A lane construction method based on lane-level trajectories

Abstract

The present invention relates to a lane construction method based on lane-level trajectories, comprising: S1, obtaining semantic object data and lane-level trajectory data; S2, performing segmentation processing on the lane-level trajectory data and the semantic object data; S3, performing longitudinal clustering processing on the segmented semantic object data to obtain a plurality of segmented semantic longitudinal clustering result sets; S4, combining the segmented lane-level trajectory data to perform vector point order optimization on each segmented semantic longitudinal clustering result set; S5, respectively performing a left-side buffer for a predetermined distance on each longitudinal element in the semantic longitudinal clustering result set to obtain a neighborhood longitudinal result set for each longitudinal element; S6, based on the neighborhood longitudinal result set of each longitudinal element, determining whether the lane construction condition is satisfied; S7, if the lane construction condition is satisfied, performing mutual projection on the lane sidelines, cutting and correcting the lane sidelines, and simultaneously constructing upper and lower lane sidelines based on the starting points of the corrected lane sidelines to complete lane construction.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent transportation, and particularly relates to a lane construction method for lane-level trajectories. Background Art

[0002] The rapid development of artificial intelligence technology has attracted much attention to autonomous driving technology. As an indispensable part of assisting autonomous driving, how to construct the topology of the road network at low cost and high precision has always been the focus of researchers. The crowdsourcing method for collection and update is a low-cost and mass-producible solution for real-time update, with very significant advantages. Among them, lane lines and lane topology connectivity are important components in road network construction.

[0003] For the construction of lanes and lane topology connectivity, through patent retrieval, it is found that: "A Lane Edge Aggregation Method Based on Trajectory Direction" (CN201911026989.2). This solution takes the trajectory segment between two adjacent trajectory points as the center, generates a buffer zone of the trajectory segment according to the set width; combines the buffer zones of each trajectory segment to generate a buffer zone list, determines whether to retain or discard the trajectory points according to whether the trajectory points are in the buffer zone list, and generates a reference trajectory according to the retained trajectory points; extends the trajectory points in the reference trajectory to both sides to generate scan segments, clusters the intersection points of the same scan segment and the original lane edge to generate each cluster point, and fits the cluster points into a line after classification to obtain the aggregation result of the lane edge. However, the elevation value is not considered in this method, and it is not suitable for constructing lane edges in areas such as overpasses or hierarchical roads. "A Road Map Generation Method, Device and Related System" (CN201811445885.0) determines the area of the road map to be generated according to the current position information; obtains all road segments in the area of the road map to be generated from the map data; for each road segment, determines the positions of each lane group unit of the road segment, and sequentially connects the lanes of each lane group unit in the driving direction into a whole to obtain the lane modeling data of the road segment; obtains the map drawing data according to the lane modeling data and performs drawing to generate the road map. Although this method constructs lane lines and road maps, it judges the road direction based on the driving direction and does not combine lane-level trajectories. There are certain errors in the specific construction process, and the generated road map cannot truly reflect the actual traffic conditions. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention proposes a lane construction method based on lane-level trajectories, which introduces lane-level trajectory data, truly reflects the driving conditions of vehicles in real situations, and can effectively serve as a basis for judging the driving direction of roads. At the same time, longitudinal clustering is performed on the semantic object data output by the cloud map based on the lane-level trajectory data output by the cloud map, and then the neighborhood longitudinal result sets of each longitudinal element are obtained according to the longitudinal clustering results, and lanes are constructed based on the minimum polygon principle.

[0005] The technical solution of the present invention is as follows:

[0006] The present invention provides a lane construction method based on lane-level trajectories, including:

[0007] Step S1, obtaining semantic object data and lane-level trajectory data output by the cloud map;

[0008] Step S2, based on the stop line information and diversion area information in the semantic object data and the trajectory information in the lane-level trajectory data, segment the lane-level trajectory data and the semantic object data, and associate the segmented lane-level trajectory data and semantic object data;

[0009] Step S3, combining the segmented lane-level trajectory data, performing longitudinal clustering processing on the segmented semantic object data, and obtaining multiple segmented semantic longitudinal clustering result sets verticals line ;

[0010] Step S4, combining the segmented lane-level trajectory data, performing vector point order optimization on each segmented semantic longitudinal clustering result set verticals line to obtain the semantic longitudinal clustering result set verticals line_ ;

[0011] Step S5, for each longitudinal element vertical line_ in the semantic longitudinal clustering result set verticals i respectively perform a left buffer for a predetermined distance to obtain the neighborhood longitudinal result set nears i that meets the lane construction conditions for each longitudinal element vertical line ;

[0012] Step S6, based on the neighborhood longitudinal result set nears i of each longitudinal element vertical line , judge whether the lane construction conditions are met;

[0013] Step S7, if the lane construction conditions are met, project the lane boundary lines onto each other, cut and correct the lane boundary lines, and at the same time construct the upper and lower lane boundary lines based on the starting points of the corrected lane boundary lines to complete the lane construction.

[0014] Preferably, step S3 includes:

[0015] For each segmented semantic object data, the following operations are performed:

[0016] Step S31, convert the surface elements in the segmented semantic object data into line elements;

[0017] Step S32, optimize the point order of the line elements in the segmented semantic object data based on the segmented lane-level trajectory data associated with the segmented semantic object data;

[0018] Step S33, connect the short lines of the line elements with optimized point order;

[0019] Step S34, connect the long solid lines of the line elements with optimized point order;

[0020] Step S35, perform secondary connection of short lines and long solid lines on the line elements with optimized point order;

[0021] Step S36, supplement and connect the unconnected data in the line elements with optimized point order based on the segmented lane-level trajectory data associated with the segmented semantic object data.

[0022] Preferably, step S5 includes:

[0023] For each longitudinal element vertical in the semantic longitudinal clustering result set verticals line_ buffer the left side by a first predetermined distance to obtain multiple neighboring longitudinal elements near of each longitudinal element vertical i ; i For each neighboring longitudinal element nears i ;

[0024] if its azimuth angle is within the predetermined azimuth angle error range and its elevation value is within the predetermined elevation value error range, record this neighboring longitudinal element near i ; i ;

[0025] Form the corresponding neighboring longitudinal result set nears of the longitudinal element vertical with the recorded neighboring longitudinal elements near i ; i line line .

[0026] Preferably, in step S7:

[0027] For each vertical element i , based on its longitudinal neighborhood result set nears line , if len(nears line ) == 1, then directly use this vertical element vertical l and its corresponding longitudinal neighborhood element near i as the lane boundary lines for mutual projection;

[0028] For each vertical element vertical i , based on its longitudinal neighborhood result set nears line , if len(nears line ) > 1, then sort the minimum distances between this vertical element vertical i and its corresponding multiple longitudinal neighborhood elements near i , and then perform minimum rectangle method min-polugon to obtain two longitudinal neighborhood elements near i that construct the minimum area as the lane boundary lines for mutual projection.

[0029] Preferably, step S31 includes:

[0030] Obtain the minimum circumscribed rectangle frame containing each surface element in the segmented semantic object data;

[0031] Calculate the change amount dist x of the minimum circumscribed rectangle frame on the X-axis Y and the change amount dist

[0032] on the Y-axis; x Based on the change amount dist Y of the minimum circumscribed rectangle frame on the X-axis and the change amount dist

[0033] on the Y-axis, determine the change direction of the minimum circumscribed rectangle frame;

[0034] Based on the change direction of the minimum circumscribed rectangle frame, extract the generating line of the minimum circumscribed rectangle frame.

[0035] Preferably, step S32 includes:

[0036] Execute for each line element in the line element set segments in the segmented semantic object data: pt pt Obtain the start point start;

[0037] Based on the lane trajectory line information associated with each line element, for the start point start of each line elementpt and the end point end pt perform the method project After processing, obtain the project start and the project end ;

[0038] If the project start > the project end , then reverse the vector points of the corresponding line elements.

[0039] Preferably, step S33 includes:

[0040] Extract the short dashed line element set segmentst from the line element set segments with optimized point order dash ;

[0041] Traverse each short dashed line element in the short dashed line element set segmentst in turn dash in the set;

[0042] Perform a vertical connection method on each short dashed line element connection to obtain the vertical connection result result of the short dashed line elements dash ;

[0043] Step S34 includes:

[0044] Extract the long solid line element set segmentst from the line element set segments with optimized point order line ;

[0045] Traverse each long solid line element in the long solid line element set segmentst in turn line in the set;

[0046] Perform a vertical connection method on each long solid line element connection to obtain the vertical connection result result of the long solid line elements line ;

[0047] Step S35 includes:

[0048] Perform a vertical connection method on the vertical connection result result of the short dashed line elements dash and the vertical connection result result of the long solid line elements line to obtain the vertical connection result result of the line type elements connection ; vertical ;

[0049] Step S36 includes:

[0050] Extract the trajectory lane change points;

[0051] The set of elements result of the buffered search element within the specified range of the trajectory change point track_ , and sort the set of adjacent objects by distance;

[0052] Analyze and judge the search element and the adjacent objects based on the azimuth angle, connect the adjacent objects that meet the azimuth angle error range class, and update to result vertical , and temporarily store it in a file.

[0053] Preferably, longitudinally connect each short dashed line element method connection The steps include:

[0054] Extract the set of short dashed line elements segmentst from the set of line elements segments with optimized point order dash ;

[0055] The set of short dashed line elements segmentst dash Connect each short dashed line element in j Buffer the second predetermined distance to obtain multiple neighborhood short dashed line elements near of each short dashed line element connect j ; j ;

[0056] Obtain the starting point connect j , midpoint connect start and end point connect mid of each short dashed line element connect end , and the starting point near j , midpoint near j and end point neart start , midpoint near mid and end point neart end ;

[0057] For each short dashed line element connect j and each of its neighborhood short dashed line elements near j Execute:

[0058] Obtain the two points point j nearest in distance between the short dashed line element connect j and the neighborhood short dashed line element near near and point connect , the first point point near is the point on the neighborhood short dashed line element near j , the second point pointconnect For the short dashed line element connect j Points on;

[0059] If point near ∈(near start , near end ) and meet

[0060] point connect ∈(connect start , connect end ), use the starting point near on the neighborhood short dashed line element near j And the midpoint near start And the midpoint near mid Calculate the azimuth angle θ of the first point point θ By the method near Of, and use the starting point connect on the short dashed line element connect near And the midpoint connect j To calculate the azimuth angle θ of the second point point start And the midpoint connect mid Calculate the azimuth angle θ of the second point point connect Of; connect ;

[0061] If abs(θ connect - θ near ) is within the predetermined error range, then according to the azimuth angle θ near And the azimuth angle θ connect Of each quadrant where they are located, judge the forward and backward connection result nextStatus of the short dashed line element connect j And the neighborhood short dashed line element near j ;

[0062] According to the forward and backward connection result nextStatus, determine the connection points of the short dashed line element connect j And the neighborhood short dashed line element near j Respectively;

[0063] According to the elevation value near of the connection point of the short dashed line element connect j And the elevation value near of the neighborhood short dashed line element near z And the elevation value connect of the connection point of the neighborhood short dashed line element near j , judge whether the short dashed line element connect z And the neighborhood short dashed line element near j And the neighborhood short dashed line element near j Meet the longitudinal connection;

[0064] If the short dashed line element connect jand the neighborhood short dashed line element near j If it meets the vertical connection, then connect the short dashed line element j The connection point and the neighborhood short dashed line element near j Connect the connection points.

[0065] The word is based on the azimuth angle θ near and the azimuth angle θ connect In the respective quadrants where they are located, determine the short dashed line element connect j and the neighborhood short dashed line element near j The steps for the front and back connection result nextStatus of are as follows:

[0066] Calculate the first point point near and the second point point connect The first offset offset in the longitude X direction x and the second offset offset in the latitude Y direction y ;

[0067] If the azimuth angle θ near and the azimuth angle θ connect are both in the first quadrant, the first offset offset x > 0 and the second offset offset y > 0, then determine that the short dashed line element connect j and the neighborhood short dashed line element near j The front and back connection result nextStatus is true;

[0068] If the azimuth angle θ near and the azimuth angle θ connect are both in the second quadrant, the first offset offset x > 0 and the second offset offset y < 0, then determine that the short dashed line element connect j and the neighborhood short dashed line element near j The front and back connection result nextStatus is true;

[0069] If the azimuth angle θ near and the azimuth angle θ connect are both in the third quadrant, the first offset offset x > 0 and the second offset offset y < 0, then determine that the short dashed line element connect j and the neighborhood short dashed line element near j The front and back connection result nextStatus is true;

[0070] If the azimuth angle θ near and the azimuth angle θ connect are both in the fourth quadrant, the first offset offset x < 0 and the second offset offset y > 0, then determine that the connection result nextStatus of the short dashed line element connect j and the neighboring short dashed line element near j is true;

[0071] In any other case except the above four cases, it is determined that the connection result nextStatus of the short dashed line element connect j and the neighboring short dashed line element near j is false.

[0072] Preferably, according to the connection result nextStatus, the steps to determine the connection points of the short dashed line element connect j and the neighboring short dashed line element near j are as follows:

[0073] If the connection result nextStatus is true, then determine the end point connect j of the short dashed line element connect end as the connection point, and the start point near j of the neighboring short dashed line element near start as the connection point;

[0074] If the connection result nextStatus is false, then determine the start point connect j of the short dashed line element connect start as the connection point, and the end point neart j of the neighboring short dashed line element near end as the connection point.

[0075] The beneficial effects of the present invention are:

[0076] The high-precision map lane construction processing method provided by the embodiments of the present disclosure, when obtaining semantic object data in the form of points, lines, and surfaces, as well as lane-level trajectory line-type high-precision map data, converts surface elements into linear elements and assigns special attribute values at the same time. Combining the attributes of linear elements, they are divided into two categories: short dashed lines and solid lines and connected longitudinally respectively to initially obtain the side lines of the constructed lane. At the same time, it can quickly obtain the point positions where the connection line element types change, assisting in the construction of the lane point connection relationship in the subsequent steps. The longitudinal line-type element connection line introduces judgment factors such as line element azimuth angle, threshold, and lane trajectory line to construct the lane side line. The combination of various conditions adapts to scenarios such as missing semantic data and incorrect positions, thereby improving the fault tolerance mechanism of lane construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 Schematic flow chart of the lane construction method based on lane-level trajectory according to the present invention;

[0078] Figure 2 Logic block diagram of longitudinal clustering according to the present invention;

[0079] Figure 3 Effect diagram of the lane construction based on lane-level trajectory according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0080] As Figure 1 , the embodiments of the present invention provide a lane construction method based on lane-level data, including:

[0081] S1: Obtain the cloud map output result data, including semantic object data and lane-level trajectory data.

[0082] After obtaining the semantic object data, it is necessary to first screen and filter the semantic object data with a confidence level less than the preset threshold, and retain the semantic object data with a confidence level greater than or equal to the preset threshold.

[0083] The semantic object data contains semantic information of markers such as stop lines, reverse flow areas, and arrows; the lane-level trajectory data records the lane trajectory line information of the vehicle at each data acquisition moment.

[0084] S2: Based on the stop line, diversion area, and trajectory information, initially identify the lane segment area, and perform segmentation processing on the lane-level trajectory data and semantic object data. At the same time, associate the trajectory id of the lane-level trajectory data with the semantic object data.

[0085] S3: Combine the segmented lane-level trajectory data, perform longitudinal clustering on the semantic object data of each segment, and temporarily store the longitudinal clustering result set verticals line for subsequent lane construction.

[0086] Further, the vertical clustering process in step S3 includes:

[0087] S31: Linearize the surface elements (such as arrows, diversion areas, stop lines, etc.) in the segmented semantic object data.

[0088] Further, the surface-to-line conversion in step S31 means: For each segmented semantic object data, obtain the minimum bounding rectangle containing each surface element in the segmented semantic data. Based on the minimum bounding rectangle of the surface element, calculate the change amounts dist x , dist y on the X-axis and Y-axis of the minimum bounding rectangle respectively. If dist x > dist y , it indicates that the minimum bounding rectangle changes based on the X-axis direction; otherwise, the minimum bounding rectangle changes based on the Y-axis. Finally, extract the generated line of the minimum bounding rectangle based on the X-axis or Y-axis change direction.

[0089] S32: Optimize the point order of the line elements in the segmented semantic data based on the lane-level trajectory data after segmentation (the line elements here include the line elements converted in the above step S31).

[0090] Further, the point order optimization in step S32 means: Traverse the set of line elements segments in the segmented semantic object data, and obtain the start point start pt , end point end pt of each line element. Perform method pt (returns the distance to the closest point along this geometry to the specified point) processing on start pt , end project based on the associated lane trajectory line information to obtain project start , project end . When project start > project end , reverse the vector points of the corresponding line element. Among them,

[0091] method project = (trackline, p): trackline.project(p)

[0092] Note: trackline is the lane trajectory line, and p is the external point to be analyzed (i.e., the start point start pt , end point end pt ) of each line element.

[0093] S33: Connect with short dashed lines.

[0094] Further, for the set of short dashed line elements segments in the line element set in step S33 dash perform vertical connection: filter through the type (e.g., type = 1 indicates that the element is a line element; type = 2 indicates that the element is a point element; type = 3 indicates that the element is a surface element) and subtype (e.g., for subtype of type = 1, if subtype = 1, it indicates that the line element is a short dashed line) in the segmented semantic object data to obtain the set of short dashed line elements segments dash . Then traverse the set of short dashed line elements segments in sequence dash , for the set of short dashed line elements segments dash perform vertical connection method on each short dashed line element in it connection . Update and temporarily store the vertical result result dash .

[0095] S34: Long solid line connection

[0096] Further, for the vertical connection of the long solid line semantic object data in step S34, where the long solid line includes a diversion line (surface to line) and long solid line semantic data. Filter the semantic data to obtain the set of long solid line elements segments line

[0097] For each long solid line element in the set of long solid line elements segments line perform vertical connection method connection . Update and temporarily store the vertical result result line .

[0098] S35: Secondary connection of short dashed line and long solid line

[0099] Further, for the secondary vertical connection of the short dashed line and the long solid line in step S35, perform vertical connection method on result dash and result line , and temporarily store the vertical result result connection . vertical .

[0100] S36: Supplementary connection of unconnected data based on segmented lane-level trajectory data

[0101] Further, in step S36, the unconnected data is supplemented and connected based on the trajectory: First, the trajectory lane-changing change points are extracted (by pairwise intersection of lane-level trajectories to obtain a set of qualified intersection points (filtering and screening whether the elevation values of the intersection points are of the same-layer trajectories), as the trajectory lane-changing change points), and the buffer search elements are the element set result within the specified range of the trajectory change points track_ , and the adjacent object set is sorted by distance. Based on the azimuth angle analysis and judgment of the search elements and adjacent objects, the adjacent objects within the azimuth angle error range are connected and updated to result vertical , and temporarily stored in a file

[0102] Among them, in the above S33 to S35, method connection The processing steps of the method are as follows:

[0103] Traverse the set of elements to be connected connect segments (The set of elements to be connected connect here segments can be the above-mentioned set of short dashed line elements segmentst dash 、the set of long solid line elements segmentst line 、and the result obtained after connecting the short dashed lines dash and the result obtained after connecting the long solid lines line jointly composed of the set of line elements), search for the neighborhood object element set near within the specified buffer distance range for each element connect m 。Obtain the starting point, midpoint, and ending point of the element connect segments , connect m , connect start , connect mid , connect end , and the starting point, midpoint, and ending point of the neighborhood element near in the neighborhood object element set near segments 。For the starting point, midpoint, and ending point of the element connect m and the neighborhood element near start , near mid , near end (For example, if the set of elements to be connected connect here segments is the above-mentioned set of short dashed line elements segmentst dash , then what is obtained is the starting point, midpoint, and ending point of each short dashed line element connect in the set of short dashed line elements connect segments and its neighborhood short dashed line element near j ), for the starting point, midpoint, and ending point of the element connect j and the neighborhood element near m of the starting point, midpoint, and ending point of the element connectm For the starting point, midpoint, and ending point, each point feature has longitude X, latitude Y, and elevation Z (the elevation Z corresponds to the elevation value of the point feature); 1. Obtain near m and connect m the two nearest points point near , point connect , point near is a point on the neighborhood object feature near m , point connect is a point on the feature connect m ;

[0104] When point near ∈(near start , near end ) and

[0105] point connect ∈(connect start , connect end ) then proceed to the next logical judgment, otherwise skip this loop 2. Calculate the azimuth angles θ θ of point near , point connect respectively through method connect , θ near . When abs(θ connect -θ near ) is within the predetermined allowable error range, then proceed to the next judgment, otherwise skip this loop 3. Judge the front-back connection result nextStatus (True for the back, False for the front) between connect connect and near near based on the azimuth angles of the two points θ i and θ i in different quadrants 4. Determine the respective connection points of the feature connect m and the neighborhood object feature near m based on nextStatus; 5. Judge whether the elevation value connect m of the connection point of the feature connect z and the elevation value near m of the connection point of the neighborhood object feature near z are both within the predetermined elevation error range. When the above conditions are met, connect m and near m meet the longitudinal connection, and record the feature connect mConnection point and neighborhood object element near m The connection point correspondence, otherwise skip this loop.

[0106] Among them, through method θ The azimuth angles θ of the above two points connect 、θ near The calculation principle is:

[0107] method θ = Geodesic.WGS84.Inverse(lat1, lon1, lat2, lon2)

[0108] Note: lat1 and lon1 are the latitude Y and longitude X of the starting point respectively, and lat2 and lon2 are the latitude Y and longitude X of the midpoint at the end respectively; when solving the azimuth angle θ connect satisfies:

[0109] θ connect = Geodesic.WGS84.Inverse(lat1, lon1, lat2, lon2)

[0110] At this time, lat1 and lon1 refer to the latitude Y and longitude X of the starting point of the element connect m ; lat2 and lon2 refer to the latitude Y and longitude X of the midpoint of the element connect m respectively.

[0111] When solving the azimuth angle θ near satisfies:

[0112] θ near = Geodesic.WGS84.Inverse(lat1, lon1, lat2, lon2)

[0113] At this time, lat1 and lon1 refer to the latitude Y and longitude X of the starting point of the neighborhood object element near m ; lat2 and lon2 refer to the latitude Y and longitude X of the midpoint of the neighborhood object element near m respectively.

[0114] Based on the situations of the azimuth angles of the two points θ connect 、θ near in different quadrants, the specific steps to determine the front and back connection result nextStatus between the element connect m and the neighborhood object element near m are as follows:

[0115] Calculate the first point point near and the second point pointconnect The first offset offset in the longitude X direction x and the second offset offset in the latitude Y direction y ;

[0116] If the azimuth angle θ near and the azimuth angle θ connect are both in the first quadrant, the first offset offset x > 0 and the second offset offset y > 0, then determine that the element connect m and the neighborhood object element near m The front-back connection result nextStatus is true;

[0117] If the azimuth angle θ near and the azimuth angle θ connect are both in the second quadrant, the first offset offset x > 0 and the second offset offset y < 0, then determine that the element connect m and the neighborhood object element near m The front-back connection result nextStatus is true;

[0118] If the azimuth angle θ near and the azimuth angle θ connect are both in the third quadrant, the first offset offset x > 0 and the second offset offset y < 0, then determine that the element connect m and the neighborhood object element near m The front-back connection result nextStatus is true;

[0119] If the azimuth angle θ near and the azimuth angle θ connect are both in the fourth quadrant, the first offset offset x < 0 and the second offset offset y > 0, then determine that the element connect m and the neighborhood object element near m The front-back connection result nextStatus is true;

[0120] In any other case except the above four cases, it is determined that the element connect m and the neighborhood object element near m The front-back connection result nextStatus is false.

[0121] Determine the element connect based on nextStatus m and the neighborhood object element near m The specific process of their respective connection points is as follows:

[0122] If the front - back connection result nextStatus is true, then determine the end point connect m of the element connect end as the connection point, and the starting point near m of the neighborhood object element near start as the connection point;

[0123] If the front - back connection result nextStatus is false, then determine the starting point connect m of the element connect start as the connection point, and the end point neart m of the neighborhood object element near end as the connection point.

[0124] S4: Lane construction pre - processing

[0125] Furthermore, the lane construction pre - processing in step S4: Read the longitudinal clustering result, the segmented trajectory data. Based on the segmented trajectory data, optimize the vector point order of the longitudinal clustering connection result verticals line for the longitudinal clustering connection result verticals

[0126] S5: Traverse the semantic longitudinal clustering result set verticals line_ to obtain the adjacent lines within the specified range on the left.

[0127] Furthermore, the adjacent line acquisition process in step S5: For each longitudinal element vertical line_ in the semantic longitudinal clustering result set verticals i perform a left - hand buffer by a specified distance to obtain the neighborhood longitudinal result set nears line . Among them, for the neighborhood longitudinal result set nears line it is necessary to perform logical filtering on the azimuth angle and elevation value. When the azimuth angle and elevation value are within the error range, record the element. (For each longitudinal element vertical line_ in the semantic longitudinal clustering result set verticals i perform a left - hand buffer by the first predetermined distance respectively to obtain multiple neighborhood longitudinal elements near i for each longitudinal element vertical i For each neighborhood longitudinal element nears i, if its azimuth is within the predetermined azimuth error range and its elevation value is within the predetermined elevation error range, record this longitudinal neighborhood element near i ; For each recorded longitudinal neighborhood element near i , form the corresponding longitudinal element vertical i of the longitudinal neighborhood result set nears line ). If len(nears line ) = 0, skip the current loop continue; if len(nears line ) = 1, perform lane construction; if len(nears vertical ) > 1, then sort by the minimum distance between the adjacent object near i and vertical i , respectively perform minimum rectangularization method min-polygon , obtain the near i side line with the minimum area as the lane side line, and then perform lane construction.

[0128] Note: The minimum rectangularization method min-polygon is to project the starting points of two lines onto the other side line respectively, and cut the side line through the projection points to construct a polygon.

[0129] S6: Lane construction.

[0130] Further, for the lane construction process in step S6, project the lane side lines onto each other, cut and correct the lane lines, and at the same time construct the upper and lower lane side lines based on the starting points of the side lines.

[0131] Figure 2 : It is the longitudinal clustering connection logic block diagram of the present invention. Using the lane-level trajectory as the basis for judging the traffic direction to optimize the semantic data vector point order. Classify the semantic data, cluster the short dashed lines, cluster the long solid lines, perform clustering connection on the short dashed line clustering result set and the long solid line clustering result set, and perform supplementary connection on the unconnected data based on the trajectory.

[0132] Figure 3 : It is the lane construction effect diagram based on the trajectory of the present invention. Taking the embodiment as an example, the left half of the figure is the map learning semantic output data. The right half is the result of constructing the lane after longitudinally clustering the semantic data with the lane-level trajectory as the basis for judging the lane direction.

[0133] The above embodiments should be understood as only for illustrating the present invention and not for limiting the protection scope of the present invention. After reading the content recorded in the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A lane construction method based on lane-level trajectories, characterized in that, Including: Step S1: Obtain semantic object data and lane-level trajectory data output by the cloud map; Step S2: Based on the stop line information and diversion area information in the semantic object data and the trajectory information in the lane-level trajectory data, perform segmentation processing on the lane-level trajectory data and semantic object data, and associate the segmented lane-level trajectory data and semantic object data; Step S3: Combine the lane-level trajectory data after segmentation, and perform vertical clustering processing on the segmented semantic object data respectively to obtain multiple segmented semantic vertical clustering result sets verticals line ; Step S4: Combine the lane-level trajectory data after segmentation, and perform vector point sequence optimization on each segmented semantic vertical clustering result set verticals line to obtain the semantic vertical clustering result set verticals line_soft ; Step S5, for each vertical element vertical in the semantic vertical clustering result set verticals line_soft perform a left buffer of a predetermined distance respectively to obtain the neighborhood vertical result set nears i of each vertical element vertical i ; line ​ Step S6, based on each vertical element vertical i of the neighborhood vertical result set nears line , determine whether the lane construction condition is satisfied; Step S7: If the lane construction conditions are met, project the lane boundaries onto each other, cut and correct the lane boundaries, and simultaneously construct the upper and lower lane boundaries based on the starting points of the corrected lane boundaries to complete lane construction; Step S3 includes: For each piece of segmented semantic object data, perform the following: Step S31: Convert the surface elements in the segmented semantic object data into line elements; Step S32: Based on the segmented lane-level trajectory data associated with the segmented semantic object data, optimize the point order of the line elements in the segmented semantic object data; Step S33: Connect short lines to the line elements after point order optimization; Step S34: Connect long solid lines to the line elements after point order optimization; Step S35: Perform secondary connection of short lines and long solid lines to the line elements after point order optimization; Step S36: Supplement and connect the unconnected data in the line elements after point order optimization based on the segmented lane-level trajectory data associated with the segmented semantic object data; Step S5 includes: For each vertical element vertical in the semantic vertical clustering result set verticals line_soft perform a left buffer of a predetermined distance respectively to obtain multiple neighborhood vertical elements near i of each vertical element vertical i ; i ; For each neighborhood longitudinal element nears i If its azimuth angle is within the predetermined azimuth angle error range and its elevation value is within the predetermined elevation value error range, then record this neighborhood longitudinal element near i ; The recorded longitudinal elements near each neighborhood i are used to form the corresponding longitudinal elements vertical i to obtain the longitudinal result set nears of the neighborhood line ; In Step S7: For each vertical element i , based on its longitudinal neighborhood result set nears line , if len(nears line ) == 1, then directly use this vertical element vertical i and its corresponding longitudinal neighborhood element near i as the lane boundary lines for mutual projection; For each vertical element i , based on its longitudinal neighborhood result set nears line , if len(nears line ) > 1, then sort the minimum distances between this vertical element vertical i and its corresponding multiple longitudinal neighborhood elements near i , and then perform minimum rectangularization method min-polygon to obtain two longitudinal neighborhood elements near i that construct the minimum area and project them onto each other as lane boundaries.

2. The method according to claim 1, characterized in that, Step S31 includes: Obtain the minimum bounding rectangle containing each surface element in the segmented semantic object data; Calculate the change amount dist of the minimum circumscribed rectangle in the X-axis x and the change amount dist in the Y-axis Y ; Based on the change amount dist of the minimum bounding rectangle in the X-axis x and the change amount dist in the Y-axis Y , determine the change direction of the minimum bounding rectangle; Based on the change direction of the minimum bounding rectangle, extract the generating line of the minimum bounding rectangle.

3. The method according to claim 2, wherein Step S32 includes: For each line element in the set of line elements segments in the segmented semantic object data, perform the following: Obtain the start point start of each line element pt and the end point end pt ; Based on the lane trajectory line information associated with each line element, the start point of each line element pt and the end point pt are processed project to obtain project start and project end ; If project start > project end , then reverse the vector points of the corresponding line elements.

4. The method according to claim 2, characterized in that, Step S33 includes: Extract the set of dash-dot line elements segmentst from the set of line elements segments with optimized point order dash ; Traverse each short dashed line element in the short dashed line element set segmentst in sequence dash in it; Vertically connect each short dashed line element method connection to obtain the result of vertically connecting short dashed line elements result dash ; Step S34 includes: Extract the set of long solid line elements segmentst from the set of line elements segments with optimized point order line ; Traverse each long solid line element in the long solid line element set segmentst in sequence line in it; Vertically connect each long solid line element method connection to obtain the result of vertically connecting long solid line elements result line ; Step S35 includes: The connection result of the short dashed vertical elements result dash and the connection result of the long solid vertical elements result line are vertically connected by method connection to obtain the vertical connection result of the line type elements result vertical ; Step S36 includes: Extract the trajectory lane change points; The element set result of the buffered search elements within the specified range of the trajectory change point track_change , and sort the adjacent object set by distance; Analyze and judge the search element and adjacent objects based on the azimuth angle, connect the adjacent objects that meet the azimuth angle error range, and update them to result vertical , and temporarily store them in a file 5. The method according to claim 4, wherein Vertically connect each short dashed line element method connection The steps include: Extract the set of dash-dot line elements segmentst from the set of line elements segments with optimized point order dash ; Buffering each short dashed line element in the set of short dashed line elements segmentst dash by a second predetermined distance to obtain, respectively, multiple neighboring short dashed line elements near j for each short dashed line element connect j ; j ; Obtain the starting points of each short dashed line element connect j the starting point of connect start the midpoint of connect mid and the ending point of connect end and for each short dashed line element connect j the neighboring short dashed line elements near j the starting point of near start the midpoint of near mid and the ending point of near end ; For each short dashed line element connect j and each neighboring short dashed line element near j execute the following: Obtain the two points with the shortest distance between the short dashed line element connect j and the neighborhood short dashed line element near j ; the first point point near and point connect , the first point point near is a point on the neighborhood short dashed line element near j , and the second point point connect is a point on the short dashed line element connect j ; If point is satisfied near ∈(near start ,near end ) and satisfies point connect ∈(connect start , connect end ), using the neighborhood short dashed line feature near j Starting point onnear start and midpoint near mid By method θ Method calculates the first point point near The azimuth angle θ near , and use short dashed line features to connect j Starting point on connect start Connect to the midpoint mid Calculate the second point connect The azimuth angle θ connect ; If abs(θ connect - θ near ) is within a predetermined error range, then according to the quadrants where the azimuth angles θ near and the azimuth angle θ connect are located respectively, judge the sequential connection result nextStatus of the short dashed line element connect j and the neighboring short dashed line element near j ; Determine the short dashed line element connect according to the next connection result nextStatus j and the neighborhood short dashed line element near j their respective connection points; According to the short dashed line element connect j The elevation value of the connection point near z And the neighboring short dashed line element near j The elevation value of the connection point connect z , judge the short dashed line element connect j And the neighboring short dashed line element near j Whether it meets the longitudinal connection; If the short dashed line element connect j and the neighborhood short dashed line element near j meet the vertical connection condition, then connect the connection point of the short dashed line element connect j to the connection point of the neighborhood short dashed line element near j together.

6. The method according to claim 5, characterized in that, According to the azimuth angle θ near and the azimuth angle θ connect in their respective quadrants, the steps of judging the successive result nextStatus of the short dashed line element connect j and the neighboring short dashed line element near j include: Calculate the first point point near and the second point point connect The first offset offset in the longitude X direction x and the second offset offset in the latitude Y direction y ; If the azimuth angle θ near and the azimuth angle θ connect are both in the first quadrant, the first offset offset x > 0 and the second offset offset y > 0, then determine that the connection result nextStatus of the short dashed line element connect j and the neighboring short dashed line element near j is true; If the azimuth angle θ near and the azimuth angle θ connect are both in the second quadrant, the first offset offset x > 0 and the second offset offset y < 0, then determine that the connection result nextStatus of the short dashed line element connect j and the neighboring short dashed line element near j is true; If the azimuth angle θ near and the azimuth angle θ connect are both in the third quadrant, the first offset offset x > 0 and the second offset offset y < 0, then determine that the connection result nextStatus of the short dashed line element connect j and the neighboring short dashed line element near j is true; If the azimuth angle θ near and the azimuth angle θ connect are both in the fourth quadrant, the first offset offset x < 0 and the second offset offset y > 0, then determine that the connection result nextStatus of the short dashed line element connect j and the neighboring short dashed line element near j is true; For any other cases other than the above four cases, the connection result nextStatus of the short dashed line element connect j and the neighborhood short dashed line element near j is determined to be false.

7. The method according to claim 4, characterized in that, Determine the short dashed line element connect according to the successive result nextStatus before and after j and the neighborhood short dashed line element near j The steps for the respective connection points are as follows: If the consecutive result nextStatus is true, determine the end point connect of the short dashed line element j as the connection point, and the start point near of the neighboring short dashed line element near end is the connection point; j as the connection point; start is the connection point; If the consecutive result nextStatus is false, determine the starting point connect of the short dashed line element j as the connection point, and determine the ending point neart of the neighboring short dashed line element near start as the connection point. j end ​​

Images