CN109285163A - Lane line based on laser point cloud or so contour line interactive mode extracting method - Google Patents
Lane line based on laser point cloud or so contour line interactive mode extracting method Download PDFInfo
- Publication number
- CN109285163A CN109285163A CN201811032540.2A CN201811032540A CN109285163A CN 109285163 A CN109285163 A CN 109285163A CN 201811032540 A CN201811032540 A CN 201811032540A CN 109285163 A CN109285163 A CN 109285163A
- Authority
- CN
- China
- Prior art keywords
- point
- lane line
- line
- lane
- cloud
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30241—Trajectory
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
- G06T2207/30256—Lane; Road marking
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of lane line based on laser point cloud or so contour line interactive mode extracting methods comprising following steps: S1, on laser point cloud job platform, specifies N number of lane line starting point manually;S2, each starting point successively execute S3~S6 respectively;S3, with the polygon cut-point cloud;Point cloud after S4, filter segmentation;S5, the adjustment in conjunction with baseline reflectance intensity and standard lane line width to point to coordinate is carried out;If S6, the lane line acquired or so contour line form point quantity are 0, just terminate the calculating that current starting point corresponds to lane line, starts to execute next starting point step S3~S6 until the calculating of all starting points finishes;Otherwise with the midpoint of the lane line that currently acquires or so contour line tail point pair, as new starting point, continue to calculate, until lane line of acquisition or so contour line form point quantity is 0, and terminate the calculating that current starting point corresponds to lane line, start to execute next starting point S3~S6 until the calculating of all starting points finishes.
Description
Technical field
High-precision electronic map of the present invention field, in particular to a kind of lane line based on laser point cloud or so contour line are handed over
Mutual formula extracting method.
Background technique
In recent years, automatic Pilot technology is fast-developing.Necessary sensor of the high-precision electronic map as autonomous driving vehicle
One of, high-precision lane grade road shape information, POI information, ADAS information, topology information etc. are provided for it.Wherein road
It can be used for high-precision positioning with the shape of POI, location information, and construct topology information and ADAS information.Road network is opened up
Flutterring information can be used for path planning, and the topology information of road and POI can be used for POI retrieval on the way, and ADAS information is for height
Grade driving assistance system.The shape of the road of high-precision map and POI, position mostly come from the extraction of laser point cloud at present.
Laser point cloud is the set that the massive point of object space distribution and surface characteristic is obtained using high-precision laser scanner technique.
Lane line based on laser point cloud or so contour line extraction is the left and right contour line shape spy that lane line is obtained from laser point cloud
Point is levied, these characteristic points orderly, with true location information, data volume are far smaller than the data volume of original laser point cloud.
Lane line acquisition is the link of an important time-consuming during high-precision electronic cartography, and guarantees that map prolongs
The high-precision important step of continuous laser point cloud.The existing lane line extracting method based on laser point cloud, including pure artificial hand
A cloud is converted to image by dynamic acquisition, is extracted using Hough transformation, a cloud is converted to image, using deep learning into
Row is extracted, is extracted based on point cloud reflected intensity filtering, then based on connectivity identification.The pure method manually acquired,
Plane precision relative to cloud is high, but since acquisition form point interval is uncertain, can cause the uncertainty of height accuracy, and
It takes a long time, the lane line precision of acquisition and the sense of responsibility of operating personnel have great relevance, higher cost.A cloud is converted into figure
Picture, then on the image, using the edge detection means of traditional image recognition, the method for extracting lane line using Hough transformation,
It can not be effectively applied on bend, and image coordinate and point cloud true coordinate have essence to a certain extent when converting mutually
Degree loss.A cloud is converted into image, the method for recycling deep learning extracts lane line, has by force to the mark quality of lane line
It relies on, needs to obtain a large amount of sample, the production of these sample datas needs a large amount of artificial investment, and can not avoid image seat
Loss of significance problem when mark and point cloud true coordinate are converted mutually.Directly on cloud, filtered based on reflected intensity, then be based on
Connectivity identifies the method extracted, and the lane line of acquisition provides a lane line width in the case where effect is good
Precision.
Summary of the invention
In view of this, the present invention proposes a kind of lane line based on laser point cloud or so contour line interactive mode extracting method.
A kind of lane line based on laser point cloud or so contour line interactive mode extracting method comprising following steps:
S1, on laser point cloud job platform, specify N number of lane line starting point manually, N >=1, select manual guidance acquisition
Or locus guiding acquisition;Source data is needed to be accompanied with and put the matched track data of cloud as pilot point when selecting locus guiding
Collection;A pilot point along lane line extending direction need to be inputted again when selecting manual guidance acquisition;
S2, each starting point successively execute S3~S6 respectively;
S3, search the pilot point nearest apart from starting point, obtain pilot point course, from starting point along course forward, vertical course
Left side guide line and right side guide line are calculated separately out to the left and to the right, and left side guide line and right side guide line are connected into one
Polygon, with the polygon cut-point cloud;
S4, to the point cloud after segmentation, calculate the probability distribution of reflected intensity, take the larger reflected intensity conduct of certain probability
Baseline reflectance intensity, for the point cloud after filter segmentation;
Form point on S5, traversal left side guide line and right side guide line, puts cloud after retrieving the segmentation that S4 step obtains respectively
On closest approach, obtain the point pair on a lane line on left and right sides contour line;In conjunction with baseline reflectance intensity and standard vehicle diatom
Adjustment of the width to point to coordinate is carried out;
If S6, the lane line acquired or so contour line form point quantity are 0, just terminate the meter that current starting point corresponds to lane line
It calculates, starts to execute next starting point step S3~S6 until the calculating of all starting points finishes;Otherwise the lane line currently to acquire
The midpoint of left and right contour line tail point pair continues to calculate as new starting point, until lane line of acquisition or so contour line form point number
Amount is 0, and terminates the calculating that current starting point corresponds to lane line, starts to execute next starting point S3~S6 until all starting point meters
It finishes;If current guidance mode is manual guidance, need the midpoint gather of collected left and right contour line point pair
As new guidance point set;
S7, the smooth of plane and elevation is carried out to each lane line of acquisition or so contour line, then calculates separately left and right wheels
Profile length, takes the highest lane line of confidence level as benchmark, and the position of other lane lines head and the tail point is projected to the lane line
On, and the benchmark lane line part polishing long compared with other projections is intercepted to other lane line tail ends;
S8, the central point for calculating each lane line or so the tail point of contour line execute S2 until each as new starting point
It is terminated when maximum length is less than preset length in the calculated lane contour line of starting point.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S1 includes:
S11, judge the laser point cloud of current region with the presence or absence of matched track;If it exists and the linear and rail of lane line
The case where mark is similar, i.e., lane expansion or traffic lane contraction is not present in road herein, and lane change is not present in track herein, selects track
Lane line acquisition button is guided to carry out locus guiding acquisition;Selected if not meeting aforementioned condition manual guidance lane line acquisition by
Button carries out manual guidance acquisition;Lane line is N, N >=1;
S12, for locus guiding acquisition when, a point is made a call in the white line of each lane line of laser point cloud;For manually
It when guidance acquisition, is got ready in the white line of each lane line of laser point cloud, and on the white line where the last one point, along lane
Line extending direction makes a call to a point, and calculates the course h of most latter two point.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S3 includes:
S31, with the starting point of input P1 as basic point, search a pilot point G1 of the nearest P1 of distance in pilot point, obtain
The course h1 of pilot point G1;
S32, distance sd calculates separately left side reference point L1, right side reference point to vertical course h1 to the left and to the right at P1
R1;Along course h1, fd calculates next basic point P2 forward, searches a pilot point G2 of the nearest P2 of distance in pilot point, and acquisition is drawn
Lead the course h2 of point G2;
S33, it is successively executed M times forward along pilot point by the way of S32, obtains left side guide line form point L1~LM, right side
Guide line form point R1~RM;And in order by left side guide line form point, right side guide line form point links together by inverted sequence, building
Polygon polygon;
S34, PCL points cloud processing library is called to obtain a cloud polygonPC using polygon cut-point cloud.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S4 includes:
S41, each point for traversing polygonPC, statistics reflected intensity are divided into the section of interval at each
Interior quantity, the probability value divided by the sum of point, i.e., in each reflected intensity section;According to statistical value, acquisition probability maximum area
Between, the reflected intensity of non-lane line point cloud as on road, negating and penetrating maximum intensity section initial value is maximum reflection intensity
Maxi is searched from reflected intensity maximum section to smallest interval, and lane line point is regarded as in section of first probability greater than prob
The reflected intensity whitei of cloud takes the larger value of the reflected intensity of whitei and basic point P1 as baseline reflectance intensity basei;
S42, PCL points cloud processing library is called, using lower limit basei, upper limit maxi, reflected intensity is carried out to polygonPC
Filtering, obtain a cloud intensityPC.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S5 includes:
Form point L1~LM on S51, traversal left side guide line, is called PCL, is retrieved using KdTreeFLANN
Closest approach PL1~PLM in the upper three-dimensional space of intensityPC;Form point R1~RM on the guide line of traversal right side, calls PCL,
Use closest approach PR1~PRM in three-dimensional space on KdTreeFLANN retrieval intensityPC;
S52, calculate point to (PLi, PRi), wherein i=1~M, distance lrWidth, if lrWidth is less than standard vehicle
Road line width stdWidth, then calculate PLi, the central point Ci of PRi, searches the course hi of the nearest pilot point of distance Ci, to vertical
The left and right in course calculates separately calcPLi, calcPRi;Enable PLi=calcPLi, PRi=calcPRi;
S53, PLi, the central point Ci of PRi are calculated, if the reflected intensity of Ci is less than lane line reflected intensity whitei, compared
Compared with PLi, the reflection intensity values of PRi, if the reflected intensity of PLi is less than the reflected intensity of PRi, by PLi, PRi two o'clock is to PRi's
Move wdist in direction;If the reflected intensity of PLi is greater than the reflected intensity of PRi, by PLi, PRi two o'clock is mobile to the direction of PLi
wdist。
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S6 includes:
If S61, the lane line acquired or so contour line form point quantity are 0, just terminate the calculating of current lane line, and start
Step S3~S6 is executed until the calculating of all starting points is finished to next starting point;Otherwise S62 is executed;
If S62, current guidance mode are manual guidance, by the midpoint gather of collected left and right contour line point pair
Point set is guided as being packed into;If guidance mode is locus guiding, do not need;
S63, the lane line that currently acquires of calculating or so contour line tail point are to PLM, the midpoint CM of PRM, using CM as newly
Starting point, execute S3~S5, until lane line of acquisition or so contour line form point quantity be 0, terminate current lane line calculating, and
Start to execute next starting point step S3~S6 until the calculating of all starting points finishes.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S7 includes:
S71, step S72 is executed to each lane line of acquisition;
Form point on S72, traversal lane line calculates the flat of a upper form point adjacent thereto since second form point
Identity distance from, if be less than minDist if reject current form point;
S73: it is that traversal S72 is handled to calculate a upper form point adjacent thereto as a result, since second form point, it is next
The angle for the vector that a form point separately constitutes, if not rejecting current form point then within section [π-θ, π+θ];
S74: it is that traversal S73 is handled as a result, since second form point, calculate the z value of a upper form point adjacent thereto
Difference modifies the z of current form point if difference is greater than zthreshold, so that difference is zthreshold;
S75: calculating the length of each left-hand lane line wheel profile, take wherein longest one be benchmark line, calculate other vehicles
The tail point of diatom divides reference line to the subpoint of reference line with subpoint, obtains subpoint to the portion between reference line tail portion
Point, in addition the coordinate shift amount of subpoint and protocercal tail point, copies former lane line tail portion to.
Implement the lane line provided by the invention based on laser point cloud or so contour line interactive mode extracting method and existing skill
Art compares the rapid computations for having the advantages that and having used PCL point cloud open source library to carry out a cloud, it is ensured that point cloud fortune
The high efficiency of calculation, and do not need that cloud will be put and be converted to image can avoid a cloud and loss of significance when image is converted mutually
Problem.And the left and right contour line of lane line, it is reflected intensity changing features visibility point, extracts left and right contour line, for protects
The relative position for demonstrate,proving lane line and point cloud is significant.
Detailed description of the invention
Fig. 1 is lane line based on laser point cloud or so contour line interactive mode extracting method flow chart;
Fig. 2 is lane line or so contour line and form point schematic diagram;
Fig. 3 is user's input point and locus guiding point schematic diagram;
Fig. 4 is user's input point and manual guidance point schematic diagram;
Fig. 5 point cloud segmentation polygon and left and right sides reference point schematic diagram;
Fig. 6 is a plurality of lane line or so contour line schematic diagram on a cloud;
Fig. 7 is that the fine tuning of lane line or so contour line form point calculates schematic diagram.
Specific embodiment
As shown in Fig. 1 to 7, a kind of lane line based on laser point cloud or so contour line interactive mode extracting method comprising
Following steps:
S1, on laser point cloud job platform, specify N number of lane line starting point manually, N >=1, select manual guidance acquisition
Or locus guiding acquisition;Source data is needed to be accompanied with and put the matched track data of cloud as pilot point when selecting locus guiding
Collection;A pilot point along lane line extending direction need to be inputted again when selecting manual guidance acquisition;
S2, each starting point successively execute S3~S6 respectively;
S3, search the pilot point nearest apart from starting point, obtain pilot point course, from starting point along course forward, vertical course
Left side guide line and right side guide line are calculated separately out to the left and to the right, and left side guide line and right side guide line are connected into one
Polygon, with the polygon cut-point cloud;
The angle of vector and direct north that pilot point and adjacent next pilot point are constituted, as pilot point course.
S4, to the point cloud after segmentation, calculate the probability distribution of reflected intensity, take the larger reflected intensity conduct of certain probability
Baseline reflectance intensity, for the point cloud after filter segmentation;
Probability is less than 50% in certain probability.
Form point on S5, traversal left side guide line and right side guide line, puts cloud after retrieving the segmentation that S4 step obtains respectively
On closest approach, obtain the point pair on a lane line on left and right sides contour line;In conjunction with baseline reflectance intensity and standard vehicle diatom
Adjustment of the width to point to coordinate is carried out;
If S6, the lane line acquired or so contour line form point quantity are 0, just terminate the meter that current starting point corresponds to lane line
It calculates, starts to execute next starting point step S3~S6 until the calculating of all starting points finishes;Otherwise the lane line currently to acquire
The midpoint of left and right contour line tail point pair continues to calculate as new starting point, until lane line of acquisition or so contour line form point number
Amount is 0, and terminates the calculating that current starting point corresponds to lane line, starts to execute next starting point S3~S6 until all starting point meters
It finishes;If current guidance mode is manual guidance, need the midpoint gather of collected left and right contour line point pair
As new guidance point set;
S7, the smooth of plane and elevation is carried out to each lane line of acquisition or so contour line, then calculates separately left and right wheels
Profile length, takes the highest lane line of confidence level as benchmark, and the position of other lane lines head and the tail point is projected to the lane line
On, and the benchmark lane line part polishing long compared with other projections is intercepted to other lane line tail ends;
Elevation refers to the z value of the contour line obtained from laser point cloud.Confidence level highest, which refers to, takes the longest lane of length
Line is as benchmark.
S8, the central point for calculating each lane line or so the tail point of contour line execute S2 until each as new starting point
It is terminated when maximum length is less than preset length in the calculated lane contour line of starting point.Preset length can be for less than 1m.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S1 includes:
S11, judge the laser point cloud of current region with the presence or absence of matched track;If it exists and the linear and rail of lane line
The case where mark is similar, i.e., lane expansion or traffic lane contraction is not present in road herein, and lane change is not present in track herein, selects track
Lane line acquisition button is guided to carry out locus guiding acquisition;Selected if not meeting aforementioned condition manual guidance lane line acquisition by
Button carries out manual guidance acquisition;Lane line is N, N >=1;It is similar to refer to that rule is parallel with the lane track on point cloud.
S12, for locus guiding acquisition when, a point is made a call in the white line of each lane line of laser point cloud;For manually
It when guidance acquisition, is got ready in the white line of each lane line of laser point cloud, and on the white line where the last one point, along lane
Line extending direction makes a call to a point, and calculates the course h of most latter two point.
Making a call to a point in white line in step s 12 is starting point, when manual guidance acquires, except the last one prolongs along lane line
The point in direction is stretched, remaining has the double action of starting point and pilot point.The point beaten along lane line extending direction and the last one edge
Lane line extending direction point be used to calculate course h.It is assigned to each starting point.Each starting point just has course attribute.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S3 includes:
S31, with the starting point of input P1 as basic point, search a pilot point G1 of the nearest P1 of distance in pilot point, obtain
The course h1 of pilot point G1;
S32, distance sd calculates separately left side reference point L1, right side reference point to vertical course h1 to the left and to the right at P1
R1;Along course h1, fd calculates next basic point P2 forward, searches a pilot point G2 of the nearest P2 of distance in pilot point, and acquisition is drawn
Lead the course h2 of point G2;
S33, it is successively executed M times forward along pilot point by the way of S32, obtains left side guide line form point L1~LM, right side
Guide line form point R1~RM;And in order by left side guide line form point, right side guide line form point links together by inverted sequence, building
Polygon polygon;
S34, PCL points cloud processing library is called to obtain a cloud polygonPC using polygon cut-point cloud.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S4 includes:
S41, each point for traversing polygonPC, statistics reflected intensity are divided into the section of interval at each
Interior quantity, the probability value divided by the sum of point, i.e., in each reflected intensity section;According to statistical value, acquisition probability maximum area
Between, the reflected intensity of non-lane line point cloud as on road, negating and penetrating maximum intensity section initial value is maximum reflection intensity
Maxi is searched from reflected intensity maximum section to smallest interval, and first probability is greater than prob, and (optionally, prob value is
10%) the reflected intensity whitei of lane line point cloud is regarded as in section, takes the larger of the reflected intensity of whitei and basic point P1
Value is used as baseline reflectance intensity basei;
S42, PCL points cloud processing library is called, using lower limit basei, upper limit maxi, reflected intensity is carried out to polygonPC
Filtering, obtain a cloud intensityPC.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S5 includes:
Form point L1~LM on S51, traversal left side guide line, is called PCL, is retrieved using KdTreeFLANN
Closest approach PL1~PLM in the upper three-dimensional space of intensityPC;Form point R1~RM on the guide line of traversal right side, calls PCL,
Use closest approach PR1~PRM in three-dimensional space on KdTreeFLANN retrieval intensityPC;
S52, calculate point to (PLi, PRi), wherein i=1~M, distance lrWidth, if lrWidth is less than standard vehicle
Road line width stdWidth, then calculate PLi, the central point Ci of PRi, searches the course hi of the nearest pilot point of distance Ci, to vertical
The left and right in course calculates separately calcPLi, calcPRi;Enable PLi=calcPLi, PRi=calcPRi;
S53, PLi, the central point Ci of PRi are calculated, if the reflected intensity of Ci is less than lane line reflected intensity whitei, compared
Compared with PLi, the reflection intensity values of PRi, if the reflected intensity of PLi is less than the reflected intensity of PRi, by PLi, PRi two o'clock is to PRi's
Move wdist in direction;If the reflected intensity of PLi is greater than the reflected intensity of PRi, by PLi, PRi two o'clock is mobile to the direction of PLi
wdist。
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S6 includes:
If S61, the lane line acquired or so contour line form point quantity are 0, just terminate the calculating of current lane line, and start
Step S3~S6 is executed until the calculating of all starting points is finished to next starting point;Otherwise S62 is executed;
If S62, current guidance mode are manual guidance, by the midpoint gather of collected left and right contour line point pair
Point set is guided as being packed into;If guidance mode is locus guiding, do not need;
S63, the lane line that currently acquires of calculating or so contour line tail point are to PLM, the midpoint CM of PRM, using CM as newly
Starting point, execute S3~S5, until lane line of acquisition or so contour line form point quantity be 0, terminate current lane line calculating, and
Start to execute next starting point step S3~S6 until the calculating of all starting points finishes.
In lane line of the present invention based on laser point cloud or so contour line interactive mode extracting method,
The step S7 includes:
S71, step S72 is executed to each lane line of acquisition;
Form point on S72, traversal lane line calculates the flat of a upper form point adjacent thereto since second form point
Identity distance from, if be less than minDist if reject current form point;MinDist is less than for the distance of current point to a upper neighbor point to be given
When determining threshold value, current point is rejected, optionally, minDist value can be 1m.
S73: it is that traversal S72 is handled to calculate a upper form point adjacent thereto as a result, since second form point, it is next
The angle for the vector that a form point separately constitutes, if not rejecting current form point then within section [π-θ, π+θ];π is pi, θ
It is an angle threshold.Value can be π/6.The angle of vector refers to the vector of current point and former and later two neighbor points composition, folder
When angle is more than given threshold value, it is believed that it is linear herein to have wedge angle, then reject current point.
S74: it is that traversal S73 is handled as a result, since second form point, calculate the z value of a upper form point adjacent thereto
Difference modifies the z of current form point if difference is greater than zthreshold, so that difference is zthreshold;Zthreshold is one
The threshold value of a z value.The z value of current point and a upper neighbor point modifies the z value of current point when being more than given threshold value.
S75: calculating the length of each left-hand lane line wheel profile, take wherein longest one be benchmark line, calculate other vehicles
The tail point of diatom divides reference line to the subpoint of reference line with subpoint, obtains subpoint to the portion between reference line tail portion
Point, in addition the coordinate shift amount of subpoint and protocercal tail point, copies former lane line tail portion to.
The embodiment of the present invention takes the interactive method automatically extracted, be because the existing method automatically extracted not yet
It can be applied on a large scale in engineering project.Road is ever-changing, has the extraction method of manual intervention, can simplify
The computational complexity of lane line drawing, and it is aided with flexible UI interaction, it can be convenient operator and pay no attention in algorithm detection effect
The position thought quickly is modified, and the efficient combination manually with algorithm is realized, and can preferably be handled in practical high-precision map producing
The various road conditions encountered.The rapid computations of a cloud are carried out present invention uses PCL point cloud open source library, it is ensured that point cloud operation
High efficiency, and do not need that cloud will be put and be converted to image, a loss of significance when cloud is converted mutually with image can be avoided and asked
Topic.And the left and right contour line of lane line, it is reflected intensity changing features visibility point, extracts left and right contour line, for guarantees
The relative position of lane line and point cloud is significant.
It is understood that for those of ordinary skill in the art, can do in accordance with the technical idea of the present invention
Various other changes and modifications out, and all these changes and deformation all should belong to the protection model of the claims in the present invention
It encloses.
Claims (7)
1. a kind of lane line based on laser point cloud or so contour line interactive mode extracting method, which is characterized in that it includes as follows
Step:
S1, on laser point cloud job platform, specify N number of lane line starting point manually, N >=1, select manual guidance acquisition or rail
Mark guidance acquisition;Source data is needed to be accompanied with and put the matched track data of cloud as guidance point set when selecting locus guiding;
A pilot point along lane line extending direction need to be inputted again when selecting manual guidance acquisition;
S2, each starting point successively execute S3~S6 respectively;
S3, search the pilot point nearest apart from starting point, obtain pilot point course, from starting point along course forward, vertical course is to the left
Calculate separately out left side guide line and right side guide line to the right, by left side guide line and right side guide line connect into one it is polygon
Shape, with the polygon cut-point cloud;
S4, to the point cloud after segmentation, calculate the probability distribution of reflected intensity, take the larger reflected intensity of certain probability as benchmark
Reflected intensity, for the point cloud after filter segmentation;
Form point on S5, traversal left side guide line and right side guide line is put on cloud after retrieving the segmentation that S4 step obtains respectively
Closest approach obtains the point pair on a lane line on left and right sides contour line;In conjunction with baseline reflectance intensity and standard lane line width pair
Adjustment of the point to coordinate is carried out;
If S6, the lane line acquired or so contour line form point quantity are 0, just terminate the calculating that current starting point corresponds to lane line, open
Begin to execute next starting point step S3~S6 until the calculating of all starting points finishes;Otherwise the lane line left and right wheels currently to acquire
The midpoint of profile tail point pair continues to calculate as new starting point, until lane line of acquisition or so contour line form point quantity be 0,
And terminate the calculating that current starting point corresponds to lane line, start to execute next starting point S3~S6 until all starting points have been calculated
Finish;If current guidance mode be manual guidance, need using the midpoint gather of collected left and right contour line point pair as
New guidance point set;
S7, the smooth of plane and elevation is carried out to each lane line of acquisition or so contour line, then calculates separately left and right contour line
Length takes the highest lane line of confidence level as benchmark, the position of other lane lines head and the tail point is projected on the lane line, and
The benchmark lane line part polishing long compared with other projections is intercepted to other lane line tail ends;
S8, the central point for calculating each lane line or so the tail point of contour line execute S2 until each starting point as new starting point
It is terminated when maximum length is less than preset length in calculated lane contour line.
2. lane line based on laser point cloud or so contour line interactive mode extracting method, feature exist as described in claim 1
In,
The step S1 includes:
S11, judge the laser point cloud of current region with the presence or absence of matched track;If it exists and the linear and track phase of lane line
Seemingly, i.e., the case where lane expansion or traffic lane contraction is not present in road herein, and lane change is not present in track herein, selects locus guiding
Lane line acquires button and carries out locus guiding acquisition;Selected if not meeting aforementioned condition manual guidance lane line acquire button into
The acquisition of row manual guidance;Lane line is N, N >=1;
S12, for locus guiding acquisition when, a point is made a call in the white line of each lane line of laser point cloud;For manual guidance
It when acquisition, gets ready in the white line of each lane line of laser point cloud, and on the white line where the last one point, prolongs along lane line
It stretches direction and makes a call to a point, and calculate the course h of most latter two point.
3. lane line based on laser point cloud or so contour line interactive mode extracting method, feature exist as claimed in claim 2
In,
The step S3 includes:
S31, with the starting point of input P1 as basic point, search a pilot point G1 of the nearest P1 of distance in pilot point, obtain guidance
The course h1 of point G1;
S32, distance sd calculates separately left side reference point L1, right side reference point R1 to vertical course h1 to the left and to the right at P1;Edge
Fd calculates next basic point P2 to course h1 forward, searches a pilot point G2 of the nearest P2 of distance in pilot point, obtains pilot point
The course h2 of G2;
S33, it is successively executed M times forward along pilot point by the way of S32, obtains left side guide line form point L1~LM, right side guidance
Linear point R1~RM;And in order by left side guide line form point, right side guide line form point links together by inverted sequence, constructs polygon
Shape polygon;
S34, PCL points cloud processing library is called to obtain a cloud polygonPC using polygon cut-point cloud.
4. lane line based on laser point cloud or so contour line interactive mode extracting method, feature exist as claimed in claim 3
In,
The step S4 includes:
S41, each point for traversing polygonPC count reflected intensity in the section that each is divided into interval
Quantity, the probability value divided by the sum of point, i.e., in each reflected intensity section;According to statistical value, acquisition probability maximum section, i.e.,
For the reflected intensity of lane line point cloud non-on road, negating and penetrating maximum intensity section initial value is maximum reflection intensity maxi, by
Reflected intensity maximum section is searched to smallest interval, and the reflection of lane line point cloud is regarded as in section of first probability greater than prob
Intensity whitei takes the larger value of the reflected intensity of whitei and basic point P1 as baseline reflectance intensity basei;
S42, PCL points cloud processing library is called, using lower limit basei, upper limit maxi, the mistake of reflected intensity is carried out to polygonPC
Filter, obtains a cloud intensityPC.
5. lane line based on laser point cloud or so contour line interactive mode extracting method, feature exist as claimed in claim 4
In,
The step S5 includes:
Form point L1~LM on S51, traversal left side guide line, calls PCL, using on KdTreeFLANN retrieval intensityPC
Closest approach PL1~PLM in three-dimensional space;Form point R1~RM on the guide line of traversal right side, calls PCL, uses
KdTreeFLANN retrieves closest approach PR1~PRM on intensityPC in three-dimensional space;
S52, calculate point to (PLi, PRi), wherein i=1~M, distance lrWidth, if lrWidth is less than standard vehicle diatom
Wide stdWidth then calculates PLi, the central point Ci of PRi, the course hi of the nearest pilot point of distance Ci is searched, to vertical course
Left and right calculate separately calcPLi, calcPRi;Enable PLi=calcPLi, PRi=calcPRi;
S53, PLi, the central point Ci of PRi are calculated, if the reflected intensity of Ci is less than lane line reflected intensity whitei, compared
The reflection intensity values of PLi, PRi, if the reflected intensity of PLi is less than the reflected intensity of PRi, by PLi, PRi two o'clock is to the side of PRi
To mobile wdist;If the reflected intensity of PLi is greater than the reflected intensity of PRi, by PLi, PRi two o'clock is mobile to the direction of PLi
wdist。
6. lane line based on laser point cloud or so contour line interactive mode extracting method, feature exist as claimed in claim 5
In,
The step S6 includes:
If S61, the lane line acquired or so contour line form point quantity are 0, just terminate the calculating of current lane line, and start under
One starting point executes step S3~S6 until the calculating of all starting points finishes;Otherwise S62 is executed;
If S62, current guidance mode be manual guidance, using the midpoint gather of collected left and right contour line point pair as
It is packed into guidance point set;If guidance mode is locus guiding, do not need;
S63, the lane line that currently acquires of calculating or so contour line tail point are to PLM, the midpoint CM of PRM, using CM as new starting point,
S3~S5 is executed, until lane line of acquisition or so contour line form point quantity is 0, terminates the calculating of current lane line, and starts
Step S3~S6 is executed until the calculating of all starting points is finished to next starting point.
7. lane line based on laser point cloud or so contour line interactive mode extracting method, feature exist as claimed in claim 6
In,
The step S7 includes:
S71, step S72 is executed to each lane line of acquisition;
Form point on S72, traversal lane line calculates the plane separation of a upper form point adjacent thereto since second form point
From, if be less than minDist if reject current form point;
S73: it is that traversal S72 is handled as a result, since second form point, calculate a upper form point adjacent thereto, next shape
The angle for the vector that point separately constitutes, if then rejecting current form point not within section [π-θ, π+θ];
S74: it is that traversal S73 is handled to calculate the z value difference of a upper form point adjacent thereto as a result, since second form point, if
Difference is greater than zthreshold, then modifies the z of current form point, so that difference is zthreshold;
S75: calculating the length of each left-hand lane line wheel profile, take wherein longest one be benchmark line, calculate other lane lines
Tail point to reference line subpoint, with subpoint divide reference line, obtain subpoint to the part between reference line tail portion, add
The coordinate shift amount of upper subpoint and protocercal tail point copies former lane line tail portion to.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811032540.2A CN109285163B (en) | 2018-09-05 | 2018-09-05 | Laser point cloud based lane line left and right contour line interactive extraction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811032540.2A CN109285163B (en) | 2018-09-05 | 2018-09-05 | Laser point cloud based lane line left and right contour line interactive extraction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109285163A true CN109285163A (en) | 2019-01-29 |
CN109285163B CN109285163B (en) | 2021-10-08 |
Family
ID=65183550
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811032540.2A Active CN109285163B (en) | 2018-09-05 | 2018-09-05 | Laser point cloud based lane line left and right contour line interactive extraction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109285163B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110060266A (en) * | 2019-04-24 | 2019-07-26 | 百度在线网络技术(北京)有限公司 | Lane line extracting method and device, server and computer-readable medium |
CN110363771A (en) * | 2019-07-15 | 2019-10-22 | 武汉中海庭数据技术有限公司 | A kind of isolation guardrail form point extracting method and device based on three dimensional point cloud |
CN112487123A (en) * | 2020-12-05 | 2021-03-12 | 武汉中海庭数据技术有限公司 | Road connectivity testing method and system based on large-range high-precision map |
CN112598075A (en) * | 2020-12-29 | 2021-04-02 | 武汉中海庭数据技术有限公司 | Crowdsourcing data multi-road segment elevation processing method and device |
CN115953752A (en) * | 2023-03-07 | 2023-04-11 | 中汽创智科技有限公司 | Lane reference line extraction method and device, electronic equipment and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101192307A (en) * | 2006-11-17 | 2008-06-04 | 鸿富锦精密工业(深圳)有限公司 | Point cloud triangular topological relations construction method |
CN105488498A (en) * | 2016-01-15 | 2016-04-13 | 武汉光庭信息技术股份有限公司 | Lane sideline automatic extraction method and lane sideline automatic extraction system based on laser point cloud |
US20180012400A1 (en) * | 2013-01-16 | 2018-01-11 | Microsoft Technology Licensing, Llc | Continuous and dynamic level of detail for efficient point cloud object rendering |
US20180081039A1 (en) * | 2016-09-20 | 2018-03-22 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Data acquiring method and apparatus applied to driverless vehicle |
CN108088445A (en) * | 2016-11-22 | 2018-05-29 | 广州映博智能科技有限公司 | 3 d grid map path planning system and method based on octree representation |
CN108133226A (en) * | 2017-11-27 | 2018-06-08 | 西北工业大学 | One kind is based on the improved three-dimensional point cloud feature extracting methods of HARRIS |
-
2018
- 2018-09-05 CN CN201811032540.2A patent/CN109285163B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101192307A (en) * | 2006-11-17 | 2008-06-04 | 鸿富锦精密工业(深圳)有限公司 | Point cloud triangular topological relations construction method |
US20180012400A1 (en) * | 2013-01-16 | 2018-01-11 | Microsoft Technology Licensing, Llc | Continuous and dynamic level of detail for efficient point cloud object rendering |
CN105488498A (en) * | 2016-01-15 | 2016-04-13 | 武汉光庭信息技术股份有限公司 | Lane sideline automatic extraction method and lane sideline automatic extraction system based on laser point cloud |
US20180081039A1 (en) * | 2016-09-20 | 2018-03-22 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Data acquiring method and apparatus applied to driverless vehicle |
CN108088445A (en) * | 2016-11-22 | 2018-05-29 | 广州映博智能科技有限公司 | 3 d grid map path planning system and method based on octree representation |
CN108133226A (en) * | 2017-11-27 | 2018-06-08 | 西北工业大学 | One kind is based on the improved three-dimensional point cloud feature extracting methods of HARRIS |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110060266A (en) * | 2019-04-24 | 2019-07-26 | 百度在线网络技术(北京)有限公司 | Lane line extracting method and device, server and computer-readable medium |
CN110060266B (en) * | 2019-04-24 | 2021-04-13 | 百度在线网络技术(北京)有限公司 | Lane line extraction method and apparatus, server, and computer-readable medium |
CN110363771A (en) * | 2019-07-15 | 2019-10-22 | 武汉中海庭数据技术有限公司 | A kind of isolation guardrail form point extracting method and device based on three dimensional point cloud |
CN110363771B (en) * | 2019-07-15 | 2021-08-17 | 武汉中海庭数据技术有限公司 | Isolation guardrail shape point extraction method and device based on three-dimensional point cloud data |
CN112487123A (en) * | 2020-12-05 | 2021-03-12 | 武汉中海庭数据技术有限公司 | Road connectivity testing method and system based on large-range high-precision map |
CN112598075A (en) * | 2020-12-29 | 2021-04-02 | 武汉中海庭数据技术有限公司 | Crowdsourcing data multi-road segment elevation processing method and device |
CN115953752A (en) * | 2023-03-07 | 2023-04-11 | 中汽创智科技有限公司 | Lane reference line extraction method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN109285163B (en) | 2021-10-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109285163A (en) | Lane line based on laser point cloud or so contour line interactive mode extracting method | |
CN111462275B (en) | Map production method and device based on laser point cloud | |
CN106570468B (en) | A method of rebuilding LiDAR original point cloud contour of building line | |
Zhang | Towards an operational system for automated updating of road databases by integration of imagery and geodata | |
CN108286979B (en) | Method and device for acquiring high-precision navigation path data, navigation system and driving system | |
CN110220521B (en) | High-precision map generation method and device | |
CN109376586A (en) | Lane boundary line interactive mode extraction method based on laser point cloud | |
CN106780524A (en) | A kind of three-dimensional point cloud road boundary extraction method | |
JP2019135544A (en) | Method and apparatus for identifying intersection in electronic map | |
CN110992473B (en) | Tree branch modeling method and system based on vehicle-mounted laser scanning point cloud | |
CN109118574A (en) | A kind of fast reverse modeling method extracted based on three-dimensional feature | |
CN111897365B (en) | Autonomous vehicle three-dimensional path planning method for contour line guide line | |
CN101975952A (en) | Semi-automatic graph measurement method for digital line graph in onboard LIDAR single-chip mode | |
CN106951905A (en) | Apple identification and localization method on a kind of tree based on TOF camera | |
CN105956542B (en) | High-resolution remote sensing image road extraction method based on statistical matching of structural wire harnesses | |
Haala | Detection of buildings by fusion of range and image data | |
CN110986877B (en) | Railway engineering clearance detection method based on high-precision vehicle-mounted laser mobile measurement system | |
CN108256218A (en) | A kind of subterranean communication tunnel fine modeling method based on actual measurement stringcourse data | |
CN109859317A (en) | A kind of 3DGIS relief model fast modeling method based on CASS and CATIA | |
Zhang et al. | Knowledge-based image analysis for 3D edge extraction and road reconstruction | |
CN112435336B (en) | Curve type identification method and device, electronic equipment and storage medium | |
CN106203278B (en) | The method and device of two-track road on a kind of extraction map | |
CN117423000B (en) | MARS-based intelligent road longitudinal section linear reconstruction method | |
CN107545601B (en) | Automatic generation method for tree height section of overhead transmission line | |
Bulatov et al. | Vectorization of road data extracted from aerial and uav imagery |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |