CN108898672A - A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line - Google Patents
A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line Download PDFInfo
- Publication number
- CN108898672A CN108898672A CN201810393115.XA CN201810393115A CN108898672A CN 108898672 A CN108898672 A CN 108898672A CN 201810393115 A CN201810393115 A CN 201810393115A CN 108898672 A CN108898672 A CN 108898672A
- Authority
- CN
- China
- Prior art keywords
- point
- lane
- road
- cloud
- ground
- 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
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 230000009466 transformation Effects 0.000 claims abstract description 7
- 238000003708 edge detection Methods 0.000 claims abstract description 4
- 230000011218 segmentation Effects 0.000 claims description 15
- 239000000284 extract Substances 0.000 claims description 14
- 238000012886 linear function Methods 0.000 claims description 9
- 238000013461 design Methods 0.000 claims description 7
- 238000013459 approach Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 208000012868 Overgrowth Diseases 0.000 claims description 3
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 10
- 238000004458 analytical method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 241000283070 Equus zebra Species 0.000 description 1
- 241000150100 Margo Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000010937 topological data analysis Methods 0.000 description 1
- 238000009966 trimming Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
-
- G06T5/70—
-
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/181—Segmentation; Edge detection involving edge growing; involving edge linking
-
- 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
- G06T2207/30108—Industrial image inspection
Abstract
The invention discloses a kind of semi-automatic cloud methods for making three-dimensional high-definition mileage chart lane line, include the following steps：S1, pretreatment carry out the removal of coordinate system transformation and non-ground points to point cloud data, obtain ground point；S2, road-edge detection is carried out according to the ground point of acquisition, obtains road boundary and road surface point；S3, the extraction that trade line is clicked through according to the road surface of acquisition, the road markings includes arrow, symbol and lane markings；S4, the lane markings based on extraction, determine the center line in lane；S5, creation have the 3D route map of road boundary, lane markings and lane center.Method provided by the present invention, can reduce the computational complexity and time cost of road surface identification extraction, and improve the precision of creation automatic driving vehicle 3D high definition route map.
Description
Technical field
The present invention relates to pilotless automobile navigation field, specifically a kind of production three-dimensional high-definition mileage chart lane line
Semi-automatic cloud method.
Background technique
Automobile industry is rapidly developing, and strides forward to automatic driving vehicle.And 3D high definition route map is to aid in automatic Pilot
The most powerful tool of the correct driving strategy of vehicle planning.By helping the computer of automobile to estimate travel route, and improve
The understandability for the case where vehicle faces it, 3D high definition route map can mitigate the burden of automatic driving vehicle independent navigation.
Different from traditional route figure, 3D high definition route map can provide the lane grade navigation of centimetre class precision.However now still without or
Few algorithms can be concentrated from large-scale data generate it is accurate effective as a result, and most methods still need some hands
It starts building to make.Therefore, creating 3D high definition route map for automatic driving vehicle is still very big challenge.
Summary of the invention
The purpose of the present invention is to provide a kind of semi-automatic cloud method for making three-dimensional high-definition mileage chart lane line, the party
Method purpose to be achieved mainly has the following：(1) semi-automated algorithm is developed, for generating from a large amount of MLS data reduction curbs
Road boundary and extraction road surface；(2) semi-automated algorithm that lane markings extract is developed by semantic and topological analysis roadway characteristic,
Lane grade navigation is provided for automatic driving vehicle；(3) estimate lane center, design track appropriate for automatic driving vehicle；
(4) 3D high definition route graph model is provided.
To achieve the above object, the present invention uses following technical scheme：
A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line, includes the following steps：
S1, pretreatment carry out the removal of coordinate system transformation and non-ground points to point cloud data, obtain ground point；
S2, road-edge detection is carried out according to the ground point of acquisition, obtains road boundary and road surface point；
S3, the extraction that trade line is clicked through according to the road surface of acquisition, the road markings includes arrow, symbol and vehicle
Road mark；
S4, the lane markings based on extraction, determine the center line in lane；
S5, creation have the 3D route map of road boundary, lane markings and lane center.
Further, specific step is as follows by the S1：
S11, coordinate system transformation；
Front of the x-axis towards vehicle is defined, y-axis selects two companies towards the top of vehicle towards the right side of vehicle, z-axis
Continuous tracing point is as a reference point, to calculate the orientation of rotation angle and fixed x-axis and y-axis；
S12, non-ground points are identified and removed using voxel-based overgrowth method：
Firstly, in XOY plane, by original point cloud according to width WbIt is divided into a cloud mass set (Bi, i=1,2,
3,…,Nb)；
Then, voxel segmentation is carried out to institute's invocation point cloud mass using the method that grows up, obtains the voxel V for being located at segmentation toph
With the voxel V for being located at segmentation bottoml, define local height HlFor VhAnd VlDifference in height, define global height HgFor VhWith entirety
The difference of minimum point in point cloud；
Finally, setting one local Ground Vibration threshold value TlWith a global Ground Vibration threshold value TgIf the H of a voxelg
Less than Tg, HlLess than Tl, then otherwise this voxel is marked as non-ground voxel labeled as ground voxel.
Further, specific step is as follows by the S2：
S21, data segmentation：First by ground point along x-axis with user-defined interval LtIt is divided into multiple localized masses；
S22, the detection of curb point：For each localized mass, candidate curb point is obtained based on two standards of slope and difference in height,
Then have:
Wherein, SslopeIndicate the slope of two continuous phase adjoint points, STIndicate predefined slope threshold value, GiIt is specified point pi
The depth displacement of point adjacent thereto, GminAnd GmaxIndicate minimum and maximum height threshold；
S23, candidate curb point is denoised to obtain curb point：Extracted candidate curb point is eliminated using RANSAC algorithm
In exceptional value, the upper limit of RANSAC algorithm iteration is 1000；
S24, road boundary fitting：
A curb point is respectively extracted in the two sides of each localized mass, according to the coordinate of curb point, calculating passes through any two
The linear function of continuous road sign point, then will generate the road boundary of part between the continuous curb point of any two；
In XOY plane, linear function is described as：Y=kx+b, wherein k indicates the slope of line, and b is the intercept of y；
In three dimensions, linear function is described as：
Wherein p0(x0,y0,z0) it is point on line, vectorIt is the direction vector parallel with line；
S25, road surface are extracted：The point for being located at the inside of corresponding road boundary in each localized mass is determined as road surface point and is mentioned
It takes.
Further, specific step is as follows by S3：
S31, road sign is extracted using intensity threshold and condition Euclid's clustering algorithm, the road sign includes arrow, symbol
And lane markings, specially：
S311, using intensity threshold algorithm, identified from the point of road surface and extract candidate road sign point；
Exceptional value in S312, the candidate road sign point of removal：
K neighbours' point is found to certain point, k is the user-defined threshold value according to average dot density, if defined
Without enough neighbours' points in radius, then the point is considered as exceptional value, and removes from cloud；
In addition, the average distance from point to its neighbour is calculated if establishing the closest approach of a point in a particular area,
Assuming that the distribution of the average distance of all the points should follow Gaussian Profile, then determine by global distance average and standard deviation definition
Interval except point be outlier, and removed from cloud；
S313, the road sign point that cluster is obtained using condition Euclid's clustering procedure；
Sparse point is divided into different clusters according to the Euclidean distance between certain point and its closest approach, if between them
Euclidean distance be less than given threshold value dc, then these points will be assigned in identical cluster, threshold value dcIt is close by the point of test data
Degree and resolution ratio determine；
S32, lane markings extract:Design parameter based on lane markings carries out geometry filtering to the lane markings of acquisition.
Further, specific step is as follows by S4:
S41, the point of the lane markings in each lane is divided into hundreds of pieces along the x axis, extracts two in an individual block
The coordinate of the central point of the lane markings of side, coordinate are calculated by following formula：
Wherein Xmax, Xmin, Ymax, Ymin, Zmax, ZminPoint in each piece respectively x, y and z directionss maximum value and most
Small value；
S42, calculate each piece of two sides lane markings central point coordinate average value to determine lane center point
Coordinate.
After adopting the above technical scheme, compared with the background technology, the present invention, having the following advantages that：
1, the present invention proposes improved curb detection and road surface partitioning algorithm, and the algorithm based on segmentation is for detecting curb
Point.It is different from the curb detection method previously based on segmentation, carry out first it is voxel-based grow up algorithm, to be examined on road surface
Non-ground points are removed before surveying.The data complexity of further process can be substantially reduced and calculate cost.
2, the invention proposes one kind based on local highway layout standard, arrives margo threshold based on intensive analysis and distance
The new algorithm of value detection lane markings, overcomes apart from the uncertainty to track dependent thresholds method.The threshold value that the algorithm uses
It is determined by region specified link design standard, greatly reduces the time loss of optimal threshold estimation.
3, the present invention estimates the lane center of automatic driving vehicle, and provide 3D by extracting different roadway characteristics
The example of high definition route map.
Detailed description of the invention
Fig. 1 is a kind of work flow diagram for semi-automatic cloud method for making three-dimensional high-definition mileage chart lane line of the present invention；
Fig. 2 (a) is original point cloud schematic diagram, and Fig. 2 (b) is that Fig. 2 (a) removes the result schematic diagram after non-ground points；
Fig. 3 is the result schematic diagram that road surface extraction is carried out to Fig. 2 (b)；
Fig. 4 is the result schematic diagram that road markings extraction is carried out to Fig. 3；
Fig. 5 is the result schematic diagram removed after noise spot to Fig. 4；
Fig. 6 is for the discrete and sparse point of the present invention by successful division at the schematic diagram of different Semantic Clusterings；
Fig. 7 is the schematic diagram for the lane middle line that the present invention calculates.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, not
For limiting the present invention.
Embodiment
Before the present embodiment is described in detail, it should be pointed out that production provided by the present embodiment is three-dimensional high
Semi-automatic cloud method of clear mileage chart lane line, is a kind of automanual mode, utilizes the link creation road in three-dimensional point cloud
Line chart.
Referring to Fig. 1, the present invention provides a kind of semi-automatic cloud method for making three-dimensional high-definition mileage chart lane line, packet
Include following steps：
S1, pretreatment carry out the removal of coordinate system transformation and non-ground points to point cloud data, obtain ground point.
The step is realized especially by following steps：
S11, coordinate system transformation.MLS system, which is utilized one, has the right hand of any user-defined orientation and starting point just
Hand over coordinate system.Such occasional frame makes the relative position of tracing point become more difficult.In order to reduce further place
The complexity of the explanation in step is managed, the orientation of vehicle frame is fixed, so that x-axis is towards the front of vehicle, y-axis is towards the right side of vehicle
Side, and z-axis is towards the top of vehicle.Select two continuous tracing points as a reference point, to calculate rotation angle and fixed x
The orientation of axis and y-axis.
In addition, when the road of processing is curved road, the step for without carrying out coordinate system conversion.
S12, non-ground points (such as eaves paulin, building, traffic mark are identified and removed using voxel-based overgrowth method
Will etc.).
It is voxel-based that grow up method key step as follows：
Firstly, in XOY coordinate plane, by original point cloud according to width WbIt is divided into a cloud mass set (Bi, i=1,
2,3,…,Nb)；
Then, voxel segmentation is carried out to institute's invocation point cloud mass using the method that grows up, obtains the voxel V for being located at segmentation toph
With the voxel V for being located at segmentation bottoml, define local height HlFor VhAnd VlDifference in height, define global height HgFor VhWith entirety
The difference of minimum point in point cloud；
Finally, setting one local Ground Vibration threshold value TlWith a global Ground Vibration threshold value TgIf the H of a voxelg
Less than Tg, HlLess than Tl, then this voxel is otherwise marked as non-labeled as ground voxel (voxel comprising ground point)
Ground voxel (voxel not comprising ground point).
As shown in Fig. 2 (b), after non-ground points remove step, the point cloud of trees and street lamp in Fig. 2 (a) is filtered out.
S2, road-edge detection is carried out according to the ground point of acquisition, obtains road boundary and road surface point.
The step is realized especially by following steps：
S21, data segmentation.First by data along x-axis with user-defined interval (Lt) it is divided into multiple localized masses.
Different from tracing point segmentation some previous studies on road surface are used, this method can allow user to define each piece
The sum of length or block because track data be along the sparse and fixed point of the certain intervals of track, rather than directly handle
Entire data set, the process points in each localized mass can greatly reduce the negative effect of road surface fluctuation.Most of all, for
More blocks will distribute less point in each piece.Interval appropriate is set, can be examined in every side of this block so as to each piece
At least one roadside candidate point is measured, the length of block otherwise should be increased.According to the size and feature of test data set, total block data is logical
Often in the range of 30 to 500.
In the step, if track data is available, data subregion can also be carried out by perpendicular to tracing point.
S22, the detection of curb point.
After data segmentation, origin cloud is divided into different localized masses, for each localized mass, using slope and difference in height two
Standard extracts candidate curb point.According to the design of the urban road network of many countries and construction reference, pavement is higher than road, i.e.,
The height of kerbstone is 3 to 30 centimetres, in addition, road surface and the slope of road boundary are greater than the gradient of road continuity point；Then for
Specified point piHave,
Wherein SslopeIndicate the slope of two continuous phase adjoint points, STIndicate predefined slope threshold value, GiIt is specified point piWith
The depth displacement of its consecutive points, GminAnd GmaxIndicate minimum and maximum height threshold.Slope threshold value, minimum and maximum height threshold root
It is determined according to urban road network's design and construction reference.
S23, candidate curb point is denoised to obtain curb point.
The curb point of all extractions is it is expected that be the bottom of curb or guardrail；However, some extraction points still inevitably go out
Existing deviation, although most of offsets are very small, the boundary point extracted is there is still a need for being modified, because lane markings extract
Performance place one's entire reliance upon extraction curb point accuracy.Therefore, it is proposed to which a kind of be based on RANdom unanimity of samples
(RANSAC) curb optimization algorithm eliminates abnormal point.
RANSAC algorithm is a kind of iterative algorithm, can be used for extracting shape and estimation from one group of data with exceptional value
The parameter of mathematical model.RANSAC algorithm draws a subset from initial data by random iteration to realize this mesh
Mark.The minimal set of selection is considered as the interior point assumed, and defines a model according to point in these.In addition, model
All parameters are all determined by interior point.It compares every other data to test the model or shape of estimation, how many to be determined
Point is also suitble to candidate family, if model certain approximate point well, is regarded as interior point.One fairly good model should wrap
Inlier is used as containing enough points.In order to refine candidate family, when series of points is classified as new interior, the parameter of model
It is reevaluated.Finally, assessing model, score is calculated according to internal error.As a result, having rejected first internal insufficient
Model, followed by score lower model.This process is repeated, after the test of certain number, comprising at most interior value and most
The model of high score is extracted.
However, the shortcomings that RANSAC is to estimate that the calculating time of optimal models does not have the upper limit.It will be continued searching preferably
Model, until reaching user-defined limitation quantity.In invention, the number due to extracting curb point is typically not greater than 300,
Without or rare exceptional value, so the upper bound of iteration be 1000.
S24, road boundary fitting：A curb point is respectively extracted in the two sides of each localized mass, according to the coordinate of curb point,
The linear function for passing through the continuous road sign point of any two is calculated, then will generate the road roadside of part between the continuous curb point of any two
Boundary.
In the xy plane, linear function is described as：Y=kx+b；Wherein k indicates the slope of line, and b is y-intercept.
In three dimensions, lines are described with the parametric form of linear function：
Wherein p0(x0,y0,z0) it is point on line, vectorIt is the direction vector parallel with line；
S25, road surface are extracted：The point for being located at the inside of corresponding road boundary in each localized mass is determined as road surface point and is mentioned
It takes.As shown in Fig. 3 the road surface figure extracted.
S3, the extraction that trade line is clicked through according to the road surface of acquisition, the road markings includes arrow, symbol and vehicle
Road mark.
The step is realized especially by following steps：
S31, road sign is extracted using intensity threshold and condition Euclid's clustering algorithm, the road sign includes arrow, symbol
(character etc.) and lane markings, specially：
S311, using intensity threshold algorithm, identified from the point of road surface and extract candidate road sign point；
Since the reflectivity of lane markings is much higher than unvarnished road surface, so strength information is to can be used for lane markings inspection
The key criterion of survey.Using intensity threshold algorithm, road sign is adaptively identified according to strength information.It is as shown in Figure 4 to extract result.
Exceptional value in S312, the candidate road sign point of removal；
Most of noise spot is isolated point and the point mutation in some areas, therefore, can be accordingly from the lane markings of extraction
Remove exceptional value.In order to identify exceptional value, first attempt to find k neighbours' point from certain point.K is the use according to average dot density
The threshold value that family defines.If without enough neighbours in defined radius, the point is considered as exceptional value, and from point Yun Zhongyi
It removes.In addition, if establishing the closest approach (with the point apart from nearest point) of a point in a particular area, then will calculate from point
To the average distance of its neighbours' (all and neighbouring point of the point), it is assumed that the distribution of the average distance of all the points should follow Gauss point
Cloth, and outlier is considered by the point except the interval of global distance average and standard deviation definition and from data
Concentrate trimming.If Fig. 5 is the result schematic diagram after removing exceptional value.
S313, the road sign point that cluster is obtained using condition Euclid's clustering procedure；
Sparse point is divided into different clusters according to the Euclidean distance between certain point and its closest approach.
Firstly, randomly choosing a point P as starting point, and it is classified as to cluster point first；Meanwhile creating a sky
List C, and the queue Q of point for needing to check；
Secondly, being d in radiusthSphere in Searching point P point neighbours (Pi, i=1,2 ..., n), and by PiIt is added to
Current queue Q；For each neighbours Pi, calculate point neighbours PiTo the Europe of P it is several in distance, if its Euclidean distance is less than
Given threshold value dc, then it is added into the list of cluster C, threshold value dcIt is determined by the dot density and resolution ratio of test data；If
All point neighbours P are handledi, then Q is reset into sky list.
New cluster point is considered as new starting point, repeats identical process.When in sphere there is no when non-agglomerated point, weight
The list of new setting cluster C, and a new starting point is randomly choosed from remaining non-cluster point.
As shown in fig. 6, working as all the points end processed, algorithm is terminated, and discrete and sparse point is by successful division at difference
Semantic Clustering.
S32, lane markings extract.According to highway layout standard, lane is extracted by part strength analysis and distance threshold
Mark.In general, there are five types of the road markings of type：Solid line, dotted line, zebra stripes, Chinese character and rotation arrow.According to " city
Roadmarking coating standard in road signs design specification (GB51038-2015) " is generated for constraining lane markings
The smallest bounding rectangles, the model of lane markings point is extracted using the geometric parameter (length and width) of minimum boundary rectangle
It encloses, forms lane markings.
S4, the lane markings based on extraction, determine the center line in lane；
The step is realized especially by following steps：
According to the center line in every lane of position estimation of the lane markings of extraction.By taking the road of the rightmost side as an example, by
Being found on the right side of road apart from its right margin distance is (d2+d1+d3/ 2) point substantially estimates lane center, wherein d1For road
Shoulder width on the right side of road, d2For the width of right-hand lane mark, d3For lane width.
Whether lane center in order to check estimation is accurate, is suitble to navigation, and the present embodiment is according to the positions of lane markings
Lane center is estimated again, to generate more accurate result.
S41, the point of the lane markings in each lane is divided into hundreds of pieces along the x axis, extracts two in an individual block
The coordinate of the central point of the lane markings of side, coordinate are calculated by following formula：
Wherein Xmax, Xmin, Ymax, Ymin, Zmax, ZminPoint in each piece respectively x, y and z directionss maximum value and most
Small value；
S42, calculate each piece of two sides lane markings central point coordinate average value to determine lane center point
Coordinate.
S41, the point of the lane markings in each lane is divided into hundreds of pieces along the x axis, extracts two in an individual block
The coordinate of the central point of the lane markings of side, coordinate are calculated by following formula：
Wherein Xmax, Xmin, Ymax, Ymin, Zmax, ZminPoint in each piece respectively x, y and z directionss maximum value and most
Small value；
S42, similarly calculates the coordinate average value of the central point of the lane markings of each piece of two sides to determine lane center
The coordinate of line point.As shown in fig. 7, being the schematic diagram of the lane middle line extracted.
S5, creation 3D route map：Comprehensive extracted road boundary, lane markings and lane center, creation have essence
The 3D route map of true road boundary, lane markings and the lane center of estimation.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with scope of protection of the claims
Subject to.
Claims (5)
1. a kind of semi-automatic cloud method for making three-dimensional high-definition mileage chart lane line, it is characterised in that：Include the following steps：
S1, pretreatment carry out the removal of coordinate system transformation and non-ground points to point cloud data, obtain ground point；
S2, road-edge detection is carried out according to the ground point of acquisition, obtains road boundary and road surface point；
S3, the extraction that trade line is clicked through according to the road surface of acquisition, the road markings include arrow, symbol and lane mark
Know；
S4, the lane markings based on extraction, determine the center line in lane；
S5, creation have the 3D route map of road boundary, lane markings and lane center.
2. a kind of semi-automatic cloud method for making three-dimensional high-definition mileage chart lane line according to claim 1, feature
It is：Specific step is as follows by the S1：
S11, coordinate system transformation；
Front of the x-axis towards vehicle is defined, y-axis is towards the right side of vehicle, and for z-axis towards the top of vehicle, selection two continuous
Tracing point is as a reference point, to calculate the orientation of rotation angle and fixed x-axis and y-axis；
S12, non-ground points are identified and removed using voxel-based overgrowth method：
Firstly, in XOY plane, by original point cloud according to width WbIt is divided into a cloud mass set Bi, i=(1,2,3 ...,
Nb)；
Then, voxel segmentation is carried out to institute's invocation point cloud mass using the method that grows up, obtains the voxel V for being located at segmentation tophBe located at
Divide the voxel V of bottoml, define local height HlFor VhAnd VlDifference in height, define global height HgFor VhIn whole point cloud
The difference of minimum point；
Finally, setting one local Ground Vibration threshold value TlWith a global Ground Vibration threshold value TgIf the H of a voxelgIt is less than
Tg, HlLess than Tl, then otherwise this voxel is marked as non-ground voxel labeled as ground voxel.
3. a kind of semi-automatic cloud method for making three-dimensional high-definition mileage chart lane line according to claim 1, feature
It is：Specific step is as follows by the S2：
S21, data segmentation：First by ground point along x-axis with user-defined interval LtIt is divided into multiple localized masses；
S22, the detection of curb point：For each localized mass, candidate curb point is obtained based on two standards of slope and difference in height, then
Have:
Wherein, SslopeIndicate the slope of two continuous phase adjoint points, STIndicate predefined slope threshold value, GiIt is specified point piWith it
The depth displacement of consecutive points, GminAnd GmaxIndicate minimum and maximum height threshold；
S23, candidate curb point is denoised to obtain curb point：It is eliminated in extracted candidate curb point using RANSAC algorithm
Exceptional value, the upper limit of RANSAC algorithm iteration are 1000；
S24, road boundary fitting：
A curb point is respectively extracted in the two sides of each localized mass, according to the coordinate of curb point, is calculated continuous by any two
The linear function of road sign point will then generate the road boundary of part between the continuous curb point of any two；
In XOY plane, linear function is described as：Y=kx+b, wherein k indicates the slope of line, and b is the intercept of y；
In three dimensions, linear function is described as：
Wherein p0(x0,y0,z0) it is point on line, vector v→<M, n, p>It is the direction vector parallel with line；
S25, road surface are extracted：The point for being located at the inside of corresponding road boundary in each localized mass is determined as road surface point and is extracted.
4. a kind of semi-automatic cloud method for making three-dimensional high-definition mileage chart lane line according to claim 1, feature
It is：Specific step is as follows by S3：
S31, extract road sign using intensity threshold and condition Euclid's clustering algorithm, the road sign include arrow, symbol and
Lane markings, specially：
S311, using intensity threshold algorithm, identified from the point of road surface and extract candidate road sign point；
Exceptional value in S312, the candidate road sign point of removal：
K neighbours' point is found to certain point, k is the user-defined threshold value according to average dot density, if in defined radius
Interior no enough neighbours' points, then the point is considered as exceptional value, and removes from cloud；
In addition, calculating the average distance from point to its neighbour if establishing the closest approach of a point in a particular area, it is assumed that
The distribution of the average distance of all the points should follow Gaussian Profile, then determine by between global distance average and standard deviation definition
Point except is outlier, and is removed from cloud；
S313, the road sign point that cluster is obtained using condition Euclid's clustering procedure；
Sparse point is divided into different clusters according to the Euclidean distance between certain point and its closest approach, if the Europe between them
Family name's distance is less than given threshold value dc, then these points will be assigned in identical cluster, threshold value dcBy test data dot density and
Resolution ratio determines；
S32, lane markings extract:Design parameter based on lane markings carries out geometry filtering to the lane markings of acquisition.
5. a kind of semi-automatic cloud method for making three-dimensional high-definition mileage chart lane line according to claim 1, feature
It is：Specific step is as follows by S4:
S41, the point of the lane markings in each lane is divided into hundreds of pieces along the x axis, extracts two sides in an individual block
The coordinate of the central point of lane markings, coordinate are calculated by following formula：
Wherein Xmax, Xmin, Ymax, Ymin, Zmax, ZminIt is maximum value and minimum of the point in each piece in x, y and z directionss respectively
Value；
S42, calculate each piece of two sides lane markings central point coordinate average value to determine the coordinate of lane center point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810393115.XA CN108898672A (en) | 2018-04-27 | 2018-04-27 | A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810393115.XA CN108898672A (en) | 2018-04-27 | 2018-04-27 | A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108898672A true CN108898672A (en) | 2018-11-27 |
Family
ID=64342526
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810393115.XA Pending CN108898672A (en) | 2018-04-27 | 2018-04-27 | A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108898672A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109738910A (en) * | 2019-01-28 | 2019-05-10 | 重庆邮电大学 | A kind of curb detection method based on three-dimensional laser radar |
| CN109798903A (en) * | 2018-12-19 | 2019-05-24 | 广州文远知行科技有限公司 | A kind of method and device obtaining road information from map datum |
| CN110097620A (en) * | 2019-04-15 | 2019-08-06 | 西安交通大学 | High-precision map creation system based on image and three-dimensional laser |
| CN110110678A (en) * | 2019-05-13 | 2019-08-09 | 腾讯科技（深圳）有限公司 | Determination method and apparatus, storage medium and the electronic device of road boundary |
| CN110378175A (en) * | 2018-08-16 | 2019-10-25 | 北京京东尚科信息技术有限公司 | The recognition methods of road edge and device |
| CN110807412A (en) * | 2019-10-30 | 2020-02-18 | 驭势科技（北京）有限公司 | Vehicle laser positioning method, vehicle-mounted equipment and storage medium |
| CN110866449A (en) * | 2019-10-21 | 2020-03-06 | 北京京东尚科信息技术有限公司 | Method and device for identifying target object in road |
| WO2020154967A1 (en) * | 2019-01-30 | 2020-08-06 | Baidu.Com Times Technology (Beijing) Co., Ltd. | Map partition system for autonomous vehicles |
| CN112325809A (en) * | 2021-01-06 | 2021-02-05 | 广东技术师范大学 | Method for detecting flatness of flange |
| CN112441022A (en) * | 2019-09-02 | 2021-03-05 | 华为技术有限公司 | Lane center line determining method and device |
| CN112560747A (en) * | 2020-12-23 | 2021-03-26 | 苏州工业园区测绘地理信息有限公司 | Vehicle-mounted point cloud data-based lane boundary interactive extraction method |
| CN112561944A (en) * | 2020-11-27 | 2021-03-26 | 中央财经大学 | Lane line extraction method based on vehicle-mounted laser point cloud |
| CN113632035A (en) * | 2019-02-25 | 2021-11-09 | 德卡产品有限公司 | System and method for surface feature detection and traversal |
| CN113837124A (en) * | 2021-09-28 | 2021-12-24 | 中国有色金属长沙勘察设计研究院有限公司 | Automatic extraction method of geotextile routing inspection route of mud discharging warehouse |
| CN116153057A (en) * | 2022-09-12 | 2023-05-23 | 东北林业大学 | Method for estimating lane width based on laser radar point cloud |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106157219A (en) * | 2016-06-30 | 2016-11-23 | 北京工业大学 | Road axis extracting method based on vehicle-mounted scanning system and device |
| CN106570446A (en) * | 2015-10-12 | 2017-04-19 | 腾讯科技（深圳）有限公司 | Lane line extraction method and device |
| CN106646508A (en) * | 2016-11-24 | 2017-05-10 | 中国科学院自动化研究所 | Slope angle estimation method for slope region based on multiline laser radar |
| CN106842231A (en) * | 2016-11-08 | 2017-06-13 | 长安大学 | A kind of road edge identification and tracking |
| CN106846272A (en) * | 2017-01-18 | 2017-06-13 | 西安工程大学 | A kind of denoising compressing method of point cloud model |
-
2018
- 2018-04-27 CN CN201810393115.XA patent/CN108898672A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570446A (en) * | 2015-10-12 | 2017-04-19 | 腾讯科技（深圳）有限公司 | Lane line extraction method and device |
| CN106157219A (en) * | 2016-06-30 | 2016-11-23 | 北京工业大学 | Road axis extracting method based on vehicle-mounted scanning system and device |
| CN106842231A (en) * | 2016-11-08 | 2017-06-13 | 长安大学 | A kind of road edge identification and tracking |
| CN106646508A (en) * | 2016-11-24 | 2017-05-10 | 中国科学院自动化研究所 | Slope angle estimation method for slope region based on multiline laser radar |
| CN106846272A (en) * | 2017-01-18 | 2017-06-13 | 西安工程大学 | A kind of denoising compressing method of point cloud model |
Non-Patent Citations (5)
| Title |
|---|
| 房华乐: "面向对象的移动激光扫描点云数据提取城市道路信息方法研究", 《中国优秀硕士学位论文全文数据库 (基础科学辑)》 * |
| 李明辉 等: "基于车载三维激光扫描的道路线提取研究", 《测绘与空间地理信息》 * |
| 李鹏飞 等: ""点云模型的噪声分类去噪算法"", 《计算机工程与应用》 * |
| 涂宏斌 等: "《基于机器学习的行为识别技术研究》", 30 September 2016, 知识产权出版社 * |
| 赵成伟: "地面激光点云分割技术的研究", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378175A (en) * | 2018-08-16 | 2019-10-25 | 北京京东尚科信息技术有限公司 | The recognition methods of road edge and device |
| CN109798903A (en) * | 2018-12-19 | 2019-05-24 | 广州文远知行科技有限公司 | A kind of method and device obtaining road information from map datum |
| CN109798903B (en) * | 2018-12-19 | 2021-03-30 | 广州文远知行科技有限公司 | Method and device for acquiring road information from map data |
| CN109738910A (en) * | 2019-01-28 | 2019-05-10 | 重庆邮电大学 | A kind of curb detection method based on three-dimensional laser radar |
| KR102334641B1 (en) * | 2019-01-30 | 2021-12-03 | 바이두닷컴 타임즈 테크놀로지(베이징) 컴퍼니 리미티드 | Map Partitioning System for Autonomous Vehicles |
| WO2020154967A1 (en) * | 2019-01-30 | 2020-08-06 | Baidu.Com Times Technology (Beijing) Co., Ltd. | Map partition system for autonomous vehicles |
| KR20200096726A (en) * | 2019-01-30 | 2020-08-13 | 바이두닷컴 타임즈 테크놀로지(베이징) 컴퍼니 리미티드 | Map partitioning system for autonomous vehicles |
| CN111771206A (en) * | 2019-01-30 | 2020-10-13 | 百度时代网络技术（北京）有限公司 | Map zoning system for autonomous vehicles |
| US11468690B2 (en) | 2019-01-30 | 2022-10-11 | Baidu Usa Llc | Map partition system for autonomous vehicles |
| CN113632035A (en) * | 2019-02-25 | 2021-11-09 | 德卡产品有限公司 | System and method for surface feature detection and traversal |
| CN110097620A (en) * | 2019-04-15 | 2019-08-06 | 西安交通大学 | High-precision map creation system based on image and three-dimensional laser |
| CN110110678A (en) * | 2019-05-13 | 2019-08-09 | 腾讯科技（深圳）有限公司 | Determination method and apparatus, storage medium and the electronic device of road boundary |
| CN110110678B (en) * | 2019-05-13 | 2021-05-14 | 腾讯科技（深圳）有限公司 | Method and apparatus for determining road boundary, storage medium, and electronic apparatus |
| CN112441022A (en) * | 2019-09-02 | 2021-03-05 | 华为技术有限公司 | Lane center line determining method and device |
| CN110866449A (en) * | 2019-10-21 | 2020-03-06 | 北京京东尚科信息技术有限公司 | Method and device for identifying target object in road |
| CN110807412A (en) * | 2019-10-30 | 2020-02-18 | 驭势科技（北京）有限公司 | Vehicle laser positioning method, vehicle-mounted equipment and storage medium |
| CN110807412B (en) * | 2019-10-30 | 2022-09-23 | 驭势科技（北京）有限公司 | Vehicle laser positioning method, vehicle-mounted equipment and storage medium |
| CN112561944A (en) * | 2020-11-27 | 2021-03-26 | 中央财经大学 | Lane line extraction method based on vehicle-mounted laser point cloud |
| CN112560747A (en) * | 2020-12-23 | 2021-03-26 | 苏州工业园区测绘地理信息有限公司 | Vehicle-mounted point cloud data-based lane boundary interactive extraction method |
| CN112325809A (en) * | 2021-01-06 | 2021-02-05 | 广东技术师范大学 | Method for detecting flatness of flange |
| CN113837124A (en) * | 2021-09-28 | 2021-12-24 | 中国有色金属长沙勘察设计研究院有限公司 | Automatic extraction method of geotextile routing inspection route of mud discharging warehouse |
| CN113837124B (en) * | 2021-09-28 | 2023-12-05 | 中国有色金属长沙勘察设计研究院有限公司 | Automatic extraction method for geotechnical cloth inspection route of sludge discharging warehouse |
| CN116153057A (en) * | 2022-09-12 | 2023-05-23 | 东北林业大学 | Method for estimating lane width based on laser radar point cloud |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108898672A (en) | A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line | |
| CN108845569A (en) | Generate semi-automatic cloud method of the horizontal bend lane of three-dimensional high-definition mileage chart | |
| CN108062517B (en) | Vehicle-mounted laser point cloud-based method for automatically extracting boundary line of unstructured road | |
| WO2018068653A1 (en) | Point cloud data processing method and apparatus, and storage medium | |
| CN112801022A (en) | Method for rapidly detecting and updating road boundary of unmanned mine card operation area | |
| CN111985322A (en) | Road environment element sensing method based on laser radar | |
| CN111079611B (en) | Automatic extraction method for road surface and marking line thereof | |
| Ibrahim et al. | Curb-based street floor extraction from mobile terrestrial LiDAR point cloud | |
| CN106780524A (en) | A kind of three-dimensional point cloud road boundary extraction method | |
| CN112200171B (en) | Road point cloud extraction method based on scanning lines | |
| CN108961758B (en) | Road junction widening lane detection method based on gradient lifting decision tree | |
| CN105844629A (en) | Automatic segmentation method for point cloud of facade of large scene city building | |
| CN112184736B (en) | Multi-plane extraction method based on European clustering | |
| CN110798805B (en) | Data processing method and device based on GPS track and storage medium | |
| CN109584294A (en) | A kind of road surface data reduction method and apparatus based on laser point cloud | |
| CN107808524B (en) | Road intersection vehicle detection method based on unmanned aerial vehicle | |
| CN112561944A (en) | Lane line extraction method based on vehicle-mounted laser point cloud | |
| CN114488073A (en) | Method for processing point cloud data acquired by laser radar | |
| CN109870458B (en) | Pavement crack detection and classification method based on three-dimensional laser sensor and bounding box | |
| CN115435798A (en) | Unmanned vehicle high-precision map road network generation system and method | |
| CN114170149A (en) | Road geometric information extraction method based on laser point cloud | |
| Yadav et al. | An automatic hybrid method for ground filtering in mobile laser scanning data of various types of roadway environments | |
| Qin et al. | A voxel-based filtering algorithm for mobile LiDAR data | |
| CN113920483A (en) | Method and device for classifying objects in road point cloud, electronic equipment and storage medium | |
| Ma et al. | Automated extraction of driving lines from mobile laser scanning point clouds |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181127 |