CN-122024467-A - Intersection lane structure identification method and system based on thunder data

CN122024467ACN 122024467 ACN122024467 ACN 122024467ACN-122024467-A

Abstract

The invention relates to the technical field of intersection lane structure identification, in particular to an intersection lane structure identification method and system based on thunder data, wherein the method comprises the steps of acquiring the thunder monitoring data of an intersection and determining the intersection center point of the intersection; the method comprises the steps of generating track data of each traffic target, identifying the direction of each entrance road of an intersection based on the spatial distribution characteristics of the track data, determining the attribution and steering behavior of the entrance road of each track, and identifying the number of lanes of the entrance road, the steering function of each lane and the center line of the lanes by spatial clustering through a series of cross sections perpendicular to the entrance road based on the track data attributed to the same entrance road. According to the method and the device, the radar monitoring data are fully utilized, the clustering identification of the direction of the entrance road is carried out based on the track space distribution characteristics, the accurate identification of the lane structure and the steering function is realized, the data acquisition cost and the updating period can be remarkably reduced, and meanwhile, the accuracy and the reliability of lane-level identification are ensured.

Inventors

CHENG YU
YANG YUNLIN
ZHANG TENG
CHEN KAIYING
SHA ZHIREN

Assignees

佳都科技集团股份有限公司
广州佳都方纬交通科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251208

Claims (10)

1. An intersection lane structure identification method based on thunder data is characterized by comprising the following steps: Acquiring the thunder-visual monitoring data of an intersection, wherein the thunder-visual monitoring data comprises time sequence position information of a plurality of traffic targets; Determining an intersection center point of the intersection based on the radar monitoring data; Generating track data of each traffic target based on the radar monitoring data; Identifying the directions of all the entrance ways of the intersection by a clustering method based on the spatial distribution characteristics of the track data; determining the attribution and steering behavior of the entrance way of each piece of track data according to the position relation between the track data and the intersection central point and the entrance way direction; Based on the track data belonging to the same entrance way, the number of lanes of the entrance way, the steering function of each lane and the lane center line are identified by spatial clustering through a series of cross sections perpendicular to the entrance way.
2. The method of claim 1, wherein the determining an intersection center point of the intersection based on the radar monitoring data comprises: extracting longitude and latitude coordinates belonging to the same intersection identifier in a preset historical time period; Calculating an arithmetic average value of the longitude and latitude coordinates, and taking the calculated longitude and latitude coordinate point as an intersection center point of the intersection.
3. The method of claim 1, wherein identifying each entrance lane direction of the intersection by a clustering method based on the spatially distributed features of the trajectory data comprises: Screening out a track data subset of which the distance between the starting point and the central point of the intersection exceeds a first preset distance threshold value from the track data; For each piece of track data in the track data subset, calculating a north deflection angle of a direction vector pointing to the central point of the intersection from the starting point of the track data subset to form a north deflection angle set; And performing cluster analysis on the north-bias angle set by adopting a DBSCAN clustering method, and respectively determining the obtained cluster centers as the directions of all the entrance ways of the intersection.
4. A method according to claim 3, wherein said determining, for each piece of trajectory data, its entrance way attribution and steering behavior based on its positional relationship with the intersection center point and the entrance way direction comprises: Calculating a start point direction angle based on a start point position of the track data for each piece of track data, and calculating an end point direction angle based on an end point position of the track data; matching the starting point direction angle with each entrance way direction to determine the entrance way attribution of the track data; Matching the end point direction angle with the determined outlet channel direction corresponding to the inlet channel attribution, wherein the outlet channel direction is the opposite direction of the inlet channel direction corresponding to the inlet channel attribution; and determining the steering behavior of the track data according to the inlet channel attribution and the outlet channel direction obtained by matching.
5. The method of claim 4, wherein calculating a starting point direction angle based on the starting point position of the trajectory data comprises: When the distance between the starting point of the track data and the central point of the intersection is smaller than a second preset distance threshold value, extending along the opposite direction of the vector direction from the starting point to the nearest passing point, and correcting the direction angle of the starting point based on the connecting line direction of the extended point and the central point of the intersection; the calculating the end point direction angle based on the end point position of the track data comprises the following steps: And when the distance between the end point of the track data and the central point of the intersection is smaller than a third preset distance threshold value, extending along the positive direction of the vector direction from the nearest passing point to the end point, and correcting the direction angle of the end point based on the connecting line direction of the extended point and the central point of the intersection.
6. A method according to claim 1 or 3, wherein said spatially clustering by a series of cross-sections perpendicular to said entrance way based on trajectory data belonging to the same entrance way comprises: Determining a vertical cross-sectional direction based on the direction of the inlet channel; setting a plurality of sampling positions in a preset range along the direction of the inlet channel; Constructing a cross section perpendicular to the direction of the inlet channel at each of the sampling locations; based on the constructed series of cross sections, spatial clustering operations are performed on trajectory data intersecting the cross sections.
7. The method of claim 6, wherein identifying the number of lanes of the entrance lane comprises: for each cross section, the following is performed: Calculating the transverse offset distance of the intersection point of each piece of track data belonging to the entrance track and the current cross section to form an offset distance set; performing cluster analysis on the offset distance set by adopting a DBSCAN clustering method; Taking the number of clusters obtained by clustering as the number of lanes at the current cross section; The final number of lanes for the entrance lane is determined statistically based on the number of lanes at all cross sections.
8. The method of claim 6, wherein the identifying the steering function of each lane comprises: Allocating a lane mark to each cluster at each cross section, wherein the lane marks are numbered sequentially from inside to outside according to the transverse position; Screening out track data with the same lane mark distributed at all cross sections as unchanged track data; Counting steering behaviors of the unchanged track data under each lane mark; and determining the steering function of the lane corresponding to each lane mark based on the statistical result.
9. The method of claim 6, wherein the identifying a lane centerline comprises: Extracting the position coordinates of the clustering center of each lane at each cross section for each lane in the finally determined number of lanes; And connecting the position coordinates of the clustering centers of the same lane at all cross sections according to the sampling sequence to form the lane center line of the lane.
10. An intersection lane structure recognition system based on radar data, the system comprising: The system comprises a radar data acquisition module, a radar data processing module and a data processing module, wherein the radar data acquisition module is used for acquiring radar monitoring data of an intersection, and the radar monitoring data comprises time sequence position information of a plurality of traffic targets; the central line point determining module is used for determining an intersection central point of the intersection based on the thunder monitoring data; The track data generation module is used for generating track data of each traffic target based on the radar monitoring data The entrance way direction identification module is used for identifying the directions of the entrance ways of the intersections through a clustering method based on the spatial distribution characteristics of the track data; the steering behavior recognition module is used for determining the attribution and steering behavior of the entrance road of each piece of track data according to the position relation between the track data and the intersection central point and the entrance road direction; The lane structure identification module is used for identifying the number of lanes of the entrance lane, the steering function of each lane and the lane center line through spatial clustering of a series of cross sections perpendicular to the entrance lane based on the track data belonging to the same entrance lane.

Description

Intersection lane structure identification method and system based on thunder data Technical Field The disclosure relates to the technical field of intersection lane structure identification, in particular to an intersection lane structure identification method and system based on thunder data. Background In the fields of automatic driving, high-precision map acquisition, intelligent traffic management and the like, accurate identification of a road junction lane structure is a key technical basis for realizing accurate positioning of vehicles, lane-level path planning and traffic flow accurate analysis. The method has the core task of accurately acquiring geometric and functional information such as the number of lanes of an entrance lane in the range of the intersection, the steering function of each lane, the center line of the lane and the like, thereby providing reliable priori knowledge of road structures for an automatic driving system and a traffic management platform. In the related art, the traditional intersection lane structure acquisition mode mainly comprises professional acquisition vehicle field measurement, inversion identification based on crowdsourcing (Global Positioning System, GPS) data and an aerial photogrammetry method. However, the professional acquisition vehicle mode has high cost, low efficiency and long updating period, the crowdsourcing data mode is limited by low precision and uneven quality of original data, and is difficult to meet the lane-level positioning requirement, and the aerial photogrammetry has the problems of high cost and slow updating, the operation is limited by airspace and weather conditions, and the data of the shielded area is difficult to effectively acquire. Therefore, on the premise of ensuring the data precision and the updating frequency, the acquisition cost is obviously reduced, and the automation and high-reliability identification of complex lane structures including straight running, steering and shared lanes are realized, so that the method is a technical problem to be solved urgently. Disclosure of Invention In view of the above, the present disclosure provides a method and a system for identifying an intersection lane structure based on radar data, so as to solve the problems of significantly reducing the acquisition cost and realizing the automatic and high-reliability identification of complex lane structures including straight running, steering and shared lanes thereof on the premise of ensuring data precision and update frequency. In one aspect, the disclosure provides a method for identifying an intersection lane structure based on radar data, the method comprising: Acquiring the thunder-visual monitoring data of an intersection, wherein the thunder-visual monitoring data comprises time sequence position information of a plurality of traffic targets; Determining an intersection center point of the intersection based on the thunder monitoring data; Generating track data of each traffic target based on the radar monitoring data; identifying the directions of all the entrance ways of the intersections by a clustering method based on the spatial distribution characteristics of the track data; determining the attribution and steering behavior of an entrance road of each piece of track data according to the position relation between the track data and the central point of the intersection and the direction of the entrance road; based on the track data belonging to the same entrance way, the number of lanes of the entrance way, the steering function of each lane and the lane center line are identified by spatial clustering through a series of cross sections perpendicular to the entrance way. Another aspect of the present disclosure also provides an intersection lane structure recognition system based on radar data, the system comprising: The system comprises a radar data acquisition module, a radar data processing module and a data processing module, wherein the radar data acquisition module is used for acquiring radar monitoring data of an intersection, and the radar monitoring data comprises time sequence position information of a plurality of traffic targets; The central line point determining module is used for determining the central point of the intersection based on the thunder monitoring data; The track data generation module is used for generating track data of each traffic target based on the radar monitoring data; The entrance way direction identification module is used for identifying the directions of the entrance ways of the intersections through a clustering method based on the spatial distribution characteristics of the track data; the steering behavior recognition module is used for determining the attribution and steering behavior of the entrance road of each piece of track data according to the position relation between the track data and the center point of the intersection and the direction of the entrance road; the lane str