CN109583365A - Method for detecting lane lines is fitted based on imaging model constraint non-uniform B-spline curve - Google Patents
Method for detecting lane lines is fitted based on imaging model constraint non-uniform B-spline curve Download PDFInfo
- Publication number
- CN109583365A CN109583365A CN201811427546.XA CN201811427546A CN109583365A CN 109583365 A CN109583365 A CN 109583365A CN 201811427546 A CN201811427546 A CN 201811427546A CN 109583365 A CN109583365 A CN 109583365A
- Authority
- CN
- China
- Prior art keywords
- line
- image
- lane line
- edge
- coordinate system
- 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
- 238000003384 imaging method Methods 0.000 title claims abstract description 54
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000001514 detection method Methods 0.000 claims abstract description 42
- 238000001914 filtration Methods 0.000 claims abstract description 16
- 238000003708 edge detection Methods 0.000 claims abstract description 11
- 230000003287 optical effect Effects 0.000 claims abstract description 10
- 230000006870 function Effects 0.000 claims description 25
- 230000008569 process Effects 0.000 claims description 15
- 230000001629 suppression Effects 0.000 claims description 9
- 230000000007 visual effect Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000009499 grossing Methods 0.000 claims description 5
- 208000001491 myopia Diseases 0.000 claims description 5
- 238000006467 substitution reaction Methods 0.000 claims description 5
- 239000006002 Pepper Substances 0.000 claims description 4
- 230000002708 enhancing effect Effects 0.000 claims description 4
- 230000009977 dual effect Effects 0.000 claims description 3
- 238000009434 installation Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 239000013589 supplement Substances 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 2
- 230000006872 improvement Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000013178 mathematical model Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 241000270295 Serpentes Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000004379 myopia Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
- G06T2207/20192—Edge enhancement; Edge preservation
-
- 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)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
It is fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve, median filtering and histogram equalization processing are carried out to image first, obtain enhanced lane line image；Secondly edge detection is carried out to image using Canny operator, obtains lane line edge image；Then Hough transform straight-line detection is carried out to edge image, reduces background interference edge while improving edge continuity；Be again based on " camera optical axis is parallel with road plane " and " left and right lane line is parallel " it is assumed that on the basis of camera geometry imaging model, derive the control point appraising model under lane line-camera imaging model constraint；Finally lane line edge pixel location information is combined to solve non-uniform B-spline curve model parameter, realizes lane line fitting.The present invention can effectively improve raising control spot placement accuracy and lane detection accuracy, improve the lane detection algorithm based on curve matching to the robustness of background interference.
Description
Technical field
The invention belongs to traffic video detection fields, and in particular to one kind is bent based on imaging model constraint uniform B-Spline
Line is fitted method for detecting lane lines.
Background technique
Current unmanned technology is the research hotspot of intelligent transportation field, lot of domestic and foreign scientific research institutions, enterprise it is big
Amount science research input advances the fast development of unmanned technology significantly.Lane line carries out lane holding, lane change etc. as vehicle
The necessary information of driving behavior is an important environmental data in the perception of unmanned vehicle environment, therefore lane detection side
The performance superiority and inferiority of method has and can not neglect to the safety of unmanned vehicle context aware systems performance or even entire unmanned vehicle control loop
Depending on influence.
The main purpose of lane detection is that the location information of lane line is extracted from video image.Currently used lane
Line detecting method is broadly divided into based on region, based on feature and based on model three classes, and wherein with the lane line based on model
Detection method is the most universal.Such methods be typically based on structured road lane line trend can with specific mathematical model come
This thought is approached, uses straight line, parabola, hyperbolic for the lane line of the different trends such as linear type, parabolic type, snake type
The mathematical models such as line, spline curve are fitted, to substantially reduce testing cost while guaranteeing lane detection accuracy.
Wherein spline curve is expressed by piecewise polynomial, can Accurate Curve-fitting arbitrary shape curve, therefore obtained in lane detection
It is widely applied.The determination at control point known to analysis correlative study is the key that B-spline curves fitting lane line, however vehicle hides
The interference such as gear, tree shade, building effects, road surface breakage increases many difficulty for the extraction of model cootrol point, to influence vehicle
The accuracy of diatom fitting even results in fitting failure.Therefore, Control point extraction how is effectively improved to background interference (vehicle
Block, tree shade, building effects, road surface other identifier, road surface breakage etc.) robustness, improve the same of control spot placement accuracy
When take into account algorithm time cost, become improve the method for detecting lane lines efficiency based on spline curve model critical issue.
Summary of the invention
It is an object of the invention to overcome traditional method for detecting lane lines based on spline curve model vulnerable to background interference,
Lead to the problem of controlling inaccurate point location or failure, provides based on imaging model constraint non-uniform B-spline curve fitting lane line
Detection method, this method are taken into account algorithm time cost while capable of effectively improving control spot placement accuracy, are preferably examined
It surveys as a result, effectively improving lane detection efficiency.
To achieve the goals above, the present invention takes following technical solution to be achieved:
Method for detecting lane lines is fitted based on imaging model constraint non-uniform B-spline curve, comprising the following steps:
Step 1: image preprocessing；
Original lane line image I is obtained from the lane line standard picture library in Ka Neijimeilong image data base, to original
Beginning lane line image I carries out median filtering and removes salt-pepper noise, then carries out the brightness and comparison of histogram equalization enhancing image
Degree keeps edge feature prominent, obtains enhanced lane line image I1；
Step 2: edge detection；
To enhanced lane line image I1Edge detection is carried out using Canny operator, obtains original lane line edge graph
As I2；
Step 3: Hough straight-line detection；
To original lane line edge image I obtained in step 22Hough straight-line detection is carried out, retains and is examined comprising straight line
The edge for surveying result, removes remaining Clutter edge, obtains edge image I3；
Step 4: imaging model constraint condition is derived；
The imaging model constraint condition being derived by are as follows:
Length is the length along path in the corresponding image coordinate system of lane line line segment of Δ Y in u column in world coordinate system
Spend Δ v are as follows:
In world coordinate system width be in the corresponding image coordinate system of left and right lane line spacing of Δ X in v row between
Away from width Delta u are as follows:
Step 5: non-uniform B-spline curve Control point extraction；
Setting scan line, every scanning are constrained by imaging model in the picture to correspond to the length of Δ Y in world coordinate system
The intersection point at line and left and right lane edge is a pair of control point；
Step 6: lane line fitting；
Non-uniform B-spline curve control point information is obtained by step 5, recycles NUBS interpolation method to left and right lane line
It is fitted, completes the detection to lane line.
A further improvement of the present invention lies in that in step 1, when carrying out median filtering, using median filtering function f^ (x,
Y) are as follows:
Wherein f^ (x, y) is median filtering output, SxyExpression center is at (x, y), the rectangle subgraph window having a size of M × N
The set of coordinates of mouth, f (a, b) are the grey scale pixel value that coordinate is (a, b).
A further improvement of the present invention lies in that in step 1, when carrying out histogram equalization, histogram equalization function sk
Are as follows:
Wherein skFor histogram equalization output, rkRepresent discrete gray levels, 0≤rk≤ 255, k=0,1,2 ..., n-1,
niTo occur gray scale r in imageiPixel number, n is the sum of all pixels in image,It is exactly the frequency in probability theory.
A further improvement of the present invention lies in that specific step is as follows for step 2:
(1) Gaussian filter smoothed image I is used1；
Gaussian smoothing function G (x, y) are as follows:
With G (x, y) and enhanced lane line image I1Convolution is carried out, smoothed image f is obtained1；
f1(x, y)=I1(x,y)*G(x,y) (4)
(2) amplitude and the direction that gradient is calculated with the finite difference of single order local derviation, obtain gradient image f2；
First differential convolution mask
(3) non-maxima suppression is carried out to gradient magnitude, obtains non-maxima suppression image f3；
In gradient image f2Every bit on, by the center pixel S of 8 neighborhoods compared with along the two of gradient line pixel；
If the gradient value of S is big unlike two adjacent pixel gradient values along gradient line, S=0 is enabled；
(4) edge is detected and connected with dual threashold value-based algorithm；
To non-maxima suppression image f3Two threshold value T are set1And T2, T1=0.4T2, gradient value is less than T1Pixel
Gray value is set as 0, obtains image f4；Then gradient value is less than T2The gray value of pixel be set as 0, obtain image f5；With image
f5Based on, with image f4For supplement, links the edge of image, obtain original lane line edge image I2。
A further improvement of the present invention lies in that specific step is as follows for step 3:
(1) Hough straight-line detection；
For any point A (x in rectangular coordinate system0,y0), the straight line of passing point A meets
Y=kx+l (5)
Wherein k is slope, and l is intercept, then crosses point A (x in X-Y plane0,y0) straight line cluster use formula (5) indicate, for
It can not then be indicated perpendicular to the straight slope infinity of X-axis；Therefore rectangular coordinate system is transformed into polar coordinate system；
The equation that straight line is indicated in polar coordinate system is ρ=xcos θ+ysin θ (6)
Wherein ρ is normal distance of the origin to straight line, and θ is the positive angle of normal and X-axis；Then a bit in image space
A sine curve in corresponding polar coordinate system ρ-θ；By detecting the intersection point in the space ρ-θ come straight in detection image space
Line；By ρ, θ discretization calculates the value of corresponding parameter ρ, so in the corresponding each value of parameter θ according to formula (6) respectively
Add 1 in corresponding parameter summing elements afterwards；The value for finally counting each summing elements then thinks greater than preset threshold value H
This group of parameter is the parameter of the straight line in image space, to mark straight line in the picture；
(2) Clutter edge is removed；
Each of the straight line marked to step (1) edge pixel, whole edge of the search comprising the pixel are simultaneously protected
It stays, edge of the straight line without common pixel point with label is rejected, to obtain edge image I3。
A further improvement of the present invention lies in that the detailed process of step 4 are as follows:
Assuming that camera optical axis is parallel with vehicle driving road plane and left and right lane line is parallel；
The known world coordinate system (X, Y, Z) and image coordinate system (U, V), camera maximum horizontal visual angle are α, maximum perpendicular view
Angle is β, and coordinate of the camera installation site in world coordinate system is C (d, 0, h), and wherein h is camera mounting height, i.e. camera exists
Value on world coordinate system Z axis, d are that camera installs horizontal-shift, i.e. value of the camera in world coordinate system in X-axis；Camera light
Axis is parallel with vehicle driving road plane, is γ with lane line angle；According to camera geometry imaging model, world coordinate system (X,
Y, Z) on road surface in certain point P (x, y, 0) and image coordinate system (U, V) relative coordinate point Q (u, v) mapping model are as follows:
H in formulaI, WIThe level, vertical resolution of image respectively after camera imaging；
According to camera imaging principle, lane line line segment length is with lane line line in world coordinate system in image after imaging
Section and the increase of camera distance and shorten, similarly, the identical left and right lane line spacing on road surface in world coordinate system, in myopia
The lane line spacing being imaged in is wider, and the lane line spacing being imaged in far visual field is relatively narrow；In conjunction with geometry phase
Machine imaging model derives that length is during u is arranged in the corresponding image coordinate system of lane line line segment of Δ Y in world coordinate system
Line segment length Δ v are as follows:
In world coordinate system width be in the corresponding image coordinate system of left and right lane line spacing of Δ X in v row between
Away from width Delta u are as follows:
A further improvement of the present invention lies in that the detailed process of step 5 are as follows:
Since lane line edge image bottom, in viRow setting horizontal scanning line Line i, m≤i≤n, in Line i
Dominating pair of vertices (Li, R are obtained with the point of intersection of left and right lane linei), wherein LiCoordinate be (ui,vi), RiCoordinate is (ui',vi)；
V is defined according to imaging model constraint conditioniCalculation formula are as follows:
It is derived according to formula (8), (10), (11):
Wherein v1, Δ v1For preset value；Formula (14) substitution formula (13) is successively found out into viValue；Thus i-th scanning is found out
Control point ordinate determined by line Line i is equal to vi；Marginal point is searched for the left and right sides respectively from scan line midpoint, is obtained
First pair of scan line and the intersection point of left and right lane line be control point, so that it is determined that the coordinate (u of a pair of control pointi,vi) and
(u′i,vi)。
A further improvement of the present invention lies in that the detailed process of step 5 are as follows:
Assuming that left and right lane line is parallel, control point L determined by i-th scan line Line i is solvedi、RiAbscissa ui
With u 'i；Δ u is derived according to imaging model constraint condition and formula (14)i+1With Δ uiRelational expression are as follows:
u′i=ui+Δui (16)
Wherein camera optical axis is calculated as follows with lane line angle γ:
The case where causing control point to be lost lane line edge missing, in adjacent control points to L1、L2Known situation
Under, the abscissa u at the control point is calculated according to formula (15)-(17)2；The wrong control points as caused by false edge are determined
It the case where position, verifies all adjacent control points and whether formula (15) is met to spacing width ratio, to detect error control point seat
It marks and it is relocated according to formula (15)-(17).
A further improvement of the present invention lies in that the detailed process of step 6 are as follows:
Assuming that B-spline curves S is by n+1 control point set { P0,P1,...PnConstitute, then each point on curve S meets:
Wherein Bi,mIt (o) is basic B-spline function, 2≤m≤n+1, tmin≤u≤tmax, tjFor node, j=0 ..., i+
M, as each node tjBetween it is equidistant when, which is referred to as Uniform B-Spline Curve, is otherwise non-uniform B-spline curve；Root
According to NUBS interpolation method, if known m is to control point, m >=3, then lane line is fitted using m-1 rank multinomial function；If can
It determines 4 pairs of control points, then carries out NUBS interpolation using three rank multinomial functions to fit lane line；If being only determined 3 pairs
Control point then uses second order polynomial Function Fitting lane line.
Compared with prior art, the invention has the benefit that the present invention carries out median filtering and straight to image first
Square figure equalization processing obtains enhanced lane line image；Secondly edge detection is carried out to image using Canny operator, obtained
To lane line edge image；Then Hough transform straight-line detection is carried out to edge image, is reduced while improving edge continuity
Background interference edge；Finally lane line edge pixel location information is combined to solve non-uniform B-spline curve model parameter, realizes vehicle
Diatom fitting.The present invention can effectively improve raising control spot placement accuracy and lane detection accuracy, improve quasi- based on curve
Robustness of the lane detection algorithm of conjunction to background interference.
Further, based on " camera optical axis is parallel with road plane " and " left and right lane line is parallel " it is assumed that in phase
On the basis of machine geometry imaging model, the control point appraising model under lane line-camera imaging model constraint is derived；It can reduce
Occlusion, tree shade, building effects, road surface breakage and various non-lane line pavement strips are done caused by determining to control point
It disturbs, improves the robustness of non-uniform B-spline curve Control point extraction method, improve lane detection accuracy.
Detailed description of the invention
Fig. 1 is position view of the camera in world coordinate system；
Fig. 2 is the control point determination process schematic diagram in the continuous situation of lane line；
Fig. 3 is lane detection result one；
Fig. 4 is the control point determination process schematic diagram in the discontinuous situation of lane line；
Fig. 5 is lane detection result two；
Fig. 6 is that the non-uniform B-spline curve based on imaging model constraint is fitted lane detection algorithm flow chart.
Specific embodiment
The present invention is described in detail with reference to the accompanying drawing.
It is provided by the invention based on imaging model constraint non-uniform B-spline curve be fitted method for detecting lane lines, including with
Lower step:
Step 1: image preprocessing；
Original lane line image I is obtained from the lane line standard picture library in Ka Neijimeilong image data base, to original
Beginning lane line image I carries out median filtering and removes salt-pepper noise, then carries out the brightness and comparison of histogram equalization enhancing image
Degree keeps edge feature prominent, obtains enhanced lane line image I1。
Median filtering function f^ (x, y)
Wherein f^ (x, y) is median filtering output, SxyExpression center is at (x, y), the rectangle subgraph window having a size of M × N
The set of coordinates of mouth, f (a, b) are the grey scale pixel value that coordinate is (a, b).
Histogram equalization function sk
Wherein skFor histogram equalization output, rkRepresent discrete gray levels (0≤rk≤ 255, k=0,1,2 ..., n-1),
niTo there is r in imageiThe pixel number of this gray scale, n are the sum of all pixels in image,It is exactly the frequency in probability theory.
Step 2: edge detection；
To enhanced lane line image I1Edge detection is carried out using Canny operator, obtains original lane line edge graph
As I2。
Specific step is as follows:
(1) Gaussian filter smoothed image I is used1；
Gaussian smoothing function G (x, y)
With G (x, y) and image I1Convolution is carried out, smoothed image f is obtained1
f1(x, y)=I1(x,y)*G(x,y) (4)
(2) amplitude and the direction that gradient is calculated with the finite difference of single order local derviation, obtain gradient image f2；
First differential convolution mask
(3) non-maxima suppression is carried out to gradient magnitude, obtains non-maxima suppression image f3；
In gradient image f2Every bit on, by the center pixel S of 8 neighborhoods compared with along the two of gradient line pixel.
If the gradient value of S is big unlike two adjacent pixel gradient values along gradient line, S=0 is enabled.
(4) edge is detected and connected with dual threashold value-based algorithm.
To non-maxima suppression image f3Two threshold value T are set1And T2, T1=0.4T2.Gradient value is less than T1Pixel
Gray value is set as 0, obtains image f4.Then gradient value is less than T2The gray value of pixel be set as 0, obtain image f5.With image
f5Based on, with image f4Link the edge of image to supplement, obtains edge image I2。
Step 3: Hough straight-line detection；
To lane line edge image I obtained in step 22Hough straight-line detection is carried out, only retaining those includes straight line
The edge of testing result removes remaining Clutter edge, obtains edge image I3.Specific step is as follows:
(1) Hough straight-line detection；
For any point A (x in rectangular coordinate system0,y0), the straight line of passing point A meets
Y=kx+l (5)
Wherein k is slope, and l is intercept.So point A (x is crossed in X-Y plane0,y0) straight line cluster can use formula (5) table
Show, but the straight slope infinity perpendicular to X-axis can not then be indicated.Therefore rectangular coordinate system is transformed into polar coordinates
System just can solve the special circumstances.
The equation that straight line is indicated in polar coordinate system is ρ=xcos θ+ysin θ (6)
Wherein ρ is normal distance of the origin to straight line, and θ is the positive angle of normal and X-axis.Then a bit in image space
A sine curve in corresponding polar coordinate system ρ-θ.By detecting the intersection point in the space ρ-θ come straight in detection image space
Line.By ρ, θ discretization calculates the value of corresponding parameter ρ, so in the corresponding each value of parameter θ according to formula (6) respectively
Add 1 in corresponding parameter summing elements afterwards.The value for finally counting each summing elements is considered as greater than preset threshold value H
This group of parameter is the parameter of the straight line in image space, to mark straight line in the picture；
(2) Clutter edge is removed；
Each of the straight line marked to step (1) edge pixel, whole edge of the search comprising the pixel are simultaneously protected
It stays, rejects those with the straight line of label without the edge of common pixel point, to obtain edge image I3。
Step 4: imaging model constraint condition derives；
Based on camera geometry imaging model, in " camera optical axis is parallel with vehicle driving road plane " and " left and right lane
Line is parallel " hypothesis on the basis of, derive imaging model constraint condition.
The known world coordinate system (X, Y, Z) and image coordinate system (U, V), camera maximum horizontal visual angle are α, maximum perpendicular view
Angle is β, and coordinate of the camera installation site in world coordinate system is C (d, 0, h), and wherein h is camera mounting height, i.e. camera exists
Value on world coordinate system Z axis, d are that camera installs horizontal-shift, i.e. value of the camera in world coordinate system in X-axis.Camera light
Axis is parallel with vehicle driving road plane, is γ with lane line angle.According to camera geometry imaging model, it is known that world coordinate system
In (X, Y, Z) on road surface in certain point P (x, y, 0) and image coordinate system (U, V) relative coordinate point Q (u, v) mapping model are as follows:
H in formulaI, WIThe level, vertical resolution of image respectively after camera imaging.
According to camera imaging principle it is found that lane line line segment length is with lane in world coordinate system in image after imaging
The increase of line line segment and camera distance and shorten, i.e., in world coordinate system on road surface equal length lane line line segment,
The lane line line segment being imaged in near-sighted field is longer, and the lane line line segment being imaged in far visual field is shorter.Similarly, exist
Identical left and right lane line spacing on road surface in world coordinate system, the lane line spacing being imaged in near-sighted field is wider, and
The lane line spacing being imaged in far visual field is relatively narrow.It, can be in conjunction with aforementioned camera imaging model according to this imaging fact
Derive that length is the line segment length in the corresponding image coordinate system of lane line line segment of Δ Y in u column in world coordinate system
Δ v are as follows:
And width is in v row in the corresponding image coordinate system of left and right lane line spacing of Δ X in world coordinate system
Spacing width Δ u are as follows:
Step 5: non-uniform B-spline curve Control point extraction；
It in order to solve NUBS curve model parameter, needs to determine suitable control point, method first are as follows: from lane line edge
Image base starts, with " length of Δ Y in corresponding world coordinate system " in the picture by imaging model constraint setting scan line, often
The intersection point at scan line and left and right lane edge is a pair of control point.The mathematical description of the process is as follows: from lane line edge
Image base starts, in viRow setting horizontal scanning line Line i (m≤i≤n), in the point of intersection of Line i and left and right lane line
Obtain dominating pair of vertices (Li, Ri), wherein LiCoordinate be (ui,vi), RiCoordinate is (ui',vi).Since lane line is whole in the picture
Body extends trend in longitudinal, and the ordinate of each dominating pair of vertices is sequentially reduced, the trend not camera subject optical axis and lane line angle γ
It influences, therefore process ignores the value of γ to simplify the calculation, according to imaging model constraint definition viCalculation formula are as follows:
It is derived according to formula (8), (10), (11):
Wherein v1, Δ v1For preset value.Formula (14) substitution formula (13) can successively be found out into viValue.It is possible thereby to find out
Control point ordinate determined by i-th scan line Line i is equal to vi.Side is searched for the left and right sides respectively from scan line midpoint
The intersection point of edge point, first pair of obtained scan line and left and right lane line is control point, and then it can be seen that the dominating pair of vertices cross
Coordinate, so that it is determined that its coordinate (ui,vi) and (u 'i,vi)。
Above-mentioned control point determination process is at the control point that assuming that left and right lane line is unobstructed, i.e. every scan line determines
It is exactly the intersection point of scan line Yu left and right lane line.However, actually detected obtained lane line often exists due to occlusion, tree
Clutter edge caused by shade, building effects, road surface breakage or discontinuous edge.In addition, empty lane line edge is also discontinuous
's.The intersection point that will result in scan line and left-lane line perhaps right-lane line in this way is not actual correct intersection point or does not deposit
In intersection point.
In view of the above-mentioned problems, the present invention is on the basis of aforementioned control point determines method, in conjunction with " left and right lane line is parallel "
It is assumed that solve i-th scan line Line i determined by control point Li、RiAbscissa uiWith u 'i.About according to imaging model
Beam and formula (12) derive Δ ui+1With Δ uiRelational expression are as follows:
u′i=ui+Δui (16)
Wherein camera optical axis is calculated as follows with lane line angle γ:
It the case where causing control point to be lost lane line edge missing, can be in adjacent control points to L1、L2Known feelings
Under condition, with the abscissa u for calculating the control point according to formula (15)-(17)2；For the mistake of the control point as caused by false edge
The case where accidentally positioning, can verify all adjacent control points and whether meet formula (15) to spacing width ratio, to detect mistake
Control point coordinates simultaneously relocate it according to formula (15)-(17).
Step 6: lane line fitting.
The control point as described in step 5 determines that method has obtained lane line Edge position control point information, so that it may be inserted using NUBS
Value method is fitted left and right lane line.B-spline curves mathematical model are as follows:
Assuming that B-spline curves S is by n+1 control point set { P0,P1,...PnConstitute, then each point on curve S meets:
Wherein Bi,mIt (o) is basic B-spline function, 2≤m≤n+1, tmin≤u≤tmax, tj(j=0 ..., i+m) it is section
Point, as each node tjBetween it is equidistant when, which is referred to as Uniform B-Spline Curve, is otherwise non-uniform B-spline curve.Root
According to known to NUBS interpolation method: if known m (m >=3), to control point, lane line can be fitted with m-1 rank multinomial function.If
It can determine 4 pairs of control points, then three rank multinomial functions can be used and carry out NUBS interpolation to fit lane line；If only determining
3 pairs of control points, then can be used second order polynomial Function Fitting lane line.
Control point coordinates determined by step 5 are substituted into formula (18), solve spline curve S (u) and in original car diatom
It is shown in image I, completes the detection to lane line.
It is illustrated below by a specific embodiment.
Referring to Fig. 6, method of the invention is W to a width sizeI×HIThe lane line image of (240 × 256) carries out edge inspection
It surveys, imaged model constraint determines control point in conjunction with lane line edge position information, to solve non-uniform B-spline curve ginseng
Number realizes lane detection.
Specifically realized using following steps:
Step 1: original lane line image is obtained from the lane line standard picture library in Ka Neijimeilong image data base
I carries out median filtering to image I and removes salt-pepper noise, then carries out the brightness and contrast of histogram equalization enhancing image, makes
Edge feature is prominent, obtains enhanced lane line image I1.3 × 3 rectangle subgraph window is taken in the present invention when median filtering
Mouthful, gray scale discrete parameter r when histogram equalizationkRange is 0≤rk≤255。
Step 2: to enhanced original lane line image I1Edge detection is carried out using Canny operator, obtains initial vehicle
Diatom edge image I2.Smoothing parameter σ=1 of the Gaussian smoothing function used when carrying out edge detection in the present invention, single order are micro-
Divide convolution maskDual threshold T1And T2For default value, meet T1=0.4T2。
Step 3: to original lane line edge image I obtained in step 22Hough straight line is carried out using Hough algorithm
Detection, to each of the straight line marked edge pixel, whole edge of the search comprising the pixel simultaneously retains, and rejects those
Straight line with label is without the edge of common pixel point, to obtain edge image I3.The parameter of Hough algorithm is in the present invention
Default value；
Step 4: from lane line edge image I3Bottom starts, vi row be arranged horizontal scanning line Line i (2≤i≤
6) dominating pair of vertices (Li, R, are obtained in the point of intersection of Line i and left and right lane linei), wherein LiCoordinate be (ui,vi), RiCoordinate
For (ui',vi).V in the present invention1=0, Δ v1=20 be preset value.Formula (14) substitution formula (13) can successively be found out into viValue.
It is possible thereby to find out control point ordinate determined by i-th scan line Line i equal to vi.To the left from scan line midpoint difference
Marginal point is searched on right both sides, and first pair of obtained scan line and the intersection point of left and right lane line are control point, and then it can be seen that should
The abscissa of dominating pair of vertices, so that it is determined that its coordinate (ui,vi) and (u 'i,vi)。
Actually detected obtained lane line is made due to occlusion, tree shade, building effects, road surface breakage or empty lane line
At Clutter edge or discontinuous edge.It the case where causing control point to be lost lane line edge missing, can be in adjacent control
System point is to L1、L2In known situation, with the abscissa u for calculating the control point according to formula (15)-(17)2；For by falseness
The case where wrong control points caused by edge position, can verify all adjacent control points and whether meet formula to spacing width ratio
(15), to detect error control point coordinate and be relocated according to formula (15)-(17) to it.
Step 5: the control point as described in step 4 determines that method has obtained lane line Edge position control point information, so that it may benefit
Left and right lane line is fitted with NUBS interpolation method.If can determine 4 pairs of control points, three rank multinomial functions can be used
NUBS interpolation is carried out to fit lane line；If 3 pairs of control points have only been determined, second order polynomial Function Fitting vehicle can be used
Diatom.Control point coordinates determined by step 4 are substituted into formula (18), solve spline curve S (u) and in original lane line chart
As showing in I, the detection to lane line is completed.
The lane line image for being 240 × 256 for a width size carries out median filtering and histogram equalization operation, uses
The edge feature of Canny edge detection operator extraction lane line.Then Hough straight-line detection is carried out to lane line edge image:
In conjunction with the characteristics of " the lane line edge of near-sighted field is of a straight line type ", Hough straight-line detection is carried out to lane line edge image, is only protected
Staying those includes the edge of straight-line detection result, further eliminates those accordingly by Background Buildings, tree shade, barrier, road surface hole
The Clutter edge of the formation such as hole, crack.
Fig. 1 is that position view of the camera in world coordinate system according to camera imaging principle derives imaging model such as
Formula (11)-(12).
Fig. 2 is the control point determination process schematic diagram in the case of lane line continuous edge.Default v1=0, Δ v1=20, it will
Formula (14), which substitutes into formula (13), can successively find out viValue: v2=167, v3=117, v4=78, v5=46, v6=21.From lane
Line edge image bottom starts, in viRow setting horizontal scanning line Line i (2≤i≤6), in Line i and left and right lane line
Point of intersection obtains dominating pair of vertices (Li, Ri), wherein LiCoordinate be (ui,vi), RiCoordinate is (u 'i,vi).It is possible thereby to find out
Control point ordinate determined by i scan line Line i is equal to vi.Edge is searched for the left and right sides respectively from scan line midpoint
The intersection point of point, first pair of obtained scan line and left and right lane line is control point, and then it can be seen that the dominating pair of vertices horizontal seat
Mark, so that it is determined that its coordinate (ui,vi) and (ui',vi).5 groups of dominating pair of vertices coordinates { (92,21) (117,21) } are obtained accordingly,
{ (80,46) (145,46) }, { (67,78) (175,78) }, { (51,117) (212,117) }, { (33,167) (254,167) }.
Then according to NUBS interpolation algorithm, NUBS interpolation is carried out using fourth order polynomial function, obtained matched curve is shown in lane
In line image as shown in Figure 3.
Fig. 4 is the control point determination process schematic diagram in the discontinuous situation in lane line edge.Equally default v1=0, Δ v1=
20, formula (14) substitution formula (13) can successively be found out into viValue: v2=167, v3=117, v4=78, v5=46, v6=21.From
Lane line edge image bottom starts, in viHorizontal scanning line Line i (2≤i≤6) are arranged in row, wherein Line 5 and Line 6
Lacked with the intersection point of left-lane line, only primarily determine out 3 left-lane line control points be respectively (15,167), (39,117),
(59,78)}；5 right-lane line control points be respectively (253,167), (213,117), (181,78), (155,46), (134,
21)}.Δ u is estimated according to formula (10)-(12)5=80, u5=75；Δu6=46, u6=88.So that it is determined that left-lane line lacks out
Two control point coordinates lost are respectively (75,46), (88,21).Then according to NUBS interpolation algorithm, using fourth order polynomial letter
Number carries out NUBS interpolation, and obtained matched curve is shown in the line image of lane as shown in Figure 5.
From Fig. 3 and 5 as can be seen that carrying out lane detection according to the method described above, preferable testing result is realized.This reality
It applies example and shows that the solution of the present invention can effectively improve the control point positional accuracy and success rate of non-uniform B-spline curve model,
Calculation amount is combined, there is preferable Detection accuracy and real-time.
Specific embodiments of the present invention given above, it should be noted that the invention is not limited to implement in detail below
Example, all same transformation done on the basis of application scheme each fall within protection scope of the present invention.
Claims (9)
1. being fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve, which is characterized in that including following step
It is rapid:
Step 1: image preprocessing；
Original lane line image I is obtained from the lane line standard picture library in Ka Neijimeilong image data base, to original car
Road line image I carries out median filtering and removes salt-pepper noise, then carries out the brightness and contrast of histogram equalization enhancing image, makes
Edge feature is prominent, obtains enhanced lane line image I1；
Step 2: edge detection；
To enhanced lane line image I1Edge detection is carried out using Canny operator, obtains original lane line edge image I2；
Step 3: Hough straight-line detection；
To original lane line edge image I obtained in step 22Hough straight-line detection is carried out, retaining includes straight-line detection result
Edge, remove remaining Clutter edge, obtain edge image I3；
Step 4: imaging model constraint condition is derived；
The imaging model constraint condition being derived by are as follows:
Length is the line segment length Δ v in the corresponding image coordinate system of lane line line segment of Δ Y in u column in world coordinate system
Are as follows:
Width is that the spacing in the corresponding image coordinate system of left and right lane line spacing of Δ X in v row is wide in world coordinate system
Spend Δ u are as follows:
Step 5: non-uniform B-spline curve Control point extraction；
Constrain setting scan line by imaging model in the picture to correspond to the length of Δ Y in world coordinate system, every scan line with
The intersection point at left and right lane edge is a pair of control point；
Step 6: lane line fitting；
Non-uniform B-spline curve control point information is obtained by step 5, NUBS interpolation method is recycled to carry out left and right lane line
The detection to lane line is completed in fitting.
2. according to claim 1 be fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve,
It is characterized in that, in step 1, when carrying out median filtering, using median filtering function f^ (x, y) are as follows:
Wherein f^ (x, y) is median filtering output, SxyExpression center at (x, y), the rectangle subgraph window having a size of M × N
Set of coordinates, f (a, b) are the grey scale pixel value that coordinate is (a, b).
3. according to claim 1 be fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve,
It is characterized in that, in step 1, when carrying out histogram equalization, histogram equalization function skAre as follows:
Wherein skFor histogram equalization output, rkRepresent discrete gray levels, 0≤rk≤ 255, k=0,1,2 ..., n-1, niFor figure
There is gray scale r as iniPixel number, n is the sum of all pixels in image,It is exactly the frequency in probability theory.
4. according to claim 1 be fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve,
It is characterized in that, specific step is as follows for step 2:
(1) Gaussian filter smoothed image I is used1；
Gaussian smoothing function G (x, y) are as follows:
With G (x, y) and enhanced lane line image I1Convolution is carried out, smoothed image f is obtained1；
f1(x, y)=I1(x,y)*G(x,y) (4)
(2) amplitude and the direction that gradient is calculated with the finite difference of single order local derviation, obtain gradient image f2；
First differential convolution mask
(3) non-maxima suppression is carried out to gradient magnitude, obtains non-maxima suppression image f3；
In gradient image f2Every bit on, by the center pixel S of 8 neighborhoods compared with along the two of gradient line pixel；If S
Gradient value it is big unlike two adjacent pixel gradient values along gradient line, then enable S=0；
(4) edge is detected and connected with dual threashold value-based algorithm；
To non-maxima suppression image f3Two threshold value T are set1And T2, T1=0.4T2, gradient value is less than T1Pixel gray scale
Value is set as 0, obtains image f4；Then gradient value is less than T2The gray value of pixel be set as 0, obtain image f5；With image f5For
Basis, with image f4For supplement, links the edge of image, obtain original lane line edge image I2。
5. according to claim 1 be fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve,
It is characterized in that, specific step is as follows for step 3:
(1) Hough straight-line detection；
For any point A (x in rectangular coordinate system0,y0), the straight line of passing point A meets
Y=kx+l (5)
Wherein k is slope, and l is intercept, then crosses point A (x in X-Y plane0,y0) straight line cluster use formula (5) indicate, for vertical
It can not then be indicated in the straight slope infinity of X-axis；Therefore rectangular coordinate system is transformed into polar coordinate system；
The equation of expression straight line is in polar coordinate system
ρ=xcos θ+ysin θ (6)
Wherein ρ is normal distance of the origin to straight line, and θ is the positive angle of normal and X-axis；It is then a little corresponding in image space
A sine curve in polar coordinate system ρ-θ；By detecting the intersection point in the space ρ-θ come the straight line in detection image space；It will
ρ, θ discretization calculate the value of corresponding parameter ρ, then in phase in the corresponding each value of parameter θ according to formula (6) respectively
Add 1 in the parameter summing elements answered；The value for finally counting each summing elements then thinks that the group is joined greater than preset threshold value H
Number is the parameter of the straight line in image space, to mark straight line in the picture；
(2) Clutter edge is removed；
Each of the straight line marked to step (1) edge pixel, whole edge of the search comprising the pixel simultaneously retain, pick
Edge except the straight line with label without common pixel point, to obtain edge image I3。
6. according to claim 1 be fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve,
It is characterized in that, the detailed process of step 4 are as follows:
Assuming that camera optical axis is parallel with vehicle driving road plane and left and right lane line is parallel；
The known world coordinate system (X, Y, Z) and image coordinate system (U, V), camera maximum horizontal visual angle are α, and maximum perpendicular visual angle is
β, coordinate of the camera installation site in world coordinate system are C (d, 0, h), and wherein h is camera mounting height, i.e., camera is in the world
Value on coordinate system Z axis, d are that camera installs horizontal-shift, i.e. value of the camera in world coordinate system in X-axis；Camera optical axis with
Vehicle driving road plane is parallel, is γ with lane line angle；According to camera geometry imaging model, world coordinate system (X, Y, Z)
The mapping model of certain point P (x, y, 0) and relative coordinate point Q (u, v) in image coordinate system (U, V) on middle road surface are as follows:
H in formulaI, WIThe level, vertical resolution of image respectively after camera imaging；
According to camera imaging principle, after imaging in image lane line line segment length be in world coordinate system lane line line segment with
The increase of camera distance and shorten, similarly, the identical left and right lane line spacing on road surface in world coordinate system, in near-sighted field
The lane line spacing that imaging obtains is wider, and the lane line spacing being imaged in far visual field is relatively narrow；In conjunction with geometry camera at
As model, derive that length is the line in the corresponding image coordinate system of lane line line segment of Δ Y in u column in world coordinate system
Segment length Δ v are as follows:
Width is that the spacing in the corresponding image coordinate system of left and right lane line spacing of Δ X in v row is wide in world coordinate system
Spend Δ u are as follows:
7. according to claim 6 be fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve,
It is characterized in that, the detailed process of step 5 are as follows:
Since lane line edge image bottom, in viRow setting horizontal scanning line Line i, m≤i≤n, in Line i and left and right
The point of intersection of lane line obtains dominating pair of vertices (Li, Ri), wherein LiCoordinate be (ui,vi), RiCoordinate is (u 'i,vi)；According at
As model constraint condition defines viCalculation formula are as follows:
It is derived according to formula (8), (10), (11):
Wherein v1, Δ v1For preset value；Formula (14) substitution formula (13) is successively found out into viValue；Thus i-th scan line is found out
Control point ordinate determined by Line i is equal to vi；Marginal point is searched for the left and right sides respectively from scan line midpoint, is obtained
The intersection point of first pair of scan line and left and right lane line is control point, so that it is determined that the coordinate (u of a pair of control pointi,vi) and
(u′i,vi)。
8. according to claim 7 be fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve,
It is characterized in that, the detailed process of step 5 are as follows:
Assuming that left and right lane line is parallel, control point L determined by i-th scan line Line i is solvedi、RiAbscissa uiWith u 'i；
Δ u is derived according to imaging model constraint condition and formula (14)i+1With Δ uiRelational expression are as follows:
u′i=ui+Δui (16)
Wherein camera optical axis is calculated as follows with lane line angle γ:
The case where causing control point to be lost lane line edge missing, in adjacent control points to L1、L2In known situation, root
The abscissa u at the control point is calculated according to formula (15)-(17)2；For the positioning of the wrong control points as caused by false edge
Situation verifies all adjacent control points and whether meets formula (15) to spacing width ratio, to detect error control point coordinate simultaneously
It is relocated according to formula (15)-(17).
9. according to claim 1 be fitted method for detecting lane lines based on imaging model constraint non-uniform B-spline curve,
It is characterized in that, the detailed process of step 6 are as follows:
Assuming that B-spline curves S is by n+1 control point set { P0,P1,...PnConstitute, then each point on curve S meets:
Wherein Bi,mIt (o) is basic B-spline function, 2≤m≤n+1, tmin≤u≤tmax, tjFor node, j=0 ..., i+m, when each
Node tjBetween it is equidistant when, which is referred to as Uniform B-Spline Curve, is otherwise non-uniform B-spline curve；According to NUBS
Interpolation method, if known m is to control point, m >=3, then lane line is fitted using m-1 rank multinomial function；If can determine 4 pairs
Control point then carries out NUBS interpolation using three rank multinomial functions to fit lane line；If 3 pairs of control points have only been determined,
Then use second order polynomial Function Fitting lane line.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811427546.XA CN109583365B (en) | 2018-11-27 | 2018-11-27 | Method for detecting lane line fitting based on imaging model constrained non-uniform B-spline curve |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811427546.XA CN109583365B (en) | 2018-11-27 | 2018-11-27 | Method for detecting lane line fitting based on imaging model constrained non-uniform B-spline curve |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109583365A true CN109583365A (en) | 2019-04-05 |
| CN109583365B CN109583365B (en) | 2022-07-26 |
Family
ID=65924533
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811427546.XA Active CN109583365B (en) | 2018-11-27 | 2018-11-27 | Method for detecting lane line fitting based on imaging model constrained non-uniform B-spline curve |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109583365B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110135252A (en) * | 2019-04-11 | 2019-08-16 | 长安大学 | A kind of adaptive accurate lane detection and deviation method for early warning for unmanned vehicle |
| CN112654998A (en) * | 2020-10-22 | 2021-04-13 | 华为技术有限公司 | Lane line detection method and device |
| CN112818873A (en) * | 2021-02-04 | 2021-05-18 | 苏州魔视智能科技有限公司 | Lane line detection method and system and electronic equipment |
| CN113450380A (en) * | 2021-07-17 | 2021-09-28 | 普达迪泰（天津）智能装备科技有限公司 | Track calibration method based on airport runway scribed lines |
| WO2022082574A1 (en) * | 2020-10-22 | 2022-04-28 | 华为技术有限公司 | Lane line detection method and apparatus |
| CN114912159A (en) * | 2022-07-18 | 2022-08-16 | 中国铁路设计集团有限公司 | Method for fitting geometric line shape of rail transit line plane |
| CN115123218A (en) * | 2022-09-02 | 2022-09-30 | 小米汽车科技有限公司 | Vehicle detection method and device and electronic equipment thereof |
| CN115311314A (en) * | 2022-10-13 | 2022-11-08 | 深圳市华汉伟业科技有限公司 | Resampling method, system and storage medium for line laser contour data |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102103747A (en) * | 2009-12-16 | 2011-06-22 | 中国科学院电子学研究所 | Method for calibrating external parameters of monitoring camera by adopting reference height |
| US20120078867A1 (en) * | 2010-09-29 | 2012-03-29 | Navteq North America, Llc | Multi-Dimensional Road Representation |
| EP2602744A1 (en) * | 2011-12-08 | 2013-06-12 | Delphi Technologies, Inc. | Method for detecting and tracking lane markings |
| CN103177246A (en) * | 2013-03-26 | 2013-06-26 | 北京理工大学 | Dual-model lane line identification method based on dynamic area division |
| CN108280450A (en) * | 2017-12-29 | 2018-07-13 | 安徽农业大学 | A kind of express highway pavement detection method based on lane line |
-
2018
- 2018-11-27 CN CN201811427546.XA patent/CN109583365B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102103747A (en) * | 2009-12-16 | 2011-06-22 | 中国科学院电子学研究所 | Method for calibrating external parameters of monitoring camera by adopting reference height |
| US20120078867A1 (en) * | 2010-09-29 | 2012-03-29 | Navteq North America, Llc | Multi-Dimensional Road Representation |
| EP2602744A1 (en) * | 2011-12-08 | 2013-06-12 | Delphi Technologies, Inc. | Method for detecting and tracking lane markings |
| CN103177246A (en) * | 2013-03-26 | 2013-06-26 | 北京理工大学 | Dual-model lane line identification method based on dynamic area division |
| CN108280450A (en) * | 2017-12-29 | 2018-07-13 | 安徽农业大学 | A kind of express highway pavement detection method based on lane line |
Non-Patent Citations (2)
| Title |
|---|
| 穆柯楠: "基于车—路视觉协同的行车环境感知方法研究", 《中国优秀博硕士学位论文全文数据库(博士)》 * |
| 陈龙等: "基于成像模型的车道线检测与跟踪方法", 《中国公路学报》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110135252A (en) * | 2019-04-11 | 2019-08-16 | 长安大学 | A kind of adaptive accurate lane detection and deviation method for early warning for unmanned vehicle |
| CN112654998A (en) * | 2020-10-22 | 2021-04-13 | 华为技术有限公司 | Lane line detection method and device |
| WO2022082571A1 (en) * | 2020-10-22 | 2022-04-28 | 华为技术有限公司 | Lane line detection method and apparatus |
| WO2022082574A1 (en) * | 2020-10-22 | 2022-04-28 | 华为技术有限公司 | Lane line detection method and apparatus |
| CN112818873A (en) * | 2021-02-04 | 2021-05-18 | 苏州魔视智能科技有限公司 | Lane line detection method and system and electronic equipment |
| CN112818873B (en) * | 2021-02-04 | 2023-05-26 | 苏州魔视智能科技有限公司 | Lane line detection method and system and electronic equipment |
| CN113450380A (en) * | 2021-07-17 | 2021-09-28 | 普达迪泰（天津）智能装备科技有限公司 | Track calibration method based on airport runway scribed lines |
| CN114912159A (en) * | 2022-07-18 | 2022-08-16 | 中国铁路设计集团有限公司 | Method for fitting geometric line shape of rail transit line plane |
| CN114912159B (en) * | 2022-07-18 | 2022-09-13 | 中国铁路设计集团有限公司 | Method for fitting geometric line shape of rail transit line plane |
| CN115123218A (en) * | 2022-09-02 | 2022-09-30 | 小米汽车科技有限公司 | Vehicle detection method and device and electronic equipment thereof |
| CN115311314A (en) * | 2022-10-13 | 2022-11-08 | 深圳市华汉伟业科技有限公司 | Resampling method, system and storage medium for line laser contour data |
| CN115311314B (en) * | 2022-10-13 | 2023-02-17 | 深圳市华汉伟业科技有限公司 | Resampling method, system and storage medium for line laser contour data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109583365B (en) | 2022-07-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109583365A (en) | Method for detecting lane lines is fitted based on imaging model constraint non-uniform B-spline curve | |
| CN107679520B (en) | Lane line visual detection method suitable for complex conditions | |
| CN110569704B (en) | Multi-strategy self-adaptive lane line detection method based on stereoscopic vision | |
| CN110136182B (en) | Registration method, device, equipment and medium for laser point cloud and 2D image | |
| CN108280450B (en) | Expressway pavement detection method based on lane lines | |
| CN104217427B (en) | Lane line localization method in a kind of Traffic Surveillance Video | |
| CN100494886C (en) | Three-dimensional scanning system circular index point detection method | |
| CN109583280A (en) | Lane detection method, apparatus, equipment and storage medium | |
| CN108597009B (en) | Method for detecting three-dimensional target based on direction angle information | |
| CN110307791B (en) | Vehicle length and speed calculation method based on three-dimensional vehicle boundary frame | |
| US7848571B2 (en) | Computer-implemented method for efficient image segmentation using automated saddle-point detection | |
| Youjin et al. | A robust lane detection method based on vanishing point estimation | |
| CN111145228A (en) | Heterogeneous image registration method based on local contour point and shape feature fusion | |
| CN109886124A (en) | One kind describing the matched texture-free metal parts grasping means of subgraph based on harness | |
| CN106875430B (en) | Single moving target tracking method and device based on fixed form under dynamic background | |
| Miksch et al. | Automatic extrinsic camera self-calibration based on homography and epipolar geometry | |
| CN116503818A (en) | Multi-lane vehicle speed detection method and system | |
| CN113793309B (en) | Subpixel level ellipse detection method based on morphological characteristics | |
| CN108846363A (en) | A kind of subregion vehicle bottom shadow detection method based on divergence expression scanning | |
| CN117576219A (en) | Camera calibration equipment and calibration method for single shot image of large wide-angle fish-eye lens | |
| CN117611525A (en) | Visual detection method and system for abrasion of pantograph slide plate | |
| Sihombing et al. | Perspective rectification in vehicle number plate recognition using 2D-2D transformation of Planar Homography | |
| CN105930813B (en) | A method of detection composes a piece of writing this under any natural scene | |
| CN110488320B (en) | Method for detecting vehicle distance by using stereoscopic vision | |
| CN112991372A (en) | 2D-3D camera external parameter calibration method based on polygon matching |
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 |