CN105261020B - A kind of express lane line detecting method - Google Patents
A kind of express lane line detecting method Download PDFInfo
- Publication number
- CN105261020B CN105261020B CN201510671147.8A CN201510671147A CN105261020B CN 105261020 B CN105261020 B CN 105261020B CN 201510671147 A CN201510671147 A CN 201510671147A CN 105261020 B CN105261020 B CN 105261020B
- Authority
- CN
- China
- Prior art keywords
- lane line
- image
- line
- value
- curve
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000001514 detection method Methods 0.000 claims abstract description 17
- 230000009466 transformation Effects 0.000 claims abstract description 14
- 238000001914 filtration Methods 0.000 claims abstract description 7
- 238000013507 mapping Methods 0.000 claims abstract description 6
- 238000012216 screening Methods 0.000 claims abstract description 6
- 238000006243 chemical reaction Methods 0.000 claims abstract description 4
- 230000008569 process Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 8
- 238000005452 bending Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000003287 optical effect Effects 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000000007 visual effect Effects 0.000 claims description 3
- 208000001491 myopia Diseases 0.000 claims description 2
- 230000009977 dual effect Effects 0.000 claims 1
- 230000005012 migration Effects 0.000 abstract description 2
- 238000013508 migration Methods 0.000 abstract description 2
- 206010039203 Road traffic accident Diseases 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 102100034112 Alkyldihydroxyacetonephosphate synthase, peroxisomal Human genes 0.000 description 2
- 101000799143 Homo sapiens Alkyldihydroxyacetonephosphate synthase, peroxisomal Proteins 0.000 description 2
- 238000000848 angular dependent Auger electron spectroscopy Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 210000003462 vein Anatomy 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 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
- 235000004522 Pentaglottis sempervirens Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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
Abstract
The present invention discloses a kind of express lane line detecting method, its road front view that will acquire first is by inverse perspective mapping method migration at intuitive birds-eye view, the denoising of corresponding gaussian filtering and binary conversion treatment, which are done, for birds-eye view then uses the related lane line region of fast Hough transformation detection, and group areas is carried out to it to reduce operand, lane line screening and fitting are carried out to related lane line region finally by improved RANSAC algorithm.It is an advantage of the invention that accurate in complex environment lane line can be extracted, and real-time is preferable, while having both bend identification function.The present invention accurate in complex environment can extract lane line, and real-time is preferable, while having both bend identification function.
Description
Technical field
The invention belongs to digital image understanding technical fields, and in particular to a kind of express lane line detecting method.
Background technique
Universal and traffic accident with automobile takes place frequently, and vehicle security drive is increasingly concerned by people.And most traffic accident
It is all as caused by the bad habit of driver.Therefore, Senior Officer's auxiliary system (Advanced Driver
Assistance System, ADAS) proposition can effectively reduce the occurrence probability of traffic accident.Deviation, which detects, simultaneously is
Senior Officer is a ring important in Senior Officer's auxiliary system, therefore fast and accurately identifies that the lane line on road is aobvious
It obtains particularly important.
In ADAS, various sensor equipments, such as laser radar, GPS etc. are often used.But pass through
Visual sensor can not only greatly save cost using computer vision methods detection lane line, and detection effect is also very
It is good.Currently, including the lane line of the identification based on characteristics of image by the method that image recognition technology detects lane line
The method for detecting lane lines two major classes of detection method and the identification based on model.
The method for detecting lane lines of identification based on feature be broadly divided into identification in grayscale image based on edge feature and
Identification etc. based on road color and vein in cromogram.Edge Gradient Feature in grayscale image, it is most of all to use suddenly
Husband converts to carry out the extraction in lane, but the computation complexity of Hough transformation is bigger, therefore the speed of service is slow, simultaneously
It is difficult for the extraction comparison of bend.Most of identification based on road color and vein is to convert RGB color to HSI face
Then the colour space models the coloration of pixel and brightness respectively.But it is operated on each pixel, therefore to road
On light and shade it is more sensitive.
The method for detecting lane lines of identification then based on model is divided into straight line model, spline curve model, Linear Parabolic line
Model, conic model and hyperbolic model etc..These models are all to belong to geometrical model, and simply fixed model cannot be very
Good is fitted for lane complicated and changeable, but the more flexible model such as similar B-spline curves, conic section can be with
Effectively lane is fitted, but the identification based on model is only used only, lane breakage and picture noise will lead to
Fitting failure.
Summary of the invention
The technical problem to be solved by the present invention is to deficiencies existing for existing method for detecting lane lines, provide a kind of quick vehicle
Road line detecting method accurate in complex environment can extract lane line, and real-time is preferable, while having both curved
Road identification function.
To solve the above problems, the present invention is achieved by the following technical solutions:
A kind of express lane line detecting method, specifically comprises the following steps:
Step 1, the image that road is got by the camera being installed on vehicle, and record relevant parameter, i.e.,
Terrain clearance, yaw angle, helical angle, the picture size of focal length, optical centre coordinate and acquired image of camera;
Step 2, for the image and parameter got, setting and the inverse perspective mapping of area-of-interest are carried out, by image
Be converted to birds-eye view;
Step 3 denoises birds-eye view using 2-d gaussian filters processing, and the image after denoising is carried out two-value
Change processing;
Step 4 obtains one group of candidate's vehicle first with Hough transformation progress straight-line detection for the picture after binaryzation
Then diatom carries out simple screening to candidate lane line using distance weighting formula and left and right candidate lane line is grouped, obtains a left side
Candidate lane line region and right candidate lane line region;
Step 5, using RANSAC algorithm to respectively to left candidate lane line region and right candidate lane line area
Domain is screened, and is fitted respectively to left candidate lane line region and right candidate lane line region straight line and curve, is obtained
Final left-lane line and right-lane line.
Above-mentioned steps 2 are specially:
Firstly, according to the reasonable area-of-interest of the image setting got;
Then, each pixel in the area-of-interest is done into matrixing, image coordinate system is converted into the world and is sat
Mark system;
Finally, world coordinate system to be converted to the image coordinate system of birds-eye view.
In above-mentioned steps 3, when carrying out 2-d gaussian filters processing denoising to birds-eye view, need first that birds-eye view is discrete
It turns to after pixel, then denoising is carried out to these pixels.
In step 3, the process for carrying out 2-d gaussian filters processing denoising to birds-eye view is included in vertical direction using Gauss
The process of low-pass filtering treatment and the process handled in the horizontal direction using second order Gauss differential filtering.
In above-mentioned steps 4, the process for carrying out straight-line detection using Hough transformation is as follows:
Firstly, the step-size in search of setting Hough transformation and the range of search space r and φ, i.e. rmin≤r≤rmax, φmin≤
φ≤φmax;
Then, two dimension ballot accumulator A (r, φ) is established, each element is set as zero in initial accumulator;
Then, the non-zero points (x after a certain number of binaryzations of stochastic inputs in imagei,yi), i=1,2,3...n, and
Using step-length by the value of independent variable φ discretization within its scope, from φminStart according to the step-length of setting in hough space according to
It is secondary to take φ value from small to large, therefore each non-zero points and the different φ values got can be according to the calculating pair of determining polar equation
The r value answered just adds 1 in corresponding two dimension ballot accumulator A (r, φ) of r value and φ value once calculating r;
Finally, the final accumulated value of two dimension ballot accumulator A (r, φ) is Hough value, respective image plane collinear is found out
The local maximum of the accumulator of point, when local maximum meets the minimum threshold of setting, then the straight line to detect.
In above-mentioned steps 4, the process being grouped to candidate lane line is specially:
Firstly, select straight line from the straight line detected as essentially linear, and calculate separately remaining straight line with should
Essentially linear distance value d;
Then, all straight lines detected are grouped, i.e., when distance value d is less than or equal to the distance threshold of setting,
Then it is divided into one group;When distance value d is greater than the distance threshold of setting, then it is divided into another set;
Finally, the left and right lane line group that rectangle frame will divide is arranged according to the intersection point of two groups of straight lines and image edge
It frames, to obtain left candidate lane line region and right candidate lane line region.
Above-mentioned steps 5 are specially:
Step 5.1, the left candidate lane line region that will acquire respectively and right candidate lane line region from centre be divided into
Preceding and two sub-regions backward, and sample point is obtained at random from subregion forward and backward subregion;
Step 5.2, the fitting for carrying out straight line and curve to the sample point in region using least square method;Utilize entirety
Region is subregion and backward the sample point progress straight line fitting of subregion forward, while being clicked through using the sample of subregion forward
Row curve matching;
The degree of fitting of step 5.3, calculated curve;
If minimum requirements is not achieved in degree of fitting in step 5.4, step 5.3, will be randomly selected since step 5.1 again
Sample point.
Compared with prior art, the beneficial effects of the invention are as follows:
1, the present invention obtains birds-eye view by inverse perspective mapping, can intuitively see the shape of lane line in image, and
And it can effectively exclude relevant interference information.
2, a kind of method that the present invention introduces group areas after Hough transformation processing, is effectively reduced algorithm operation
Amount.
3, the present invention is by improving general RANSAC (Random Sample Consensus, stochastical sampling are consistent)
Relevant degree of fitting calculation formula is arranged in algorithm, improves the accuracy of detection and reduces operand.
Detailed description of the invention
Fig. 1 is the flow chart of express lane line detection designed by the present invention.
Fig. 2 is the stochastical sampling point region of candidate lane line.
Fig. 3 is Bezier.
Specific embodiment
It elaborates below in conjunction with drawings and examples to the present invention:The present embodiment before being with technical solution of the present invention
Put the example implemented, give detailed embodiment and process, but protection scope of the present invention should not necessarily be limited by it is following
Embodiment.
A kind of express lane line detecting method, as shown in Figure 1, specifically comprising the following steps:
Step 1 gets the image of road by being mounted on the camera of vehicle front, and records relevant parameter.
The present invention gets road image according to the camera for being mounted on vehicle front, and at the same time recording camera
The ginsengs such as focal length, optical centre coordinate, the camera height from the ground of installation, yaw angle, helical angle, the dimension of picture that gets
Number.
It is (350,360) that camera focal length, which decomposes the coordinate in reference axis, in present example, and optical centre coordinate is
(320,240), distance is 1.3 meters to the camera of installation from the ground, and yaw angle is 0 degree, and helical angle is -9 degree, carries out lane line knowledge
Other dimension of picture size is 640 × 480.
Step 2, for the image and parameter got, setting and the inverse perspective mapping of area-of-interest are carried out, by forward sight
Figure is converted to birds-eye view.
According to the reasonable area-of-interest of the image setting got, each pixel in the region is done into matrix change
It changes, the image coordinate system of front view is converted into world coordinate system, then world coordinate system is converted to the image of birds-eye view again
Coordinate system.Therefore, front view is converted into birds-eye view.
Step 3 denoises birds-eye view using 2-d gaussian filters processing, and the image after denoising is carried out two-value
Change processing.
The present invention is using the incorgruous gaussian filtering process of two dimension.Gassian low-pass filter, Convolution Formula are used in vertical direction
For formula (1), wherein ρyIt is configured according to specific lane line situation and (is set as 9 in implementation process of the present invention);In the horizontal direction then
Using second order Gauss differential filtering, Convolution Formula is formula (2), wherein ρx(this hair is configured according to specific lane line situation
It is set as 23) in bright implementation process.
In practical operation, the discrete operation turned to pixel is filtered in these by the present invention, thus be effect more
Add obviously, i.e., then passing bay line carries out picture pixels at binaryzation according to formula (3) compared with seeming brighter under dark background
Reason.
Wherein, κ is the parameter set according to image concrete condition, and T is the maximum value in all pixels, and B is all pixels
Middle minimum value.
Step 4 is directed to the picture after binaryzation, carries out straight-line detection using fast Hough transformation, by the way that corresponding search is arranged
The size of step-length and hough space obtains one group of candidate lane line, is then carried out using distance weighting formula to candidate lane line
Simple screening and the grouping of left and right lane line.
Two points can determine straight line in image space, according to polar equation (4) form, by calculating also just
R and φ in polar equation has been determined, has obtained the straight line of a polar equation form in x-y coordinate system, according to dotted line antithesis
Property, then it is equivalent to point (r, φ) and this straight line has been determined.But converted when by coordinate space, it will be the sky of reference axis by x and y
Between be converted by r and φ be reference axis hough space, then polar equation (4) in hough space be shown one just
Chord curve.Therefore each point (x in image spacei,yi), it substitutes into equation (4), is being the Hough of reference axis by r and φ
As soon as show to be a sine curve in space, and sinusoidal meet at of how many item a little represents that how many point can be total to
Line indicates there are a certain number of points point-blank in image when reaching certain amount.
The present invention will be using the form (4) of the polar equation in standard Hough transformation:
R=x sin φ+y cos φ (4)
Wherein:R be in image space coordinate origin to the straight line distance, and the coordinate origin in image refer to image a left side
First, upper angle pixel coordinate;φ is the angle of straight line and axis.
According to above-mentioned principle, the present invention detects straight line, and specific step is as follows:
Firstly, the range of setting r and φ, i.e.,:rmin≤r≤rmax, φmin≤φ≤φmax.Its step-size in search is 3, setting
A length of (rmax-rmin)/3, width are (φmax-φminThe search space of)/3.
Then, two dimension ballot accumulator A (r, φ) is established, each element is set as zero in initial accumulator.
Then, the non-zero points (x after a certain number of binaryzations of stochastic inputs in imagei,yi), i=1,2,3...n, and
Using step-length by the value of independent variable φ discretization within its scope, from φminStart according to the step-length of setting in hough space according to
It is secondary to take φ value from small to large, therefore each non-zero points and the different φ values got can be counted according to determining polar equation (4)
Corresponding r value is calculated, once calculating r, 1 is just added in r value and φ value corresponding A (r, φ) unit accumulator.
A (r, φ)=A (r, φ)+1
Finally, two dimension ballot accumulator A (r, φ) accumulated value be Hough value, represent be how many point be conllinear
, and this conllinear straight line be the polar equation controlled by parameter r and φ.And the Hough value of local maxima, that is, it indicates
The polar equation of parameter r and φ control is part most multi-point and common-line.Find out the accumulator of respective image plane collinear point
Local maximum, when local maximum meets its minimum threshold, then the straight line to detect.
The above-mentioned process being grouped to candidate lane line is specific as follows:
Every straight line that detected has its parameter r and φ.Therefore the present invention with wherein straight line be it is essentially linear,
Then formula (5) are utilized, calculates remaining straight line and the essentially linear distance value.
Assuming that first line segment detected is (r1, φ1), Article 2 line segment be (r2, φ2), then this two linear distances
The calculation formula of relationship is
D=| r1-r2|+λ|φ1-φ2| (5)
Wherein λ is weight coefficient, and
In addition, guaranteeing parameter phi by formula (6)iJack per line, to calculate more acurrate.
When distance threshold of the d value less than setting, then it is divided into one group, greater than the distance threshold of setting, is then divided into another set.
According to the intersection point of each group of straight line and image edge, rectangle frame is set and frames the left and right lane line group divided, thus
Obtain the relevant range of candidate lane line.
Step 5 occurs in order to prevent by left and right lane detection to situation together, therefore will be to detecting associated straight lines area
It is as a result more accurate after domain is grouped, then when taking RANSAC algorithm.The present invention utilizes improved RANSAC (random sampling
Consistency) algorithm screens candidate lane line, is finally fitted with straight line and three rank Beziers to lane line.
Step 5.1 randomly selects non-zero sample point.In lane line front view, near-sighted field is namely close to picture bottom
Region be approximately lane line linearity region, if there is bend to depend on long sight field areas, i.e. end point region in road picture,
Therefore it in order to reduce calculation amount, only carries out curve fitting to far visual field, to the ROI region straight line fitting integrally chosen.According to
The candidate lane line region got, is divided into two sub-regions from centre and forwardly and rearwardly obtains corresponding sample point at random, such as
Fig. 2.So the sample point of upper area is for curve matching, all sample points of overall region are for straight line fitting.
Step 5.2, the fitting for doing straight line and curve to corresponding region lane line simultaneously using least square method.
(1) fitting of straight line
According to the Hough transformation in previous step, several candidate lane lines are obtained, are converted to standard straight-line equation y=ax
The fitting of+b form.And according to the sample point in the candidate lane line region randomly selected, the distance that point arrives the straight line is calculated, it is false
If candidate lane line shares N item, if kth straight line y=akx+bk, according to n sample point is taken in previous step, then calculate a little
(xi,yi) to straight line apart from size:
Then all distances are summed, i.e.,
If dkLess than setting threshold value d*, on the contrary then be erroneous detection then be exactly determining form of straight lines lane line.If Quan Weida
It arrives, then again since step 5.1.
(2) fitting of curve
Curve matching is mainly according to three rank Bezier forms (8):
Wherein, [0,1] t ∈, P (0)=P0, P (1)=P3.One three rank Bezier basic configuration is exactly by P1And P2
Control.
Therefore what is calculated focuses on calculating P1And P2.Assuming that there is n sample point, midpoint can be expressed as qi=(ui,
vi), if ti∈ [0,1] is qiTo q1Euclidean distance ratio is accumulated, specific formula is as follows:
Wherein, q0=q1, andAnd t1=0 and tn=1 indicates
The beginning and end of curve.
It defines simultaneously:
Pass through P=(TM)*Q, wherein (TM)*It is the pseudo inverse matrix of (TM).To calculate all control points.All controls
System point can be fitted after calculating with the formula of Bezier.
The degree of fitting of step 5.3, calculated curve, degree of fitting formula are formula (10)
In formula:Score is curve matching degree, and s is the parameter (the pixel value summation of curve) of primitive curve;L=l/v-1, l
It is length of curve, which can calculate according to Bezier property, and v is picture height, therefore, if L=0 is meaned
Be a very long curve, mean that the curve is very short if L=-1.θ indicates embroidery, be defined as θ=θ '/
2, θ ' between control point lines bending diversity factor, i.e. θ '=cos (θ1-θ2), θ1And θ2It, can be with for bending angle before and after curve
By being obtained in the curve that detects, such as Fig. 3, wherein m1And m2It is the limitation parameter according to corresponding situation sets itself.
If minimum requirements is not achieved in degree of fitting in step 5.4, step 5.3, will be randomly selected since step 5.1 again
Sample point.
The invention discloses a kind of based on the express lane line detection algorithms for getting a bird's eye view map analysis and improvement RANSAC.It is first
The road front view that will acquire, at intuitive birds-eye view, is corresponding Gauss for birds-eye view by inverse perspective mapping method migration
Then filtering and noise reduction and binary conversion treatment use the related lane line region of fast Hough transformation detection, and carry out group areas to it
To reduce operand, lane line screening and fitting are carried out to related lane line region finally by improved RANSAC algorithm.This
Invention has preferable robustness and real-time, and has both certain bend identification function, to realize in complicated road
More accurate rapidly extracting is to lane line in environment, to prevent vehicle from deviation occurring in the process of moving.
Claims (4)
1. a kind of express lane line detecting method, characterized in that specifically comprise the following steps:
Step 1, the image that road is got by the camera being installed on vehicle, and record relevant parameter, that is, it images
Terrain clearance, the yaw angle, helical angle, the image size of focal length, optical centre coordinate and acquired image of head;
Step 2, for the image and parameter got, carry out setting and the inverse perspective mapping of area-of-interest, image converted
For birds-eye view;
Step 3 denoises birds-eye view using 2-d gaussian filters processing, uses Gassian low-pass filter in vertical direction,
Horizontal direction then uses second order Gauss differential filtering, and the image after denoising can make lane line highlight its vehicle under darker background
Then the feature of diatom carries out binary conversion treatment again, lead to lane line and pavement of road difference because light influences in order to prevent
Property it is too small, therefore using binary conversion treatment of overall importance is considered, i.e.,:
Wherein, κ is the parameter set according to image concrete condition, and T is the maximum value in all pixels, B be in all pixels most
Small value;
Step 4 obtains one group of candidate lane line first with Hough transformation progress straight-line detection for the image after binaryzation,
Then simple screening is carried out to candidate lane line using distance weighting formula and left and right candidate lane line is grouped, i.e.,:
Assuming that first line segment detected is (r1, φ1), Article 2 line segment be (r2, φ2), wherein riIt is to be sat in image space
Mark distance of the origin to line segment, φiIt is the angle of line segment and axis, and the two parameters can directly be obtained by Hough transformation
It arrives, then, the calculation formula of this two linear distance relationships is d=| r1-r2|+λ|φ1-φ2|
Wherein λ is weight coefficient, and
In addition, guaranteeing parameter phi by above-mentioned formulaiJack per line, to calculate more acurrate;
When distance threshold of the d value less than or equal to setting, then it is divided into one group, greater than the distance threshold of setting, is then divided into another set,
According to the intersection point of each group of straight line and image border, rectangle frame is set and frames the left and right lane line group divided, thus
Obtain the relevant range of candidate lane line;
Step 5 respectively carries out left candidate lane line region and right candidate lane line region using RANSAC algorithm
Screening, and is fitted in the part in each lane line region, is fitted in the near-sighted field in region with straight line model, according to
The sample point in candidate lane line region that machine is chosen calculates the distance that point arrives straight line, it is assumed that and candidate lane line shares N item, if
Kth straight line y=akx+bk, according to n sample point is taken in previous step, then calculate point (xi,yi) to straight line distance it is big
It is small:
Then all distances are summed, i.e.,
If dkLess than setting threshold value d*, on the contrary then be erroneous detection then be exactly determining form of straight lines lane line;If complete be not up to,
Again sample point is chosen;
Far visual field then uses Bezier to be fitted, and curve matching is mainly according to three rank Bezier forms:
Wherein, [0,1] t ∈, P (0)=P0, P (1)=P3;One three rank Bezier basic configuration is exactly by P1And P2Control;
Therefore what is calculated focuses on calculating P1And P2;Assuming that there is n sample point, midpoint can be expressed as qi=(ui,vi), if
ti∈ [0,1] is qiTo q1Euclidean distance ratio is accumulated, specific formula is as follows:
Wherein, q0=q1, andAnd t1=0 and tn=1 indicates curve
Beginning and end;
It defines simultaneously:
By P=(TM) * Q, wherein (TM) * is the pseudo inverse matrix of (TM);To calculate all control points;All control points
It can be fitted with the formula of Bezier after calculating;
The degree of fitting of calculated curve is defined simultaneously, and degree of fitting formula is
In formula:Score is curve matching degree, and s is parameter, that is, curve pixel value summation of primitive curve;L=l/v-1, l are bent
Line length, the length can be calculated according to Bezier property, and v is picture altitude, therefore, if L=0 is meant to be
One very long curve means that the curve is very short if L=-1;θ indicates embroidery, is defined as θ=θ '/2, θ '
The bending diversity factor of lines between control point, i.e. θ '=cos (θ1-θ2), θ1And θ2For bending angle before and after curve, can pass through
It is obtained in the curve detected, wherein m1And m2It is the limitation parameter according to corresponding situation sets itself;
Erroneous detection can be effectively reduced using dual model fitting, to more accurately obtain final left-lane line and right-lane line.
2. a kind of express lane line detecting method according to claim 1, characterized in that step 2 is specially:
Firstly, according to the reasonable area-of-interest of the image setting got;
Then, each pixel in the area-of-interest is done into matrixing, image coordinate system is converted into world coordinate system;
Finally, world coordinate system to be converted to the image coordinate system of birds-eye view.
3. a kind of express lane line detecting method according to claim 1, characterized in that in step 3, carried out to birds-eye view
2-d gaussian filters handle denoising when, need first by birds-eye view is discrete turn to pixel after, then to these pixels carry out
Denoising.
4. a kind of express lane line detecting method according to claim 1, characterized in that in step 4, utilize Hough transformation
The process for carrying out straight-line detection is as follows:
Firstly, the step-size in search of setting Hough transformation and the range of search space r and φ, i.e. rmin≤r≤rmax, φmin≤φ≤
φmax;
Then, two dimension ballot accumulator A (r, φ) is established, each element is set as zero in initial accumulator;
Then, the non-zero points (x after a certain number of binaryzations of stochastic inputs in imagei,yi), i=1,2,3...n, and utilize
Step-length is by the value of independent variable φ discretization within its scope, from φminStart according to the step-length of setting in hough space successively from
It is small to take φ value to big, therefore each non-zero points and the different φ values got can be corresponding according to the calculating of determining polar equation
R value just adds 1 in corresponding two dimension ballot accumulator A (r, φ) of r value and φ value once calculating r;
Finally, the final accumulated value of two dimension ballot accumulator A (r, φ) is Hough value, respective image plane collinear point is found out
The local maximum of accumulator, when local maximum meets the minimum threshold of setting, then the straight line to detect.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510671147.8A CN105261020B (en) | 2015-10-16 | 2015-10-16 | A kind of express lane line detecting method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510671147.8A CN105261020B (en) | 2015-10-16 | 2015-10-16 | A kind of express lane line detecting method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105261020A CN105261020A (en) | 2016-01-20 |
CN105261020B true CN105261020B (en) | 2018-11-20 |
Family
ID=55100692
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510671147.8A Active CN105261020B (en) | 2015-10-16 | 2015-10-16 | A kind of express lane line detecting method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105261020B (en) |
Families Citing this family (41)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105718916A (en) * | 2016-01-27 | 2016-06-29 | 大连楼兰科技股份有限公司 | Lane line detection method based on Hough transform |
JP6256508B2 (en) * | 2016-03-30 | 2018-01-10 | マツダ株式会社 | Electronic mirror control device |
CN106407893B (en) * | 2016-08-29 | 2019-11-22 | 东软集团股份有限公司 | A kind of method, apparatus and equipment detecting lane line |
CN106529505A (en) * | 2016-12-05 | 2017-03-22 | 惠州华阳通用电子有限公司 | Image-vision-based lane line detection method |
CN106778661A (en) * | 2016-12-28 | 2017-05-31 | 深圳市美好幸福生活安全系统有限公司 | A kind of express lane line detecting method based on morphological transformation and adaptive threshold |
CN106778668B (en) * | 2016-12-30 | 2019-08-09 | 明见(厦门)技术有限公司 | A kind of method for detecting lane lines of robust that combining RANSAC and CNN |
CN106874875A (en) * | 2017-02-17 | 2017-06-20 | 武汉理工大学 | A kind of vehicle-mounted lane detection system and method |
CN107045629B (en) * | 2017-04-19 | 2020-06-26 | 南京理工大学 | Multi-lane line detection method |
CN108873876A (en) * | 2017-05-09 | 2018-11-23 | 恒大法拉第未来智能汽车(广东)有限公司 | Unmanned vehicle wheelpath method and device for planning |
CN107194941A (en) * | 2017-05-23 | 2017-09-22 | 武汉科技大学 | A kind of unmanned plane independent landing method, system and electronic equipment based on monocular vision |
CN107609486A (en) * | 2017-08-16 | 2018-01-19 | 中国地质大学(武汉) | To anti-collision early warning method and system before a kind of vehicle |
CN107704801B (en) * | 2017-08-18 | 2021-08-06 | 电子科技大学 | Curve lane line detection method based on segmented straight line and segmented Bezier curve |
CN109583271B (en) * | 2017-09-29 | 2020-11-06 | 杭州海康威视数字技术股份有限公司 | Method, device and terminal for fitting lane line |
CN109753841B (en) * | 2017-11-01 | 2023-12-12 | 比亚迪股份有限公司 | Lane line identification method and device |
CN107862290B (en) * | 2017-11-10 | 2021-09-24 | 智车优行科技(北京)有限公司 | Lane line detection method and system |
CN108038416B (en) * | 2017-11-10 | 2021-09-24 | 智车优行科技(北京)有限公司 | Lane line detection method and system |
CN107895375B (en) * | 2017-11-23 | 2020-03-31 | 南京莱斯电子设备有限公司 | Complex road route extraction method based on visual multi-features |
CN107978165A (en) * | 2017-12-12 | 2018-05-01 | 南京理工大学 | Intersection identifier marking and signal lamp Intellisense method based on computer vision |
CN108052904B (en) * | 2017-12-13 | 2021-11-30 | 辽宁工业大学 | Method and device for acquiring lane line |
CN107918775B (en) * | 2017-12-28 | 2020-04-17 | 聊城大学 | Zebra crossing detection method and system for assisting safe driving of vehicle |
CN108256446B (en) | 2017-12-29 | 2020-12-11 | 百度在线网络技术(北京)有限公司 | Method, device and equipment for determining lane line in road |
CN108256445B (en) * | 2017-12-29 | 2020-11-06 | 北京华航无线电测量研究所 | Lane line detection method and system |
CN110163039B (en) * | 2018-03-15 | 2020-11-24 | 北京航空航天大学 | Method, apparatus, storage medium, and processor for determining vehicle driving state |
CN108828518B (en) * | 2018-03-29 | 2022-11-18 | 上海大学 | Method for positioning stacking car in wagon carriage |
CN110399762A (en) * | 2018-04-24 | 2019-11-01 | 北京四维图新科技股份有限公司 | A kind of method and device of the lane detection based on monocular image |
CN108764253A (en) * | 2018-05-15 | 2018-11-06 | 杭州电子科技大学 | Pointer instrument digitizing solution |
CN108932472A (en) * | 2018-05-23 | 2018-12-04 | 中国汽车技术研究中心有限公司 | A kind of automatic Pilot running region method of discrimination based on lane detection |
CN109214334B (en) * | 2018-09-03 | 2020-12-29 | 百度在线网络技术(北京)有限公司 | Lane line processing method and device |
CN110490033B (en) * | 2018-10-29 | 2022-08-23 | 毫末智行科技有限公司 | Image processing method and device for lane detection |
CN109572715A (en) * | 2018-11-01 | 2019-04-05 | 北京海纳川汽车部件股份有限公司 | Lane departure warning method, device and the automatic driving vehicle of automatic driving vehicle |
CN109886131B (en) * | 2019-01-24 | 2023-05-02 | 淮安信息职业技术学院 | Road curve recognition method and device |
CN109886175B (en) * | 2019-02-13 | 2021-06-01 | 合肥思艾汽车科技有限公司 | Method for detecting lane line by combining straight line and circular arc |
CN109902758B (en) * | 2019-03-11 | 2022-05-31 | 重庆邮电大学 | Deep learning-based lane area identification data set calibration method |
CN110287779B (en) * | 2019-05-17 | 2021-10-15 | 百度在线网络技术(北京)有限公司 | Method, device and equipment for detecting lane line |
CN110176000B (en) * | 2019-06-03 | 2022-04-05 | 斑马网络技术有限公司 | Road quality detection method and device, storage medium and electronic equipment |
CN111563412B (en) * | 2020-03-31 | 2022-05-17 | 武汉大学 | Rapid lane line detection method based on parameter space voting and Bessel fitting |
CN111611862B (en) * | 2020-04-22 | 2022-09-09 | 浙江众合科技股份有限公司 | Curve fitting-based semi-automatic labeling method for subway rail |
CN111798434A (en) * | 2020-07-08 | 2020-10-20 | 哈尔滨体育学院 | Martial arts competition area detection method based on Ranpac model |
CN112418123B (en) * | 2020-11-30 | 2021-08-03 | 西南交通大学 | Hough transformation-based engineering drawing line and line type identification method |
CN113011293B (en) * | 2021-03-05 | 2022-09-30 | 郑州天迈科技股份有限公司 | Real-time extraction method for lane line parameters |
CN117392631B (en) * | 2023-12-08 | 2024-03-08 | 吉咖智能机器人有限公司 | Road boundary extraction method and device, electronic equipment and storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102722705A (en) * | 2012-06-12 | 2012-10-10 | 武汉大学 | Method for detecting multi-lane line on basis of random sample consensus (RANSAC) algorithm |
CN103593671A (en) * | 2013-11-25 | 2014-02-19 | 中国航天科工集团第三研究院第八三五七研究所 | Wide-range lane line vision inspection method based on collaboration of three cameras |
CN104517111A (en) * | 2013-09-27 | 2015-04-15 | 比亚迪股份有限公司 | Lane line detection method and system, and lane deviation early warning method and system |
-
2015
- 2015-10-16 CN CN201510671147.8A patent/CN105261020B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102722705A (en) * | 2012-06-12 | 2012-10-10 | 武汉大学 | Method for detecting multi-lane line on basis of random sample consensus (RANSAC) algorithm |
CN104517111A (en) * | 2013-09-27 | 2015-04-15 | 比亚迪股份有限公司 | Lane line detection method and system, and lane deviation early warning method and system |
CN103593671A (en) * | 2013-11-25 | 2014-02-19 | 中国航天科工集团第三研究院第八三五七研究所 | Wide-range lane line vision inspection method based on collaboration of three cameras |
Non-Patent Citations (3)
Title |
---|
无人驾驶车基于双视觉系统的城市道路检测;唐威;《中国优秀硕士学位论文全文数据库》;20121015(第10期);第I138-2775页 * |
用于车道识别的分段切换车道模型;郭磊等;《公路交通科技》;20061130;第23卷(第11期);第90-94页 * |
车前障碍物检测及车道线识别研究;王海涛;《中国优秀硕士学位论文全文数据库》;20140315(第3期);第I138-884页 * |
Also Published As
Publication number | Publication date |
---|---|
CN105261020A (en) | 2016-01-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105261020B (en) | A kind of express lane line detecting method | |
CN105488454B (en) | Front vehicles detection and ranging based on monocular vision | |
Wu et al. | Lane-mark extraction for automobiles under complex conditions | |
CN110647850A (en) | Automatic lane deviation measuring method based on inverse perspective principle | |
Kong et al. | Generalizing Laplacian of Gaussian filters for vanishing-point detection | |
CN104134209B (en) | A kind of feature extracting and matching method and system in vision guided navigation | |
CN109657632B (en) | Lane line detection and identification method | |
CN107330376A (en) | A kind of Lane detection method and system | |
CN107367310A (en) | A kind of river level remote monitoring method based on computer vision | |
EP2958054A2 (en) | Hazard detection in a scene with moving shadows | |
Li et al. | Nighttime lane markings recognition based on Canny detection and Hough transform | |
CN109738910A (en) | A kind of curb detection method based on three-dimensional laser radar | |
CN104143098B (en) | Night pedestrian recognition method based on far-infrared camera head | |
Yang et al. | Fast and accurate vanishing point detection and its application in inverse perspective mapping of structured road | |
CN108280450A (en) | A kind of express highway pavement detection method based on lane line | |
CN103902985B (en) | High-robustness real-time lane detection algorithm based on ROI | |
CN108052904B (en) | Method and device for acquiring lane line | |
CN106407924A (en) | Binocular road identifying and detecting method based on pavement characteristics | |
CN103996198A (en) | Method for detecting region of interest in complicated natural environment | |
CN108171695A (en) | A kind of express highway pavement detection method based on image procossing | |
CN112906583B (en) | Lane line detection method and device | |
CN104915642B (en) | Front vehicles distance measuring method and device | |
CN106778551A (en) | A kind of fastlink and urban road Lane detection method | |
KR20130053980A (en) | Obstacle detection method using image data fusion and apparatus | |
CN108171154A (en) | A kind of method that target detection is combined with lane detection in automatic Pilot |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20160120 Assignee: Guilin ruiweisaide Technology Co.,Ltd. Assignor: GUILIN University OF ELECTRONIC TECHNOLOGY Contract record no.: X2023980046266 Denomination of invention: A Fast Lane Line Detection Method Granted publication date: 20181120 License type: Common License Record date: 20231108 |
|
EE01 | Entry into force of recordation of patent licensing contract |