CN103902985A - High-robustness real-time lane detection algorithm based on ROI - Google Patents

High-robustness real-time lane detection algorithm based on ROI Download PDF

Info

Publication number
CN103902985A
CN103902985A CN201410148832.8A CN201410148832A CN103902985A CN 103902985 A CN103902985 A CN 103902985A CN 201410148832 A CN201410148832 A CN 201410148832A CN 103902985 A CN103902985 A CN 103902985A
Authority
CN
China
Prior art keywords
image
lane line
real
track
roi
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201410148832.8A
Other languages
Chinese (zh)
Other versions
CN103902985B (en
Inventor
陈孟元
柴灿
郎朗
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Anhui Polytechnic University
Original Assignee
Anhui Polytechnic University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Anhui Polytechnic University filed Critical Anhui Polytechnic University
Priority to CN201410148832.8A priority Critical patent/CN103902985B/en
Publication of CN103902985A publication Critical patent/CN103902985A/en
Application granted granted Critical
Publication of CN103902985B publication Critical patent/CN103902985B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention provides a high-robustness real-time lane detection algorithm based on ROI. In order to achieve curve detection, an image is divided into an upper portion and a lower portion, namely a close shot area and a long shot area, and solutions are obtained through Hough and a hyperbolic curve pair model respectively. The whole image is mainly filtered through a transverse gradient operator so that the calculation speed can be increased and the purpose of real-time detection can be achieved; an area of gradient direction angles of a lane boundary image is counted through a sliding ROI window strategy, boundary noise of the abnormal gradient direction angles is eliminated, and therefore the accuracy of lane detection is guaranteed; in the hyperbolic curve pair model adopted in the long shot area, parameters in a close shot model are mainly used as initial parameters, a parameter K is finally determined through a search strategy, and the curve portion is detected. The high-robustness real-time lane detection algorithm is good in robustness in the complex road condition and the complex environment and lane counseling information can be provided in real time.

Description

The real-time track detecting of a kind of strong robustness based on ROI algorithm
Technical field
The present invention relates to driving and use safety monitoring field, be specifically related to the real-time track detecting of a kind of strong robustness based on ROI algorithm.
Background technology
Recently, what ITS became comes into vogue, because people more pay close attention to vehicle safety aspect.There are much research topics based on vision about ITS, comprise that obstacle detection, pedestrian dodge, lane line departure warning, anticollision etc.In these challenging tasks, lane detection is one of most important part of ITS.The main body of this technology is from complex environment, to isolate lane line consulting by some characteristics; But most lane detection technology having existed is very sensitive on the impact of inclement weather and woods shade.Lane detection is an inner important component part of intelligent transportation system (ITS), has proposed a kind of bend detection system under complex environment, for example: on tree shade, complex road condition, road, have road sign.
Current research has realized the detection of lane line by different technology, for example: the algorithm based on color, the algorithm based on edge, the algorithm based on histogram, the algorithm based on visual angle conversion, the algorithm based on model.But the algorithm based on color indicates with only having used an overall threshold value or specific colouring information to remove to extract lane line, but this may receive impact in having shade or bad weather.For the algorithm based on edge, difficult point is the edge that abates the noise.Discuss whole problem in detail although there is document to propose algorithm, also deposited a lot of not enough.For the algorithm based on column, they only can not have could move in other vehicle and road sign thereof before car.Based on the algorithm of visual angle conversion, difficult point is to correct camera, and this algorithm is on slope and inefficent have wind in the situation that.Algorithm based on model replaces lane line with model, and it has better robustness when the consulting of more and lane line disappears for noise.In order to receive more reliable model parameter, apply better noise and reject algorithm.
Summary of the invention
In order to address the above problem, the invention provides the real-time track detecting of a kind of strong robustness based on ROI algorithm, first detect first image to extract adaptive threshold, use subsequently with the transverse gradients operator of adaptive threshold and lane line information is filtered and use the fuzzy strategy of searching to repair lane line information, consider Hough operand, need to get rid of lane line noise and redundancy information, adopt the method based on slip ROI and image association to extract effective wagon diatom; Subsequently, in close shot region, adopt the Hough conversion method of fast zoom table that straight way is carried out search arithmetic and shown; In distant view region, use hyperbolic curve to calculate hyperbolic curve to parameter to model the result of accepting Hough computing, adopt search strategy to determine K parameter values simultaneously, simulate bend track, last output display entirety track.Can there is the accurate lane line information of extracting under the complex situations such as woods shade, other track mark, other vehicle in this method.
In order to achieve the above object, the present invention is achieved by the following technical solutions:
The invention provides the real-time track of a kind of strong robustness based on ROI detecting algorithm, described algorithm specifically comprises the following steps, and step 1, first detects first image to extract initial adaptive threshold;
Step 2 is also set up coordinate system on the image of picked-up, and the upper left corner is initial point, is laterally x axle, is longitudinally y axle; Part near image bottom is called close shot region, away from part be called distant view region;
Step 3, employing are carried out the filtration of lane line information with the transverse gradients operator of adaptive threshold to image greyscale figure, first carry out track information extraction: before carrying out track rim detection, convert RGB image to GTG figure, the formula that changes into GTG figure from RGB model is: ; Wherein R, G, B is the component of numeric distribution in [0,255]; Then adopt the mode of the shade of transverse gradients operator to carry out lane line detecting, threshold value adopts adaptive threshold, and its shade is:
Figure 288977DEST_PATH_IMAGE002
Step 4, offer ROI and extract effective track information:
The Grad of step 4.1 image and direction indication thereof are:
Figure 344658DEST_PATH_IMAGE003
Figure 98987DEST_PATH_IMAGE004
Wherein,
Figure 898316DEST_PATH_IMAGE005
with
Figure 481089DEST_PATH_IMAGE006
the gradient of representative laterally and longitudinally respectively, its computing shade is as follows:
Figure 910933DEST_PATH_IMAGE007
Figure 214875DEST_PATH_IMAGE008
Step 4.2 will
Figure 552316DEST_PATH_IMAGE009
scope is set to
Figure 258104DEST_PATH_IMAGE010
, and be quantified as 180 angles, each angle is
Figure 593270DEST_PATH_IMAGE011
, calculated direction angle
Figure 384508DEST_PATH_IMAGE009
if, do not exist
Figure 23617DEST_PATH_IMAGE012
in, seeked advice from and eliminated as boom car diatom;
The pixel that step 4.3 utilization retains carries out preliminary hough conversion, preliminary definite lane line information, and several is set has a window of the slip of certain ROI scope along preliminary lane line information, and to the execution step of moving window one by one 4.2, analyze that to get rid of may be the isolated noise beyond lane line;
Step 4.4 is since second image, adopt lane line consulting to open the little characteristic of image meta amount of movement in front and back, this image and front several images are compared, extract the most effective lane line information to reduce the operand of Hough conversion in the track matching of follow-up straight line;
Step 5, straight way and bend matching:
(1) straight way matching
Adopt Hough to carry out matching to straight line track in image close shot region, its formula is as follows:
Figure 857581DEST_PATH_IMAGE013
Matching adopts Hough conversion method, and concrete steps are as follows:
A) model ,
Figure 753042DEST_PATH_IMAGE015
each angle and
Figure 170772DEST_PATH_IMAGE016
database table, search for Hough conversion, to reduce operand;
B) obtain lane line information figure E and data set S, , wherein m is lane line information pixel number;
C) arrange
Figure 613572DEST_PATH_IMAGE018
as initial search point, as d) calculation of parameter, and counter is set, calculate next point by row subsequently
Figure 96506DEST_PATH_IMAGE019
;
If d) current
Figure 631392DEST_PATH_IMAGE020
meet
Figure 744842DEST_PATH_IMAGE021
,
Figure 732390DEST_PATH_IMAGE022
, arrange , until
Figure 80511DEST_PATH_IMAGE024
;
E) retain maximum
Figure 427179DEST_PATH_IMAGE025
and E, fitting a straight line is complete;
(2) bend matching
In image distant view region, adopt hyperbolic curve to model, its mathematic(al) representation is:
Figure 902023DEST_PATH_IMAGE026
Wherein, for the ordinate of local horizon in picture plane, K, with for bend parameter;
According to the standard form of Hough space cathetus
Figure 408558DEST_PATH_IMAGE030
have:
Figure 974669DEST_PATH_IMAGE031
Contrast can draw:
Figure 338654DEST_PATH_IMAGE032
,
Figure 761545DEST_PATH_IMAGE033
,
Figure 148664DEST_PATH_IMAGE034
,
Figure 580783DEST_PATH_IMAGE035
Each parameter can calculate by straight line Hough conversion portion, arranges
Figure 533695DEST_PATH_IMAGE036
Figure 330750DEST_PATH_IMAGE037
if, matched curve point
Figure 1903DEST_PATH_IMAGE038
with moving window central point meet
Figure 251323DEST_PATH_IMAGE040
(unit is pixel),
Figure 281596DEST_PATH_IMAGE041
;
(3) export entirety lane line in real time: by step (1) and step (2) gained straight line track and the combination of bending track demonstration, the real-time lane line of output entirety;
Step 6, calculate each moving window parameter and obtain adaptive threshold: since second image, offer ROI scope according to moving window as image subregion, and with this scope, the GTG after changing is desired to make money or profit and carried out adaptive threshold extraction by the method for Otsu, equation expression is:
Figure 440045DEST_PATH_IMAGE042
Wherein t is required adaptive threshold,
Figure 151649DEST_PATH_IMAGE043
be i background ratio row,
Figure 79153DEST_PATH_IMAGE044
be i background mean value, be i object ratio, be i object average,
Figure 238105DEST_PATH_IMAGE047
be i image subregion image average, i is moving window number, and parameter calculation formula is as follows:
Figure 223379DEST_PATH_IMAGE048
Figure 595454DEST_PATH_IMAGE049
Figure 725566DEST_PATH_IMAGE050
Figure 778972DEST_PATH_IMAGE051
Figure 681069DEST_PATH_IMAGE052
Figure 224046DEST_PATH_IMAGE053
Wherein,
Figure 578804DEST_PATH_IMAGE054
be the ratio of the pixel that in i moving window, GTG numerical value is k,
Figure 498219DEST_PATH_IMAGE055
be i the pixel number that moving window GTG numerical value is k,
Figure 192505DEST_PATH_IMAGE056
it is the total number of pixel in i moving window;
Threshold operation is:
Figure 906383DEST_PATH_IMAGE057
Wherein, Bin(x, y) be imaging point, f(x, y) be the respective coordinates of original pixel;
Step 7, since second image, first every image performs step six, then performs step three to step 5, exports in real time overall lane line.
Further improvement of the present invention is: described algorithm also comprises: in step 3, to using the lane line information after transverse gradients operator filters to search for repair strategy generally, search for the shade that repair strategy adopts generally and be:
Figure 14016DEST_PATH_IMAGE058
Taking current image lane line information data set as , last is
Figure 85582DEST_PATH_IMAGE060
if A, B, C are non-vanishing, D, E wherein have one non-vanishing and
Figure 173624DEST_PATH_IMAGE061
upgrade
Figure 502974DEST_PATH_IMAGE059
.
As the real-time track detecting of a kind of strong robustness based on ROI algorithm, the present invention has following beneficial effects:
1. realized the real-time detecting of the lane line in embedded system, operand has had significant raising than current most algorithms.
2. robust performance is strong, for example, at complicated road environment: have extremely strong stability in the situation such as tree shade, road sign, wagon flow are many, bend, and also have the same new energy of stablizing under weather environment at rainy day and night etc.
3. algorithm is not limited to current main flow algorithm, adopts the algorithm less to processor load, and the greatly lower energy consumption of processor is simplified and stable whole system.
In real time straight bend detection algorithm under complex environment provided by the invention has good robustness in complicated road conditions and environment, and track information can be provided in real time.
Brief description of the drawings
Fig. 1 entire system algorithm flow chart.
Fig. 2 adaptive thresholding algorithm process flow diagram.
Fig. 3 carries the Embedded System Structure figure of algorithm.
Embodiment
In order to deepen the understanding of the present invention, below in conjunction with drawings and Examples, the present invention is described in further detail, this embodiment, only for explaining the present invention, does not form and limits protection scope of the present invention.
As Figure 1-3, the present embodiment provides the real-time track detecting of a kind of strong robustness based on ROI algorithm, and described algorithm specifically comprises the following steps:
Step 1, first detect first image to extract initial adaptive threshold;
Step 2 is also set up coordinate system on the image of picked-up, and the upper left corner is initial point, is laterally x axle, is longitudinally y axle; Part near image bottom is called close shot region, away from part be called distant view region; Because lane line in real process only there will be in certain regional extent, at this, we only process a part for image, to reduce operand.In close shot region, in step subsequently, can carry out Hough search arithmetic, and draw straight line track; In distant view region, can and search for K parameter values with reference to the parameter of Hough computing, to draw bend track.
Step 3, employing are carried out the filtration of lane line information with the transverse gradients operator of adaptive threshold to image greyscale figure, before carrying out track rim detection, need first to convert RGB image to GTG figure, traffic lane line is white, yellow or red normally, so that with road surface forms a sharp contrast, in order to preserve this characteristic of traffic lane line, the employing formula that changes into GTG figure from RGB model is:
Figure 764191DEST_PATH_IMAGE001
; Wherein R, G, B is the component of numeric distribution in [0,255];
The performance of rim detection is important for system, and algorithm must have robustness under atrocious weather and poor environment, potential lane line message can be remained; Simultaneously because this algorithm is to be used in embedded system, so algorithm can not take too much operand, in order to avoid real-time is exerted an influence.The reason that does not adopt canny algorithm is that its load to processor is excessive, adopts and claims the mode of the shade that becomes transverse gradients operator to carry out lane line detecting at this, and threshold value adopts adaptive threshold, and its shade is:
Figure 229808DEST_PATH_IMAGE002
For obtaining the lane line consulting of maximum, improve system robustness, to using the lane line information after transverse gradients operator filters to search for repair strategy generally, search for the shade that repair strategy adopts generally and be:
Figure 551067DEST_PATH_IMAGE058
Taking current image lane line information data set as
Figure 367714DEST_PATH_IMAGE059
, last is
Figure 104726DEST_PATH_IMAGE060
if A, B, C are non-vanishing, D, E wherein have one non-vanishing and
Figure 690428DEST_PATH_IMAGE061
upgrade
Figure 182589DEST_PATH_IMAGE059
.
Step 4, offer ROI and extract effective track information:
The Grad of step 4.1 image and direction indication thereof are:
Figure 158635DEST_PATH_IMAGE003
Figure 758725DEST_PATH_IMAGE004
Wherein,
Figure 464513DEST_PATH_IMAGE005
with
Figure 861996DEST_PATH_IMAGE006
the gradient of representative laterally and longitudinally respectively, its computing shade is as follows:
Figure 732049DEST_PATH_IMAGE008
Step 4.2 will
Figure 354660DEST_PATH_IMAGE009
scope is set to , and be quantified as 180 angles, each angle is , calculated direction angle
Figure 84085DEST_PATH_IMAGE009
if,
Figure 501815DEST_PATH_IMAGE009
do not exist in, seeked advice from and eliminated as boom car diatom;
The pixel that step 4.3 utilization retains carries out preliminary hough conversion, preliminary definite lane line information, and several is set has a window of the slip of certain ROI scope along preliminary lane line information, and to the execution step of moving window one by one 4.2, analyze that to get rid of may be the isolated noise beyond lane line;
Step 4.4 is since second image, adopt lane line consulting to open the little characteristic of image meta amount of movement in front and back, this image and front several images are compared, extract the most effective lane line information to reduce the operand of Hough conversion in the track matching of follow-up straight line;
Step 5, straight way and bend matching:
(1) straight way matching
Adopt Hough to carry out matching to straight line track in image close shot region, its formula is as follows:
Figure 944615DEST_PATH_IMAGE013
Matching adopts Hough conversion method, and concrete steps are as follows:
A) model
Figure 427549DEST_PATH_IMAGE014
,
Figure 962436DEST_PATH_IMAGE015
each angle and database table, search for Hough conversion, to reduce operand;
B) obtain lane line information figure E and data set S,
Figure 860170DEST_PATH_IMAGE017
, wherein m is lane line information pixel number;
C) arrange as initial search point, as d) calculation of parameter, and counter is set, calculate next point by row subsequently
Figure 473871DEST_PATH_IMAGE019
;
If d) current
Figure 820539DEST_PATH_IMAGE062
meet
Figure 233066DEST_PATH_IMAGE021
,
Figure 54873DEST_PATH_IMAGE022
, arrange
Figure 298772DEST_PATH_IMAGE023
, until
Figure 754024DEST_PATH_IMAGE024
;
E) retain maximum
Figure 716164DEST_PATH_IMAGE025
and E, fitting a straight line is complete;
(2) bend matching
In image distant view region, adopt hyperbolic curve to model, its mathematic(al) representation is:
Figure 344592DEST_PATH_IMAGE026
Wherein,
Figure 442998DEST_PATH_IMAGE027
for the ordinate of local horizon in picture plane, K,
Figure 69151DEST_PATH_IMAGE028
with for bend parameter;
According to the standard form of Hough space cathetus have:
Figure 841301DEST_PATH_IMAGE031
Contrast can draw:
Figure 700673DEST_PATH_IMAGE032
,
Figure 374755DEST_PATH_IMAGE033
,
Figure 548247DEST_PATH_IMAGE034
,
Figure 621246DEST_PATH_IMAGE035
Each parameter can calculate by straight line Hough conversion portion, arranges
Figure 651518DEST_PATH_IMAGE036
Figure 809967DEST_PATH_IMAGE037
if, matched curve point
Figure 318309DEST_PATH_IMAGE038
with moving window central point
Figure 449076DEST_PATH_IMAGE039
meet
Figure 650250DEST_PATH_IMAGE040
(unit is pixel),
Figure 295995DEST_PATH_IMAGE041
, reduce the searching times to curve with this, and make the matched curve actual lane line information point of better fitting;
(3) export entirety lane line in real time: by step (1) and step (2) gained straight line track and the combination of bending track demonstration, the real-time lane line of output entirety;
Step 6, calculate each moving window parameter and obtain adaptive threshold: since second image, offer ROI scope according to moving window as image subregion, and with this scope, the GTG after changing is desired to make money or profit and carried out adaptive threshold extraction by the method for Otsu, equation expression is:
Figure 545711DEST_PATH_IMAGE042
Wherein t is required adaptive threshold,
Figure 593301DEST_PATH_IMAGE043
be i background ratio row,
Figure 985885DEST_PATH_IMAGE044
be i background mean value,
Figure 243560DEST_PATH_IMAGE045
be i object ratio, be i object average,
Figure 323697DEST_PATH_IMAGE047
be i image subregion image average, i is moving window number, and parameter calculation formula is as follows:
Figure 804357DEST_PATH_IMAGE048
Figure 159115DEST_PATH_IMAGE049
Figure 775746DEST_PATH_IMAGE051
Figure 489624DEST_PATH_IMAGE052
Figure 597257DEST_PATH_IMAGE053
Wherein,
Figure 992466DEST_PATH_IMAGE054
be the ratio of the pixel that in i moving window, GTG numerical value is k,
Figure 603576DEST_PATH_IMAGE055
be i the pixel number that moving window GTG numerical value is k,
Figure 753935DEST_PATH_IMAGE056
it is the total number of pixel in i moving window;
Threshold operation is:
Figure 20968DEST_PATH_IMAGE057
Wherein, Bin(x, y) be imaging point, f(x, y) be the respective coordinates of original pixel;
Step 7, since second image, first every image performs step six, then performs step three to step 5, exports in real time overall lane line.
The present embodiment is a kind of in real time straight bend detection method under complex environment, first adopt with the transverse gradients operator of adaptive threshold image greyscale figure is carried out to the filtration of lane line information, and adopt the lane line consulting of searching for generally after strategy repairing is filtered, improve the accuracy of system robustness and lane detection; Then adopt slip ROI window policy statistics lane boundary information figure gradient direction angle, the border noise information of rejecting abnormalities gradient direction angle, an associated front and back image lane line information, reduces redundancy lane line border information, to reduce Hough translation operation amount simultaneously.Straight line track part proposes a kind of Hough conversion method of fast zoom table, detects for straight line; Bend part adopts hyperbolic curve to model, and its initial parameter value is undertaken in straight line track part, and finally determines its K parameter by search strategy, realizes the detecting of bend part.This algorithm has good robustness in complicated road conditions and environment, and track information can be provided in real time.

Claims (2)

1. the real-time track detecting of the strong robustness based on a ROI algorithm, is characterized in that: described algorithm specifically comprises the following steps:
Step 1, first detect first image to extract initial adaptive threshold;
Step 2 is also set up coordinate system on the image of picked-up, and the upper left corner is initial point, is laterally x axle, is longitudinally y axle; Part near image bottom is called close shot region, away from part be called distant view region;
Step 3, employing are carried out the filtration of lane line information with the transverse gradients operator of adaptive threshold to image greyscale figure, first carry out track information extraction: before carrying out track rim detection, convert RGB image to GTG figure, the formula that changes into GTG figure from RGB model is:
Figure 88677DEST_PATH_IMAGE001
; Wherein R, G, B is the component of numeric distribution in [0,255]; Then adopt the mode of the shade of transverse gradients operator to carry out lane line detecting, threshold value adopts adaptive threshold, and its shade is:
Figure 819873DEST_PATH_IMAGE002
Step 4, offer ROI and extract effective track information:
The Grad of step 4.1 image and direction indication thereof are:
Wherein,
Figure 691861DEST_PATH_IMAGE005
with
Figure 215246DEST_PATH_IMAGE006
the gradient of representative laterally and longitudinally respectively, its computing shade is as follows:
Figure 441828DEST_PATH_IMAGE007
Figure 745771DEST_PATH_IMAGE008
Step 4.2 will
Figure 348790DEST_PATH_IMAGE009
scope is set to
Figure 788999DEST_PATH_IMAGE010
, and be quantified as 180 angles, each angle is
Figure 452061DEST_PATH_IMAGE011
, calculated direction angle
Figure 915404DEST_PATH_IMAGE009
if, do not exist in, seeked advice from and eliminated as boom car diatom;
The pixel that step 4.3 utilization retains carries out preliminary hough conversion, preliminary definite lane line information, and several is set has a window of the slip of certain ROI scope along preliminary lane line information, and to the execution step of moving window one by one 4.2, analyze that to get rid of may be the isolated noise beyond lane line;
Step 4.4 is since second image, adopt lane line consulting to open the little characteristic of image meta amount of movement in front and back, this image and front several images are compared, extract the most effective lane line information to reduce the operand of Hough conversion in the track matching of follow-up straight line;
Step 5, straight way and bend matching:
(1) straight way matching
Adopt Hough to carry out matching to straight line track in image close shot region, its formula is as follows:
Figure 453723DEST_PATH_IMAGE013
Matching adopts Hough conversion method, and concrete steps are as follows:
A) model
Figure 466678DEST_PATH_IMAGE014
,
Figure 349184DEST_PATH_IMAGE015
each angle and database table, search for Hough conversion, to reduce operand;
B) obtain lane line information figure E and data set S,
Figure 768849DEST_PATH_IMAGE017
, wherein m is lane line information pixel number;
C) arrange
Figure 269101DEST_PATH_IMAGE018
as initial search point, as d) calculation of parameter, and counter is set, calculate next point by row subsequently
Figure 17614DEST_PATH_IMAGE019
;
If d) current
Figure 224605DEST_PATH_IMAGE020
meet
Figure 400371DEST_PATH_IMAGE021
,
Figure 408426DEST_PATH_IMAGE022
, arrange
Figure 695051DEST_PATH_IMAGE023
, until
Figure 84444DEST_PATH_IMAGE024
;
E) retain maximum
Figure 103216DEST_PATH_IMAGE025
and E, fitting a straight line is complete;
(2) bend matching
In image distant view region, adopt hyperbolic curve to model, its mathematic(al) representation is:
Wherein,
Figure 668375DEST_PATH_IMAGE027
for the ordinate of local horizon in picture plane, K, with
Figure 367527DEST_PATH_IMAGE029
for bend parameter;
According to the standard form of Hough space cathetus
Figure 64087DEST_PATH_IMAGE030
have:
Figure 630198DEST_PATH_IMAGE031
Contrast can draw:
Figure 994183DEST_PATH_IMAGE032
,
Figure 685583DEST_PATH_IMAGE033
,
Figure 869440DEST_PATH_IMAGE034
,
Figure 239241DEST_PATH_IMAGE035
Each parameter can calculate by straight line Hough conversion portion, arranges
Figure 457733DEST_PATH_IMAGE036
Figure 317105DEST_PATH_IMAGE037
if, matched curve point
Figure 925941DEST_PATH_IMAGE038
with moving window central point
Figure 896171DEST_PATH_IMAGE039
meet (unit is pixel),
Figure 999442DEST_PATH_IMAGE041
;
(3) export entirety lane line in real time: by step (1) and step (2) gained straight line track and the combination of bending track demonstration, the real-time lane line of output entirety;
Step 6, calculate each moving window parameter and obtain adaptive threshold: since second image, offer ROI scope according to moving window as image subregion, and with this scope, the GTG after changing is desired to make money or profit and carried out adaptive threshold extraction by the method for Otsu, equation expression is:
Figure 95574DEST_PATH_IMAGE042
Wherein t is required adaptive threshold,
Figure 869495DEST_PATH_IMAGE043
be i background ratio row,
Figure 794070DEST_PATH_IMAGE044
be i background mean value,
Figure 995244DEST_PATH_IMAGE045
be i object ratio,
Figure 578672DEST_PATH_IMAGE046
be i object average,
Figure 156284DEST_PATH_IMAGE047
be i image subregion image average, i is moving window number, and parameter calculation formula is as follows:
Figure 248054DEST_PATH_IMAGE049
Figure 381095DEST_PATH_IMAGE050
Figure 496818DEST_PATH_IMAGE051
Figure 336598DEST_PATH_IMAGE052
Figure 879575DEST_PATH_IMAGE053
Wherein, be the ratio of the pixel that in i moving window, GTG numerical value is k, be i the pixel number that moving window GTG numerical value is k,
Figure 850964DEST_PATH_IMAGE056
it is the total number of pixel in i moving window;
Threshold operation is:
Figure 830421DEST_PATH_IMAGE057
Wherein, Bin(x, y) be imaging point, f(x, y) be the respective coordinates of original pixel;
Step 7, since second image, first every image performs step six, then performs step three to step 5, exports in real time overall lane line.
2. according to the real-time track detecting of a kind of strong robustness based on ROI shown in claim 1 algorithm, it is characterized in that: described algorithm also comprises: in step 3, to using the lane line information after transverse gradients operator filters to search for repair strategy generally, search for the shade that repair strategy adopts generally and be:
Figure 672475DEST_PATH_IMAGE058
Taking current image lane line information data set as
Figure 67684DEST_PATH_IMAGE059
, last is
Figure 944374DEST_PATH_IMAGE060
if A, B, C are non-vanishing, D, E wherein have one non-vanishing and
Figure 829153DEST_PATH_IMAGE061
upgrade
Figure 158503DEST_PATH_IMAGE059
.
CN201410148832.8A 2014-04-15 2014-04-15 High-robustness real-time lane detection algorithm based on ROI Active CN103902985B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410148832.8A CN103902985B (en) 2014-04-15 2014-04-15 High-robustness real-time lane detection algorithm based on ROI

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410148832.8A CN103902985B (en) 2014-04-15 2014-04-15 High-robustness real-time lane detection algorithm based on ROI

Publications (2)

Publication Number Publication Date
CN103902985A true CN103902985A (en) 2014-07-02
CN103902985B CN103902985B (en) 2017-02-15

Family

ID=50994297

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410148832.8A Active CN103902985B (en) 2014-04-15 2014-04-15 High-robustness real-time lane detection algorithm based on ROI

Country Status (1)

Country Link
CN (1) CN103902985B (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105353373A (en) * 2015-12-16 2016-02-24 武汉大学 Hough transformation based ground penetrating radar target extraction method and device
CN106778668A (en) * 2016-12-30 2017-05-31 明见（厦门）技术有限公司 A kind of method for detecting lane lines of the robust of joint RANSAC and CNN
CN107180228A (en) * 2017-05-02 2017-09-19 开易（北京）科技有限公司 A kind of grad enhancement conversion method and system for lane detection
CN107392139A (en) * 2017-07-18 2017-11-24 海信集团有限公司 A kind of method for detecting lane lines and terminal device based on Hough transformation
CN107590438A (en) * 2017-08-16 2018-01-16 中国地质大学（武汉） A kind of intelligent auxiliary driving method and system
CN107901907A (en) * 2017-09-30 2018-04-13 惠州市德赛西威汽车电子股份有限公司 A kind of method for detecting lane lines based on color lump statistics
CN108830165A (en) * 2018-05-22 2018-11-16 南通职业大学 A kind of method for detecting lane lines considering front truck interference
CN109087326A (en) * 2018-09-18 2018-12-25 辽宁工业大学 Otsu algorithm based on local auto-adaptive
CN109543493A (en) * 2017-09-22 2019-03-29 杭州海康威视数字技术股份有限公司 A kind of detection method of lane line, device and electronic equipment

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5835028A (en) * 1997-05-05 1998-11-10 Bender; Lee Lane marker position sensor and alarm
US20050209748A1 (en) * 2004-03-12 2005-09-22 Toyota Jidosha Kabushiki Kaisha Lane boundary detector
CN101639983A (en) * 2009-08-21 2010-02-03 任雪梅 Multilane traffic volume detection method based on image information entropy
CN101826257A (en) * 2010-03-29 2010-09-08 北京市公安局公安交通管理局 Method for detecting motor way travel condition in real time
CN103366584A (en) * 2013-06-20 2013-10-23 银江股份有限公司 Real-time traffic flow detection-based self-adaptive tide lane control method
CN103617412A (en) * 2013-10-31 2014-03-05 电子科技大学 Real-time lane line detection method

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5835028A (en) * 1997-05-05 1998-11-10 Bender; Lee Lane marker position sensor and alarm
US20050209748A1 (en) * 2004-03-12 2005-09-22 Toyota Jidosha Kabushiki Kaisha Lane boundary detector
CN101639983A (en) * 2009-08-21 2010-02-03 任雪梅 Multilane traffic volume detection method based on image information entropy
CN101826257A (en) * 2010-03-29 2010-09-08 北京市公安局公安交通管理局 Method for detecting motor way travel condition in real time
CN103366584A (en) * 2013-06-20 2013-10-23 银江股份有限公司 Real-time traffic flow detection-based self-adaptive tide lane control method
CN103617412A (en) * 2013-10-31 2014-03-05 电子科技大学 Real-time lane line detection method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
洪敏等: "基于线性抛物线模型的车道检测与跟踪方法", 《长江大学学报（自然科学版）》 *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105353373B (en) * 2015-12-16 2018-04-20 武汉大学 One kind is based on Hough transform Ground Penetrating Radar target extraction method and device
CN105353373A (en) * 2015-12-16 2016-02-24 武汉大学 Hough transformation based ground penetrating radar target extraction method and device
CN106778668A (en) * 2016-12-30 2017-05-31 明见（厦门）技术有限公司 A kind of method for detecting lane lines of the robust of joint RANSAC and CNN
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
CN107180228A (en) * 2017-05-02 2017-09-19 开易（北京）科技有限公司 A kind of grad enhancement conversion method and system for lane detection
CN107392139A (en) * 2017-07-18 2017-11-24 海信集团有限公司 A kind of method for detecting lane lines and terminal device based on Hough transformation
CN107392139B (en) * 2017-07-18 2020-10-20 海信集团有限公司 Lane line detection method based on Hough transform and terminal equipment
CN107590438A (en) * 2017-08-16 2018-01-16 中国地质大学（武汉） A kind of intelligent auxiliary driving method and system
CN109543493A (en) * 2017-09-22 2019-03-29 杭州海康威视数字技术股份有限公司 A kind of detection method of lane line, device and electronic equipment
CN107901907A (en) * 2017-09-30 2018-04-13 惠州市德赛西威汽车电子股份有限公司 A kind of method for detecting lane lines based on color lump statistics
CN107901907B (en) * 2017-09-30 2019-12-20 惠州市德赛西威汽车电子股份有限公司 Lane line detection method based on color block statistics
CN108830165A (en) * 2018-05-22 2018-11-16 南通职业大学 A kind of method for detecting lane lines considering front truck interference
CN109087326A (en) * 2018-09-18 2018-12-25 辽宁工业大学 Otsu algorithm based on local auto-adaptive

Also Published As

Publication number Publication date
CN103902985B (en) 2017-02-15

Similar Documents

Publication Publication Date Title
CN103902985A (en) High-robustness real-time lane detection algorithm based on ROI
CN105678285B (en) A kind of adaptive road birds-eye view transform method and road track detection method
CN103235938B (en) The method and system of car plate detection and indentification
CN105261020B (en) A kind of express lane line detecting method
CN104008645B (en) One is applicable to the prediction of urban road lane line and method for early warning
Li et al. Nighttime lane markings recognition based on Canny detection and Hough transform
CN104657727B (en) A kind of detection method of lane line
CN103198705B (en) Parking place state automatic detection method
CN102298693B (en) Expressway bend detection method based on computer vision
CN105426864A (en) Multiple lane line detecting method based on isometric peripheral point matching
CN104050450A (en) Vehicle license plate recognition method based on video
CN102982304B (en) Utilize polarized light image to detect the method and system of vehicle location
WO2020103892A1 (en) Lane line detection method and apparatus, electronic device, and readable storage medium
CN103324930A (en) License plate character segmentation method based on grey level histogram binaryzation
CN110386065A (en) Monitoring method, device, computer equipment and the storage medium of vehicle blind zone
CN103996030A (en) Lane line detection method
JP2020064583A (en) Vehicle detection method, nighttime vehicle detection method based on dynamic light intensity, and system for the same
CN105224909A (en) Lane line confirmation method in lane detection system
CN105117726A (en) License plate positioning method based on multi-feature area accumulation
Premachandra et al. Image based automatic road surface crack detection for achieving smooth driving on deformed roads
CN107578046B (en) Auxiliary vehicle driving method based on image binarization processing
CN110276318A (en) Nighttime road rains recognition methods, device, computer equipment and storage medium
CN105787912A (en) Classification-based step type edge sub pixel localization method
CN103886609A (en) Vehicle tracking method based on particle filtering and LBP features
CN110033425B (en) Interference area detection device and method and electronic equipment

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant