclosure uses trajectory data with a higher update frequency for data mining, has a shorter mining cycle, and can identify changes in motor vehicle signal lights more timely.

[0168] Compared with the existing motor vehicle signal light recognition method based on trajectory data for data mining, the method provided by the embodiment of the present disclosure can not only determine whether the motor vehicle signal light exists, but also further mine the type of the motor vehicle signal light when the motor vehicle signal light exists.

[0169] The method provided by the embodiment of the present disclosure performs data mining based on road network data and vehicle travel trajectory data, extracts multiple representative features related to motor vehicle signal lights for use in training machine learning models, thereby achieving the determination of whether a motor vehicle signal light exists. In addition, when the method provided by the embodiment of the present disclosure determines that a motor vehicle signal light exists, it further determines the type of the motor vehicle signal light based on the road network data and trajectory data in a specific travel direction. The method of the embodiment of the present disclosure improves the accuracy of identifying motor vehicle signal lights, saves the manual operation cost for subsequent verification, and the traffic solution provided based on more accurate motor vehicle signal light identification significantly improves the user experience.

[0170] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of the code contains at least one executable instruction for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0171] In general, various example embodiments of the present disclosure may be implemented in hardware or dedicated circuits, software, firmware, logic, or any combination thereof. Certain aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device. When various aspects of the embodiments of the present disclosure are illustrated or described as block diagrams, flow charts, or using some other graphical representation, it will be understood that the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general purpose hardware or controllers or other computing devices, or some combination thereof as non-limiting examples.

[0172] The exemplary embodiments of the present disclosure described in detail above are merely illustrative and not restrictive. It should be understood by those skilled in the art that various modifications and combinations may be made to these embodiments or their features without departing from the principles and spirit of the present disclosure, and such modifications should fall within the scope of the present disclosure.

Claims

1. A trajectory data processing method, comprising: Acquiring trajectory data and road network data, wherein the trajectory data includes vehicle trajectory point data within a predetermined range of the intersection; Extracting a road condition feature of the intersection from the trajectory data, the road condition feature including a trajectory point density feature indicating a distribution pattern of vehicle trajectory points in a direction traveling toward the intersection; Based on the road network data, obtaining the complexity characteristics of the intersection; as well as Determining whether there is a motor vehicle signal light at the intersection based on the road condition characteristics and complexity characteristics of the intersection; Wherein, obtaining the complexity feature of the intersection based on the road network data includes: Based on the road network data, determine the topological structure and special attributes of the intersection, wherein the topological structure includes at least one other intersection connected to the intersection and a road between the intersection and the intersection, and the special attributes are related to a specific physical entity at the intersection, and the specific physical entity includes at least one of a crosswalk, an electronic eye, and a speed limit sign; and Based on the topological structure and special attributes of the intersection, the complexity characteristics of the intersection are obtained through a pre-trained graph embedding model.

2. The method of claim 1, wherein: Extracting the traffic condition features of the intersection from the trajectory data includes: Based on the trajectory data, determining a distribution of trajectory points in a direction of travel toward the intersection; Determining, based on the trajectory point distribution, a trajectory point density feature and a vehicle kinematic feature within a predetermined range of the intersection, the vehicle kinematic feature indicating a motion state of a vehicle passing through the intersection; and Determining basic attribute characteristics of the intersection based on the trajectory point distribution and the road network data, the basic attribute characteristics indicating a relationship between lanes at the intersection and the number of trajectories passing through the intersection; Among them, the road condition characteristics of the intersection also include the vehicle kinematic characteristics and the basic attribute characteristics.

3. The method of claim 2, wherein: The trajectory data is vehicle trajectory point data acquired at each sampling time point within a predetermined time range; Wherein, determining the distribution of trajectory points in the direction of traveling toward the intersection based on the trajectory data includes: Adding the vehicle trajectory point data acquired at each sampling time point within a predetermined time range to obtain an accumulation result of the trajectory point data within the predetermined time range; Extracting the trajectory point data accumulation result in the direction of traveling toward the intersection based on the trajectory point data accumulation result; and The trajectory point distribution is determined based on the accumulation result of the trajectory point data in the direction of traveling toward the intersection.

4. The method of claim 2, wherein: The basic attribute feature is used to reflect at least one of the vehicle congestion level at the intersection and the vehicle congestion level of each lane at the intersection.

5. The method of claim 2, wherein: Determining the basic attribute characteristics of the intersection according to the trajectory point distribution and the road network data includes: Determining the number of lanes at the intersection based on the road network data; Determining the number of trajectories at the intersection and the number of trajectories on each lane at the intersection based on the trajectory point distribution and the number of lanes at the intersection; and Based on the number of trajectories at the intersection and the number of trajectories on each lane at the intersection, basic attribute characteristics of the intersection are determined.

6. The method of claim 1, wherein: Determining whether there is a motor vehicle signal light at the intersection based on the road condition characteristics and complexity characteristics of the intersection includes: Based on the traffic characteristics and complexity characteristics of the intersection, determining whether there is a motor vehicle signal light at the intersection through a pre-trained machine learning model; Among them, the machine learning model is obtained by pre-training based on a sample set of trajectory data and road network data. The machine learning model takes the road condition characteristics and complexity characteristics extracted from the sample set of trajectory data and road network data as input and uses the determination of whether there are motor vehicle traffic lights as output.

7. The method of claim 6, wherein: The machine learning model is an extreme gradient boosting model, Among them, the extreme gradient boosting model takes the traffic characteristics and complexity characteristics of the intersection as input, constructs a new combined feature based on the traffic characteristics and complexity characteristics of the intersection, and outputs a determination of whether there is a motor vehicle traffic light at the intersection based on the constructed combined feature and the traffic characteristics and complexity characteristics of the intersection.

8. The method of claim 1, further comprising: In the case where it is determined that there is a motor vehicle signal light at the intersection, determining whether there is a direction indicator signal light at the intersection based on the road network data and at least a portion of the trajectory data; wherein at least a portion of the trajectory data is associated with at least one trajectory along a specific trajectory direction at the intersection, and the direction indicator light corresponds to the specific trajectory direction; The specific trajectory direction includes at least one of a right turn, a left turn and a U-turn.

9. The method of claim 8, wherein: Determining whether there is a direction indicator light at the intersection based on the road network data and at least a portion of the trajectory data includes: Based on the road network data, determining whether there is a lane at the intersection corresponding to the specific trajectory direction; and In a case where it is determined that there is a lane corresponding to the specific trajectory direction, it is determined whether there is a direction indicator light at the intersection based on trajectory data associated with the at least one trajectory in the trajectory data and the complexity characteristics of the intersection.

10. The method of claim 9, wherein: Determining whether there is a direction indicator light at the intersection based on the trajectory data associated with the at least one trajectory in the trajectory data and the complexity feature of the intersection includes: determining, based on trajectory data associated with the at least one trajectory in the trajectory data, a distribution of vehicle trajectory points associated with the specific trajectory direction in a direction of travel toward the intersection; determining, based on the trajectory point distribution, a trajectory point density feature and a vehicle kinematic feature of the intersection associated with the at least one trajectory, the vehicle kinematic feature indicating a kinematic feature of a vehicle passing through the intersection along the specific trajectory direction; Determining basic attribute characteristics of the intersection according to the trajectory point distribution and the road network data, the basic attribute characteristics indicating the relationship between the lanes at the intersection along the specific trajectory direction and the number of trajectories passing through the intersection along the specific trajectory direction, wherein the road condition characteristics of the intersection also include the vehicle kinematic characteristics and the basic attribute characteristics; and Based on the road condition characteristics and complexity characteristics of the intersection associated with the specific trajectory direction, a pre-trained machine learning model is used to determine whether there is a direction indicator light at the intersection.

11. The method of claim 10, wherein: The machine learning model is an extreme gradient boosting model, The extreme gradient boosting model takes the road condition characteristics and complexity characteristics of the intersection associated with the specific trajectory direction as input, and takes the determination of whether there is a direction indicator light at the intersection as output.

12. A trajectory data processing device, comprising: A data acquisition module is configured to acquire trajectory data and road network data, wherein the trajectory data includes vehicle trajectory point data within a predetermined range of the intersection; A road intersection feature extraction module is configured to extract road condition features of the road intersection from the trajectory data, wherein the road condition features include trajectory point density features indicating a distribution pattern of vehicle trajectory points in a direction of travel toward the road intersection; A complexity feature extraction module is configured to obtain the complexity feature of the intersection based on the road network data; as well as A signal light determination module is configured to determine whether there is a motor vehicle signal light at the intersection based on the road condition characteristics and complexity characteristics of the intersection; Wherein, obtaining the complexity feature of the intersection based on the road network data includes: Based on the road network data, determine the topological structure and special attributes of the intersection, wherein the topological structure includes at least one other intersection connected to the intersection and a road between the intersection and the intersection, and the special attributes are related to a specific physical entity at the intersection, and the specific physical entity includes at least one of a crosswalk, an electronic eye, and a speed limit sign; and Based on the topological structure and special attributes of the intersection, the complexity characteristics of the intersection are obtained through a pre-trained graph embedding model.

13. A trajectory data processing device, comprising: one or more processors; as well as One or more memories, wherein a computer executable program is stored, and when the computer executable program is executed by the processor, the method of any one of claims 1-11 is performed.

14. A computer program product, comprising computer instructions, which, when executed by a processor, cause a computer device to perform the method of any one of claims 1 to 11.

15. A computer-readable storage medium having computer-executable instructions stored thereon, wherein the instructions are used to implement the method according to any one of claims 1 to 11 when executed by a processor.