CN115824236A - Lane construction method based on lane-level track - Google Patents
Lane construction method based on lane-level track Download PDFInfo
- Publication number
- CN115824236A CN115824236A CN202211491363.0A CN202211491363A CN115824236A CN 115824236 A CN115824236 A CN 115824236A CN 202211491363 A CN202211491363 A CN 202211491363A CN 115824236 A CN115824236 A CN 115824236A
- Authority
- CN
- China
- Prior art keywords
- short
- connect
- line
- dashed
- lane
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a lane construction method based on lane-level tracks, which comprises the following steps: s1, obtaining semantic object data and lane-level track data; s2, performing segmentation processing on the lane-level track 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 track data, and performing vector point sequence optimization on each segmented semantic longitudinal clustering result set; s5, performing left side buffering preset distance on each longitudinal element in the semantic longitudinal clustering result set respectively to obtain a neighborhood longitudinal result set of each longitudinal element; s6, judging whether lane construction conditions are met or not based on the neighbor longitudinal result set of each longitudinal element; and S7, if the lane construction conditions are met, mutually projecting lane sidelines, cutting and correcting the lane sidelines, and constructing an upper lane sideline and a lower lane sideline based on the corrected lane sideline starting point to complete lane construction.
Description
Technical Field
The invention belongs to the technical field of intelligent traffic, and particularly relates to a lane construction method of a lane-level track.
Background
The rapid development of artificial intelligence technology has attracted much attention, and as an essential part of assisting automatic driving, a high-precision map has always been a focus of researchers how to construct a topology of a road network at low cost and high precision. The crowdsourcing mode collection and updating is a scheme which is low in cost and capable of realizing real-time updating and is suitable for mass production, and the crowdsourcing mode collection and updating has very remarkable advantages. Wherein, the topological connection of the lane lines and the lanes is an important component in the construction of the road network.
The method comprises the following steps of constructing topological communication of lanes and lanes, and finding out: a lane sideline aggregation method based on track direction (CN 201911026989.2) adopts a track line segment between two adjacent track points as a center, and generates a buffer zone of the track line segment according to a set width; combining the buffers of all the track segments to generate a buffer list, judging whether track points are reserved or abandoned in the buffer list according to whether the track points are in the buffer list, and generating a reference track according to all the reserved track points; and extending the track points in the reference track to two sides to generate scanning line segments, clustering the intersection points of the same scanning line segment and the original lane sidelines to generate each clustering point, classifying each clustering point, and fitting into a line to obtain an aggregation result of the lane sidelines. However, this method does not take the elevation value into consideration, and is not suitable for constructing a lane boundary in an area such as an intersection or a level road. A road map generation method, a device and a related system (CN 201811445885.0) determine a road map area to be generated according to current position information; acquiring all road segments of a road map area to be generated from map data; determining the position of each lane group unit of each road segment according to each road segment, and sequentially connecting the lanes of each lane group unit into a whole according to the traveling direction to obtain lane modeling data of the road segments; and obtaining map drawing data according to the lane modeling data, drawing, and generating a road map. Although the method constructs the lane lines and the road map, the road direction is judged based on the traveling direction without combining lane level tracks. Certain errors exist in the specific construction process, and the generated road map cannot truly reflect the real traffic situation.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a lane construction method based on lane level tracks, lane level track data are introduced, the driving condition of a vehicle under the actual condition is reflected truly, and the lane construction method can be effectively used as the judgment basis of the driving direction of a road. And meanwhile, the semantic object data output by the cloud map is longitudinally clustered based on the lane-level track data output by the cloud map, then a neighborhood longitudinal result set of each longitudinal element is obtained according to the longitudinal clustering result, and a lane is constructed based on the minimum polygon principle.
The technical scheme of the invention is as follows:
the invention provides a lane construction method based on lane-level tracks, which comprises the following steps:
the method comprises the following steps of S1, obtaining semantic object data and lane-level track data output by a cloud map;
s2, based on stop line information, diversion area information and track information in lane-level track data in the semantic object data, performing segmentation processing on the lane-level track data and the semantic object data, and associating the segmented lane-level track data with the semantic object data;
s3, combining the segmented lane-level track data, performing longitudinal clustering processing on the segmented semantic object data to obtain a plurality of segmented semantic longitudinal clustering result sets verticals line ；
S4, combining the segmented lane-level track data to collect vertical clustering results of each segmented semantic meaning line Optimizing vector point sequence to obtain semantic longitudinal clustering result set verticals line_ ；
Step S5, the semantic vertical clustering result sets verticals line_ Vertical elements of (1) i Respectively buffering the left side for a preset distance to obtain the vertical elements i Neighbor longitudinal result set near meeting lane construction conditions line ；
Step S6, vertical element vertical based on each vertical element i Neighbor vertical result set of (NEars) line Judging whether the lane construction condition is met;
and S7, if the lane construction conditions are met, mutually projecting lane sidelines, cutting and correcting the lane sidelines, and constructing an upper lane sideline and a lower lane sideline based on the corrected lane sideline starting point to finish lane construction.
Preferably, step S3 comprises:
for each segmented semantic object data after segmentation, executing:
step S31, converting the surface elements in the segmented semantic object data into line elements;
step S32, performing point sequence optimization on line elements in the segmented semantic object data based on segmented lane-level track data associated with the segmented semantic object data;
step S33, short-circuit wire connection is carried out on the wire elements subjected to the dot sequence optimization;
step S34, carrying out long and solid line connection on the line elements subjected to point sequence optimization;
step S35, carrying out secondary connection of short-circuit lines and long solid lines on the line elements subjected to point sequence optimization;
and S36, performing supplementary connection on unconnected data in the elements passing through the point sequence optimization line based on the segmented lane-level track data associated with the segmented semantic object data.
Preferably, step S5 includes:
clustering result sets verticals longitudinally to semantics line_ Vertical elements of (1) i Performing left buffering for a first predetermined distance to obtain vertical components i Of a plurality of neighborhoods longitudinal element near i ；
For each neighbor longitudinal element, ranks i If the azimuth angle is within the predetermined azimuth angle error range and the elevation value is within the predetermined elevation value error range, recording the neighbor longitudinal element near i ；
Each neighborhood longitudinal element near to be recorded i Forming a corresponding vertical element i Neighbor longitudinal result set near line 。
Preferably, in step S7:
vertical for each longitudinal element i Neighbor longitudinal result set near based thereon line If len (nears) is satisfied line ) If =1, the vertical element is used directly l And a neighbor longitudinal element near corresponding thereto i Projecting the lane sidelines mutually;
vertical for each longitudinal element i Based onIts neighbor longitudinal result set near line If len (nears) is satisfied line ) Vertical for this longitudinal element > 1 i And a plurality of neighborhood longitudinal elements near corresponding to the same i Is sorted by the minimum distance, and then is subjected to minimum rectangle method min-polugon To obtain two neighbor longitudinal elements near with minimum area i Are projected onto each other as lane borders.
Preferably, step S31 includes:
acquiring a minimum external rectangular frame containing each surface element in the segmented semantic object data;
calculating the variation dist of the minimum circumscribed rectangle frame on the X axis x And a variation dist in the Y axis Y ；
Variable quantity dist based on minimum circumscribed rectangle frame on X axis x And a variation dist in the Y axis Y Determining the change direction of the minimum circumscribed rectangular frame;
and extracting a generating line of the minimum circumscribed rectangle frame based on the change direction of the minimum circumscribed rectangle frame.
Preferably, step S32 includes:
executing for each line element in the line element set segments in the segmented semantic object data:
acquiring a start of each line element pt And end point end pt ；
Starting point start for each line element based on lane trace line information associated with each line element pt And end point end pt Carrying out the method project Obtaining project after processing start And project end ；
If project start ＞project end Then the vector points of the corresponding line elements are processed in reverse order.
Preferably, step S33 includes:
extracting short-dashed line element set segments from line element set segments subjected to point order optimization dash ；
Traversing the short-dashed element set segment in turn dash Each short dashed element in (a);
longitudinally connecting method for each short-dashed element connection Obtaining a short-dashed line longitudinal element connection result dash ；
Step S34 includes:
extracting long solid line element set segments from the line element set segments subjected to point sequence optimization line ；
Traversing long solid line element set segment in turn line Each long solid line element in (1);
longitudinally connecting method for each long solid line element connection Obtaining the long solid line longitudinal element connection result line ；
Step S35 includes:
connecting results result to short-dashed vertical elements dash Result connected with long solid line longitudinal element line Carrying out longitudinal linking method connection Obtaining the longitudinal connection result of the linear elements vertical ；
Step S36 includes:
extracting track change points;
buffering element set result of searching elements in specified range of track change point track_ And sorting the set of nearby objects by distance;
analyzing and judging the search element and the adjacent object based on the azimuth angle, connecting the adjacent objects which accord with the azimuth angle error range class, and updating to result vertical And temporarily stored in a file.
Preferably, a longitudinal connecting method is performed for each short-dashed element connection Comprises the following steps:
extracting short-dashed line element set segments from line element set segments subjected to point order optimization dash ；
Segment the short-dashed element set dash Each short-dashed line element connect in (1) j Buffering for a second predetermined distance to obtain each short-dashed element connect j Multiple neighbor short dashed line elements near j ；
Obtaining each short-dashed element connect j Starting point connect of start Midpoint connect mid And an end point connect end And each short-dashed element connect j Each neighbor short dashed line element near j The starting point near of (1) start Middle point near mid And an end point near end ；
For each short-dashed element connect j And each adjacent short-dashed element near thereof j All the steps are carried out as follows:
obtaining short-dashed element connect j And neighbor short dashed line element near j Two points with the closest distance between them near And point connect First point of near Is the neighbor short dashed line element near j Point of (3), point of (ii) connect Is a short-dashed element connect j A point on;
if point is satisfied near ∈(near start ,near end ) And satisfy
point connect ∈(connect start ,connect end ) Using neighbor short dashed line element near j The starting point near of start And midpoint near mid By means of methods θ The method calculates the first point near Azimuth angle theta of near And connect using a short-dashed element j Origin connect of start And midpoint connect mid Calculate the second point connect Azimuth angle theta of connect ；
If abs (. Theta.) (θ) connect -θ near ) Within a predetermined error range, according to the azimuth angle theta near And azimuth angle theta connect Judging the short-dashed element connect in the quadrant in which the element is positioned j And neighbor short dashed line element near j The front and back continuing result nextStatus;
determining short-dashed element connect according to the result nextStatus of the front and back connections j And neighbor short dashed line element near j Respective connection points;
connect according to the short-dashed element j Elevation value near of joint z And neighbor short dashed line element near j Altitude connect of splice point z Judging the short-dashed element connect j And neighbor short dashed line element near j Whether longitudinal connection is met;
if the short-dashed element connect j And neighbor short dashed line element near j If the connection is vertical, the short-dashed element is connected j Connection point and adjacent short dashed line element near j The connection points are connected.
Word according to azimuth angle theta near And azimuth angle theta connect Judging the short-dashed element connect in the quadrant in which the element is positioned j And neighbor short dashed line element near j The step of pre-post continuing the result nextStatus comprises the following steps:
calculate the first point near And a second point connect First offset in longitudinal X direction x And a second offset in the latitude Y direction y ；
If the azimuth angle theta near And azimuth angle theta connect Are all in the first quadrant and the first offset x > 0 and a second offset y If > 0, the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is true;
if the azimuth angle theta near And azimuth angle theta connect Are all in the second quadrant and the first offset x > 0 and a second offset y If < 0, the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is true;
if the azimuth angle theta near And azimuth angle theta connect Are all in the third quadrant and the first offset x > 0 and a second offset y If < 0, the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is true;
if the azimuth angle theta near And azimuth angle theta connect Are all in the fourth quadrant and the first offset x ＜0And a second offset y If > 0, the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is true;
in any case other than the above four cases, the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is false.
Preferably, the short-dashed element connect is determined according to the tandem result nextStatus j And neighbor short dashed line element near j The respective connection point steps are:
if the nextStatus is true, determining the short-dashed element connect j End point connect of (2) end As a connection point, a neighbor short-dashed element near j The starting point near of (1) start Is a connection point;
if the nextStatus is false, determining the short-dashed element connect j Starting point connect of start As a connection point, a neighbor short-dashed element near j End point of (9) near end Is the splice point.
The invention has the beneficial effects that:
according to the lane construction processing method of the high-precision map, when semantic object data in the form of points, lines and surfaces and lane-level track line type high-precision map data are acquired, surface elements are converted into linear elements and special attribute values are given to the linear elements. The method is divided into a short-dashed line and a solid line by combining the attributes of the linear elements and is respectively longitudinally connected, so that the sideline of the constructed lane is preliminarily obtained, meanwhile, the point position of the type change of the connecting line element can be quickly obtained, and the construction of the subsequent lane point connection relation is assisted. Judgment factors such as line element azimuth angles, threshold values and lane track lines are introduced into the longitudinal line type element connecting lines and used for constructing lane sidelines, and scenes such as semantic data missing, position errors and the like are adapted by combining conditions in multiple aspects, so that a fault-tolerant mechanism for lane construction is improved.
Drawings
FIG. 1 is a schematic flow chart of a lane construction method based on lane-level tracks according to the present invention;
FIG. 2 is a block diagram of the vertical clustering logic of the present invention;
fig. 3 is an effect diagram of the lane construction based on the lane-level track according to the present invention.
Detailed Description
As shown in fig. 1, an embodiment of the present invention provides a lane construction method based on lane-level data, including:
s1: and acquiring output result data of the cloud map, wherein the output result data comprises semantic object data and lane-level track data.
After the semantic object data are acquired, the semantic object data with the confidence level smaller than a preset specified threshold value need to be screened and filtered, and the semantic object data with the confidence level larger than or equal to the preset specified threshold value are reserved.
The semantic object data comprises semantic information of markers such as stop lines, backward flow areas, arrows and the like; the lane-level track data records lane track line information of the vehicle at each data acquisition time.
S2: based on the stop line, the flow guide area and the track information, the lane segmentation area is initially identified, lane-level track data and semantic object data are segmented, and meanwhile the track id of the lane-level track data is associated with the semantic object data.
S3: combining the segmented lane-level track data, longitudinally clustering the segmented semantic object data, and collecting verticals of the segmented longitudinal clustering result set line The temporary storage is carried out, so that the construction of subsequent lanes is facilitated.
Further, the vertical clustering process in step S3 includes:
s31: and performing linearization processing on surface elements (such as arrows, guide areas, stop lines and the like) in each segmented semantic object data.
Furthermore, the surface transition in step S31 is to obtain a minimum bounding rectangle containing each surface element in the segmented semantic data for each segmented semantic object data. Respectively calculating to obtain the variable quantity dist of the minimum external rectangular frame on the X axis and the Y axis based on the minimum external rectangular frame of the surface element x ，dist y . If dist x >dist y The minimum bounding rectangle is described to be changed based on the direction of the X axis, whereas the minimum bounding rectangle is described to be changed based on the Y axis. And finally, extracting a generating line of the minimum circumscribed rectangular frame based on the change direction of the X axis or the Y axis.
S32: the line elements in the segmented semantic data (the line elements here include the line elements converted in step S31) are optimized in point order based on the segmented lane-level trajectory data.
Further, the dot order optimization in step S32 means: traversing the line element set segments in the segmented semantic object data to obtain the starting start of each line element pt End point end pt . For start pt ，end pt Method based on associated lane trajectory information project (returning the distance along the geometry to the point closest to the specified point) the project is obtained after processing start ，project end . When project is detected start >project end Then, the vector points of the corresponding line elements are processed in reverse order. Wherein the content of the first and second substances,
method project ＝(trackline,p):trackline.project(p)
remarking: trackline is the lane trajectory line, and p is the external point to be analyzed (i.e., the starting point start of each line element) pt End point end pt )。
S33: the short dashed lines connect.
Further, the short-dashed element set segment in the line element set in step S33 dash The vertical connection is carried out, wherein the short-dashed element set segments are obtained by filtering type (for example: type =1, which represents that the element is a line element; type =2, which represents that the element is a point element; type =3, which represents that the element is a plane element) and subtype (for example: subtype for type =1, if subtype =1, which represents that the line element is a short-dashed line) in the segmentation semantic object data dash . Then sequentially traversing the short-dashed element set segments dash For short-dashed element set segments dash Each short-dashed element in (1) is longitudinally connected with the method connection . To the longitudinal directionResult dash Updating and temporarily storing.
S34: the long solid line connects.
Further, the long solid line semantic object data in step S34 are connected longitudinally, where the long solid line includes the flow guiding line (surface turning line) and the long solid line semantic data. Filtering semantic data to obtain long solid line element set segments line
For long solid line element set segments line The longitudinal connecting method is performed for each long solid line element in the connection . For longitudinal result line Updating and temporarily storing.
S35: the short dashed line is secondarily connected to the long solid line.
Further, in step S35, the short dashed line is longitudinally connected with the long solid line for a second time to reset dash 、result line Carrying out longitudinal connection method connection For longitudinal results result vertical And performing temporary storage.
S36: and performing supplementary connection on the unconnected data based on the segmented lane-level track data.
Further, in step S36, the unconnected data are subjected to supplementary connection based on the track, namely track change points are extracted (crossing track level tracks are intersected pairwise to obtain a qualified intersection point set (whether the elevation value of a filtering and screening intersection point is a same-layer track) serving as the track change points), and element set result of elements in the specified range of the track change points is searched in a buffering mode track_ And sorting the set of nearby objects by distance. Analyzing and judging the search element and the adjacent object based on the azimuth angle, connecting the adjacent objects which accord with the azimuth angle error range class, and updating to result vertical And temporarily stored in a file.
Wherein method in S33 to S35 connection The method comprises the following processing steps:
traversing element set connect to be connected segments (here, the element set to be connected connect segments The short-dashed element set segment may be the above-mentioned one dash Long solid line element set segment line And result from the connection of the short dashed lines dash Result after connection with the long solid line line A line element set composed of all elements), each element connect is searched for m Buffering neighbor object element set near within a specified distance range segments . Obtaining element connect m Starting point, middle point, and ending point connect of elements start ,connect mid ,connect end And neighbor object element set near segments Middle neighbor element near m The start, midpoint and end near of an element start 、near mid 、near end (for example, if the element set to be connected here is connect segments Is the short-dashed element set segment dash Then the short-dashed element set connect is obtained segments Each short-dashed line element connect in (1) j And its neighbor short dashed line element near j Start point, midpoint, and end point) of the element connect, and for the element connect m Start point, middle point, end point and neighborhood element near of (c) m Each point element has a longitude X, a latitude Y, and an elevation Z (the elevation Z corresponds to the elevation value of the point element); 1. obtaining near m And connect m Two nearest point of near ,point connect ,point near Is neighbor object element near m Point of (3) connect Is element connect m A point on;
when point near ∈(near start ,near end ) And is
point connect ∈(connect start ,connect end ) The next logic decision is made, otherwise, the loop 2 is skipped and the method is passed θ Calculate point separately near ,point connect Azimuth angle theta of connect 、θ near . When abs (theta) connect -θ near ) Within the predetermined allowable error range, the next judgment is made, otherwise, the loop 3 is skipped and two points theta are passed connect 、θ near Judging connect under the condition that the azimuth angles are in different quadrants i And near i Next to last result nextStatus (True is rear, false is front) 4, and element connect is determined based on nextStatus m And neighbor object element near m Respective connection points; 5. judgment element connect m Altitude connect of splice point z And neighbor object element near m Elevation value near of joint z Whether all the elevation errors are within a preset elevation error range or not, and if the elevation errors meet the above conditions, connecting m And near m Conforming to the vertical link and recording the element connect m Connection point and neighbor object element near m And the continuation point corresponds to the point, otherwise, the loop is skipped.
Wherein, by means of θ Azimuth angle theta of the two points connect 、θ near The principle of calculation is as follows:
method θ ＝Geodesic.WGS84.Inverse(lat1,lon1,lat2,lon2)
remarking: lat1 and lon1 are latitude Y and longitude X of the starting point respectively, and lat2 and lon2 end points are latitude Y and longitude X of the middle point respectively; at solution azimuth angle theta connect And then, satisfy:
θ connect ＝Geodesic.WGS84.Inverse(lat1,lon1,lat2,lon2)
in this case, lat1 and lon1 are the elements connect m Latitude Y and longitude X of the origin; lat2, lon2 is an element connect m Latitude Y and longitude X of the midpoint.
At solution azimuth angle theta near And then, satisfy:
θ near ＝Geodesic.WGS84.Inverse(lat1,lon1,lat2,lon2)
in this case, lat1 and lon1 are neighbor object elements near m Latitude Y and longitude X of the origin; lat2, lon2 is neighbor object element near m Latitude Y and longitude X of the midpoint.
Passing through two points theta connect 、θ near Determining the element connect under the condition that the azimuth angles of the two lines are in different quadrants m Neighbor object element near m The concrete steps of the front and back connection result nextStatus are as follows:
calculate the first point near And a second point connect First offset in longitudinal X direction x And a second offset in the latitude Y direction y ；
If the azimuth angle theta near And azimuth angle theta connect Are all in the first quadrant and the first offset x > 0 and a second offset y If > 0, determining element connect m Neighbor object element near m The nextStatus is true;
if the azimuth angle theta near And azimuth angle theta connect Are all in the second quadrant and the first offset x > 0 and a second offset y If < 0, the element connect is determined m Neighbor object element near m The nextStatus is true;
if the azimuth angle theta near And azimuth angle theta connect Are all in the third quadrant and the first offset x > 0 and a second offset y If < 0, the element connect is determined m Neighbor object element near m The nextStatus is true;
if the azimuth angle theta near And azimuth angle theta connect Are all in the fourth quadrant and the first offset x < 0 and second offset y If > 0, determining element connect m Neighbor object element near m The nextStatus is true;
in any case other than the above four cases, the element connect is determined m Neighbor object element near m The nextStatus is false.
Determining an element connect based on nextStatus m And neighbor object element near m The specific process of each connection point is as follows:
if the nextStatus is true, determining the element connect m End point connect of end As a continuation point, neighbor object element near m The starting point near of (1) start Is a connection point;
if the nextStatus is false, determining element connect m Starting point connect of start As a continuation point, neighbor object element near m End point of (9) near end Is the splice point.
S4: lane construction preprocessing
Further, the lane construction preprocessing in the step S4 is: and reading the longitudinal clustering result and the track data after segmentation. Longitudinal clustering connection result verticals based on segmented track data line And (5) carrying out vector dot sequence optimization.
S5: traversing semantic vertical clustering result set verticals line_ And acquiring the adjacent line of the left specified range.
Further, the obtaining of the adjacent line in step S5: clustering result sets verticals longitudinally to semantics line_ Vertical elements of (1) i Performing left buffer to specify the distance to obtain neighbor longitudinal result set near line . Wherein pairs of neighbor longitudinal result sets near line And (4) logical filtering of azimuth and elevation values is required, and when the azimuth and elevation values are within an error range, the element is recorded. (Vertics for semantic vertical clustering result set line_ Vertical elements of (1) i Performing left buffering for a first predetermined distance to obtain vertical components i Of a plurality of neighborhoods longitudinal element near i For each neighbor longitudinal element, ranks i If the azimuth angle is within the predetermined azimuth angle error range and the elevation value is within the predetermined elevation value error range, recording the neighbor longitudinal element near i (ii) a Each neighborhood longitudinal element near to be recorded i Forming a corresponding vertical element i Neighbor longitudinal result set near line ). If len (nears) line ) If =0, skipping the current cycle continue; if len (nears) line ) If =1, constructing the lane; if len (nears) vertical )>1, then pass through the near object near i Vertical and vertical i Are sorted by the minimum distance of (a), respectively, the minimum rectangle method min-polygon Obtaining near for constructing the minimum area i And the sidelines are used as the sidelines of the lanes, and then the lanes are constructed.
Remarking: method of minimum rectangularization min-polygon The starting points of the two lines are respectively projected with the other side line, and the side line is cut through the projected points to construct polygon.
S6: and (5) constructing a lane.
And further, performing lane construction processing in the step S6, performing mutual projection on lane sidelines, cutting and correcting the lane line, and constructing an upper lane sideline and a lower lane sideline based on the starting point of the sideline.
FIG. 2: the invention relates to a longitudinal clustering connection logic block diagram. And (4) taking the lane-level track as a passing direction to judge and optimize the semantic data vector point sequence. The method comprises the steps of classifying semantic data, clustering short dotted lines, clustering long solid lines, clustering and connecting short dotted line clustering result sets and long solid line clustering result sets, and performing supplementary connection on unconnected data based on tracks.
FIG. 3: and constructing an effect graph for the track-based lane. Taking the embodiment as an example, the left half of the map is the map learning semantic output data. And the right half part is used for constructing a lane result after longitudinally clustering semantic data by taking the lane-level track as a lane direction judgment basis.
The above examples are to be construed as merely illustrative and not limitative of the remainder of the disclosure. After reading the description of the invention, the skilled person can make various changes or modifications to the invention, and these equivalent changes and modifications also fall into the scope of the invention defined by the claims.
Claims (10)
1. A lane construction method based on a lane-level track is characterized by comprising the following steps:
the method comprises the following steps of S1, obtaining semantic object data and lane-level track data output by a cloud map;
s2, based on stop line information, diversion area information and track information in lane-level track data in the semantic object data, performing segmentation processing on the lane-level track data and the semantic object data, and associating the segmented lane-level track data with the semantic object data;
s3, combining the segmented lane-level track data, and respectively carrying out longitudinal clustering processing on the segmented semantic object data to obtain a plurality of segmented semantic longitudinal clustering result sets verticals line ；
S4, combining the segmented lane-level track data to collect vertical clustering results of each segmented semantic meaning line Optimizing vector point sequence to obtain semantic longitudinal clustering result set verticals line_soft ；
Step S5, the semantic vertical clustering result sets verticals line_soft Vertical elements of (1) i Respectively buffering the left side for a preset distance to obtain the vertical elements i Neighbor longitudinal result set near line ；
Step S6, vertical element vertical based on each vertical element i Neighbor longitudinal result set near line Judging whether the lane construction condition is met;
and S7, if the lane construction conditions are met, mutually projecting lane sidelines, cutting and correcting the lane sidelines, and constructing an upper lane sideline and a lower lane sideline based on the corrected lane sideline starting point to finish lane construction.
2. The method according to claim 1, wherein step S3 comprises:
for each segmented semantic object data after segmentation, executing:
step S31, converting the surface elements in the segmented semantic object data into line elements;
step S32, performing point sequence optimization on line elements in the segmented semantic object data based on segmented lane-level track data associated with the segmented semantic object data;
step S33, short-circuit connection is carried out on the line elements subjected to the point sequence optimization;
step S34, carrying out long and solid line connection on the line elements subjected to point sequence optimization;
step S35, carrying out secondary connection of short-circuit lines and long solid lines on the line elements subjected to point sequence optimization;
and S36, performing supplementary connection on unconnected data in the elements passing through the point sequence optimization line based on the segmented lane-level track data associated with the segmented semantic object data.
3. The method according to claim 1, wherein step S5 comprises:
clustering result sets verticals longitudinally to semantics line_soft Vertical elements of (1) i Respectively buffering the left side for a preset distance to obtain the vertical elements i Of a plurality of neighborhoods longitudinal element near i ；
For each neighbor vertical element, ranks i If the azimuth angle is within the predetermined azimuth angle error range and the elevation value is within the predetermined elevation value error range, recording the neighbor longitudinal element near i ；
Each neighborhood longitudinal element near to be recorded i Forming a corresponding vertical element vertical i Neighbor longitudinal result set near line 。
4. The method according to claim 1, characterized in that in step S7:
vertical for each longitudinal element i Neighbor longitudinal result set near based thereon line If len (nears) is satisfied line ) If =1, the vertical element is used directly i And a neighbor longitudinal element near corresponding thereto i Projecting the lane sidelines mutually;
vertical for each longitudinal element i Neighbor longitudinal result set near based thereon line If len (nears) is satisfied line ) Vertical for this longitudinal element > 1 i And a plurality of neighborhood longitudinal elements near corresponding to the same i Is sorted by the minimum distance, and then is subjected to minimum rectangle method min-polygon To obtain two neighbor longitudinal elements near with minimum area i Are projected onto each other as lane borders.
5. The method according to claim 2, wherein step S31 comprises:
acquiring a minimum external rectangular frame containing each surface element in the segmented semantic object data;
calculating the variation dist of the minimum circumscribed rectangle frame on the X axis x And a variation dist in the Y axis Y ；
Variable quantity dist based on minimum circumscribed rectangle frame on X axis x And a variation dist in the Y axis Y Determining the change direction of the minimum circumscribed rectangular frame;
and extracting a generating line of the minimum circumscribed rectangular frame based on the change direction of the minimum circumscribed rectangular frame.
6. The method of claim 5, wherein step S32 comprises:
executing for each line element in the line element set segments in the segmented semantic object data:
acquiring a start of each line element pt And end point end pt ；
Starting point start for each line element based on lane trace line information associated with each line element pt And end point end pt Carrying out the method project Obtaining project after processing start And project end ；
If project start ＞project end Then the vector points of the corresponding line elements are processed in reverse order.
7. The method of claim 5, wherein step S33 comprises:
extracting short-dashed line element set segments from line element set segments subjected to point order optimization dash ；
Sequentially traversing the short-dashed element set segment dash Each short dashed element in (a);
longitudinally connecting method for each short-dashed element connection Obtaining a short-dashed line longitudinal element connection result dash ；
Step S34 includes:
extracting long solid line element set segments from the line element set segments subjected to point sequence optimization line ；
Sequentially traversing long solid line element set segment line Each long solid line element in (1);
longitudinally connecting method for each long solid line element connection Obtaining the long solid line longitudinal element connection result line ；
Step S35 includes:
connecting results result to short-dashed vertical elements dash Result connected with long solid line longitudinal element line Carrying out longitudinal linking method connection Obtaining the longitudinal connection result of the linear elements vertical ；
Step S36 includes:
extracting track lane change points;
buffering element set result of searching elements in specified range of track change point track_change And sorting the set of nearby objects by distance;
analyzing and judging the search element and the adjacent object based on the azimuth angle, connecting the adjacent objects according with the azimuth angle error range class, and updating to result vertical And temporarily stored in a file.
8. The method of claim 7, wherein each short-dashed element is longitudinally connected by means of a method connection Comprises the following steps:
extracting short-dashed line element set segments from line element set segments subjected to point order optimization dash ；
Segment the short-dashed element set dash Each short-dashed line element connect in (1) j Buffering for a second predetermined distance to obtain each short-dashed element connect j Multiple neighbor short dashed line elements near j ；
Obtaining each short-dashed element connect j Starting point connect of start Midpoint connect mid And an end point connect end And each short-dashed element connect j Each neighbor short dashed line element near j The starting point near of (1) start Middle point near mid And an end point near end ；
For each short-dashed element connect j And each adjacent short-dashed element near thereof j All the steps are carried out as follows:
obtaining short-dashed element connect j And neighbor short dashed line element near j Two points with the closest distance between near And point connect First point of near Is the neighbor short dashed line element near j Point of (3), point of (ii) connect Is a short-dashed element connect j A point on;
if point is satisfied near ∈(near start ，near end ) And satisfy point connect ∈(connect start ，connect end ) Using neighbor short dashed line element near j The starting point near of start And midpoint near mid By means of methods θ The method calculates the first point near Azimuth angle theta of near And connect using a short-dashed element j Origin connect of start And midpoint connect mid Calculate the second point connect Azimuth angle theta of connect ；
If abs (. Theta.) (θ) connect -θ near ) Within a predetermined error range, according to the azimuth angle theta near And azimuth angle theta connect Judging the short-dashed element connect in the quadrant in which the element is positioned j And neighbor short dashed line element near j The front and back continuing result nextStatus;
determining short-dashed element connect according to the result nextStatus of the front and back connections j And neighbor short dashed line element near j Respective connection points;
connect according to the short-dashed element j Elevation value near of joint z And neighbor short dashed line element near j Altitude connect of splice point z Judging the short-dashed element connect j And neighbor short dashed line element near j Whether longitudinal connection is met;
if the short-dashed element connect j And neighbor short dashed line element near j If the connection is vertical, the short-dashed element is connected j Connection point and adjacent short-dashed element near j The connection points are connected.
9. The method of claim 8, wherein θ is based on an azimuth angle near And azimuth angle theta connect Judging the short-dashed element connect in the quadrant in which the element is positioned j And neighbor short dashed line element near j The step of pre-post continuing the result nextStatus comprises the following steps:
calculate the first point near And a second point connect First offset in longitudinal X direction x And a second offset in the latitude Y direction y ；
If the azimuth angle theta near And azimuth angle theta connect Are all in the first quadrant and the first offset x > 0 and a second offset y If > 0, the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is true;
if the azimuth angle theta near And azimuth angle theta connect Are all in the second quadrant and the first offset x > 0 and a second offset y If < 0, the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is true;
if the azimuth angle theta near And azimuth angle theta connect Are all in the third quadrant and the first offset x > 0 and a second offset y If < 0, the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is true;
if the azimuth angle theta near And azimuth angle theta connect Are all in the fourth quadrant and the first offset x < 0 and second offset y If > 0, then the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is true;
in any case other than the above four cases, the short-dashed element connect is determined j And neighbor short dashed line element near j The nextStatus is false.
10. The method as claimed in claim 8, wherein the short dash element connect is determined according to a tandem result nextStatus j And neighbor short dashed line element near j The steps of the respective connection points are:
if the nextStatus is true, determining the short-dashed element connect j End point connect of end As a connection point, a neighbor short-dashed element near j The starting point near of (1) start Is a connection point;
if the nextStatus is false, determining the short-dashed element connect j Starting point connect of start As a connection point, a neighbor short-dashed element near j End point of (9) near end Is the splice point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211491363.0A CN115824236A (en) | 2022-11-25 | 2022-11-25 | Lane construction method based on lane-level track |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211491363.0A CN115824236A (en) | 2022-11-25 | 2022-11-25 | Lane construction method based on lane-level track |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115824236A true CN115824236A (en) | 2023-03-21 |
Family
ID=85531620
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211491363.0A Pending CN115824236A (en) | 2022-11-25 | 2022-11-25 | Lane construction method based on lane-level track |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115824236A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116860906A (en) * | 2023-09-05 | 2023-10-10 | 高德软件有限公司 | Track generation method, track generation device, track generation equipment and storage medium |
-
2022
- 2022-11-25 CN CN202211491363.0A patent/CN115824236A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116860906A (en) * | 2023-09-05 | 2023-10-10 | 高德软件有限公司 | Track generation method, track generation device, track generation equipment and storage medium |
| CN116860906B (en) * | 2023-09-05 | 2023-11-28 | 高德软件有限公司 | Track generation method, track generation device, track generation equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fernandes et al. | Point-cloud based 3D object detection and classification methods for self-driving applications: A survey and taxonomy | |
| CN109635748B (en) | Method for extracting road characteristics in high-resolution image | |
| KR102309711B1 (en) | Cnn-based learning method, learning device for selecting useful training data and test method, test device using the same | |
| CN110751099B (en) | Unmanned aerial vehicle aerial video track high-precision extraction method based on deep learning | |
| CN111461209B (en) | Model training device and method | |
| CN111210621B (en) | Signal green wave coordination route optimization control method and system based on real-time road condition | |
| CN115824236A (en) | Lane construction method based on lane-level track | |
| CN111295666A (en) | Lane line detection method, device, control equipment and storage medium | |
| CN112837297A (en) | Progressive multi-scale craniofacial bone fracture detection method | |
| CN115393745A (en) | Automatic bridge image progress identification method based on unmanned aerial vehicle and deep learning | |
| CN113903173B (en) | Vehicle track feature extraction method based on directed graph structure and LSTM | |
| CN111611668A (en) | Road network automatic selection method considering geometric features and semantic information | |
| CN116071722A (en) | Lane geometric information extraction method, system, equipment and medium based on road section track | |
| Feng et al. | Learning multi-layer attention aggregation Siamese network for robust RGBT tracking | |
| Yang et al. | Discretization–filtering–reconstruction: railway detection in images for navigation of inspection UAV | |
| Liu et al. | Multi-lane detection by combining line anchor and feature shift for urban traffic management | |
| CN112833891A (en) | Road data and lane-level map data fusion method based on satellite film recognition | |
| CN117036733A (en) | Urban road scene characteristic line extraction method | |
| CN115309846B (en) | Road network structure identification method based on parallel coefficients | |
| CN113793502B (en) | Pedestrian crossing prediction method under no-signal-lamp control | |
| CN114462609A (en) | Floating car data track reduction method based on hidden Markov model | |
| CN114863411A (en) | License plate recognition method and device | |
| Schmitz et al. | Large-Scale Curb Extraction Based on 3D Deep Learning and Iterative Refinement Post-Processing | |
| Jiang et al. | RailSeg: Learning Local-Global Feature Aggregation with Contextual Information for Railway Point Cloud Semantic Segmentation | |
| Ho et al. | Unweighted Bipartite Matching For Robust Vehicle Counting |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |