CN104657735B - Method for detecting lane lines, system, lane departure warning method and system - Google Patents
Method for detecting lane lines, system, lane departure warning method and system Download PDFInfo
- Publication number
- CN104657735B CN104657735B CN201310593577.3A CN201310593577A CN104657735B CN 104657735 B CN104657735 B CN 104657735B CN 201310593577 A CN201310593577 A CN 201310593577A CN 104657735 B CN104657735 B CN 104657735B
- Authority
- CN
- China
- Prior art keywords
- image
- lane
- lane line
- line
- inward flange
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention provides a kind of method for detecting lane lines, including：S1, collection image, and coloured image is converted into gray level image；S2, interesting image regions are chosen, and be divided into left-side images and image right；S3, the binarization of gray value threshold value of left-side images and image right per a line is asked for using binaryzation, extract the pixel point set that gray value in left-side images and image right is more than or equal to binarization of gray value threshold value；S4, using one-dimensional sobel operators the edge image of left-side images and image right is obtained, ask for its edge binary-state threshold, extract the inward flange point set of left side edge image and right side edge image；S5, the above-mentioned pixel point set in selection left-side images and above-mentioned inward flange point set common factor are left-lane line inward flange point, and the common factor for choosing above-mentioned pixel point set and above-mentioned inward flange point set in image right is right-lane line inward flange point；S6, left-lane line and right-lane line obtained according to left-lane line inward flange point and right-lane line inward flange point.
Description
Technical field
The present invention relates to vehicle safety auxiliary driving technology field, more particularly to a kind of method for detecting lane lines, system,
Lane departure warning method and system.
Background technology
With the development of society, automobile has become the popular vehicles, because fatigue driving or notice point
Dissipate, accident caused by automotive run-off-road line is also on the increase, and usual speed is higher during the generation of such accident, therefore is endangered
Property is higher.
Research shows, if potential traffic accident gives driver's early warning in 1 second before occurring, can avoid the overwhelming majority
Similar traffic accident.Therefore, detect lane line in real time, identification vehicle whether run-off-road, do not carry out lane change operation in driver
But vehicle tends to remind driver in time during run-off-road, can greatly improve travel safety.
Cause the situation of automotive run-off-road line many in driving procedure, such as driving habit, fatigue driving, dispersion attention
Deng.Deviation caused by driving habit can actively be avoided by driver, and track caused by fatigue driving, dispersion attention
Deviateing can not rely on driver actively to avoid, and often cause traffic accident.
In order to overcome above mentioned problem, Lane Departure Warning System arises at the historic moment, the most crucial portion of Lane Departure Warning System
It is exactly lane detection system to divide.Traditional lane detection system work process is as follows：First, included by camera shooting
The image of lane line where vehicle；The lane line inward flange point in above-mentioned image, last profit are extracted followed by binarization method
Straight line is extracted with Hough transformation, so as to obtain lane line.
In above-mentioned method for detecting lane lines, an image binaryzation is used only in track edge target point and extracts car
Diatom edge inward flange point, it is not actual lane line inward flange point that the lane line inward flange point of extraction, which has a large portion,
Hot spot present in image and the larger noise of other brightness are likely to be, and then influences the accurate of lane line inward flange point detection
Property so that lane detection is not accurate enough.
The content of the invention
The technical problems to be solved by the invention be for existing method for detecting lane lines it is computationally intensive caused by be
A kind of the problem of system response delay, there is provided method for detecting lane lines.
Technical scheme is used by the present invention solves above-mentioned technical problem, there is provided a kind of method for detecting lane lines, including
Following steps：
The image of the left and right lane line of S1, collection comprising track where vehicle, and coloured image is converted into gray level image；
S2, the interesting image regions that there may be left and right lane line are chosen in above-mentioned gray level image, by above-mentioned image
Area-of-interest is divided into left-side images and image right；
S3, ask for the binarization of gray value threshold of left-side images and image right per a line respectively using image binaryzation method
Value, and the pixel point set that gray value in left-side images and image right is more than or equal to binarization of gray value threshold value is extracted respectively；
S4, the edge image that left-side images and image right are calculated using one-dimensional sobel operators respectively, are asked for respectively
The edge binary-state threshold of left side edge image and right side edge image, and using above-mentioned edge binary-state threshold to left side edge
Image and right side edge image carry out binaryzation, then extract the inward flange point of left side edge image and right side edge image respectively
Set；
S5, choose step S3 is extracted in left-side images above-mentioned pixel point set and the above-mentioned inward flange point of step S4 extractions
Intersection of sets integrates as left-lane line inward flange point, chooses step S3 is extracted in image right above-mentioned pixel point set and step S4
The common factor of the above-mentioned inward flange point set of extraction is right-lane line inward flange point；
S6, left-lane line and right car obtained according to the left-lane line inward flange point and right-lane line inward flange point of selection respectively
Diatom.
Further, after step S1, image preprocessing step is also included before step S2：
Denoising and smoothing processing are carried out to above-mentioned image using Gaussian filter.
Further, step S1 is specially：
The left and right track in track where vehicle front includes vehicle is shot by forward sight camera or the camera of left and right two
The image of line, and picture signal is inputed into Video Decoder；
Video Decoder inputs control unit after picture signal is decoded, control unit passes through video input interface thereon
Above-mentioned picture signal is gathered, and the picture signal collected is converted into gray level image storage in memory.
Further, the interesting image regions chosen in step S2 by do not include the figure of sky portion in collection image
Picture.
Further, the method tool of the binarization of gray value threshold value of left-side images and image right per a line is asked in step S3
Body is：
The binarization of gray value threshold value of left-side images and image right per a line is asked for by histogram Two-peak method respectively；
Pixel quantity according to binarization of gray value threshold value is met in the binary image of left-side images and image right is fitted
When binarization of gray value threshold value is decreased or increased；
By the binarization of gray value threshold value of present frame in the binary image of left-side images and image right and the ash of former frame
Degree binary-state threshold is compared, and the row of preset range is exceeded to binarization of gray value changes of threshold, then is combined and worked as according to different specific weight
The binarization of gray value threshold value of previous frame and former frame sets the final binarization of gray value threshold value of the row.
Further, step S4 is specially：
The edge image of left-side images and image right is calculated respectively using one-dimensional sobel operators, passes through histogram
Two-peak method asks for the edge binary-state threshold of left side edge image and right side edge image respectively, and uses above-mentioned left side edge figure
The edge binary-state threshold of picture and the edge binary-state threshold of right side edge image are respectively to left side edge image and right side edge
Image carries out binaryzation；
For meeting that the pixel (x, y) of following condition is extracted as left-lane line inward flange after left side edge image binaryzation
Point：
Y (x, y)=255；
Y (x-1, y)=255；
Y (x+1, y)=0；
Wherein (x-1, y) is the left neighbor pixel of (x, y), and (x+1, y) is the right neighbor pixel of (x, y), and Y is represented
Gray value after the pixel binaryzation；
For meeting that the pixel (x, y) of following condition is extracted as in initial right-lane line after right side edge image binaryzation
Marginal point：
Y (x, y)=0；
Y (x-1, y)=255；
Y (x+1, y)=255；
Wherein (x-1, y) is the left neighbor pixel of (x, y), and (x+1, y) is the right neighbor pixel of (x, y), and Y is represented
Gray value after the pixel binaryzation.
Further, step S6 is specially：
Become to change commanders using Radon and become comprising the rectangular coordinate plane of left-lane line inward flange point and right-lane line inward flange point
Pole coordinate parameter space plane (ρ, θ) is changed to, obtains straight line existing for left-side images and image right respectively；
Polar coordinates ρ-the θ for representing parameter space are quantized into the small lattice of multiple identicals, sat according to expression lane line at right angle
Each point coordinates in mark system X-Y（X, y）, according to formula ρ=xcos θ+ysin θ to the area within 0-180 ° in parameter space
Domain calculates each polar diameter ρ values, gained polar diameter ρ values are fallen into some small lattice, just with the progressive each polar angle θ values of the step-length of small lattice
The summary counter of the small lattice is set to add 1；After the completion of point whole in rectangular coordinate system all calculates, small lattice are tested, selected
The small lattice of summary counter numerical value front three are taken, (ρ, θ) value of the small lattice of three corresponds to three straight lines in rectangular co-ordinate；
Summary counter numerical value corresponding to selection is more than the straight line of preset value；
Judge the straight line of above-mentioned steps selection whether in the range of default (ρ, θ) value；
The straight line confidence level for meeting above-mentioned judgement is set to 1, and preserves (ρ, θ) value corresponding to it；
Detected into next frame, the new straight line detected is compared with the straight line that former frame detects, if there are two directly
The difference of (ρ, θ) value of line within a preset range, then assert the confidence level that two straight lines are same straight line, then this straight line
Add 1；If the straight line that present frame detects, it is impossible to any bar matching line segments with former frame, then it is assumed that this straight line is new straight line,
Its confidence level is set to 1, and records its (ρ, θ) value；If all straight lines that the straight line that former frame detects detects with present frame
It can not match, then it is assumed that this straight line disappears in the current frame, subtracts 1 by its confidence level；
Repeat the above steps, keep constant if the confidence level of straight line reaches 25, and (ρ, θ) value of this straight line is updated
To (ρ, θ) value of present frame detection；
Judge the confidence level of all straight lines of record, if the confidence level of straight line is equal to 25, judge this straight line to examine
The lane line of survey；If the confidence level of certain straight line is reduced to 0, this straight line is deleted from (ρ, the θ) in record.
Further, methods described also comprises the following steps：
According to the position of lane line in the current frame image detected, car in next two field picture is predicted using Kalman filter
The position of diatom；
The left and right lane line that former frame is detected extends lane detection of 50 pixels as next frame to the left and right respectively
Region.
According to the method for detecting lane lines of the present invention, when detecting left and right lane line inward flange point, first with image two-value
Change method asks for the binarization of gray value threshold value of left-side images and image right per a line respectively, and extracts left-side images and the right side respectively
Gray value is more than or equal to the pixel point set of binarization of gray value threshold value in the image of side；One-dimensional sobel operators are recycled to count respectively
Calculation obtains the edge image of left-side images and image right, asks for the edge two of left side edge image and right side edge image respectively
Value threshold value, and binaryzation is carried out to left side edge image and right side edge image using above-mentioned edge binary-state threshold, respectively
Extract the inward flange point set of left side edge image and right side edge image；Then, choose what step S3 in left-side images was extracted
The common factor of above-mentioned pixel point set and the above-mentioned inward flange point set of step S4 extractions is left-lane line inward flange point, chooses right side
The common factor of the above-mentioned pixel point set that step S3 is extracted in image and the above-mentioned inward flange point set of step S4 extractions is right lane
Line inward flange point.So, by first time binary conversion treatment, shade present in image can be removed and other brightness are less
Noise, and the edge image by being asked for one-dimensional sobel operators carries out second of binaryzation, can remove first time two-value again
Non-edge noise in change, such as the hot spot in image and the larger noise of other brightness, thus substantially increase lane line inner edge
The accuracy of edge detection, and then make it that lane detection is more accurate, improve the security of vehicle driving.
In addition, present invention also offers a kind of lane detection system, including image taking module and image processing module,
Described image processing module includes control unit, Video Decoder and memory；
Described image taking module, for shooting the image of the left and right lane line comprising track where vehicle；
Described image processing module, including image capture module and lane detection module；
Described image acquisition module, for receiving the picture signal of image taking module photograph by Video Decoder, and
Control unit is inputted after picture signal is decoded, control unit gathers above-mentioned image by video input interface thereon and believed
Number, and the picture signal collected is converted into gray level image storage in memory；
The lane detection module, the image sense of left and right lane line is there may be for being chosen in above-mentioned gray level image
Interest region, and above-mentioned interesting image regions are divided into left-side images and image right and handled respectively；Then, image is utilized
Binarization method asks for the binarization of gray value threshold value of left-side images and image right per a line respectively, and extracts left-side images respectively
It is more than or equal to the pixel point set of binarization of gray value threshold value with gray value in image right；Distinguished using one-dimensional sobel operators
The edge image of left-side images and image right is calculated, asks for the edge of left side edge image and right side edge image respectively
Binary-state threshold, and binaryzation is carried out to left side edge image and right side edge image using above-mentioned edge binary-state threshold, so
Extract the inward flange point set in left side edge image and right side edge image respectively afterwards；Then, step in left-side images is chosen
The common factor of the above-mentioned pixel point set of S3 extractions and the above-mentioned inward flange point set of step S4 extractions is left-lane line inward flange point,
Choose the common factor of step S3 is extracted in image right above-mentioned pixel point set and the above-mentioned inward flange point set of step S4 extractions
For right-lane line inward flange point；Finally, obtained respectively according to the left-lane line inward flange point of extraction and right-lane line inward flange point
Left-lane line and right-lane line.
Further, described image processing module also includes being connected to described image acquisition module and lane detection module
Between image pre-processing module, described image pretreatment module carries out denoising and smooth using Gaussian filter to above-mentioned image
Processing.
Further, described image taking module be vehicle viewing system forward sight camera, described image taking module
Forward sight camera or the camera of left and right two for vehicle viewing system.
In addition, present invention also offers a kind of lane departure warning method, comprise the following steps：
Detect to obtain lane line according to above-mentioned method for detecting lane lines；
According to the lane line and the relative position of vehicle and the current state of vehicle detected, it is determined whether need pre-
It is alert；
In the case of it is determined that needing early warning, the early warning in the form of sound and/or light.
According to the lane departure warning method of the present invention, by first time binary conversion treatment, it can remove in image and exist
Shade and the less noise of other brightness, and the edge image by being asked for one-dimensional sobel operators carry out second of two-value
Change, the non-edge noise in first time binaryzation can be removed again, such as the hot spot in image and the larger noise of other brightness, because
And the accuracy of lane line inward flange detection is substantially increased, and then make it that lane detection is more accurate, improve vehicle row
The security of car.
In addition, present invention also offers a kind of Lane Departure Warning System, including above-mentioned lane detection system, vehicle
With lane line relative position detection module, early warning logic judgment module and warning module；
The vehicle and lane line relative position detection module, for reference to detected by the lane detection system
The position of lane line and vehicle calibration parameter, determine relative position of the vehicle currently with lane line；
The early warning logic judgment module, entered according to the relative position and vehicle's current condition of Current vehicle and lane line
Row logic judgment, it is determined whether need early warning；
The warning module, warning module shift to an earlier date according to the judged result of the early warning logic judgment module to user
The early warning of carry out sound and/or light form.
Brief description of the drawings
Fig. 1 is the block diagram for the lane departure warning method that one embodiment of the invention provides；
Fig. 2 is the block diagram for the Lane Departure Warning System that one embodiment of the invention provides.
Mark in accompanying drawing is as follows：
10th, image taking module；20th, image processing module；21st, image capture module；22nd, lane detection module；23、
Image pre-processing module；24th, vehicle and lane line relative position detection module；25th, early warning logic judgment module；30th, early warning mould
Block.
7th, the method for detecting lane lines according to claim 1 to 5 any one, it is characterised in that adopted in step S6
Left-lane line and right-lane line are obtained with least square method.
Embodiment
In order that technical problem solved by the invention, technical scheme and beneficial effect are more clearly understood, below in conjunction with
Drawings and Examples, the present invention is described in further detail.It should be appreciated that specific embodiment described herein is only
To explain the present invention, it is not intended to limit the present invention.
The method for detecting lane lines that one embodiment of the invention provides, comprises the following steps：
The image of the left and right lane line of S1, collection comprising track where vehicle, and coloured image is converted into gray level image；
S2, the interesting image regions that there may be left and right lane line are chosen in above-mentioned gray level image, by above-mentioned image
Area-of-interest is divided into left-side images and image right；
S3, ask for the binarization of gray value threshold value of left-side images and image right per a line respectively using image binaryzation method
GrayTL1 ... GrayTLn and GrayTR1 ... GrayTRn, and extract respectively gray value in left-side images and image right be more than or
Equal to the pixel point set of binarization of gray value threshold value；
S4, the edge image that left-side images and image right are calculated using one-dimensional sobel operators respectively, are asked for respectively
The edge binary-state threshold EdgeTL and EdgeTR of left side edge image and right side edge image simultaneously uses above-mentioned edge binaryzation
Threshold value EdgeTL and EdgeTR carries out binaryzation to left side edge image and right side edge image, then extracts left side edge respectively
The inward flange point set of image and right side edge image；；
S5, choose step S3 is extracted in left-side images above-mentioned pixel point set and the above-mentioned inward flange point of step S4 extractions
Intersection of sets integrates as left-lane line inward flange point, chooses step S3 is extracted in image right above-mentioned pixel point set and step S4
The common factor of the above-mentioned inward flange point set of extraction is right-lane line inward flange point；
S6, left-lane line and right car obtained according to the left-lane line inward flange point and right-lane line inward flange point of selection respectively
Diatom.
In the present embodiment, after step S1, also include image preprocessing step before step S2.Described image pretreatment step
Rapid is to carry out denoising and smoothing processing to above-mentioned image using Gaussian filter, to improve picture quality.
In the present embodiment, step S1 is specially：
The left and right track in track where vehicle front includes vehicle is shot by forward sight camera or the camera of left and right two
The image of line, and picture signal is inputed into Video Decoder；A kind of specification of camera is as follows：Effective resolution is
640x480, frame per second are 30 frames/second.During using the camera collection vehicle both sides image for being installed on left and right rearview mirror, relative to peace
There is visibility point advantage loaded on the forward sight camera at windshield.On the one hand, left and right camera may be mounted at rearview mirror
It is interior, and install obliquely（Camera lens is towards front lower place）, the influence of light is not easily susceptible to, with the dusk sunlight can be avoided straight in the morning
Connect and be irradiated on camera, also do not influenceed at night by other automobile front lamp light, ensure the picture quality gathered in real time, significantly
Improve lane identification rate；On the other hand, the left and right camera in rearview mirror is not easy to by windscreen wiper and rainwater in the rainy day
Influence.
The picture signal that Video Decoder inputs camera（Analog signal）Control is inputted after being decoded as YUV data signal
Unit processed, control unit gather above-mentioned picture signal, and the picture signal that will be collected by video input interface thereon
Gray level image is converted to be stored in flash memory Flash and/or the memory of internal memory DDR types.Control unit is preferably DSP
（Digital Signal Processing, digital signal processor）Chip.
The interesting image regions chosen in step S2 by do not include the image of sky portion in collection image.According to throwing
Shadow is theoretical, and when camera illumination is parallel to the ground or at an angle, the top in camera view is usually the back of the body such as sky
Scape, and road surface is in the latter half of image, the present invention is using image the latter half as area-of-interest.To reduce amount of calculation simultaneously
And rational area-of-interest can be selected, the present invention ensures that rational area-of-interest selects by the setting angle of camera
Select.When camera is installed, the ratio that image is accounted for by sky in camera setting angle adjustment image and ground is 4:
6, therefore system can quickly select rational interesting image regions ROI（Region Of Interest）.Work as camera
When on the rearview mirror of left and right, in the image that left and right camera obtains, track is located at the left side and the right of image respectively, because
Area-of-interest selection is finally to surround the trapezoid area of nearside lane line by this present invention.
In the present embodiment, method that the binarization of gray value threshold value of left-side images and image right per a line is asked in step S3
Specially：
The binarization of gray value threshold value of left-side images and image right per a line is asked for by histogram Two-peak method respectively
GrayTL1 ... GrayTLn and GrayTR1 ... GrayTRn；Certainly in other embodiments, binary-state threshold GrayTL1 ...
GrayTLn and GrayTR1 ... GrayTRn acquisition methods can also be P parametric methods, Da-Jin algorithm, maximum entropy threshold method or iteration
Method etc..Cross histogram Two-peak method and ask for the binarization of gray value threshold value of left-side images and image right per a line respectively, relative to whole
Width image carries out binaryzation adaptability and greatly improved, and the method can avoid the influence of the outer complex environment of vehicle significantly, improve car
Diatom endpoint detections efficiency.
Pixel quantity according to binarization of gray value threshold value is met in the binary image of left-side images and image right is fitted
When binarization of gray value threshold value GrayTL1 ... GrayTLn and GrayTR1 ... GrayTRn is decreased or increased.The method can will be to be selected
The quantity of marginal point controls within the specific limits.
By the binarization of gray value threshold value of present frame in the binary image of left-side images and image right and the ash of former frame
Degree binary-state threshold is compared, and the row of preset range is exceeded to binarization of gray value changes of threshold, then is combined and worked as according to different specific weight
The binarization of gray value threshold value of previous frame and former frame sets the final binarization of gray value threshold value of the row.
Certainly, in other embodiments, the every a line of left-side images and image right asked for by histogram Two-peak method
Binarization of gray value threshold value GrayTL1 ... GrayTLn and GrayTR1 ... GrayTRn can also directly be set as the final ash of the row
Spend binary-state threshold.Comparatively, the method computing is more simple, but more less better than above method precision.
In the present embodiment, step S4 " carries out binaryzation to the edge image of left-side images and all rows of image right respectively
In the hope of in the edge binary-state threshold EdgeTL and EdgeTR " of left side edge image and right side edge image, edge binaryzation
Threshold value EdgeTL and EdgeTR acquisition methods can be Two-peak method, P parametric methods, Da-Jin algorithm, maximum entropy threshold method or iterative method
Deng.
In the present embodiment, for meeting that the pixel (x, y) of following condition is extracted as left-lane line inner edge in left-side images
Edge point：
Y (x, y)=255；
Y (x-1, y)=255；
Y (x+1, y)=0；
Wherein (x-1, y) is the left neighbor pixel of (x, y), and (x+1, y) is the right neighbor pixel of (x, y), and Y is represented
Gray value after the pixel binaryzation.
And for meeting that the pixel (x, y) of following condition is extracted as right-lane line inward flange point in image right：
Y (x, y)=0；
Y (x-1, y)=255；
Y (x+1, y)=255；
Wherein (x-1, y) is the left neighbor pixel of (x, y), and (x+1, y) is the right neighbor pixel of (x, y), and Y is represented
Gray value after the pixel binaryzation.
In the present embodiment, step S6 is specially：
Become to change commanders using Radon and become comprising the rectangular coordinate plane of left-lane line inward flange point and right-lane line inward flange point
Pole coordinate parameter space plane (ρ, θ) is changed to, obtains straight line existing for left-side images and image right respectively；It will represent that parameter is empty
Between polar coordinates ρ-θ be quantized into the small lattice of multiple identicals, according to representing that every bit of the lane line in rectangular coordinate system X-Y sit
Mark（X, y）, it is progressive with the step-length of small lattice to the region within 0-180 ° in parameter space according to formula ρ=xcos θ+ysin θ
Each polar angle θ values, each polar diameter ρ values are calculated, gained polar diameter ρ values are fallen into some small lattice, just make the summary counter of the small lattice
Add 1；After the completion of point whole in rectangular coordinate system all calculates, small lattice are tested, choose summary counter numerical value first three
The small lattice of position, (ρ, θ) value of the small lattice of three correspond to three straight lines in rectangular co-ordinate；
Summary counter numerical value corresponding to selection is more than the straight line of preset value；
Judge the straight line of above-mentioned steps selection whether in the range of default (ρ, θ) value；
The straight line confidence level for meeting above-mentioned judgement is set to 1, and preserves (ρ, θ) value corresponding to it；
Detected into next frame, the new straight line detected is compared with the straight line that former frame detects, if there are two directly
The difference of (ρ, θ) value of line within a preset range, then assert the confidence level that two straight lines are same straight line, then this straight line
Add 1；If the straight line that present frame detects, it is impossible to any bar matching line segments with former frame, then it is assumed that this straight line is new straight line,
Its confidence level is set to 1, and records its (ρ, θ) value；If all straight lines that the straight line that former frame detects detects with present frame
It can not match, then it is assumed that this straight line disappears in the current frame, subtracts 1 by its confidence level；
Repeat the above steps, keep constant if the confidence level of straight line reaches 25, and (ρ, θ) value of this straight line is updated
To (ρ, θ) value of present frame detection；
Judge the confidence level of all straight lines of record, if the confidence level of straight line is equal to 25, judge this straight line to examine
The lane line of survey；If the confidence level of certain straight line is reduced to 0, this straight line is deleted from (ρ, the θ) in record.
The determination methods of above-mentioned lane line, by setting the confidence level of straight line, improve lane detection accuracy and
Continuity, greatly improve the alarm efficiency of system.
In the present embodiment, methods described also comprises the following steps：
According to the position of lane line in the current frame image detected, car in next two field picture is predicted using Kalman filter
The position of diatom；The lane information of former frame is incorporated using Kalman filter, eliminates the interference of noise spot.Former frame is detected
Left and right lane line extend lane detection region of 50 pixels as next frame to the left and right respectively.
Road image video is continuous（30 frames/second）, the image between successive frame is not in what track direction was mutated
Situation.Therefore detection zone of the region as next two field picture can be set according to the position for the lane line that former frame detects
Domain.The lane line that art work embodiment detects former frame in the present invention sets 50 pixels as next frame to the left and right respectively
Detection zone, the constraints effectively eliminate the influence of noise jamming point, improve the accuracy rate and efficiency of lane detection.
Method for detecting lane lines according to the above embodiment of the present invention, it is first sharp when detecting left and right lane line inward flange point
Ask for the binarization of gray value threshold value of left-side images and image right per a line respectively with image binaryzation method, and extraction is left respectively
Gray value is more than or equal to the pixel point set of binarization of gray value threshold value in side image and image right；Recycle one-dimensional sobel
The edge image of left-side images and image right is calculated in operator respectively, asks for left side edge image and right side edge figure respectively
The edge binary-state threshold of picture, and two are carried out to left side edge image and right side edge image using above-mentioned edge binary-state threshold
Value, the inward flange point set of left side edge image and right side edge image is extracted respectively；Then, step in left-side images is chosen
The common factor of the above-mentioned pixel point set of S3 extractions and the above-mentioned inward flange point set of step S4 extractions is left-lane line inward flange point,
Choose the common factor of step S3 is extracted in image right above-mentioned pixel point set and the above-mentioned inward flange point set of step S4 extractions
For right-lane line inward flange point.So, by first time binary conversion treatment, shade present in image can be removed and other are bright
Spend less noise, and the edge image by being asked for one-dimensional sobel operators carries out second of binaryzation, the can be removed again
Non-edge noise in binaryzation, such as the hot spot in image and the larger noise of other brightness, thus substantially increase car
The accuracy of diatom inward flange detection, and then make it that lane detection is more accurate, improve the security of vehicle driving.
In addition, as shown in figure 1, one embodiment of the invention additionally provides a kind of lane detection system, including image taking
Module 10 and image processing module 20, described image processing module include control unit, Video Decoder and memory；
Described image taking module 10, for shooting the image of the left and right lane line comprising track where vehicle；
Described image processing module 20, including image capture module 21 and lane detection module 22；
Described image acquisition module, for receiving the picture signal of image taking module photograph by Video Decoder, and
Control unit is inputted after picture signal is decoded, control unit gathers above-mentioned image by video input interface thereon and believed
Number, and the picture signal collected is converted into gray level image storage in memory；Memory can be DDR internal memories or
FLASH flash memories.Control unit is preferably dsp chip.
The lane detection module 22, the image of left and right lane line is there may be for being chosen in above-mentioned gray level image
Area-of-interest, and above-mentioned interesting image regions are divided into left-side images and image right；Then, image binaryzation is utilized
Method asks for the binary-state threshold GrayTL1 ... GrayTLn and GrayTR1 ... of left-side images and image right per a line respectively
GrayTRn, and the pixel that gray value in left-side images and image right is more than or equal to binarization of gray value threshold value is extracted respectively
Set；The edge image of left-side images and image right is calculated respectively using one-dimensional sobel operators, asks for left side respectively
The edge binary-state threshold EdgeTL and EdgeTR of the edge image of edge image and image right, and use above-mentioned edge binaryzation
Threshold value carries out binaryzation to left side edge image and right side edge image, then extracts left side edge image and right side edge respectively
Inward flange point set in image；Then, the above-mentioned pixel point set that step S3 is extracted in left-side images is chosen to carry with step S4
The common factor of the above-mentioned inward flange point set taken is left-lane line inward flange point, chooses the above-mentioned picture that step S3 is extracted in image right
Vegetarian refreshments set and the common factor of the above-mentioned inward flange point set of step S4 extractions are right-lane line inward flange point；Finally, according to extraction
Left-lane line inward flange point and right-lane line inward flange point obtain left-lane line and right-lane line respectively.Lane detection module
22 are integrated in dsp chip, and lane detection function is realized by writing corresponding software in dsp chip.
In the present embodiment, described image processing module 20 also includes being connected to described image acquisition module and lane detection
Image pre-processing module 23 between module, described image pretreatment module 23 are gone using Gaussian filter to above-mentioned image
Make an uproar and smoothing processing.Image pre-processing module 23 is integrated in dsp chip.
In the present embodiment, forward sight camera or left and right two shooting of the described image taking module 10 for vehicle viewing system
Head, it is preferential using two cameras in left and right.Lane detection is realized using the existing viewing system of vehicle, it is other without increasing
Equipment, advantageously reduce parts and reduce production cost.Utilize the camera collection vehicle both sides for being installed on left and right rearview mirror
During image, there is visibility point advantage relative to the forward sight camera being installed at windshield.On the one hand, left and right camera can
With in rearview mirror, and install obliquely（Camera lens is towards front lower place）, the influence of light is not easily susceptible to, in the morning and at dusk
Sunlight can be avoided to shine directly on camera, also not influenceed at night by other automobile front lamp light, ensure collection in real time
Picture quality, greatly improve lane identification rate；On the other hand, the left and right camera in rearview mirror is not easy in the rainy day
Influenceed by windscreen wiper and rainwater.
In addition, one embodiment of the invention additionally provides a kind of lane departure warning method, comprise the following steps：
Detect to obtain lane line according to above-mentioned method for detecting lane lines；
According to the lane line and the relative position of vehicle and the current state of vehicle detected, it is determined whether need pre-
It is alert；The method step is known technological means, and the present invention is no longer not described in detail.
The present embodiment detects left and right lane line respectively using the camera of left and right two, when only detecting side lane line then
Judge whether track deviates using the lane line, if detecting the lane line of both sides simultaneously, need according to both sides lane line
State comprehensive descision vehicle whether run-off-road line.Therefore, only improve vehicle deviate when alarm rate while, reduce mistake
Report rate.
In the case of it is determined that needing early warning, the early warning in the form of sound and/or light.The method step is known technology hand
Section, the present invention are no longer not described in detail.
Lane departure warning method according to the above embodiment of the present invention, by first time binary conversion treatment, it can remove
Shade present in image and the less noise of other brightness, and the edge image by being asked for one-dimensional sobel operators is carried out
Second of binaryzation, the non-edge noise in first time binaryzation can be removed again, as the hot spot in image and other brightness compared with
Big noise, thus the accuracy of lane line inward flange detection is substantially increased, and then make it that lane detection is more accurate, carry
The security of vehicle driving is risen.
In addition, as shown in Fig. 2 one embodiment of the invention additionally provides a kind of Lane Departure Warning System, including it is above-mentioned
Lane detection system, vehicle and lane line relative position detection module 24, early warning logic judgment module 25 and warning module 30；
The vehicle and lane line relative position detection module 24, early warning logic judgment module 25 are integrated in dsp chip, by
Corresponding software is write in dsp chip to realize lane detection function, i.e., vehicle and lane line relative position detection module 24 and
Early warning logic judgment module 25 is a part for image processing module.
The vehicle and lane line relative position detection module 24, for reference to detected by the lane detection system
Lane line position and vehicle calibration parameter, determine relative position of the vehicle currently with lane line；This is known technology hand
Section, the present invention are no longer not described in detail.
The early warning logic judgment module 25, according to the relative position and vehicle's current condition of Current vehicle and lane line
Carry out logic judgment, it is determined whether need early warning；This is known technological means, and the present invention is no longer not described in detail.
The warning module 30, warning module carry according to the judged result of the early warning logic judgment module to user
The early warning of preceding carry out sound and/or light form, such as early warning is sent by buzzer, or show that early warning is believed on vehicle DVD
Breath, or in instrument board liquid crystal display screen display warning information.This is known technological means, and the present invention is not no longer detailed
Description.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement made within refreshing and principle etc., should be included in the scope of the protection.
Claims (13)
1. a kind of method for detecting lane lines, it is characterised in that comprise the following steps：
The image of the left and right lane line of S1, collection comprising track where vehicle, and coloured image is converted into gray level image；
S2, the interesting image regions that there may be left and right lane line are chosen in above-mentioned gray level image, above-mentioned image sense is emerging
Interesting region segmentation is left-side images and image right；
S3, ask for the binarization of gray value threshold value of left-side images and image right per a line respectively using image binaryzation method, and
The pixel point set that gray value in left-side images and image right is more than or equal to binarization of gray value threshold value is extracted respectively；
S4, the edge image that left-side images and image right are calculated using one-dimensional sobel operators respectively, ask for left side respectively
The edge binary-state threshold of edge image and right side edge image, and using above-mentioned edge binary-state threshold to left side edge image
Binaryzation is carried out with right side edge image, then extracts the inward flange point set of left side edge image and right side edge image respectively
Close；
S5, choose step S3 is extracted in left-side images above-mentioned pixel point set and the above-mentioned inward flange point set of step S4 extractions
Common factor be left-lane line inward flange point, choose step S3 is extracted in image right above-mentioned pixel point set and step S4 and extract
The common factor of above-mentioned inward flange point set be right-lane line inward flange point；
S6, left-lane line and right lane obtained according to the left-lane line inward flange point and right-lane line inward flange point of selection respectively
Line.
2. method for detecting lane lines according to claim 1, it is characterised in that after step S1, also wrapped before step S2
Include image preprocessing step：
Denoising and smoothing processing are carried out to above-mentioned image using Gaussian filter.
3. method for detecting lane lines according to claim 1, it is characterised in that step S1 is specially：
The left and right lane line in track where vehicle front includes vehicle is shot by forward sight camera or the camera of left and right two
Image, and picture signal is inputed into Video Decoder；
Video Decoder inputs control unit after picture signal is decoded, control unit is gathered by video input interface thereon
Above-mentioned picture signal, and the picture signal collected is converted into gray level image storage in memory.
4. method for detecting lane lines according to claim 3, it is characterised in that the interesting image area chosen in step S2
Domain by do not include the image of sky portion in collection image.
5. method for detecting lane lines according to claim 4, it is characterised in that left-side images and right side are asked in step S3
The method of binarization of gray value threshold value of the image per a line be specially：
The binarization of gray value threshold value of left-side images and image right per a line is asked for by histogram Two-peak method respectively；
According to meet in the binary image of left-side images and image right binary-state threshold pixel quantity suitably reduce or
Increase binarization of gray value threshold value；
By the binarization of gray value threshold value of present frame in the binary image of left-side images and image right and the gray scale two of former frame
Value threshold value compares, and the row of preset range is exceeded to binarization of gray value changes of threshold, then according to different specific weight combination present frame
The final binarization of gray value threshold value of the row is set with the binarization of gray value threshold value of former frame.
6. method for detecting lane lines according to claim 4, it is characterised in that step S4 is specially：
The edge image of left-side images and image right is calculated respectively using one-dimensional sobel operators, it is bimodal by histogram
Method asks for the edge binary-state threshold of left side edge image and right side edge image respectively, and uses above-mentioned left side edge image
The edge binary-state threshold of edge binary-state threshold and right side edge image is respectively to left side edge image and right side edge image
Carry out binaryzation；
Pixel (x, y) for meeting following condition after left side edge image binaryzation is extracted as left-lane line inward flange point：
Y (x, y)=255；
Y (x-1, y)=255；
Y (x+1, y)=0；
Wherein (x-1, y) is the left neighbor pixel of (x, y), and (x+1, y) is the right neighbor pixel of (x, y), and Y represents the picture
Gray value after vegetarian refreshments binaryzation；
Pixel (x, y) for meeting following condition after right side edge image binaryzation is extracted as initial right-lane line inward flange
Point：
Y (x, y)=0；
Y (x-1, y)=255；
Y (x+1, y)=255；
Wherein (x-1, y) is the left neighbor pixel of (x, y), and (x+1, y) is the right neighbor pixel of (x, y), and Y represents the picture
Gray value after vegetarian refreshments binaryzation.
7. according to the method for detecting lane lines described in claim 1 to 6 any one, it is characterised in that step S6 is specially：
Become the rectangular coordinate plane comprising left-lane line inward flange point and right-lane line inward flange point of changing commanders using Radon to transform to
Pole coordinate parameter space plane (ρ, θ), obtains straight line existing for left-side images and image right respectively；
Polar coordinates ρ-the θ for representing parameter space are quantized into the small lattice of multiple identicals, according to expression lane line in rectangular coordinate system
Unite X-Y in each point coordinates (x, y), according to formula ρ=xcos θ+ysin θ to the region within 0-180 ° in parameter space with
The progressive each polar angle θ values of the step-lengths of small lattice, calculate each polar diameter ρ values, gained polar diameter ρ values are fallen into some small lattice, just make this
The summary counter of small lattice adds 1；After the completion of point whole in rectangular coordinate system all calculates, small lattice are tested, chosen tired
The small lattice of counter numerical value front three, (ρ, θ) value of the small lattice of three correspond to three straight lines in rectangular co-ordinate；
Summary counter numerical value corresponding to selection is more than the straight line of preset value；
Judge the straight line of above-mentioned steps selection whether in the range of default (ρ, θ) value；
The straight line confidence level for meeting above-mentioned judgement is set to 1, and preserves (ρ, θ) value corresponding to it；
Detected into next frame, the new straight line detected is compared with the straight line that former frame detects, if there are two straight lines
Within a preset range, then it is same straight line to assert two straight lines to the difference of (ρ, θ) value, then the confidence level of this straight line adds 1；
If the straight line that present frame detects, it is impossible to any bar matching line segments with former frame, then it is assumed that this straight line is new straight line, by it
Confidence level is set to 1, and records its (ρ, θ) value；If the equal nothing of all straight lines that the straight line that former frame detects detects with present frame
Method matches, then it is assumed that this straight line disappears in the current frame, subtracts 1 by its confidence level；
Repeat the above steps, keep constant if the confidence level of straight line reaches 25, and (ρ, θ) value renewal of this straight line is arrived and worked as
(ρ, θ) value of previous frame detection；
Judge the confidence level of all straight lines of record, if the confidence level of straight line is equal to 25, judge that this straight line will detect
Lane line；If the confidence level of certain straight line is reduced to 0, this straight line is deleted from (ρ, the θ) in record.
8. according to the method for detecting lane lines described in claim 1 to 6 any one, it is characterised in that methods described also includes
Following steps：
According to the position of lane line in the current frame image detected, lane line in next two field picture is predicted using Kalman filter
Position；
The left and right lane line that former frame is detected extends lane detection region of 50 pixels as next frame to the left and right respectively,
If select area-of-interest according to (S2) without lane line in the region.
9. a kind of lane detection system, it is characterised in that including image taking module and image processing module, at described image
Reason module includes control unit, Video Decoder and memory；
Described image taking module, for shooting the image of the left and right lane line comprising track where vehicle；
Described image processing module, including image capture module and lane detection module；
Described image acquisition module, for receiving the picture signal of image taking module photograph by Video Decoder, and will figure
Control unit is inputted after being decoded as signal, control unit gathers above-mentioned picture signal by video input interface thereon, and
The picture signal collected is converted into gray level image storage in memory；
The lane detection module, the interesting image of left and right lane line is there may be for being chosen in above-mentioned gray level image
Region, and above-mentioned interesting image regions are divided into left-side images and image right；Then, image binaryzation method point is utilized
The binarization of gray value threshold value of left-side images and image right per a line is not asked for, and is extracted respectively in left-side images and image right
Gray value is more than or equal to the pixel point set of binarization of gray value threshold value；Left side is calculated respectively using one-dimensional sobel operators
The edge image of image and image right, the edge binaryzation threshold to left side edge image and right side edge image is asked for respectively
Value, and binaryzation is carried out to left side edge image and right side edge image using above-mentioned edge binary-state threshold, then carry respectively
Take the inward flange point set in left side edge image and right side edge image；Then, choose what step S3 in left-side images was extracted
The common factor of above-mentioned pixel point set and the above-mentioned inward flange point set of step S4 extractions is left-lane line inward flange point, chooses right side
The common factor of the above-mentioned pixel point set that step S3 is extracted in image and the above-mentioned inward flange point set of step S4 extractions is right lane
Line inward flange point；Finally, left-lane line is obtained according to the left-lane line inward flange point of extraction and right-lane line inward flange point respectively
And right-lane line.
10. lane detection system according to claim 9, it is characterised in that described image processing module also includes connecting
The image pre-processing module being connected between described image acquisition module and lane detection module, described image pretreatment module profit
Denoising and smoothing processing are carried out to above-mentioned image with Gaussian filter.
11. the lane detection system according to claim 9 or 10, it is characterised in that described image taking module is car
The forward sight camera or the camera of left and right two of viewing system.
12. a kind of lane departure warning method, it is characterised in that comprise the following steps：
Method for detecting lane lines according to claim 1 to 8 any one detects to obtain lane line；
According to the lane line and the relative position of vehicle and the current state of vehicle detected, it is determined whether need early warning；
In the case of it is determined that needing early warning, the early warning in the form of sound or light.
13. a kind of Lane Departure Warning System, it is characterised in that including the lane line described in claim 9 to 11 any one
Detecting system, vehicle and lane line relative position detection module, early warning logic judgment module and warning module；
The vehicle and lane line relative position detection module, for the track with reference to detected by the lane detection system
The position of line and vehicle calibration parameter, determine relative position of the vehicle currently with lane line；
The early warning logic judgment module, patrolled according to the relative position and vehicle's current condition of Current vehicle and lane line
Collect and judge, it is determined whether need early warning；
The warning module, according to the judged result of the early warning logic judgment module, sound or light shape is carried out in advance to user
The early warning of formula.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310593577.3A CN104657735B (en) | 2013-11-21 | 2013-11-21 | Method for detecting lane lines, system, lane departure warning method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310593577.3A CN104657735B (en) | 2013-11-21 | 2013-11-21 | Method for detecting lane lines, system, lane departure warning method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104657735A CN104657735A (en) | 2015-05-27 |
| CN104657735B true CN104657735B (en) | 2018-01-23 |
Family
ID=53248835
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310593577.3A Active CN104657735B (en) | 2013-11-21 | 2013-11-21 | Method for detecting lane lines, system, lane departure warning method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104657735B (en) |
Families Citing this family (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106469436B (en) * | 2015-08-17 | 2019-11-08 | 比亚迪股份有限公司 | Image denoising system and image de-noising method |
| CN105216694B (en) * | 2015-10-14 | 2019-02-12 | 南宁学院 | A kind of Lane Departure Warning System |
| CN105539293B (en) * | 2016-02-03 | 2018-01-09 | 北京中科慧眼科技有限公司 | Lane departure warning method and device and car steering accessory system |
| CN105678791B (en) * | 2016-02-24 | 2018-07-17 | 西安交通大学 | One kind being based on the not unique lane detection and tracking method of parameter |
| CN105868696B (en) * | 2016-03-23 | 2019-06-14 | 奇瑞汽车股份有限公司 | A kind of method and apparatus detecting multilane lane line |
| CN108304749A (en) * | 2017-01-13 | 2018-07-20 | 比亚迪股份有限公司 | The recognition methods of road speed line, device and vehicle |
| CN106682646B (en) * | 2017-01-16 | 2020-12-22 | 北京新能源汽车股份有限公司 | Lane line identification method and device |
| CN107465892A (en) * | 2017-07-13 | 2017-12-12 | 湖南中南楷思科技发展有限公司 | Digital signal processing image identifying system and recognition methods |
| CN109325389A (en) * | 2017-07-31 | 2019-02-12 | 比亚迪股份有限公司 | Lane detection method, apparatus and vehicle |
| CN107729814B (en) * | 2017-09-19 | 2020-06-23 | 浙江大华技术股份有限公司 | Method and device for detecting lane line |
| CN109558765B (en) * | 2017-09-25 | 2022-02-08 | 比亚迪股份有限公司 | Automobile and lane line detection method and device |
| 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 |
| CN107909047B (en) * | 2017-11-28 | 2021-07-06 | 上海信耀电子有限公司 | Automobile and lane detection method and system applied to automobile |
| CN108345855B (en) * | 2018-02-08 | 2020-08-18 | 青岛平行智能产业管理有限公司 | Lane line pressing detection method and system |
| CN108182430B (en) * | 2018-02-27 | 2022-01-14 | 厦门蔚森智能科技有限公司 | Double-area lane line identification system and method |
| CN108509906B (en) * | 2018-03-30 | 2022-02-08 | 长安大学 | Double-threshold Radon identification method for line light point capture of unmanned aerial vehicle aerial photography roadbed |
| CN110386065B (en) * | 2018-04-20 | 2021-09-21 | 比亚迪股份有限公司 | Vehicle blind area monitoring method and device, computer equipment and storage medium |
| CN108875657A (en) * | 2018-06-26 | 2018-11-23 | 北京茵沃汽车科技有限公司 | A kind of method for detecting lane lines |
| CN109448011B (en) * | 2018-10-22 | 2021-01-15 | 浙江科技学院 | Wind-proof early warning method for photovoltaic power generation device |
| US11430226B2 (en) | 2019-01-14 | 2022-08-30 | Boe Technology Group Co., Ltd. | Lane line recognition method, lane line recognition device and non-volatile storage medium |
| CN110414385B (en) * | 2019-07-12 | 2021-06-25 | 淮阴工学院 | Lane line detection method and system based on homography transformation and characteristic window |
| CN111002997A (en) * | 2019-12-06 | 2020-04-14 | 武汉中海庭数据技术有限公司 | Lane departure prompting method and device |
| CN111881878B (en) * | 2020-08-07 | 2024-02-02 | 吉林大学 | Lane line identification method for look-around multiplexing |
| CN113815648B (en) * | 2021-10-29 | 2023-08-01 | 广州理工学院 | Control method of vehicle control system |
| CN114001706B (en) * | 2021-12-29 | 2022-04-29 | 阿里巴巴达摩院(杭州)科技有限公司 | Course angle estimation method and device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102303609A (en) * | 2011-06-16 | 2012-01-04 | 广东铁将军防盗设备有限公司 | System and method for prewarning lane deviation |
| CN102663744A (en) * | 2012-03-22 | 2012-09-12 | 杭州电子科技大学 | Complex road detection method under gradient point pair constraint |
-
2013
- 2013-11-21 CN CN201310593577.3A patent/CN104657735B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102303609A (en) * | 2011-06-16 | 2012-01-04 | 广东铁将军防盗设备有限公司 | System and method for prewarning lane deviation |
| CN102663744A (en) * | 2012-03-22 | 2012-09-12 | 杭州电子科技大学 | Complex road detection method under gradient point pair constraint |
Non-Patent Citations (1)
| Title |
|---|
| 基于改进Hough变换的车道线检测技术;杨喜宁等;《计算机测量与控制》;20100228;第18卷(第2期);第292-294,298页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104657735A (en) | 2015-05-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104657735B (en) | Method for detecting lane lines, system, lane departure warning method and system | |
| CN107554430B (en) | Vehicle blind area visualization method, device, terminal, system and vehicle | |
| CN104933409B (en) | A kind of parking stall recognition methods based on panoramic picture dotted line feature | |
| CN104517111B (en) | Method for detecting lane lines, system, lane departure warning method and system | |
| CN109460709B (en) | RTG visual barrier detection method based on RGB and D information fusion | |
| CN104751151B (en) | A kind of identification of multilane in real time and tracking | |
| CN105206109B (en) | A kind of vehicle greasy weather identification early warning system and method based on infrared CCD | |
| CN105835880B (en) | Lane following system | |
| Yan et al. | A method of lane edge detection based on Canny algorithm | |
| CN105047019B (en) | A kind of passenger stock prevent rear car overtake other vehicles after lane change determination methods and device suddenly | |
| CN107891808B (en) | Driving reminding method and device and vehicle | |
| CN109435942A (en) | A kind of parking stall line parking stall recognition methods and device based on information fusion | |
| JP4930046B2 (en) | Road surface discrimination method and road surface discrimination device | |
| Li et al. | Nighttime lane markings recognition based on Canny detection and Hough transform | |
| CN106887004A (en) | A kind of method for detecting lane lines based on Block- matching | |
| CN103523014A (en) | In-vehicle apparatus | |
| Liu et al. | Development of a vision-based driver assistance system with lane departure warning and forward collision warning functions | |
| CN109002797B (en) | Vehicle lane change detection method, device, storage medium and computer equipment | |
| CN106679633A (en) | Vehicle-mounted distance measuring system and vehicle-mounted distance measuring method | |
| KR101483742B1 (en) | Lane Detection method for Advanced Vehicle | |
| CN105021126A (en) | Truck side guard rail mounting size measurement system and method based on machine vision technology | |
| Kuo et al. | Vision-based vehicle detection in the nighttime | |
| CN112666573B (en) | Detection method for retaining wall and barrier behind mine unloading area vehicle | |
| CN107704836B (en) | Fatigue driving detection method based on object detection | |
| CN115100618A (en) | Multi-source heterogeneous perception information multi-level fusion representation and target identification method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |