Lane group division method and device, equipment, storage medium and vehicle

Abstract

The invention relates to a lane group division method and device, equipment, a storage medium and a vehicle, and relates to the technical field of data processing. The lane group division method comprises the following steps: acquiring position information of a plurality of track points in a preset area; determining the distance between each track point in the plurality of track points and a preset road center line to obtain a distance sequence, and determining k reference segmentation positions based on the position information of the plurality of track points; the reference segmentation position is used for segmenting a plurality of track points, and k is a positive integer; and determining k target segmentation positions based on the distance sequence and the k reference segmentation positions, and dividing a preset area based on the k target segmentation positions to obtain a plurality of lane groups. And dividing the lane group by using the public source track, so that the divided lane group is close to a real lane group.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, in particular to the field of trajectory data processing technology, and specifically to a lane group division method, device, equipment, storage medium and vehicle. Background Art

[0002] High-precision maps provide important support for perception, decision-making and planning technologies in the field of autonomous driving. Although professionally collected high-precision maps are highly accurate, due to production cost and efficiency issues, the update frequency often cannot keep up with road changes.

[0003] With the rapid development of sensor technology, wireless communication and network technology, people's travel will generate a large amount of spatiotemporal trajectory big data, which contains rich and detailed road information and human behavior information. The collection of trajectory data has gradually evolved from professional department measurement vehicles or professionals to non-professionals freely and voluntarily recording their travel trajectories, and data collection has begun to transform into a crowd-source model.

[0004] Crowd-source trajectory data comes from traveling vehicles equipped with global navigation satellite system (GNSS) receiving modules. It has the advantages of wide coverage of urban road areas and high timeliness, providing a high-time freshness and low-cost production method for road mapping. However, how to use crowd-source trajectories to divide lane groups is a technical problem that needs to be solved urgently. Summary of the invention

[0005] The present application provides a lane group division method, device, equipment, storage medium and vehicle, which are used to divide lane groups using multi-source trajectories so that the divided lane groups are close to the actual lane groups. The technical solution of the present application is as follows:

[0006] According to a first aspect of the present application, a lane group division method is provided, the method comprising: obtaining position information of multiple trajectory points in a preset area; determining the distance between each of the multiple trajectory points and a preset road centerline to obtain a distance sequence, and determining k reference segmentation positions based on the position information of the multiple trajectory points; the reference segmentation position is used to segment the multiple trajectory points, k being a positive integer; determining k target segmentation positions based on the distance sequence and the k reference segmentation positions, and dividing the preset area based on the k target segmentation positions to obtain multiple lane groups.

[0007] According to the above technical means, when the position information of multiple trajectory points in a preset area (non-intersection area) is obtained, the distance sequence corresponding to the multiple trajectory points and the reference segmentation position for dividing the lane group are determined. Further, the target segmentation position for dividing the preset area is determined based on the distance sequence and the reference segmentation position, and the preset area is divided based on the target segmentation position to obtain multiple lane groups. In this way, the lane groups are divided using multi-source trajectories, so that the divided lane groups are close to the actual lane groups.

[0008] In a possible implementation, the above “determining k reference segmentation positions based on the position information of multiple trajectory points” includes: determining a width sequence based on the position information of multiple trajectory points; the width sequence is a sequence of maximum widths of vehicles traveling in a preset area. Based on the width sequence and a preset sliding window algorithm, k reference segmentation positions are determined.

[0009] According to the above technical means, the reference segmentation position is determined according to the change of the width of the trajectory point to preliminarily determine the division position of the lane group.

[0010] In a possible implementation, the above-mentioned "determining k target segmentation positions based on the distance sequence and k reference segmentation positions" includes: determining k candidate segmentation positions based on the k reference segmentation positions, and dividing the distance sequence into multiple group distance sequence subsets based on the k candidate segmentation positions; calculating the sum of squared errors of the multiple group distance sequence subsets based on a preset loss function, and determining the k candidate segmentation positions as k target segmentation positions when the sum of squared errors meets a preset condition; the sum of squared errors is used to characterize the stable state of the distribution of the multiple group distance sequence subsets.

[0011] According to the above technical means, the accurate segmentation position is determined according to the reference segmentation position, so as to achieve accurate segmentation of the lane group, so that the divided lane group is close to the actual lane group.

[0012] In a possible implementation, the above-mentioned “dividing the preset area based on k target segmentation positions” includes: determining n target segmentation positions based on k target segmentation positions and a preset segmentation algorithm, and dividing the preset area into multiple fixed lane areas and lane change areas based on the n target segmentation positions; the lane change area is used to connect different fixed lane areas, and n is an integer greater than or equal to k.

[0013] According to the above technical means, when multiple lane groups are determined, the fixed lane area and the lane change area are also divided based on the target segmentation position, so as to accurately segment the lane groups so that the divided lane groups are close to the actual lane groups.

[0014] In a possible implementation, the lane group division method further includes: within the first lane group, for each preset distance between trajectory points at the cross section of the road segment, a Gaussian mixture model is used to obtain the number of lanes and lane position information at the corresponding position; the first lane group is any one of the multiple lane groups.

[0015] According to the above technical means, the number of lanes and lane position information of the fixed lane area are determined according to the Gaussian mixture model.

[0016] In a possible implementation, the lane group division method further includes: respectively determining the transfer probability of the first lane in the first lane group and each lane in the second lane group to obtain multiple transfer probabilities; the first lane group and the second lane group are two adjacent lane groups in multiple fixed lane areas, and the first lane is any lane in the first lane group; determining the target transfer probability based on the multiple transfer probabilities, and establishing a topological connection between the lane corresponding to the target transfer probability and the first lane; the target transfer probability is the largest transfer probability among the multiple transfer probabilities.

[0017] According to the above technical means, the topological connection between the fixed lane areas is determined according to the transition probability between the lanes.

[0018] In a possible implementation, the lane group division method also includes: when determining the virtual lane lines in the intersection area and the lane lines in the non-intersection area, determining the transition probability between the virtual lane lines and the lane lines, and establishing a topological connection between the intersection area and the non-intersection area based on the transition probability between the virtual lane lines and the lane lines.

[0019] According to the above technical means, the transition probability between virtual lane lines and lane lines is determined to establish a topological connection between the intersection area and the non-intersection area, so that the divided lane groups are close to the actual lane groups.

[0020] In a possible implementation, the lane group division method also includes: determining at least one candidate intersection center of the intersection area, and obtaining the heading angle of each reference trajectory in multiple reference trajectory points to obtain multiple heading angles; the distance between the reference trajectory point and the first candidate intersection center is less than a first preset threshold, and the first candidate intersection center is any one of the at least one candidate intersection center; determining the sine component and cosine component of each heading angle in the multiple heading angles to obtain multiple sine components and multiple cosine components, and clustering the multiple sine components and the multiple cosine components to obtain multiple cluster clusters; when the number of multiple cluster clusters is greater than the second preset threshold, determining the first candidate intersection center as the target intersection center.

[0021] According to the above technical means, the candidate intersection centers are screened to avoid determining the center of an ordinary curve as the intersection center, thereby improving the accuracy of intersection center recognition.

[0022] According to a second aspect of the present application, a lane group division device is provided, the device comprising: an acquisition unit, a determination unit and a division unit; the acquisition unit is used to acquire position information of multiple trajectory points in a preset area; the determination unit is used to determine the distance between each of the multiple trajectory points and a preset road centerline to obtain a distance sequence, and determine k reference segmentation positions based on the position information of the multiple trajectory points; the reference segmentation position is used to divide the multiple trajectory points, k is a positive integer; the determination unit is also used to determine k target segmentation positions based on the distance sequence and the k reference segmentation positions; the division unit is used to divide the preset area based on the k target segmentation positions to obtain multiple lane groups.

[0023] In a possible implementation, the determination unit is specifically configured to: determine a width sequence based on position information of a plurality of trajectory points; the width sequence is a sequence of maximum widths of vehicles traveling in a preset area; and determine k reference segmentation positions based on the width sequence and a preset sliding window algorithm.

[0024] In a possible implementation, the determination unit is further specifically used to: determine k candidate segmentation positions based on k reference segmentation positions, and divide the distance sequence into multiple group distance sequence subsets based on the k candidate segmentation positions; calculate the sum of squared errors of the multiple group distance sequence subsets based on a preset loss function, and determine the k candidate segmentation positions as k target segmentation positions when the sum of squared errors meets a preset condition; the sum of squared errors is used to characterize the stable state of the distribution of the multiple group distance sequence subsets.

[0025] In a possible implementation, the division unit is specifically used to: determine n target segmentation positions based on k target segmentation positions and a preset segmentation algorithm, and divide the preset area into multiple fixed lane areas and lane change areas based on the n target segmentation positions; the lane change area is used to connect different fixed lane areas, and n is an integer greater than or equal to k.

[0026] In a possible implementation, the acquisition unit is further used to: within the first lane group, obtain the number of lanes and lane position information at corresponding positions for trajectory points at the cross section of the road segment at every preset distance through a Gaussian mixture model; the first lane group is any one of the multiple lane groups.

[0027] In a possible implementation, the lane group division device further includes a processing unit. The determination unit is further used to determine the transfer probability of the first lane in the first lane group and each lane in the second lane group, respectively, to obtain multiple transfer probabilities; the first lane group and the second lane group are two adjacent lane groups in multiple fixed lane areas, and the first lane is any lane in the first lane group. The determination unit is further used to determine the target transfer probability based on the multiple transfer probabilities. The processing unit is further used to establish a topological connection between the lane corresponding to the target transfer probability and the first lane; the target transfer probability is the maximum transfer probability among the multiple transfer probabilities.

[0028] In a possible implementation, the determination unit is further used to determine the transition probability between the virtual lane line and the lane line in the intersection area and the lane line in the non-intersection area. The processing unit is further used to establish a topological connection between the intersection area and the non-intersection area based on the transition probability between the virtual lane line and the lane line.

[0029] In one possible implementation, the determination unit is further used to determine at least one candidate intersection center of the intersection area. The acquisition unit is further used to acquire the heading angle of each reference trajectory in the multiple reference trajectory points to obtain multiple heading angles; the distance between the reference trajectory point and the first candidate intersection center is less than a first preset threshold, and the first candidate intersection center is any one of the at least one candidate intersection center. The determination unit is further used to determine the sine component and cosine component of each heading angle in the multiple heading angles to obtain multiple sine components and multiple cosine components. The processing unit is further used to cluster the multiple sine components and the multiple cosine components to obtain multiple cluster clusters. The determination unit is further used to determine the first candidate intersection center as the target intersection center when the number of the multiple cluster clusters is greater than the second preset threshold.

[0030] According to a third aspect provided by the present invention, there is provided an electronic device, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to execute instructions to implement a lane group division method of any possible implementation manner in the above-mentioned first aspect.

[0031] According to the fourth aspect provided by the present invention, a computer-readable storage medium is provided. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the lane group division method of any possible implementation method in the above-mentioned first aspect.

[0032] According to a fifth aspect provided by the present invention, a computer program product is provided, the computer program product comprising computer instructions, which, when executed on an electronic device, enables the electronic device to execute a lane group division method according to any possible implementation of the first aspect described above.

[0033] According to a sixth aspect of the present invention, there is provided a vehicle, comprising the electronic device according to the third aspect.

[0034] Beneficial effects of the present invention:

[0035] (1) When the position information of multiple trajectory points in a preset area (non-intersection area) is obtained, the distance sequence corresponding to the multiple trajectory points and the reference segmentation position for dividing the lane groups are determined. Further, the target segmentation position for dividing the preset area is determined based on the distance sequence and the reference segmentation position, and the preset area is divided based on the target segmentation position to obtain multiple lane groups. In this way, the lane groups are divided using crowdsourced trajectories, so that the divided lane groups are close to the actual lane groups.

[0036] (2) According to the change of the width of the trajectory points, the reference segmentation position is determined to preliminarily determine the division position of the lane group.

[0037] (3) The accurate segmentation position is determined based on the reference segmentation position, so as to achieve accurate segmentation of the lane group, so that the divided lane group is close to the actual lane group.

[0038] (4) When multiple lane groups are determined, the fixed lane area and the lane change area are also divided based on the target segmentation position, so that the lane groups are accurately segmented, making the divided lane groups close to the actual lane groups.

[0039] (5) The number of lanes and lane position information in the fixed lane area can be accurately determined based on the Gaussian mixture model.

[0040] (6) Based on the transition probability between lanes, accurately establish the topological connection between fixed lane areas.

[0041] (7) Based on the transition probability between virtual lane lines, the topological connection between intersection areas and non-intersection areas can be accurately established, making the divided lane groups closer to the actual lane groups.

[0042] It should be noted that the technical effects brought about by any implementation method in the second to sixth aspects can refer to the technical effects brought about by the corresponding implementation method in the first aspect, and will not be repeated here.

[0043] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of the structure of a lane group division system provided in an embodiment of the present application;

[0045] Figure 2A flow chart of a lane group division method provided in an embodiment of the present application;

[0046] Figure 3 One of the schematic diagrams of a road segment trajectory projection provided in an embodiment of the present application;

[0047] Figure 4 A second schematic diagram of a road segment trajectory projection provided in an embodiment of the present application;

[0048] Figure 5 A schematic diagram of lane position extraction provided in an embodiment of the present application;

[0049] Figure 6 A schematic diagram of lane geometry fitting provided in an embodiment of the present application;

[0050] Figure 7 A schematic diagram of lane extraction for an adjacent lane group provided in an embodiment of the present application;

[0051] Figure 8 A schematic diagram of topological connection construction of adjacent lane groups provided in an embodiment of the present application;

[0052] Fig. 9 A schematic diagram of turning point detection and clustering provided in an embodiment of the present application;

[0053] Fig.10 A schematic diagram of the effect of extracting the intersection center of a trajectory area provided in an embodiment of the present application;

[0054] Fig.11 A regional OSM road network map and a schematic diagram of road network map nodes provided in an embodiment of the present application;

[0055] Fig.12 A schematic diagram of the intersection center and range detection effect provided in an embodiment of the present application;

[0056] Fig.13 A schematic diagram of the effect of extracting lane-level entrances and exits at an intersection provided in an embodiment of the present application;

[0057] Fig.14 A schematic diagram of an intersection entrance and exit matching effect provided in an embodiment of the present application;

[0058] Fig.15 A schematic diagram of the effect of extracting the centerline of a lane inside an intersection provided in an embodiment of the present application;

[0059] Fig.16 A schematic diagram of the topological connection effect of an intersection and a non-intersection provided in an embodiment of the present application;

[0060] Fig.17A schematic diagram of the structure of a lane group division device provided in an embodiment of the present application;

[0061] Fig.18 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention;

[0062] Fig.19 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to enable ordinary persons in the art to better understand the technical solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0064] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims.

[0065] First, a brief introduction to the application scenarios involved in this application is given.

[0066] High-precision maps provide important support for perception, decision-making and planning technologies in the field of autonomous driving. Although professionally collected high-precision maps are highly accurate, due to production cost and efficiency issues, the update frequency often cannot keep up with road changes.

[0067] With the rapid development of sensor technology, wireless communication and network technology, people's travel will generate a large amount of spatiotemporal trajectory big data, which contains rich and detailed road information and human behavior information. The collection of trajectory data has gradually evolved from professional department measurement vehicles or professionals to non-professionals freely and voluntarily recording their travel trajectories, and data collection has begun to transform into a crowd-source model.

[0068] Crowd-source trajectory data comes from traveling vehicles equipped with GNSS receiving modules. It has the advantages of wide coverage of urban road areas and high timeliness, providing a high-time freshness and low-cost production method for road mapping. However, how to use crowd-source trajectories to divide lane groups is a technical problem that needs to be solved urgently.

[0069] Specifically, the situation of lane groups in non-intersection areas is more complicated. There can be multiple lane groups in a section of road, and different lane groups are connected and switched with each other. Therefore, how to use multi-source trajectories to divide lane groups is a technical problem that needs to be solved urgently.

[0070] At present, the road mapping technology that provides the data basis for high-precision maps mainly collects data with professional survey vehicles and undergoes a lot of manual post-processing. At the same time, some studies focus on extracting road information from satellite remote sensing images and drone-based orthophotos. However, the acquisition of high-precision, high-quality satellite images and drone image data has problems such as extreme data sensitivity and environmental influences, resulting in low frequency of updating of large-scale urban road network data and high collection costs.

[0071] Since crowdsource trajectory data comes from traveling vehicles equipped with GNSS receiving modules, it has the advantages of wide coverage of urban road areas and high timeliness, providing a high-time freshness and low-cost production method for road mapping. Existing lane geometry information extraction methods based on crowdsource trajectory data can be divided into lane geometry information extraction methods based on local distribution of trajectory points and lane estimation methods based on trajectory clustering algorithms according to the processing objects.

[0072] In the related art, lane geometry information extraction based on local distribution of trajectory points includes: when the vehicle travels along the center line of the lane, the trajectory points solved by the on-board GNSS receiver module follow a Gaussian distribution in the direction perpendicular to the center line of the lane, and the mean center is the position of the center line of the lane. Therefore, on the cross section of the road, the distribution of trajectory points can be regarded as a Gaussian mixture distribution composed of multiple Gaussian components, and the solution of lane geometry information can be regarded as the solution of parameters in the Gaussian mixture model.

[0073] The specific algorithm flow is as follows:

[0074] A sliding window is used to select the trajectory point data within the window and project it to the vertical lane direction. The known lane width is introduced as a known parameter, the number of lanes is iterated, and the Gaussian mixture model is fitted using the trajectory point distribution to solve the center position and weight of each lane; the confidence of the solution under different lane conditions is evaluated through the risk assessment model. The number and position of lanes are determined based on the confidence.

[0075] Lane estimation methods based on trajectory clustering algorithms include: Compared with lane geometry information extraction methods that only target local GNSS point distribution, this type of method uses complete trajectories or sub-trajectories as clustering inputs, emphasizing the sequentiality of trajectory observation points and the connection relationship of trajectories. Through trajectory clustering, trajectory clusters passing through the same lane are extracted, and each group of clustered trajectory clusters is used to estimate the lane centerline. The initial cluster centerline is determined according to the distribution state of the target trajectory data. The target trajectory data is clustered based on the initial cluster centerline to obtain the target cluster centerline after clustering processing; the target lane to which the vehicle to be detected belongs is determined based on the target cluster centerline. In this way, the initial cluster centerline can be quickly determined by the distribution state of the preset number of target trajectory data collected. The target trajectory data is clustered based on the initial cluster centerline to obtain an accurate target cluster centerline, and then the target lane to which the vehicle to be detected belongs is quickly determined based on the target cluster centerline.

[0076] Trajectory data can reflect the details of the urban road network, such as the location, width and driving direction of the lanes. In urban road design, the entrance and exit lanes of the intersection, the bay parking station and the distribution lanes of the expressway are typical scenes of real lane number changes. Existing methods for extracting lane geometry information based on trajectory data, such as the k-means clustering algorithm (K-Means), kernel density estimation and Gaussian mixture model, essentially model the distribution characteristics of trajectory points within a certain range to obtain the number and location of lanes. However, this type of method relies on the statistical characteristics of trajectory data. At the same time, in the process of identifying the number of lanes and the center position of lanes, the Gaussian mixture model is sensitive to the distribution of trajectory data. The parameters often need to be manually adjusted based on prior information, and the accuracy is not high. It is impossible to extract the lane geometry and topological connectivity information of the divergence and merging areas that are common in urban road networks. The lane estimation method based on the trajectory clustering algorithm relies on trajectory data. When there are random lane change trajectories generated by free driving behavior in the trajectory data, the lanes generated by such trajectories will be inconsistent with the actual lanes in terms of position and number.

[0077] Urban road networks often have more complex lane-level topological structures, such as scenarios where the number of lanes changes and lane-level topological information at intersections. Therefore, it is difficult to use the above method to construct truly complete structured lane geometry information and intersection topological information for urban road networks. Therefore, it is still a very difficult task to accurately express the complete geometry information and topological structure of lane-level road networks in complex urban scenarios based on trajectory data.

[0078] In view of the above problems, an embodiment of the present application provides a lane group division method, which includes: obtaining the position information of multiple trajectory points in a preset area. Determining the distance between each of the multiple trajectory points and the preset road centerline to obtain a distance sequence, and determining k reference segmentation positions based on the position information of the multiple trajectory points; the reference segmentation position is used to segment the multiple trajectory points, k≥1. Based on the distance sequence and the k reference segmentation positions, k target segmentation positions are determined, and the preset area is divided based on the k target segmentation positions to obtain multiple lane groups.

[0079] In this way, when the position information of multiple trajectory points in the preset area (non-intersection area) is obtained, the distance sequence corresponding to the multiple trajectory points and the reference segmentation position for dividing the lane group are determined. Further, the target segmentation position for dividing the preset area is determined based on the distance sequence and the reference segmentation position, and the preset area is divided based on the target segmentation position to obtain multiple lane groups. In this way, the lane groups are divided using multi-source trajectories, so that the divided lane groups are close to the actual lane groups.

[0080] Secondly, the implementation environment (implementation architecture) involved in the method provided in this application is briefly introduced.

[0081] The lane group division method provided in the embodiment of the present application can be applied to the lane group division system. Figure 1 A structural schematic diagram of the lane group division system is shown. Figure 1 As shown, the lane group division system 10 includes an intersection detection module 101 , an intersection lane extraction module 102 , a non-intersection lane extraction module 103 and an intersection and non-intersection connection module 104 .

[0082] The intersection detection module 101 is used to detect intersection areas based on crowd-source data.

[0083] It should be noted that crowd-source data may include crowd-sourced data. Crowd-source data is data collected by vehicles.

[0084] The intersection lane extraction module 102 is used to extract virtual lanes based on the information of the intersection area and the trajectory points in the intersection area.

[0085] The non-intersection lane extraction module 103 is used to detect the non-intersection area based on crowd-source data, and divide the fixed lane area and the lane change area based on the non-intersection area and the trajectory points in the non-intersection area. Furthermore, the non-intersection lane extraction module 103 is also used to determine the number of lanes and lane positions in the fixed lane area based on the trajectory points in the fixed lane area and the prior Gaussian mixture model. The non-intersection lane extraction module 103 is also used to establish a topological connection between lane groups in the fixed lane area based on the transition probability of each lane in the fixed lane area.

[0086] The intersection and non-intersection connection module 104 is used to calculate the transition probability between adjacent intersection entrances and exits and non-intersection lane nodes based on the trajectory flow distribution to construct a lane topology structure, match the adjacent non-intersection lane nodes most relevant to each intersection entrance and exit, and establish a topological connection relationship between adjacent intersections and non-intersection areas.

[0087] In some embodiments, the lane group division system 10 may include a server and an electronic device. The server is used to store crowd-source data and send the crowd-source data to the electronic device.

[0088] Accordingly, the electronic device obtains the crowd-source data sent by the server, and obtains the position information of multiple track points in the preset area based on the crowd-source data. The distance between each track point in the multiple track points and the preset road centerline is determined to obtain a distance sequence, and k reference segmentation positions are determined based on the position information of the multiple track points. The reference segmentation positions are used to segment different lane groups. Based on the distance sequence and the k reference segmentation positions, k target segmentation positions are determined, and the preset area is divided based on the k target segmentation positions to obtain multiple lane groups.

[0089] In some embodiments, the lane group division system 10 may include a server. The server is used to store crowd-source data, and obtain the location information of multiple track points in a preset area based on the crowd-source data. Determine the distance between each track point in the multiple track points and the preset road centerline to obtain a distance sequence, and determine k reference segmentation positions based on the location information of the multiple track points. The reference segmentation positions are used to divide different lane groups. Based on the distance sequence and the k reference segmentation positions, k target segmentation positions are determined, and the preset area is divided based on the k target segmentation positions to obtain multiple lane groups.

[0090] In some embodiments, the preset road center line may be a line within a preset area, such as a center line within the preset area.

[0091] The terminal device may be a mobile phone, a tablet computer, a desktop computer, a portable notebook computer, etc. The server may be an independent server or a server cluster composed of multiple servers.

[0092] In order to better understand the lane group division system 10 in the embodiment of the present application, the operation of the lane group division system 10 is described below by way of example. The lane group division method executed by the server is described below.

[0093] In one design, in order to utilize multi-source trajectories to divide lane groups, the divided lane groups are made close to the actual lane groups. Figure 2 As shown, the lane group division method provided in the embodiment of the present application includes: S201-S205.

[0094] S201, obtaining location information of multiple track points in a preset area.

[0095] As a possible implementation method, the server obtains the trajectory point information of the crowd-source data, and identifies the intersection area and the non-intersection area based on the trajectory point information. The trajectory point information includes the timestamps and location information of several trajectory points. Furthermore, according to the timestamps of the trajectory points and the distance intervals of the trajectory points, several trajectory points in the intersection area and the non-intersection area are grouped respectively to obtain a trajectory group in the intersection area and a trajectory group in the non-intersection area. Subsequently, the server obtains the location information of multiple trajectory points in the non-intersection area.

[0096] It should be noted that the preset area is a non-intersection area.

[0097] S202: Determine the distance between each trajectory point in the plurality of trajectory points and a preset road centerline to obtain a distance sequence.

[0098] In some embodiments, the server virtually creates a preset road centerline in a preset area, and determines the distance between each of the plurality of trajectory points and the preset road centerline to obtain a distance sequence.

[0099] In other embodiments, when acquiring multiple trajectory points, the server projects the multiple trajectory points into a preset coordinate system, and calculates the distance between each of the multiple trajectory points and a preset reference line to obtain a distance sequence.

[0100] Specifically: In order to identify different lane group areas, the embodiment of the present application projects multiple trajectory points into the Frenet coordinate system, and calculates the lateral and longitudinal distance distribution of the multiple trajectory points from the preset road center line. The Frenet coordinate system is a local coordinate system based on the road curve, which can provide an intuitive way to represent the lateral and longitudinal distance distribution of the trajectory based on the shape and relative position of the reference line (the preset road center line). The lateral distance of the trajectory is the vertical distance from the track to the road reference line, and the longitudinal distance of the trajectory is the distance traveled along the forward direction of the path. Figure 3 As shown, multiple trajectory points are projected into the Frenet coordinate system, the x-axis coordinate is the longitudinal distance of the trajectory, and the y-axis coordinate is the lateral distance of the trajectory, and the following is obtained: Figure 4 The distribution shown. Figure 4 As shown in Figure 1, the offset distribution of the lateral distance of the trajectory reflects the lane change mode of the trajectory. When the number of lanes remains unchanged, the lateral distance distribution remains relatively stable within the lane group, but with the change of the lane group, the lateral distance distribution will change significantly. The above distance sequence is a lateral distance sequence.

[0101] Furthermore, the embodiment of the present application regards the longitudinal distance as a time function, and the distribution of the lateral distance is analogous to the representation of the time series, and transforms the lane structure change detection problem of the trajectory into the change point detection problem in the time series segmentation, and detects the position where the mean of the lateral distance in the distance sequence changes. The change point detection algorithm can find a suitable change point for the time series, so that the sum of the errors after the sequence is segmented based on the change point is minimized.

[0102] For example, the distance sequence is {d1,d2…d n}, by defining a loss function, searching for all reference segmentation positions {t1, t2…t k}, which is formulated as an optimization problem to detect all reference segmentation positions {t1, t2…t k}, select the segmentation method with the smallest loss function value, and the formula for calculating the target segmentation position is as follows:

[0103]

[0104] Among them, V(t1,t2,…,t K ) is the loss function, which is specifically expressed as follows:

[0105]

[0106] Among them, K is the number of segmentation change points, is the sum of squared errors between the sample value and the sample mean, It is the average value of each segmented sequence after segmentation.

[0107] For example, Figure 5 As shown in FIG, a lane position extraction result based on a priori Gaussian mixture model is shown. Figure 6 As shown, Figure 6 Lane geometry fitting results.

[0108] It should be noted that the more stable the distribution of a set of data is, the smaller the sum of squared errors is. Finally, by selecting the method that minimizes the total sum of squared deviations, the number and position of the change points are determined, the road segment trajectory is segmented, and the data in the same road segment is divided into lane groups with different structures through trajectory sequence segmentation, so that the internal topological structure of the lane group after segmentation is constant.

[0109] S203: Determine k reference segmentation positions based on the position information of the plurality of trajectory points.

[0110] The reference segmentation position is used to segment different lane groups, and k is a positive integer.

[0111] As a possible implementation method, the server determines a width sequence based on the position information of multiple trajectory points. The width sequence is a sequence of maximum widths of the vehicle traveling in a preset area. Further, the server determines k reference segmentation positions based on the width sequence and a preset sliding window algorithm.

[0112] It can be understood that different lane groups have different widths, and a preset sliding window algorithm is used to identify the changing widths in the width sequence to determine the reference segmentation position.

[0113] S204 : Determine k target segmentation positions based on the distance sequence and the k reference segmentation positions.

[0114] As a possible implementation method, the server determines k candidate segmentation positions based on k reference segmentation positions, and divides the distance sequence into multiple group distance sequence subsets based on the k candidate segmentation positions; calculates the sum of squared errors of the multiple group distance sequence subsets based on a preset loss function, and determines the k candidate segmentation positions as k target segmentation positions when the sum of squared errors meets preset conditions; the sum of squared errors is used to characterize the stable state of the distribution of the multiple group distance sequence subsets.

[0115] Specifically, when k reference segmentation positions are obtained, candidate segmentation positions within a preset range of each reference segmentation position are obtained according to the k reference segmentation positions to obtain k candidate segmentation positions. Further, the server divides the distance sequence into k+1 group distance sequence subsets based on the k candidate segmentation positions, and calculates the sum of square errors of multiple group distance sequence subsets based on a preset loss function. Subsequently, when the sum of square errors meets a preset condition, the k candidate segmentation positions are determined to be k target segmentation positions; the sum of square errors is used to characterize the stable distribution state of multiple group distance sequence subsets.

[0116] For example, after obtaining K segmentation positions t1, t2, ...t k In the case of , multiple sampling positions within a distance t15 meters m are obtained and one of the sampling positions is determined as a candidate segmentation position.

[0117] It should be noted that satisfying the preset condition may be that the sum of squared errors is less than or equal to a preset threshold, or may be that the sum of squared errors is the smallest total sum of squared deviations among multiple sums of squared errors.

[0118] S205 . Divide the preset area based on the k target segmentation positions to obtain multiple lane groups.

[0119] As a possible implementation method, the server divides the preset area according to the k target segmentation positions to obtain k+1 lane groups

[0120] As another possible implementation method, the server determines n target segmentation positions based on k target segmentation positions and a preset segmentation algorithm, and divides the preset area into multiple fixed lane areas and lane change areas based on the n target segmentation positions. The lane change area is used to connect different fixed lane areas, and n is an integer greater than or equal to k.

[0121] The lane group division method provided in the embodiment of the present application brings the following beneficial effects: when the position information of multiple trajectory points in a preset area (non-intersection area) is obtained, the distance sequence corresponding to the multiple trajectory points and the reference segmentation position for dividing the multiple trajectory points are determined. Further, based on the distance sequence and the reference segmentation position, the target segmentation position for dividing the preset area is determined, and the preset area is divided based on the target segmentation position to obtain multiple lane groups. In this way, the lane groups are divided using multi-source trajectories, so that the divided lane groups are close to the actual lane groups.

[0122] In one design, in order to determine the segmentation positions of different lane groups, S203 provided in the embodiment of the present application includes: S2031-S2032.

[0123] S2031. Determine a width sequence based on position information of multiple trajectory points.

[0124] The width sequence is a sequence of maximum widths of a preset vehicle driving area.

[0125] S2032: Determine k reference segmentation positions based on the width sequence and a preset sliding window algorithm.

[0126] In one design, in order to determine the segmentation positions of different lane groups, S204 provided in the embodiment of the present application includes: S2041-S2044.

[0127] S2041. Determine k candidate segmentation positions based on k reference segmentation positions.

[0128] S2042 . Divide the distance sequence into a plurality of group distance sequence subsets based on the k candidate segmentation positions.

[0129] S2043. Calculate the sum of squared errors of multiple group distance sequence subsets based on a preset loss function.

[0130] The sum of squared errors is used to characterize the stable distribution state of multiple group distance sequence subsets.

[0131] S2044: When the sum of squared errors meets a preset condition, determine k candidate segmentation positions as k target segmentation positions.

[0132] In one design, in order to accurately identify the lane group, an accurate lane group is generated. The above S205 includes: S2051-S2052.

[0133] S2051. Determine n target segmentation positions based on k target segmentation positions and a preset segmentation algorithm.

[0134] S2052: Divide the preset area into a plurality of fixed lane areas and lane change areas based on the n target segmentation positions.

[0135] In other embodiments, the server determines a plurality of fixed lane areas after determining the number and positions of the segmentation positions. Further, a lane change area of ​​a first preset distance is reserved between adjacent fixed lane areas as a lane change area for topological connection between subsequent adjacent lane groups. Thus, the trajectories belonging to each lane group and the lane change areas between adjacent lane groups are finally output.

[0136] It can be understood that the embodiment of the present application performs lane group segmentation based on the lane change structure, so that the internal topological structure of the lane group after segmentation is constant, and the lane number change area between the lane groups is reserved to facilitate subsequent lane change topological connection, which can better serve the downstream lane information extraction task.

[0137] In one design, in order to generate a lane line topology map, the lane group division method provided in the embodiment of the present application further includes: S206.

[0138] S206. In the first lane group, for each preset distance, the trajectory points at the cross section of the road segment are used to obtain the number of lanes and lane position information at the corresponding positions through a Gaussian mixture model.

[0139] The first lane group is any one of the multiple lane groups.

[0140] In some embodiments, the server takes the trajectory points of each lane group after segmentation as input, extracts the lane position points based on a priori Gaussian mixture model, and models and fits the geometric structure of the lane centerline through a spline curve.

[0141] Specifically, the a priori Gaussian mixture model used in the embodiment of the present application to extract the lane position specifically includes: starting from the starting point of the lane group, constructing a rectangular moving window, taking the maximum coverage width of the trajectory on the road surface as the side length of the window, detecting the trajectory observation points in the window, and gradually moving along the main direction of the road until the end of the road section (lane group). The main direction of the road and the road cross-sectional direction perpendicular to it are identified by the movement direction of the trajectory in the window, and all GNSS trajectory points in the window are projected in the road cross-sectional direction to obtain the projection distribution of the data in the road cross-sectional direction. Modeling is performed based on the classic Gaussian mixture model, and the probability density function of the Gaussian mixture model is constructed, as shown in the following formula three.

[0142]

[0143] The density function p(x) is used to represent the comprehensive probability of the Gaussian mixture model. x represents the trajectory point sample to be calculated projected to the road cross section direction, k is the number of Gaussian mixture components, j = 1, 2, ..., k, μ j Represents the average value of samples belonging to each Gaussian component, and also represents the position of the center line of each lane, σ j Represents the standard deviation of samples in each Gaussian component, w j is the weight of the jth Gaussian component distribution, which is determined by the proportion of the number of trajectories in each lane, w1+w2+…+w k =1.

[0144] In addition, the embodiment of the present application introduces a priori conditions of variable lane width on the basis of the traditional Gaussian mixture model, and optimizes the evaluation function of lane number selection to improve the applicability of the algorithm in lane number recognition. The evaluation function for lane number selection is shown in the following formula 4.

[0145]

[0146] Among them, for the trajectory point n represents the number of samples, k is the number of Gaussian mixture components, p(x i |θ k ) represents the sample value x i In the Gaussian mixture model parameters θ k Gaussian probability value under the condition, where the model parameter θ k It can be expressed as θ k (w k ,μ k ,σ). The first term on the right side of the equal sign is the log-likelihood function, which is used to evaluate the model fit. To prevent overfitting, λ is the regularization parameter with a default value of 1 and Δh i It represents the average distance between the center lines of adjacent lanes. A sigmoid function for lane width information constraint is introduced in the regularization term.

[0147] In some embodiments, considering that the lane width values ​​of urban roads are usually within a certain range, such as 3.25m to 3.75m, the sigmoid function maps real values ​​to the interval (0,1) to constrain the lane width values ​​in the model within a certain range. The final model selects the Gaussian mixture model with the smallest function evaluation value as the optimal model, determines the number of mixed components as the number of lanes, and automatically identifies the number of lanes and outputs the lane position information as an ordered point sequence.

[0148] Furthermore, the embodiment of the present application uses a spline curve fitting method based on the sequence of lane position points extracted by the prior Gaussian mixture model to further represent the geometric shape of the lane. The embodiment of the present application models and fits the geometric structure of the lane centerline based on the spline curve to better obtain the geometric shape of the lane to cooperate with application scenarios such as autonomous driving.

[0149] Among them, the spline curve is a smooth and continuous curve with good approximation performance and mathematical properties, and can well simulate the geometry of the actual lane. The embodiment of the present application uses cubic Hermite splines to represent the geometry of the lane. The lane curve is described by a piecewise polynomial function and parameterized in a unified form with a series of control points and their corresponding tangent vectors, which makes the representation of the lane curve flexible. Based on the definition of cubic Hermite splines, the lane curve can be expressed as the following formula 5:

[0150]

[0151] in, u i represents the lane position node, p i ,v i Respectively represent the position and tangent vector at node i, p i is the position of control point i. The cubic Hermite spline can maintain the continuity of the curve between adjacent control points, has higher accuracy when representing lane curves, and can quickly calculate important indicators such as lane curvature and curvature radius, which provides help for the subsequent vehicle's driving trajectory planning.

[0152] In one design, in order to determine the topological connection of lane groups in adjacent fixed lane areas, the lane group division method provided in the embodiment of the present application further includes: S207-S209.

[0153] S207 , respectively determine the transition probabilities of the first lane in the first lane group and each lane in the second lane group to obtain a plurality of transition probabilities.

[0154] The first lane group and the second lane group are two adjacent lane groups in the multiple fixed lane areas, and the first lane is any lane in the first lane group.

[0155] S208. Determine a target transition probability based on the multiple transition probabilities.

[0156] Among them, the target transition probability is the largest transition probability among multiple transition probabilities.

[0157] S209: Establish a topological connection between the lane corresponding to the target transfer probability and the first lane.

[0158] Specifically, Figure 7and Figure 8 As shown, a schematic diagram of the construction of the topological connection relationship between adjacent lane groups is shown. Figure 7 The lane extraction result for the adjacent lane group, Figure 8 The topological connection between adjacent lane groups is constructed. The embodiment of the present application determines the topological relationship between adjacent lanes based on the statistical information of the trajectory, ensuring that the topological connection of the lanes is consistent with the observed trajectory pattern. The lane topology structure is constructed by calculating the trajectory transition probability analysis between adjacent lanes, and the lane topological connection relationship of adjacent lane groups is automatically obtained.

[0159] The specific implementation method is as follows: Between adjacent lane groups, this embodiment constructs a lane transfer matrix to represent the transfer probability between adjacent lanes. Each lane transfer matrix is ​​a two-dimensional matrix, and its shape depends on the number of lane nodes at the end position of the previous lane group and the end position of the next lane group in the adjacent lane group. The lane node set at the end position of the previous lane group is I = {i1, i2, i3, ..., i m}, the lane node set J at the starting position of the next lane group = {j1, j2, j3, …, j n}. The lane transfer matrix is ​​as follows: Formula 6.

[0160]

[0161] Among them, each element e in the lane transfer matrix i,j Indicates the number of trajectories that transition from the lane node at the end position of the previous lane group to the lane node at the start position of the next lane group. When a trajectory passes through the lane end position to the adjacent next lane group start position along the driving direction, the algorithm automatically detects the lane transition and calculates the transfer probability between any two lane nodes based on the trajectory flow distribution. The matrix records the transfer probability between any two lane nodes and is used to describe the transfer relationship between adjacent lanes. For each lane node, the algorithm searches for the lane node of the adjacent lane group that is most relevant to it in the lane transfer matrix and establishes a connection relationship between the end and start points of the adjacent lane group.

[0162] In one design, in order to determine the intersection center of the intersection, the lane group division method provided in the embodiment of the present application further includes: S210-S214.

[0163] S210: Determine at least one candidate intersection center in the intersection area.

[0164] S211. Acquire the heading angle of each reference trajectory in the multiple reference trajectory points to obtain multiple heading angles.

[0165] The distance between the reference trajectory point and the center of the first candidate intersection is less than a first preset threshold, and the center of the first candidate intersection is any one of the at least one candidate intersection center.

[0166] S212: Determine a sine component and a cosine component of each heading angle among the multiple heading angles to obtain multiple sine components and multiple cosine components.

[0167] S213 . Cluster the multiple sine components and the multiple cosine components to obtain multiple clusters.

[0168] S214: When the number of the plurality of clusters is greater than a second preset threshold, determine the first candidate intersection center as the target intersection center.

[0169] In some embodiments, intersection information is extracted based on trajectory point information of crowd-source data, including the following steps: S1-S5.

[0170] S1. Detect and cluster the trajectory points in the crowdsource data to obtain the turning points.

[0171] For example, the trajectory points of any one of the multiple trajectories in the crowd-source data are obtained, and the heading angular velocity of each trajectory point is calculated according to the heading angle and time data (such as timestamp) of each trajectory point in the trajectory.

[0172] The heading angle of trajectory point i is θ i , the heading angle of the previous trajectory point (i-1) adjacent to trajectory point i is θ i-1 , the timestamp of trajectory point i is t i , the timestamp of trajectory point (i-1) is t i-1 For example. Considering the periodicity of angular velocity, the heading angle difference Δθ i Should be between -180° and 180°, the heading angle difference Δθ i and the angular velocity w i Calculate according to the following formula 7 and formula 8.

[0173] Δθ i =|(θ i -θ i-1 +180°)\mod360°-180°|Formula 7

[0174]

[0175] The point where the angular velocity is greater than the third preset threshold is taken as the turning point.

[0176] For each trajectory, the turning points in the trajectory sequence are clustered using the density-based spatial clustering of applications with noise (DBSCAN) to obtain continuous turning clusters. DBSCAN clustering is a density-based spatial clustering algorithm, and its steps are as follows:

[0177] (a) For each data point p, calculate all neighboring points in its neighborhood. The neighborhood is defined as follows:

[0178] N ∈ (p) = q∈D | dist(p,q)≤∈Formula 9

[0179] Where D is the dataset, dist(p,q) is the distance between two points, and ∈ is the neighborhood radius.

[0180] (b) If the number of neighbor points of p is greater than or equal to the predefined threshold min_samples, p is marked as a core point.

[0181] (c) For all core points, their neighbor points are clustered. If the neighborhood of a core point overlaps with the neighborhood of another core point, then the two neighborhoods belong to the same cluster.

[0182] (d) All points that are not classified into any cluster are considered as noise points.

[0183] It should be noted that the DBSCAN clustering algorithm is used to extract continuous turning points in the trajectory as turning clusters, and the cumulative steering angle and duration of each turning cluster are calculated. The turning clusters whose cumulative steering angle and duration are greater than the fourth preset threshold are retained to reduce the impact of non-curved turning driving behavior on subsequent intersection detection.

[0184] like Fig. 9 As shown, a schematic diagram of turning point detection and clustering effect is exemplified.

[0185] S2. Extract the intersection center.

[0186] The example of this application uses turn clustering to extract candidate intersection centers, and then uses the trajectory heading distribution within a preset range to screen the candidate intersection centers to obtain the regional intersection center.

[0187] The embodiment of the present application takes the trajectory set and the turning cluster in the intersection area as input, and uses the MeanShift algorithm to cluster the turning clusters in the area to extract the candidate intersection center. The MeanShift algorithm is a density-based parameter-free clustering algorithm, and its steps are as follows:

[0188] (e) Select the bandwidth parameter h. The bandwidth parameter determines the local search range in the data space. Considering that the spatial range of the intersection is usually about 50m, the clustering bandwidth is set to 50m.

[0189] (f) Determine the average shift vector. Determine the kernel function as a Gaussian kernel function for calculating the similarity between the turning points. The Gaussian kernel is defined as the following formula 10.

[0190]

[0191] where ||x|| represents the norm of the data point x, and for each point x of all turning clusters in the trajectory area i , construct a hypersphere (high-dimensional spherical region) with a radius of h with this point as the center, and let the set of all points in the sphere be N(x i Then, calculate the average shift vector M(x i ) as shown in the following formula 11.

[0192]

[0193] (g) Update the position of each point and move it to the position indicated by the calculated average shift vector.

[0194] (h) Repeat steps f and g until the position changes of all points are small (convergence), or the predetermined maximum number of iterations is reached, and the points that converge to the same position are grouped into a cluster to form a clustering result. The cluster center is used as the candidate intersection center.

[0195] Furthermore, due to the influence of curves and lane-changing driving behaviors, the intersection centers obtained by clustering need to be screened.

[0196] Specifically, compared with ordinary curves, the trajectory heading distribution in the intersection area is wider. Screening is performed by determining the number of main directions of the trajectory heading angle around the center point of the candidate intersection. First, the trajectory within a certain range around the center of each candidate intersection is extracted and the sine and cosine components of the trajectory heading angle are calculated. For the trajectory point P around the candidate center (such as within the fifth preset threshold range), i , let the heading angle be θ i , and its sine and cosine components are respectively expressed by the following formula 12 and formula 13.

[0197]

[0198]

[0199] Furthermore, the calculated sine and cosine components are clustered using the MeanShift clustering algorithm. If the number of clusters is greater than or equal to the sixth preset threshold, the trajectory around the candidate center point has at least three main directions, and the candidate center is determined to be the target center. On the contrary, if the number of clusters is less than the sixth preset threshold, the candidate center point is deleted.

[0200] Exemplarily, the sixth preset threshold value may be 3, or other values, which is not limited in the embodiments of the present application.

[0201] like Fig.10 As shown, a schematic diagram of the effect of extracting the intersection center of the trajectory area is exemplified.

[0202] S3. Open street map (OSM) assisted intersection detection.

[0203] The embodiment of the present application uses OSM road network data to supplement the intersection detection results. Due to the limitation of factors such as the amount of trajectory data and driving behavior, it may not be possible to fully detect all intersections in the area by relying solely on trajectory data. The intersection detection results are supplemented by OSM road network data in the area.

[0204] Get the OSM road network of the trajectory within the preset longitude and latitude range online and convert the OSM road network into a directed graph.

[0205] like Fig.11 As shown, a regional OSM road network map and a road network map node schematic diagram are exemplarily shown.

[0206] Exemplarily, nodes with degrees greater than or equal to 3 are extracted from the road network graph, and the angles formed by the connecting edges of each node are calculated. When the maximum value of the angle is in the range of 45° to 135°, the node is considered to be a candidate intersection node, otherwise the node is considered to be a junction node and is discarded.

[0207] The MeanShift clustering algorithm is used to cluster the filtered intersection nodes, and the cluster center is extracted as the OSM intersection center.

[0208] The extracted OSM intersection center is compared with the previous intersection detection results, and the center points that are too close to the detected intersections (less than the preset threshold) are removed to complete the OSM-assisted intersection center detection.

[0209] S4. Intersection range detection.

[0210] The embodiment of the present application is based on the intersection center and regional trajectory data, and uses a circular expansion and continuously calculates the average trajectory speed and turning angle density within the range to obtain the intersection range.

[0211] In some embodiments, the average speed of the trajectory points less than the seventh preset threshold value from the center point of the intersection is calculated as the speed threshold, and a certain route angular velocity threshold is set. Since the intersection range is usually about 50 meters, 30 meters is selected as the starting radius of the circular expansion. Within this radius, the average speed and average angular velocity of the trajectory points are calculated. If the average speed is greater than the set speed threshold, or the average angular velocity is less than the angular velocity threshold, the circular range is regarded as the intersection range. If the condition is not met, the circular range is expanded by 5 meters. Repeat this process until the threshold condition is met or the circular range reaches 100 meters.

[0212] like Fig.12 As shown, a schematic diagram of the intersection center and range detection effect is exemplarily shown.

[0213] S5. Extract the lane-level entrances and exits of the intersection.

[0214] This example uses the regional trajectory set and intersection detection results as input, and clusters the entry and exit trajectory points of the trajectory passing through the intersection range to obtain the lane-level entrance and exit of the intersection. According to the intersection range extracted by the previous module, the starting point and end point of the trajectory segment in each intersection are extracted. The MeanShift algorithm is used to cluster the starting point and end point of the trajectory segment separately.

[0215] For example, the lane width is 3.6 m, and 1.8 m is used as the clustering bandwidth, and the cluster center is used as the extracted lane-level entrance and exit of the intersection.

[0216] like Fig.13 As shown, a schematic diagram of the lane-level entrance and exit extraction effect of an intersection is exemplarily shown.

[0217] Construction of internal topological structure of intersection and extraction of lane centerline.

[0218] Subsequently, this example takes the trajectory data within each intersection range and the intersection lane-level entrance and exit extraction results as input, builds the intersection topology structure according to the trajectory mobility matching, and uses the main curve fitting method to lift the intersection lane centerline.

[0219] Specifically, based on the lane-level entrance and exit extraction of the intersection in the previous module, according to the trajectory corresponding to each entrance and exit pair, based on the trajectory flow, the entrances and exits passed before and after the same trajectory are matched with each other, so as to extract the connection status and turning relationship between the entrances and exits inside the intersection.

[0220] like Fig.14 As shown, a schematic diagram of the intersection entrance and exit matching effect is shown as an example;

[0221] For each matched pair of entrances and exits, the main curve fitting method is used to extract the lane centerline.

[0222] The main curve fitting algorithm steps are as follows:

[0223] (a) A trajectory line from the trajectory belonging to the same pair of entrances and exits is randomly selected as the initial lane line; (b) The cross-section line is calculated for each point of the initial lane line; (c) The geometric center of the intersection of the cross-section line and the input trajectory cluster is calculated, and the geometric center is used as the center point of the lane. The intersection points are connected and the fitted lane center line is updated; (d) The first two steps are iterated until the position of the center point is stable and unchanged as the convergence condition, and the final fitting result is obtained.

[0224] like Fig.15 As shown, a schematic diagram of the effect of extracting the center line of the lane inside the intersection is shown as an example.

[0225] In one design, in order to establish a topological connection between the non-intersection area and the intersection area, the lane group division method provided in the embodiment of the present application further includes: S215-S216.

[0226] S215. When the virtual lane line in the intersection area and the lane line in the non-intersection area are determined, determine the transition probability between the virtual lane line and the lane line.

[0227] S216: Based on the transition probability between the virtual lane lines, a topological connection between the intersection area and the non-intersection area is established.

[0228] In some embodiments, a topological connection between intersections and non-intersection areas is constructed, and the lane topology structure is constructed by calculating the transition probability between adjacent intersection entrances and exits and non-intersection lane nodes based on trajectory flow distribution, matching the adjacent non-intersection lane nodes most relevant to each intersection entrance and exit, and establishing a topological connection relationship between adjacent intersections and non-intersection areas.

[0229] like Fig.16 As shown, a schematic diagram of the topological connection effect of intersections and non-intersections is exemplified.

[0230] It can be understood that the lane group division method provided in the embodiment of the present application calculates the trajectory point heading angular velocity based on the trajectory heading data, takes the trajectory points with angular velocity greater than a certain threshold as candidate turning points, and uses density-based spatial clustering on the trajectory sequence to screen out continuous turning points to form a turning cluster. Further, the cumulative angle change value and duration of the turning cluster are calculated, and the trajectory points in the turning cluster whose angle change and duration are greater than a certain threshold are determined as turning points. Subsequently, mean shift clustering is used for the turning points in the trajectory area, and the cluster center is used as the candidate intersection center, and the intersection center is screened by calculating the number of main headings of points around the candidate center. Finally, for the intersection center obtained by screening, the intersection range is obtained by using a circular expansion method and continuously calculating the average speed and angular velocity; the intersection nodes in the road network diagram are extracted using the OSM road network in the trajectory area to supplement the intersection detection results.

[0231] In some embodiments, the lane group division method provided in the embodiments of the present application also includes an intersection internal topology structure and a lane extraction method.

[0232] Specifically, the starting point and end point of the trajectory segment in the intersection area are extracted, and mean shift clustering is used to cluster the lane-level entrances and exits of the intersection. The entrances and exits before and after the same trajectory are matched to construct the internal topological structure of the intersection. For each matched pair of entrances and exits, the main curve fitting method is used to construct the lane line to realize the extraction of the internal lane line of the intersection.

[0233] The lane geometry and topology of non-intersection areas are extracted. The following methods are used to implement the following: for non-intersection area trajectories, lane group segmentation is performed based on the lane change structure, the geometry of lane lines within the lane group is extracted, and the topological connection between adjacent lane groups is constructed to achieve the construction of lane geometry and topological relationship in non-intersection areas.

[0234] For the topological connection between non-intersection areas and intersection areas, the following methods are used to implement the topological connection: based on the trajectory flow distribution, the transition probability between adjacent intersection entrances and exits and non-intersection lane nodes is calculated to construct the lane topology structure, the adjacent non-intersection lane nodes most relevant to each intersection entrance and exit are matched, and the topological connection relationship between adjacent intersections and non-intersection areas is established.

[0235] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the method. In order to realize the above functions, the lane group division device or electronic device includes a hardware structure and / or software module corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0236] The embodiments of the present application can exemplarily divide the lane group division device or electronic device into functional modules according to the above method. For example, the lane group division device or electronic device can include various functional modules corresponding to various functional divisions, or two or more functions can be integrated into one processing module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiments of the present application is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0237] For example, an embodiment of the present application also provides a lane group division device.

[0238] Fig.17 A lane group division device is shown according to an exemplary embodiment. Fig.17 The lane group division device 30 includes an acquisition unit 301, a determination unit 302 and a division unit 303.

[0239] The acquisition unit 301 is used to acquire the position information of multiple track points in a preset area.

[0240] The determination unit 302 is used to determine the distance between each trajectory point in the multiple trajectory points and the preset road center line, obtain a distance sequence, and determine k reference segmentation positions based on the position information of the multiple trajectory points; the reference segmentation position is used to segment the multiple trajectory points, and k is a positive integer.

[0241] The determination unit 302 is further configured to determine k target segmentation positions based on the distance sequence and the k reference segmentation positions.

[0242] The division unit 303 is used to divide the preset area based on the k target segmentation positions to obtain multiple lane groups.

[0243] Optionally, the determination unit 302 is specifically configured to: determine a width sequence based on the position information of the plurality of trajectory points; the width sequence is a sequence of maximum widths of the vehicle traveling in a preset area. Based on the width sequence and a preset sliding window algorithm, determine k reference segmentation positions.

[0244] Optionally, the determination unit 302 is further specifically used to: determine k candidate segmentation positions based on k reference segmentation positions, and divide the distance sequence into multiple group distance sequence subsets based on the k candidate segmentation positions; calculate the sum of squared errors of the multiple group distance sequence subsets based on a preset loss function, and determine the k candidate segmentation positions as k target segmentation positions when the sum of squared errors meets a preset condition; the sum of squared errors is used to characterize the stable state of the distribution of the multiple group distance sequence subsets.

[0245] Optionally, the division unit 303 is specifically used to: determine n target segmentation positions based on k target segmentation positions and a preset segmentation algorithm, and divide the preset area into multiple fixed lane areas and lane change areas based on the n target segmentation positions; the lane change area is used to connect different fixed lane areas, and n is an integer greater than or equal to k.

[0246] Optionally, the acquisition unit 301 is further used to: within the first lane group, obtain the number of lanes and lane position information at corresponding positions of the trajectory points at the cross section of the road segment at each preset distance through a Gaussian mixture model; the first lane group is any one of the multiple lane groups.

[0247] Optional, such as Fig.17 As shown, the lane group division device 30 also includes a processing unit 304 .

[0248] The determination unit 302 is further used to respectively determine the transition probabilities of the first lane in the first lane group and each lane in the second lane group to obtain multiple transition probabilities; the first lane group and the second lane group are two adjacent lane groups in multiple fixed lane areas, and the first lane is any lane in the first lane group.

[0249] The determining unit 302 is further configured to determine a target transition probability based on the multiple transition probabilities.

[0250] The processing unit 304 is further configured to establish a topological connection between the lane corresponding to the target transition probability and the first lane; the target transition probability is the largest transition probability among the multiple transition probabilities.

[0251] Optionally, the determination unit 302 is further configured to determine the transition probability between the virtual lane line and the lane line in the intersection area and the lane line in the non-intersection area.

[0252] The processing unit 304 is further configured to establish a topological connection between the intersection area and the non-intersection area based on the transition probability between the virtual lane lines.

[0253] Optionally, the determination unit 302 is further configured to determine at least one candidate intersection center in the intersection area.

[0254] The acquisition unit 301 is also used to acquire the heading angle of each reference trajectory in multiple reference trajectory points to obtain multiple heading angles; the distance between the reference trajectory point and the center of the first candidate intersection is less than a first preset threshold, and the center of the first candidate intersection is any one of at least one candidate intersection center.

[0255] The determination unit 302 is further configured to determine a sine component and a cosine component of each heading angle among the plurality of heading angles, and obtain a plurality of sine components and a plurality of cosine components.

[0256] The processing unit 304 is further configured to cluster the multiple sine components and the multiple cosine components to obtain multiple clusters.

[0257] The determining unit 302 is further configured to determine the first candidate intersection center as the target intersection center when the number of the plurality of clusters is greater than a second preset threshold.

[0258] In the case of implementing the functions of the above-mentioned integrated modules in the form of hardware, the embodiment of the present application provides a possible structural diagram of the electronic device involved in the above-mentioned embodiment. Fig.18 As shown, the electronic device 40 includes a processor 401, a memory 402 and a bus 403. The processor 401 and the memory 402 may be connected via the bus 403.

[0259] The processor 401 is the control center of the communication device, which can be a processor or a general term for multiple processing elements. For example, the processor 401 can be a general-purpose central processing unit (CPU) or other general-purpose processors. Among them, the general-purpose processor can be a microprocessor or any conventional processor.

[0260] As an embodiment, the processor 401 may include one or more CPUs, such as Fig.18 CPU 0 and CPU 1 are shown in .

[0261] The memory 402 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0262] As a possible implementation, the memory 402 may exist independently of the processor 401, and the memory 402 may be connected to the processor 401 via a bus 403 to store instructions or program codes. When the processor 401 calls and executes the instructions or program codes stored in the memory 402, the sensor determination method provided in the embodiment of the present application can be implemented.

[0263] It should be noted that the processor 401 may be the same as the processor 101, or may be a different device, which is not limited in the present embodiment.

[0264] In another possible implementation, the memory 402 may also be integrated with the processor 401 .

[0265] The bus 403 may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.18 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0266] It should be pointed out that Fig.18 The structure shown does not constitute a limitation on the electronic device 40. Fig.18 In addition to the components shown, the electronic device 40 may include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0267] Optionally, the electronic device 40 provided in the embodiment of the present application may further include a communication interface 404 .

[0268] The communication interface 404 is used to connect with other devices through a communication network. The communication network may be Ethernet, wireless access network, wireless local area network (WLAN), etc. The communication interface 404 may include a receiving unit for receiving data and a sending unit for sending data.

[0269] In one design, in the electronic device provided in the embodiment of the present application, the communication interface can also be integrated into the processor.

[0270] In another hardware structure of the electronic device provided in the embodiment of the present application, Fig.19 As shown, the electronic device 50 may include a processor 501 and a memory 502. The processor 501 is connected to the memory 502 via a connection line. The memory 502 stores application code.

[0271] The functions of the processor may refer to the description of the above processor.

[0272] It should be pointed out that the above-mentioned another hardware structure does not constitute a limitation on the electronic device. In addition to the above-mentioned another hardware component, the electronic device may include more or fewer components, or a combination of certain components, or different component arrangements.

[0273] In the case of implementing the functions of the above-mentioned integrated modules in the form of hardware, the structural diagram of the middleware involved in the above-mentioned embodiments provided in the embodiments of the present application can refer to the structural diagram of the above-mentioned execution machine.

[0274] In the case of implementing the functions of the above-mentioned integrated modules in the form of hardware, the structural schematic diagram of the electronic device involved in the above-mentioned embodiment provided in the embodiment of the present application can refer to the description of the electronic device 40 and will not be repeated here.

[0275] An embodiment of the present application further provides a vehicle, comprising the above-mentioned electronic device 40.

[0276] An embodiment of the present application also provides a computer-readable storage medium, in which instructions are stored. When a computer executes the instructions, the computer executes each step in the lane group division method process shown in the above method embodiment.

[0277] An embodiment of the present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the lane group division method in the above method embodiment.

[0278] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, and a hard disk. Random Access Memory (RAM), Read-Only Memory (ROM), Erasable Programmable Read Only Memory (EPROM), registers, hard disks, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any other form of computer-readable storage medium in a suitable combination of the above, or numerical values ​​in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an Application Specific Integrated Circuit (ASIC). In the embodiments of the present application, a computer-readable storage medium may be any tangible medium that contains or stores a program, which may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0279] Since the server, user equipment, computer-readable storage medium, and computer program product in the embodiments of the present application can be applied to the above method, the technical effects that can be obtained can also refer to the above method embodiments, and the embodiments of the present application will not be repeated here.

[0280] The above are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto, and any changes or substitutions within the technical scope disclosed in the present application should be included in the protection scope of the present application.

Claims

1. A lane group division method, characterized in that: The method comprises: Obtain location information of multiple track points in a preset area; Determine the distance between each trajectory point in the plurality of trajectory points and a preset road centerline to obtain a distance sequence, and determine k reference segmentation positions based on the position information of the plurality of trajectory points; the reference segmentation positions are used to segment the plurality of trajectory points, and k is a positive integer; Based on the distance sequence and the k reference segmentation positions, k target segmentation positions are determined, and the preset area is divided based on the k target segmentation positions to obtain a plurality of lane groups.

2. The lane group division method according to claim 1, characterized in that: The step of determining k reference segmentation positions based on the position information of the plurality of trajectory points comprises: Determine a width sequence based on the position information of the plurality of trajectory points; the width sequence is a sequence of maximum widths of the vehicle traveling in the preset area; The k reference segmentation positions are determined based on the width sequence and a preset sliding window algorithm.

3. The lane group division method according to claim 1, characterized in that: The step of determining k target segmentation positions based on the distance sequence and the k reference segmentation positions includes: Determine k candidate segmentation positions based on the k reference segmentation positions, and divide the distance sequence into a plurality of group distance sequence subsets based on the k candidate segmentation positions; The sum of squared errors of the multiple group distance sequence subsets is calculated based on a preset loss function, and when the sum of squared errors meets a preset condition, the k candidate segmentation positions are determined to be the k target segmentation positions; the sum of squared errors is used to characterize the stable state of distribution of the multiple group distance sequence subsets.

4. The lane group division method according to claim 1, characterized in that: The dividing the preset area based on the k target segmentation positions includes: Based on k target segmentation positions and a preset segmentation algorithm, n target segmentation positions are determined, and based on the n target segmentation positions, the preset area is divided into multiple fixed lane areas and lane change areas; the lane change area is used to connect different fixed lane areas, and n is an integer greater than or equal to k.

5. The lane group division method according to claim 4, characterized in that: The method further comprises: In the first lane group, at each preset distance, the number of lanes and lane position information at the corresponding position are obtained through a Gaussian mixture model for trajectory points at the cross section of the road segment in the preset area; the first lane group is any one of the multiple lane groups.

6. The lane group division method according to claim 4, characterized in that: The method further comprises: Determine the transition probabilities of the first lane in the first lane group and each lane in the second lane group respectively, and obtain multiple transition probabilities; the first lane group and the second lane group are two adjacent lane groups in the multiple fixed lane areas, and the first lane is any lane in the first lane group; A target transition probability is determined based on the multiple transition probabilities, and a topological connection is established between a lane corresponding to the target transition probability and the first lane; the target transition probability is the largest transition probability among the multiple transition probabilities.

7. The lane group division method according to any one of claims 1 to 6, characterized in that: The preset area is a non-intersection area, and the method further includes: When determining the virtual lane line of the intersection area and the lane line of the non-intersection area, the transition probability between the virtual lane line and the lane line is determined, and based on the transition probability between the virtual lane line and the lane line, a topological connection between the intersection area and the non-intersection area is established.

8. The lane group division method according to any one of claims 1 to 6, characterized in that: The method further comprises: Determine at least one candidate intersection center of the intersection area, and obtain the heading angle of each reference trajectory in the multiple reference trajectory points to obtain multiple heading angles; the distance between the reference trajectory point and the first candidate intersection center is less than a first preset threshold, and the first candidate intersection center is any one of the at least one candidate intersection center; Determine a sine component and a cosine component of each of the multiple heading angles to obtain a plurality of sine components and a plurality of cosine components, and cluster the plurality of sine components and the plurality of cosine components to obtain a plurality of cluster clusters; When the number of the plurality of clusters is greater than a second preset threshold, the first candidate intersection center is determined as the target intersection center.

9. A lane group division device, characterized in that: The device comprises: an acquisition unit, a determination unit and a division unit; The acquisition unit is used to acquire the position information of multiple track points in a preset area; The determining unit is used to determine the distance between each of the plurality of trajectory points and a preset road centerline to obtain a distance sequence, and determine k reference segmentation positions based on the position information of the plurality of trajectory points; the reference segmentation positions are used to segment the plurality of trajectory points, and k is a positive integer; The determining unit is further configured to determine k target segmentation positions based on the distance sequence and the k reference segmentation positions; The division unit is used to divide the preset area based on the k target segmentation positions to obtain multiple lane groups.

10. An electronic device, characterized in that: including memory and processor; The memory is coupled to the processor; The memory is used to store computer program code, wherein the computer program code includes computer instructions; When the processor executes the computer instructions, the electronic device performs the lane group division method as described in any one of claims 1-8.

11. A computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium, characterized in that: When the instruction is executed on an electronic device, the electronic device executes the lane group division method as described in any one of claims 1 to 8.

12. A vehicle, characterized in that: Comprising the electronic device as claimed in claim 10.

Images