CN106529505A - Image-vision-based lane line detection method - Google Patents
Image-vision-based lane line detection method Download PDFInfo
- Publication number
- CN106529505A CN106529505A CN201611102913.XA CN201611102913A CN106529505A CN 106529505 A CN106529505 A CN 106529505A CN 201611102913 A CN201611102913 A CN 201611102913A CN 106529505 A CN106529505 A CN 106529505A
- Authority
- CN
- China
- Prior art keywords
- lane line
- region
- image
- lane
- line
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 10
- 238000000034 method Methods 0.000 claims abstract description 56
- 238000005070 sampling Methods 0.000 claims abstract description 25
- 238000000605 extraction Methods 0.000 claims abstract description 22
- 230000008569 process Effects 0.000 claims description 31
- 238000002372 labelling Methods 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 9
- 230000000295 complement effect Effects 0.000 claims description 5
- 230000008034 disappearance Effects 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 4
- 239000000203 mixture Substances 0.000 claims description 2
- 238000012216 screening Methods 0.000 abstract description 5
- 238000004364 calculation method Methods 0.000 abstract description 2
- 238000006243 chemical reaction Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 241001081972 Arctium medians Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 230000000007 visual effect 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
Disclosed in the invention is an image-vision-based lane line detection method. The method comprises the following steps: collecting a to-be-detected image and carrying out pretreatment on the to-be-detected image; on the basis of a geometrical feature of a lane line and a grey difference, carrying out lane line feature extraction and screening a lane line region; and sampling the lane line region and carrying out linear fitting on the collected data by using a random sampling consensus algorithm, thereby obtaining a lane line parameter by calculation. Using the method provided by the invention, the interference in a non-lane-line region in an image can be effectively filtered; the computing load is reduced substantially; the precision and the extraction speed of lane line parameters are improved; the robustness of the system is enhanced; and the accuracy is high.
Description
Technical field
A kind of the present invention relates to intelligent driving technical field, more particularly to method for detecting lane lines based on image vision.
Background technology
With developing rapidly for highway communication, traffic safety increasingly gets more and more people's extensive concerning, therefore intelligent vehicle
Research become the focus that countries in the world scholars pay close attention to, wherein Lane Departure Warning System is the one of senior DAS (Driver Assistant System)
Kind, the system can give a warning information to driver when detecting vehicle and non-active deviation current lane occur, to remind
Driver corrects in time, very the generation of the reduction vehicle accident of limits.
In Lane Departure Warning System, lane detection is key technology therein.Common method for detecting lane lines
It is divided into two steps：(1) track line feature extraction, the foundation of (2) lane line geometric model with match, wherein track line feature extraction
Precision directly affects the fitting of next step lane line geometric model and matches, therefore track line feature extraction is the most key step
Suddenly.The purpose of track line feature extraction is to retain possible track region to greatest extent, and filters out possible non-track area
Domain.Traditional region segmentation mainly using Threshold segmentation, Sobel operators or Canny operators realizing, wherein for lightness environment
Complicated situation, adaptive threshold or Da-Jin algorithm all effectively cannot be split；And general Sobel and Canny difference is calculated
Son relies solely on neighborhood template, does not account for the geometric properties such as width, the direction of lane line, not only when edge segmentation is realized
Lane line information is enhanced, while other garbages such as vehicle, pedestrian, lane markings are also enhanced, consequently, it is possible to causing car
Diatom information is submerged.When lane line geometric model is fitted with matching, Hough transform (hough transform) is also wide variety of
One of method, the method strong robustness, but substantial amounts of amount of calculation and memory space is needed, speed is slower.Therefore, how to design height
It is the problem for present needing exist for solving to imitate accurate track line feature extraction and geometric model fitting and matching algorithm.
The content of the invention
The technical problem to be solved is to provide a kind of lane detection technical scheme based on image vision,
Overcome that the edge transition that in the line feature extraction of track, produced using conventional edge detection method is split or edge is corresponding not
By force, make lane line marginal information be submerged in the defects such as the interference informations such as lane markings, vehicle, pedestrian, effective filter out image
In non-lane line region interference, reduce operand, improve lane line parameter degree of accuracy and extraction rate, the Shandong of strengthening system
Rod.
For solving above technical problem, the embodiment of the present invention provides a kind of method for detecting lane lines based on image vision,
Including：
Gathering altimetric image to be checked, and treat detection image carries out pretreatment；
Enter driveway line feature extraction based on lane line geometric properties and gray difference, filter out lane line region；
The lane line region is sampled, fitting a straight line is carried out to gathered data using stochastic sampling unification algorism,
Calculate lane line parameter.
In a kind of attainable mode, the detection image for the treatment of carries out pretreatment, including：
ROI region is set on the altimetric image to be checked；The ROI region includes vehicle front and road disappearance horizontal plane
Between road image；
Gray processing process is carried out to the image of ROI region；
Median filter process is carried out to gray level image, removes the noise during image capturing and transmitting.
Preferably, it is described to enter driveway line feature extraction based on lane line geometric properties and gray difference, filter out track
Line region, specifically includes：
The lane line of the ROI region is marked with gray difference according to the wide constraint relation of lane line, and it is right
ROI region after labelling carries out image binaryzation process；
The connectedness in the lane line region after image binaryzation process is analyzed, and according to the geometry in lane line region
Feature is screened to lane line, obtains the lane line region of target；The geometric properties include the direction of lane line, length,
Width and connected domain area.
Further, the wide constraint relation according to lane line and lane line of the gray difference to the ROI region
It is marked, and image binaryzation process is carried out to the ROI region after labelling, specifically includes：
ROI image is begun stepping through from ROI region first trip, calculate respectively the width of lane line, the gray value of areas at both sides,
The gray scale difference sum of the gray value of zone line and areas at both sides；
When the gray value of the areas at both sides of lane line is respectively less than the gray value of zone line, and track line width and two
When the gray scale difference sum in border area domain is in the range of specified threshold, the lane line in ROI region is marked；
Image binaryzation process is carried out to the ROI region after labelling.
Further, the connectedness in the lane line region after the process to image binaryzation is analyzed, and according to car
The geometric properties in diatom region are screened to lane line, are obtained the lane line region of target, are specifically included：
ROI region after image binaryzation process is scanned, the connected region of all labeled lane lines is found out
Domain simultaneously calculates the area of each connected region；
Minimum external envelope rectangular shape description and its geometric properties of all connected regions is calculated successively；
According to the binding occurrence of the geometric properties in the lane line region to target, the lane line of ROI region is screened, protected
Stay the lane line region for meeting restriction range.
Preferably, the lane line region is sampled, gathered data is carried out directly using stochastic sampling unification algorism
Line is fitted, and calculates lane line parameter, including：
Central point with ROI region enters the affiliated point set of driveway line from left and right both direction respectively and is swept as starting point
Retouch；
The point set obtained to scanning, carries out fitting a straight line using stochastic sampling unification algorism, extracts the car in ROI region
Diatom parameter.
Preferably, the central point with ROI region as starting point, from left and right, enter belonging to driveway line respectively by both direction
Point set is scanned, including：
Scanning is with the abscissa at the image center place of ROI region as starting point；
Often first gray value scanning to the left of row be 255 coordinate as left-lane line point set to be fitted, and terminate
The row is scanned；Often first gray value scanning to the right of row be 255 coordinate as right-lane line point set to be fitted, and terminate
The row is scanned；Wherein, all coordinate gray values in the lane line region of the ROI region image after image binaryzation are 255, non-
All coordinate gray values in lane line region are 0.
Preferably, the point set that described pair of scanning is obtained, carries out fitting a straight line using stochastic sampling unification algorism, extracts
Lane line parameter in ROI region, including：
Institute's pointed set in the lane line region that scanning is obtained constitutes set P, randomly selects 2 characteristic points in set P
Set S is constituted, and lane line straight line model y=k is initialized with Srx+br, wherein, kr、brRespectively according to 2 for randomly selecting
Feature point pairs straight line model carries out initializing slope and the intercept for obtaining；
Calculate complementary set ScIn point and lane line straight line model y=krx+brApart from d；Set ScIt is set S in set P
In complementary set；
Threshold tolerance d will be less than apart from diComplementary set ScIn corresponding point constitute the interior point set of set Q, set Q and set S composition
S*；
If interior point set S*In point number more than point quantity t in minimum, then using interior point set S*In point and using minimum
Square law is to lane line straight line model y=krx+brIt is updated, and stores the model parameter；
If lane line straight line model y=krx+brIt is updated over, or, interior point set S*In the number of point be not more than
Point quantity t in minimum, then in set P randomly select again 2 characteristic points and constitute set S to lane line straight line model y=krx
+brInitialized, repeating above step carries out multiple sampling；
After multiple sampling is completed, the most interior point set S of interior point number is chosen*The lane line parameter that calculates of characteristic point
As final lane line parameter.
Method for detecting lane lines based on image vision provided in an embodiment of the present invention, according to default track line width with
Grey scale difference threshold value realizes the initial extraction in lane line region, maintains lane line region to greatest extent；To owning in image
Connected domain be analyzed, further lane line region is screened with reference to the geometric properties of lane line, effectively can filter
Except the interference in non-lane line region in image, the extraction of lane line feature is completed, be that the accurate extraction of lane line parameter is established
Basis；As the present invention has considered the gray feature and geometric properties of lane line, therefore either daytime or night be all
Lane line edge can accurately be split.Therefore, the technical scheme that the present invention is provided greatly reduces operand, improves meter
Speed is calculated, accurately high, strong robustness.
Description of the drawings
Fig. 1 be the present invention provide one embodiment based on the method for detecting lane lines of image vision the step of flow process
Figure.
Fig. 2 be the present invention provide ROI region is marked and binary conversion treatment a kind of embodiment the step of flow process
Figure.
Fig. 3 is that the connectedness to the lane line region after image binaryzation process that the present invention is provided is analyzed and screens
One embodiment the step of flow chart.
Fig. 4 is that the connectedness to the lane line region after image binaryzation process that the present invention is provided is analyzed and screens
Result schematic diagram.
Fig. 5 is the step of employing stochastic sampling unification algorism that the present invention is provided obtains one embodiment of lane line parameter
Flow chart.
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described.
Referring to Fig. 1, the step of be one embodiment based on the method for detecting lane lines of image vision that the present invention is provided
Flow chart.
The method for detecting lane lines based on image vision that the present embodiment is provided, mainly includes the following steps that：
Step S1：Gathering altimetric image to be checked, and treat detection image carries out pretreatment.
When being embodied as, using the photographic head installed in front part of vehicle region (such as below rearview mirror), can be with collecting vehicle
Current frame image of the front containing lane line；The main purpose of Image semantic classification is to eliminate image sampling noise, shade, sky
Deng the impact of chaff interference, maximize and retain lane line region.
Preferably, the detection image for the treatment of carries out pretreatment, including：
A. ROI region is set on the altimetric image to be checked；The ROI region includes vehicle front and road disappearance level
Road image between face.
ROI (Region Of Interest) is also called area-of-interest.In image processing field, area-of-interest
(ROI) it is an image-region selecting from image, this region is graphical analyses emphasis of interest.In the present embodiment
In, the ROI region of selection is mainly the road image between bonnet of motor car top and road disappearance horizontal plane, is concentrated mainly on
The bottom of image.By choosing ROI region, image processing speed can be accelerated, while avoiding the interference of surrounding.In this reality
Apply in example, all algorithms are all carried out in ROI region below, also, ROI region is preferably sized to 320 × 240 pixels.
B. gray processing process is carried out to the image of ROI region.Gray processing process can be filtered some and be reduced data processing amount,
Such that it is able to further speed up image processing speed.
C. median filter process is carried out to gray level image, removes the noise during image capturing and transmitting, so as to ensure
The accuracy of data.
Step S2：Enter driveway line feature extraction based on lane line geometric properties and gray difference, filter out lane line area
Domain.
In a kind of attainable mode, step S2 specifically includes following two parts：
Step S21：Rower is entered with gray difference to the lane line of the ROI region according to the wide constraint relation of lane line
Note, and image binaryzation process is carried out to the ROI region after labelling；
Step S22：The connectedness in the lane line region after image binaryzation process is analyzed, and according to lane line area
The geometric properties in domain are screened to lane line, obtain the lane line region of target；The geometric properties include but is not limited to car
The direction of diatom, length, width and connected domain area.
As lane line feature extraction phases, the wide constraint relation and gray difference feature of lane line is primarily based on to car
Diatom carries out initial markers, marks the region of be likely to be lane line；It is then based on the geometric properties pair in lane line region
Lane line is screened, and retains the region of most strong lane line, and the stage mainly eliminates the interference in other non-lane line regions, such as：
Arrow, doubling, footpath, track prompting character etc..
Step S3：The lane line region is sampled, gathered data is carried out directly using stochastic sampling unification algorism
Line is fitted, and calculates lane line parameter.
Referring to Fig. 2, the step of be a kind of embodiment of being marked to ROI region of providing of the present invention and binary conversion treatment
Flow chart.
In the present embodiment, when the lane line to ROI region is marked and image binaryzation is processed, from ROI region
First trip begins stepping through ROI image, calculates the width of lane line, the gray value of areas at both sides, the gray value of zone line respectively
And the gray scale difference sum of areas at both sides；When the gray value of the areas at both sides of lane line is respectively less than the gray value of zone line, and
And the gray scale difference sum of track line width and areas at both sides in the range of specified threshold when, the lane line in ROI region is entered
Line flag；Image binaryzation process is carried out to the ROI region after labelling.
Specifically, as shown in Fig. 2 step S21 can be realized in the following ways：
Step S211：ROI region is traveled through, calculate respectively the width of lane line, the gray value of areas at both sides, in
Between region gray value and the gray scale difference sum of areas at both sides.
Specifically, the corresponding feature of lane line is calculated respectively using below equation：
Diff_L=I (x, y)-I (x- δ, y) (1)
Diff_R=I (x, y)-I (x+ δ, y) (2)
Diff=diff_L+diff_R- | (I (x+ δ, y)-I (x- δ, y)) | (3)
Diff_Thresh=θ * I (x, y) (4)
Wherein, in formula, I (x, y) is gray value of the image at coordinate (x, y) place before labelling, and (x- δ are y) image before labelling to I
In coordinate (x- δ, y) gray value at place；Diff_L is gray value and left side certain distance of the image in coordinate (x, y) before labelling
The difference of gray value；(x+ δ, y) for image before labelling, in coordinate, (x+ δ, the y) gray value at place, diff_R are sitting I for image before labelling
The difference of the gray value of the gray value and right side certain distance of mark (x, y)；Diff represents before labelling image one on the left of the coordinate (x, y)
The gray scale difference of set a distance and the gray scale difference sum of right side certain distance；Diff_Thresh represents the threshold value of gray scale difference sum；θ is
Scaling coefficient, span are 0<θ<1 decimal, needs to obtain according to after substantial amounts of sample image data statistical computation,
Between general value 0.25～0.75, the present embodiment is preferably 0.5；δyThe track of the y rows being located for current pixel point (x, y)
Line width, in units of pixel, its accounting equation is：
Due to transparent effect, the lane line with one fixed width is presented near big and far smaller visual effect in the picture, therefore,
In formula (5), min and max is respectively the maximal and minmal value of whole piece track line width, and height is the height of image, with pixel
For unit, ε is Error Gain, and its main purpose is that, for the interference for reducing noise, general value is 5.Wherein minima min and
Maximum max needs to be adjusted according to the size of image and the fixed position of photographic head, preferably by min values in the present embodiment
For 0, max values 13.
Step S212：Judge whether current pixel meets flag condition one by one.
According to the wide constraint of lane line, the gray scale difference of each driveway line and non-lane line region in image is taken into full account
It is different, mark all possible lane line region.Only in lane line width range, the gray value of areas at both sides is less than simultaneously
The gray value of zone line, and region of the gray scale difference sum of areas at both sides in the range of specified threshold is just marked as possibility
Lane line region.That is, work as diff_L>0, and, diff_R>0, and, threshold values diff_ of the diff more than gray scale difference sum
Thresh, current pixel can be put under the region of lane line.
Step S213：Binary conversion treatment is carried out to ROI region：
Wherein, G (x, y) be labelling after image (x, y) place gray value.Step S214：Judge the institute of current ROI region
There is pixel whether to travel through to complete, if it is not, then return to step S211 is calculated.
In a kind of attainable mode, step S22, the company in the lane line region after processing to image binaryzation
When the general character is analyzed and screens, first the ROI region after image binaryzation process is scanned, is found out all labeled
Lane line connected region and calculate the area of each connected region；Then, the minimum of all connected regions is calculated successively
External envelope rectangular shape description and its geometric properties；Finally, the pact according to the geometric properties in the lane line region to target
Beam value, screens to the lane line of ROI region, and aperture closes the lane line region of restriction range, used as the lane line of target.
Specifically, referring to Fig. 3, it is that the connectedness to the lane line region after image binaryzation process that the present invention is provided is entered
Row analysis and screen one embodiment the step of flow chart.
Step S221：Calculate the area contourArea of the connected region of current labeled lane line.
Wherein, connected region again referred to as connected domain, refers to the picture by a collection of connection in bianry image in the present embodiment
The shape that vegetarian refreshments collection is constituted.
Step S222：By the area contourArea and default area of the connected region of current labeled lane line
Threshold value minSize is compared.If when the area contourArea of certain connected region is less than set threshold value minSize
When, then delete the point set in the region, otherwise execution step S223.Wherein, chi of the span of threshold value minSize according to image
It is very little to be adjusted.
In the present embodiment, the value of threshold value minSize is preferably the 0.00015 of the product of the height and the width of image
Times, some zonule effect of noise can be eliminated through this step.
Step S223：Detection obtains the minimum external envelope rectangle of connected region, and calculates its area minAreaRect.
In the present embodiment, shape descriptor of the minimum external envelope rectangle as lane line, certain in its expression scene
Object (lane line), in order to describe and recognize the object, needs by its shape of some geometric feature descriptions, and the geometric properties are just
It is shape descriptor.Existing shape descriptor includes but is not limited to minimum circumscribed circle, minimum external envelope rectangle etc..
Step S224：The geometric properties of the minimum external envelope rectangle of the connected region are calculated, including the minimum external envelope square
The length of shape, width and the anglec of rotation (i.e. direction).
Step S225：The length of the minimum external envelope rectangle of the connected region is compared with length threshold.Work as minimum
The length (bounding_length) of external envelope rectangle meets bounding_length>During longLane, directly retain the connection
The point set in region, wherein longLane are the threshold value of minimum external envelope rectangle length, and in the present embodiment, threshold value longLane value is
0.3 times of the high height of image.
Step S226：Detect the connected region minimum external envelope rectangle the anglec of rotation residing for angular range, when most
Anglec of rotation angle_deg of little external envelope rectangle meets -70 °<angle_deg<- 10 ° or 10 °<angle_deg<70°
When, further the length and width of minimum external envelope rectangle is detected by step S227.
Step S227：As length bounding_length and the width bounding_width of the minimum external envelope rectangle
Ratio r atio=bounding_length/bounding_width >=4 when, directly retain the connected region point set；If ratio
Outside value ratio=bounding_length/bounding_width >=2 and connected region area contourArea and minimum
When the ratio of the area minAreaRect of enclosure rectangle is more than 0.75, equally retain the connected region point set.The screening of this process
Both the angle requirement of lane line had been met, it is solid line, dotted line, the phenomenon of fracture to have taken into account lane line again.
Referring to Fig. 4, it is that the connectedness to the lane line region after image binaryzation process that the present invention is provided is analyzed
With the result schematic diagram of screening.Wherein, Fig. 4 a are lane lines by the binary image after first labelling；Fig. 4 b are lane lines
The area image of connected domain；Fig. 4 c are the minimum external envelope rectangular shapes of connected domain；Fig. 4 d are that ROI fare connected domains are sieved
The target lane line obtained after choosing, it can be seen that after implementing the screening of said process, the embodiment of the present invention accurately can retain few
The lane line information of amount.
In the present embodiment, after completing the track wire tag to present frame ROI region and screening, respectively to left and right track
The point set application RANSAC algorithms in line region carry out the lane line parameter extraction of straight line model.
RANSAC is the abbreviation of Random Sample Consensus (stochastic sampling is consistent), with the thought of RANSAC,
The mathematical model parameter of data can be calculated by the method for iteration according to one group of sample data set comprising abnormal data,
Obtain the algorithm of the uncertainty of effective sample data.
Referring to Fig. 5, it is the one embodiment for the employing stochastic sampling unification algorism acquisition lane line parameter that the present invention is provided
The step of flow chart.
In the present embodiment, the lane line region is sampled, using stochastic sampling unification algorism to gathered data
Fitting a straight line is carried out, lane line parameter is calculated, including：
Step S51：Central point with ROI region enters the affiliated point of driveway line from left and right both direction respectively as starting point
Collection is scanned；
Step S52：The point set obtained to scanning, carries out fitting a straight line using stochastic sampling unification algorism, extracts ROI areas
Lane line parameter in domain.
Wherein, the central point with ROI region enters the affiliated point of driveway line from left and right both direction respectively as starting point
Collection is scanned, including：
Scanning is with the abscissa at the image center place of ROI region as starting point；Often capable first for scanning to the left is grey
Angle value is the coordinate of 255 (white points i.e. in digital picture) as left-lane line point set P to be fittedl, and terminate row scanning；Often
First gray value scanning to the right of row be 255 coordinate as right-lane line point set P to be fittedr, and terminate row scanning；
Wherein, all coordinate gray values in the lane line region of the ROI region image after image binaryzation are 255, non-lane line area
All coordinate gray values in domain are 0 (stain i.e. in digital picture).It should be noted that the present invention can also adopt other
Gray value enters the fitting of driveway line, it is only necessary to make a distinction lane line region and non-lane line region.In this approach
To be all inner side point nearest away from vehicle on track, shake will not be produced because of track line width, and be avoided to complete
Figure is scanned, and improves efficiency.
Wherein, the S52 can be realized in the following ways, including：
Step S521：Institute's pointed set in the lane line region that scanning is obtained constitutes set P, and (i.e. left-lane line treats match point
Collection PlPoint set P to be fitted with right-lane liner)
Step S522：2 characteristic points are randomly selected in set P and constitutes set S, and initialize lane line straight line mould with S
Type y=krx+br, wherein, kr、brRespectively carry out initializing what is obtained according to the 2 feature point pairs straight line models randomly selected
Slope and intercept；
Step S523：Calculate complementary set ScIn point and lane line straight line model y=krx+brApart from d；Set ScIt is set
Complementary sets of the S in set P；
Step S524：Judging distance d and threshold tolerance diSize；
Step S525：Threshold tolerance d will be less than apart from diComplementary set ScIn corresponding point constitute set Q, set Q and set S structures
Into interior point set S*；
Step S526：By interior point set S*In point number with it is minimum in point quantity t be compared.If interior point set S*In
The number of point is interior more than minimum to put quantity t, then execution step S527；
Step S527：Using interior point set S*In point and using method of least square to lane line straight line model y=krx+
brIt is updated, and stores the model parameter；
If lane line straight line model y=krx+brIt is updated over, or, interior point set S*In the number of point be not more than
Point quantity t in minimum, then return to step S522, randomly selects again 2 characteristic points in set P and constitutes set S to lane line
Straight line model y=krx+brInitialized；
Step S528：Repeating above step S522～step S527 carries out multiple sampling；
Step S529：After multiple sampling is completed, the most interior point set S of interior point number is chosen*Characteristic point calculate
Lane line parameter is used as final lane line parameter.Specifically, after completing K sampling, choose point in the most S* of interior point number
The lane line parameter that set is calculated is used as final (left or right) lane line parameter.Wherein K is frequency in sampling, and its value mainly depends on
Dependency between 2 characteristic points randomly selected in step S522.It is general in the case of known prior probability, K is by counting
Calculate average statistical to obtain, in the present invention, K is preferably 50.
Method for detecting lane lines based on image vision provided in an embodiment of the present invention, according to default track line width with
Grey scale difference threshold value realizes the initial extraction in lane line region, maintains lane line region to greatest extent；To owning in image
Connected domain be analyzed, further lane line region is screened with reference to the geometric properties of lane line, effectively can filter
Except the interference in non-lane line region in image, the extraction of lane line feature is completed, be that the accurate extraction of lane line parameter is established
Basis；As the present invention has considered the gray feature and geometric properties of lane line, therefore either daytime or night be all
Lane line edge can accurately be split.Therefore, the technical scheme that the present invention is provided greatly reduces operand, improves meter
Speed is calculated, accurately high, strong robustness.
The above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (8)
1. a kind of method for detecting lane lines based on image vision, it is characterised in that include：
Gathering altimetric image to be checked, and treat detection image carries out pretreatment；
Enter driveway line feature extraction based on lane line geometric properties and gray difference, filter out lane line region；
The lane line region is sampled, fitting a straight line is carried out using stochastic sampling unification algorism to gathered data, calculate
Go out lane line parameter.
2. the method for detecting lane lines based on image vision as claimed in claim 1, it is characterised in that described to figure to be detected
As carrying out pretreatment, including：
ROI region is set on the altimetric image to be checked；The ROI region is included between vehicle front and road disappearance horizontal plane
Road image；
Gray processing process is carried out to the image of ROI region；
Median filter process is carried out to gray level image, removes the noise during image capturing and transmitting.
3. the method for detecting lane lines based on image vision as claimed in claim 2, it is characterised in that described based on lane line
Geometric properties and gray difference enter driveway line feature extraction, filter out lane line region, specifically include：
The lane line of the ROI region is marked with gray difference according to the wide constraint relation of lane line, and to labelling
ROI region afterwards carries out image binaryzation process；
The connectedness in the lane line region after image binaryzation process is analyzed, and according to the geometric properties in lane line region
Lane line is screened, the lane line region of target is obtained；The geometric properties include the direction of lane line, length, width
With connected domain area.
4. the method for detecting lane lines based on image vision as claimed in claim 3, it is characterised in that described according to lane line
Wide constraint relation the lane line of the ROI region is marked with gray difference, and the ROI region after labelling is carried out
Image binaryzation process, specifically includes：
ROI image is begun stepping through from ROI region first trip, the width of lane line, the gray value of areas at both sides, centre is calculated respectively
The gray scale difference sum of the gray value in region and areas at both sides；
When the gray value of the areas at both sides of lane line is respectively less than the gray value of zone line, and track line width and two border areas
When the gray scale difference sum in domain is in the range of specified threshold, the lane line in ROI region is marked；
Image binaryzation process is carried out to the ROI region after labelling.
5. the method for detecting lane lines based on image vision as claimed in claim 3, it is characterised in that described to image two-value
The connectedness in the lane line region after change process is analyzed, and lane line is sieved according to the geometric properties in lane line region
Choosing, obtains the lane line region of target, specifically includes：
ROI region after image binaryzation process is scanned, the connected region of all labeled lane lines is found out simultaneously
Calculate the area of each connected region；
Minimum external envelope rectangular shape description and its geometric properties of all connected regions is calculated successively；
According to the binding occurrence of the geometric properties in the lane line region to target, the lane line of ROI region is screened, aperture
Close the lane line region of restriction range.
6. the method for detecting lane lines based on image vision as claimed in claim 2, it is characterised in that to the lane line area
Domain is sampled, and carries out fitting a straight line using stochastic sampling unification algorism to gathered data, calculates lane line parameter, including：
Central point with ROI region enters the affiliated point set of driveway line from left and right both direction respectively and is scanned as starting point；
The point set obtained to scanning, carries out fitting a straight line using stochastic sampling unification algorism, extracts the lane line in ROI region
Parameter.
7. the method for detecting lane lines based on image vision as claimed in claim 6, it is characterised in that described with ROI region
Central point be starting point, enter the affiliated point set of driveway line from left and right both direction respectively and be scanned, including：
Scanning is with the abscissa at the image center place of ROI region as starting point；
Often first gray value scanning to the left of row be 255 coordinate as left-lane line point set to be fitted, and terminate the row
Scanning；
Often first gray value scanning to the right of row be 255 coordinate as right-lane line point set to be fitted, and terminate the row
Scanning；
Wherein, all coordinate gray values in the lane line region of the ROI region image after image binaryzation are 255, non-track
All coordinate gray values in line region are 0.
8. the method for detecting lane lines based on image vision as claimed in claim 6, it is characterised in that described pair of scanning is obtained
Point set, fitting a straight line is carried out using stochastic sampling unification algorism, the lane line parameter in ROI region is extracted, including：
Institute's pointed set in the lane line region that scanning is obtained constitutes set P, randomly selects 2 characteristic points and constitute in set P
Set S, and lane line straight line model y=k is initialized with Srx+br, wherein, kr、brRespectively according to 2 features randomly selected
Point carries out initializing the slope and the intercept that obtain to straight line model；
Calculate complementary set ScIn point and lane line straight line model y=krx+brApart from d；Set ScSet S in set P more than
Collection；
Threshold tolerance d will be less than apart from diComplementary set ScIn corresponding point constitute the interior point set S of set Q, set Q and set S composition*；
If interior point set S*In point number more than point quantity t in minimum, then using interior point set S*In point and adopt least square
Method is to lane line straight line model y=krx+brIt is updated, and stores the model parameter；
If lane line straight line model y=krx+brIt is updated over, or, interior point set S*In point number be not more than minimum
Interior quantity t, then in set P randomly select again 2 characteristic points and constitute set S to lane line straight line model y=krx+brEnter
Row initialization, repeating above step carries out multiple sampling；
After multiple sampling is completed, the most interior point set S of interior point number is chosen*The lane line parameter conduct that calculates of characteristic point
Final lane line parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611102913.XA CN106529505A (en) | 2016-12-05 | 2016-12-05 | Image-vision-based lane line detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611102913.XA CN106529505A (en) | 2016-12-05 | 2016-12-05 | Image-vision-based lane line detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106529505A true CN106529505A (en) | 2017-03-22 |
Family
ID=58354871
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611102913.XA Pending CN106529505A (en) | 2016-12-05 | 2016-12-05 | Image-vision-based lane line detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106529505A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107563326A (en) * | 2017-08-31 | 2018-01-09 | 京东方科技集团股份有限公司 | One kind driving householder method, driving assistant device and vehicle |
| CN107832732A (en) * | 2017-11-24 | 2018-03-23 | 河南理工大学 | Method for detecting lane lines based on ternary tree traversal |
| CN107944388A (en) * | 2017-11-24 | 2018-04-20 | 海信集团有限公司 | A kind of method for detecting lane lines, device and terminal |
| CN108734105A (en) * | 2018-04-20 | 2018-11-02 | 东软集团股份有限公司 | Method for detecting lane lines, device, storage medium and electronic equipment |
| CN108805065A (en) * | 2018-05-31 | 2018-11-13 | 华南理工大学 | One kind being based on the improved method for detecting lane lines of geometric properties |
| CN108875607A (en) * | 2017-09-29 | 2018-11-23 | 惠州华阳通用电子有限公司 | Method for detecting lane lines, device and computer readable storage medium |
| 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 |
| CN108875589A (en) * | 2018-05-25 | 2018-11-23 | 武汉东智科技股份有限公司 | A kind of video detecting method of road area |
| CN109359632A (en) * | 2018-11-30 | 2019-02-19 | 重庆昂布瑞拉科技有限公司 | Highway sideline detection method and device |
| CN109670455A (en) * | 2018-12-21 | 2019-04-23 | 联创汽车电子有限公司 | Computer vision lane detection system and its detection method |
| CN111079541A (en) * | 2019-11-19 | 2020-04-28 | 重庆大学 | A monocular vision-based road stop line detection method |
| CN111220786A (en) * | 2020-03-09 | 2020-06-02 | 生态环境部华南环境科学研究所 | Method for rapidly monitoring organic pollution of deep water sediments |
| CN111241894A (en) * | 2018-11-28 | 2020-06-05 | 海信集团有限公司 | Method for detecting lane line and vehicle-mounted equipment |
| CN111860113A (en) * | 2020-06-01 | 2020-10-30 | 安徽奇点智能新能源汽车有限公司 | Lane line detection method and system |
| CN112597846A (en) * | 2020-12-14 | 2021-04-02 | 合肥英睿系统技术有限公司 | Lane line detection method, lane line detection device, computer device, and storage medium |
| US20210374436A1 (en) * | 2020-11-10 | 2021-12-02 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Image Processing Method, Apparatus, and Electronic Device |
| CN115314634A (en) * | 2022-06-28 | 2022-11-08 | 上海艾为电子技术股份有限公司 | Anti-shake detection method and device, terminal equipment and readable storage medium |
Citations (7)
| 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 |
| CN104318258A (en) * | 2014-09-29 | 2015-01-28 | 南京邮电大学 | Time domain fuzzy and kalman filter-based lane detection method |
| CN104573627A (en) * | 2014-11-19 | 2015-04-29 | 湖南大学 | Lane line reservation and detection algorithm based on binary image |
| CN105046235A (en) * | 2015-08-03 | 2015-11-11 | 百度在线网络技术（北京）有限公司 | Lane line recognition modeling method and apparatus and recognition method and apparatus |
| CN105261020A (en) * | 2015-10-16 | 2016-01-20 | 桂林电子科技大学 | Method for detecting fast lane line |
| CN105320927A (en) * | 2015-03-25 | 2016-02-10 | 中科院微电子研究所昆山分所 | Lane line detection method and system |
| CN105488492A (en) * | 2015-12-25 | 2016-04-13 | 北京大学深圳研究生院 | Color image preprocessing method, road identification method and related device |
-
2016
- 2016-12-05 CN CN201611102913.XA patent/CN106529505A/en active Pending
Patent Citations (7)
| 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 |
| CN104318258A (en) * | 2014-09-29 | 2015-01-28 | 南京邮电大学 | Time domain fuzzy and kalman filter-based lane detection method |
| CN104573627A (en) * | 2014-11-19 | 2015-04-29 | 湖南大学 | Lane line reservation and detection algorithm based on binary image |
| CN105320927A (en) * | 2015-03-25 | 2016-02-10 | 中科院微电子研究所昆山分所 | Lane line detection method and system |
| CN105046235A (en) * | 2015-08-03 | 2015-11-11 | 百度在线网络技术（北京）有限公司 | Lane line recognition modeling method and apparatus and recognition method and apparatus |
| CN105261020A (en) * | 2015-10-16 | 2016-01-20 | 桂林电子科技大学 | Method for detecting fast lane line |
| CN105488492A (en) * | 2015-12-25 | 2016-04-13 | 北京大学深圳研究生院 | Color image preprocessing method, road identification method and related device |
Non-Patent Citations (4)
| Title |
|---|
| DU, XINXIN 等: "vision-based approach towards lane line detection and vehicle localization", 《MACHINE VISION AND APPLICATIONS》 * |
| 西格沃特等: "《自主移动机器人导论 第2版》", 31 May 2013 * |
| 鞠乾翱 等: "基于机器视觉的快速车道线识别", 《计算机应用研究》 * |
| 龚声蓉等: "《数字图像处理与分析 第2版》", 31 May 2014 * |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN107563326A (en) * | 2017-08-31 | 2018-01-09 | 京东方科技集团股份有限公司 | One kind driving householder method, driving assistant device and vehicle |
| US10906538B2 (en) | 2017-08-31 | 2021-02-02 | Boe Technology Group Co., Ltd. | Driving assistance method, driving assistance device, and vehicle having driving assistance device |
| CN108875607A (en) * | 2017-09-29 | 2018-11-23 | 惠州华阳通用电子有限公司 | Method for detecting lane lines, device and computer readable storage medium |
| CN107832732A (en) * | 2017-11-24 | 2018-03-23 | 河南理工大学 | Method for detecting lane lines based on ternary tree traversal |
| CN107944388A (en) * | 2017-11-24 | 2018-04-20 | 海信集团有限公司 | A kind of method for detecting lane lines, device and terminal |
| CN107832732B (en) * | 2017-11-24 | 2021-02-26 | 河南理工大学 | Lane line detection method based on ternary tree traversal |
| CN108734105A (en) * | 2018-04-20 | 2018-11-02 | 东软集团股份有限公司 | Method for detecting lane lines, device, storage medium and electronic equipment |
| CN108875589A (en) * | 2018-05-25 | 2018-11-23 | 武汉东智科技股份有限公司 | A kind of video detecting method of road area |
| CN108805065A (en) * | 2018-05-31 | 2018-11-13 | 华南理工大学 | One kind being based on the improved method for detecting lane lines of geometric properties |
| CN111241894A (en) * | 2018-11-28 | 2020-06-05 | 海信集团有限公司 | Method for detecting lane line and vehicle-mounted equipment |
| CN111241894B (en) * | 2018-11-28 | 2023-06-27 | 海信集团有限公司 | Method for detecting lane line and vehicle-mounted equipment |
| CN109359632A (en) * | 2018-11-30 | 2019-02-19 | 重庆昂布瑞拉科技有限公司 | Highway sideline detection method and device |
| CN109359632B (en) * | 2018-11-30 | 2021-08-10 | 重庆子粒科技有限公司 | Road sideline detection method and device |
| CN109670455A (en) * | 2018-12-21 | 2019-04-23 | 联创汽车电子有限公司 | Computer vision lane detection system and its detection method |
| CN111079541A (en) * | 2019-11-19 | 2020-04-28 | 重庆大学 | A monocular vision-based road stop line detection method |
| CN111220786A (en) * | 2020-03-09 | 2020-06-02 | 生态环境部华南环境科学研究所 | Method for rapidly monitoring organic pollution of deep water sediments |
| CN111860113A (en) * | 2020-06-01 | 2020-10-30 | 安徽奇点智能新能源汽车有限公司 | Lane line detection method and system |
| US20210374436A1 (en) * | 2020-11-10 | 2021-12-02 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Image Processing Method, Apparatus, and Electronic Device |
| US12175719B2 (en) * | 2020-11-10 | 2024-12-24 | Beijing Baidu Netcom Science Technology Co., Ltd. | Image processing method, apparatus, and electronic device |
| CN112597846A (en) * | 2020-12-14 | 2021-04-02 | 合肥英睿系统技术有限公司 | Lane line detection method, lane line detection device, computer device, and storage medium |
| CN112597846B (en) * | 2020-12-14 | 2022-11-11 | 合肥英睿系统技术有限公司 | Lane line detection method, lane line detection device, computer device, and storage medium |
| CN115314634A (en) * | 2022-06-28 | 2022-11-08 | 上海艾为电子技术股份有限公司 | Anti-shake detection method and device, terminal equipment and readable storage medium |
| CN115314634B (en) * | 2022-06-28 | 2024-05-31 | 上海艾为电子技术股份有限公司 | Anti-shake detection method, device, terminal equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106529505A (en) | Image-vision-based lane line detection method | |
| CN112215306B (en) | Target detection method based on fusion of monocular vision and millimeter wave radar | |
| Aly | Real time detection of lane markers in urban streets | |
| CN108038416B (en) | Lane line detection method and system | |
| Kastrinaki et al. | A survey of video processing techniques for traffic applications | |
| CN105678285B (en) | A kind of adaptive road birds-eye view transform method and road track detection method | |
| CN105160309B (en) | Three lanes detection method based on morphological image segmentation and region growing | |
| Nieto et al. | Road environment modeling using robust perspective analysis and recursive Bayesian segmentation | |
| CN103065138B (en) | Recognition method of license plate number of motor vehicle | |
| Mu et al. | Lane detection based on object segmentation and piecewise fitting | |
| JP5867596B2 (en) | Three-dimensional object detection apparatus and three-dimensional object detection method | |
| CN109684921A (en) | A kind of road edge identification and tracking based on three-dimensional laser radar | |
| CN105809138A (en) | Road warning mark detection and recognition method based on block recognition | |
| CN107066986A (en) | A kind of lane line based on monocular vision and preceding object object detecting method | |
| CN102314599A (en) | Identification and deviation-detection method for lane | |
| CN105260712A (en) | Method and system for detecting pedestrians in front of a vehicle | |
| CN107895151A (en) | Method for detecting lane lines based on machine vision under a kind of high light conditions | |
| CN102682455A (en) | Front vehicle detection method based on monocular vision | |
| CN109886175B (en) | Method for detecting lane line by combining straight line and circular arc | |
| CN102419820A (en) | Method for rapidly detecting automobile logo in video and image | |
| CN108805050B (en) | Wire Detection Method Based on Local Binary Pattern | |
| KR20110001427A (en) | Lane Fast Detection Method by Extracting Region of Interest | |
| CN113516853A (en) | Multi-lane traffic flow detection method for complex monitoring scene | |
| CN104063882A (en) | Vehicle video speed measuring method based on binocular camera | |
| Wang et al. | Lane detection based on two-stage noise features filtering and clustering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170322 |