CN108205667A - Method for detecting lane lines and device, lane detection terminal, storage medium - Google Patents
Method for detecting lane lines and device, lane detection terminal, storage medium Download PDFInfo
- Publication number
- CN108205667A CN108205667A CN201810208562.3A CN201810208562A CN108205667A CN 108205667 A CN108205667 A CN 108205667A CN 201810208562 A CN201810208562 A CN 201810208562A CN 108205667 A CN108205667 A CN 108205667A
- Authority
- CN
- China
- Prior art keywords
- submodule
- straight line
- characteristic value
- pixel
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20061—Hough transform
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
- G06T2207/30256—Lane; Road marking
Abstract
Present invention is disclosed a kind of method for detecting lane lines and device, lane detection terminal, computer readable storage mediums, are related to assisting driving technology field, the program includes：The pavement image for including lane line is obtained, pavement image is divided into multiple submodule；According to the pixel value of feature pixel in submodule, the characteristic value of computational submodule；The submodule that characteristic value is higher than threshold value is filtered out from multiple submodule, the submodule filtered out is as foreground area, other regions in pavement image in addition to foreground area are as background area；Binary conversion treatment is carried out to foreground area, the pixel value zero setting of background area obtains the binary image of pavement image；Extraction represents the straight line of lane line in binary image.The program can obtain the apparent binary image of foreground pixel, small due to being interfered by background pixel based on the binary image into the detection of driveway line, can accurately extract lane line, and then can improve the accuracy that lane line deviates early warning.
Description
Technical field
The present invention relates to auxiliary driving technology field, more particularly to a kind of method for detecting lane lines and device, lane line inspection
Survey terminal, computer readable storage medium.
Background technology
Lane Departure Warning System is an important research content of automobile assistant driving, which usually utilizes video camera
Vehicle front image is shot, and lane information is obtained according to picture material.When the relative position in vehicle and track shifts
When, it can be in time to driver's early warning, so as to avoid accident.
The most important step of Lane Departure Warning System is exactly lane detection, usually will camera shooting in lane detection
The gray-scale map binaryzation of the vehicle front of machine shooting, and using straight in hough (Hough) transform method detection binary image
Line.It is projected in polar coordinate space especially by by each pixel in bianry image, formation hough matrixes, and according to
The votes search local maximum of each coordinate (ρ, θ), obtains in the corresponding binary image of maximum in hough matrixes
Straight line.
Due to real road scene complexity, under illumination condition, road pixel reflection is uneven, and road both sides is usually
There is shade distribution, the grey scale change of road in itself is larger, and prospect lane line pixel is will appear in binarization excessively or is examined
The situation of prospect lane line pixel is not detected.For example, as shown in Figure 1, left-lane shadow interference is serious on the left of image, in the middle part of image
Uneven illumination, road grey scale change itself is larger, and such case causes binaryzation greatly to interfere.To gray level image using current
Widely used global threshold binarization method (such as Da-Jin algorithm) and local threshold binarization method (such as Sauvola algorithms) are respectively
Processing, as a result as shown in Figures 2 and 3.Global threshold binarization method poor effect, entire road are all marked as foreground features
Pixel (i.e. gray value is 255), although local threshold binarization method has some improvement, prospect lane line pixel can be extracted
Out, but background sideways pixel still serious interference, to subsequent hough transformation bring greatly interference and calculation amount.
To sum up, due to real road scene complexity, for the road gray level image under night or strong illumination, lane line institute
Contrast in region and other regions is relatively low, and the pixel value of lane line is closer to the pixel value in other regions, at this time not
It is obtained by use global threshold binarization method (such as Da-Jin algorithm) or local threshold binarization method (such as Sauvola methods)
Lane line pixel and then is difficult to accurately extract by Hough transformation by the serious interference of background pixel point in binary image
To lane line.
Invention content
In order to solve for the road gray level image under night or strong illumination present in the relevant technologies, due to lane line
The contrast of region and other regions is relatively low, lane line pixel in the existing obtained binary image of binarization method
By the serious interference of background pixel point, and then the problem of be difficult to accurately extract lane line by Hough transformation, the present invention carries
A kind of method for detecting lane lines is supplied.
On the one hand, the present invention provides a kind of method for detecting lane lines, the method includes：
The pavement image for including lane line is obtained, the pavement image is divided into multiple submodule；
According to the pixel value of feature pixel in the submodule, the characteristic value of the submodule is calculated；
The submodule that characteristic value is higher than threshold value is filtered out from multiple submodules, the submodule filtered out is as prospect
Region, other regions in the pavement image in addition to foreground area are as background area；
Binary conversion treatment is carried out to the foreground area, the pixel value zero setting of background area obtains the pavement image
Binary image；
Extraction represents the straight line of lane line in the binary image.
Further, the pixel value according to feature pixel in the submodule calculates the feature of the submodule
Value, including：
According to the pixel value of pixel in the submodule, submodule feature pixel in the block is determined；
By calculating the pixel average or pixel maximum of feature pixel in the submodule, the submodule is obtained
Characteristic value.
Further, the submodule that characteristic value is higher than threshold value, the submodule filtered out are filtered out from multiple submodules
Block as foreground area, other regions in the pavement image in addition to foreground area as background area, including：
Selected characteristic value is higher than the submodule of first threshold from the multiple submodule, and the submodule of selection is alternately
Module；
Different alternate modulars according to characteristic value are clustered, are divided into different alternate modulars according to cluster result multiple
Cluster areas；
According to the characteristic value of the cluster areas Neutron module, the assemblage characteristic value of the cluster areas is determined；
The cluster areas that assemblage characteristic value is higher than second threshold is chosen from multiple cluster areas, the described of selection gathers
Class region is as foreground area, and other regions in the pavement image in addition to foreground area are as background area, and described first
Threshold value is less than second threshold.
Further, it is described that different alternate modulars are clustered according to characteristic value, it will be different alternative according to cluster result
Module is divided into multiple cluster areas and includes：
By the pavement image it is longitudinally divided be at least two target areas；
According to position of the alternate modular in the pavement image, the alternate modular is divided to affiliated target
Region；
Alternate modular in the target area is clustered according to characteristic value, by the alternative mould in the target area
Block is divided into multiple cluster areas according to cluster result.
Further, the straight line for representing lane line is extracted in the binary image, including：
Straight-line detection is carried out to the binary image by Hough transformation, obtains a plurality of candidate straight line；
According to candidate the straight line number of occupied submodule, feature of occupied submodule in the pavement image
The straight line votes of value and the candidate straight line determine the weight of the candidate straight line；
According to the weight of the candidate straight line, weight selection meets the candidate of preset condition from a plurality of candidate straight line
Straight line obtains representing the straight line of lane line in the binary image as target line.
Further, according to the weight of the candidate straight line, weight selection meets default from a plurality of candidate straight line
The candidate straight line of condition obtains representing the straight line of lane line in the binary image as target line, including：
According to position of a plurality of candidate straight line in the pavement image, to a plurality of candidate straight line according to residing
Position is grouped, and forms several straight line set；
Candidate straight line quantity and lane line quantity to be extracted according to included in straight line set, from different straight lines
The candidate straight line of identical quantity is chosen in set as target line, weight of the target line in residing straight line set is most
Height, the target line is as the lane line in the pavement image.
On the other hand, the present invention also provides a kind of lane detection device, described device includes：
The pavement image for obtaining the pavement image for including lane line, is divided into multiple sons by image division module
Module；
Characteristic value calculating module for the pixel value according to feature pixel in the submodule, calculates the submodule
Characteristic value；
Prospect screening module, for filtering out the submodule that characteristic value is higher than threshold value, screening from multiple submodules
The submodule gone out is as foreground area, other regions in the pavement image in addition to foreground area are as background area；
Binarization block, for carrying out binary conversion treatment to the foreground area, the pixel value zero setting of background area obtains
The binary image of the pavement image；
Lines detection module, for extracting the straight line for representing lane line in the binary image.
Further, the prospect screening module includes：
Alternate modular selection unit, for from the multiple submodule selected characteristic value be higher than first threshold submodule
Block, the submodule of selection alternately module；
Region clustering unit, will be different according to cluster result for being clustered to different alternate modulars according to characteristic value
Alternate modular is divided into multiple cluster areas；
Regional characteristic value determination unit for the characteristic value according to the cluster areas Neutron module, determines the cluster
The assemblage characteristic value in region；
Region selection unit, for choosing the cluster that assemblage characteristic value is higher than second threshold from multiple cluster areas
Region, the cluster areas of selection is as foreground area, other region conducts in the pavement image in addition to foreground area
Background area, the first threshold are less than second threshold.
In addition, the present invention also provides a kind of lane detection terminal, the terminal includes：
Processor；
For storing the memory of processor-executable instruction；
Wherein, the processor is configured as performing any one above-mentioned method for detecting lane lines.
In addition, the present invention also provides a kind of computer readable storage medium, the computer-readable recording medium storage
There is computer program, the computer program can be performed any one above-mentioned method for detecting lane lines by processor.
The technical solution that the embodiment of the present invention provides can include the following benefits：
Technical solution provided by the invention by the way that pavement image is divided into multiple submodule, and calculates each submodule
Characteristic value, filter out the higher submodule of characteristic value as foreground area carry out binary conversion treatment, the relatively low submodule of characteristic value
Block, so as to which background area will not interfere the binaryzation of foreground area, is improved as the direct zero setting of background area pixels value
The contrast of pixel and other pixels, can obtain lane line pixel more obvious where foreground area lane line
Binary image, it is small due to being interfered by background pixel point based on the binary image into the detection of driveway line, so as to
Lane line is accurately extracted, and then the accuracy that lane line deviates early warning can be improved.
It should be understood that above general description and following detailed description is only exemplary, this can not be limited
Invention.
Description of the drawings
Attached drawing herein is incorporated into specification and forms the part of this specification, shows the implementation for meeting the present invention
Example, and in specification together principle for explaining the present invention.
The road surface gray level image that Fig. 1 is shot when being by illumination effect；
Fig. 2 is the result schematic diagram for carrying out binaryzation to the road surface gray level image of Fig. 1 using existing Da-Jin algorithm；
Fig. 3 is the result schematic diagram for carrying out binaryzation to the road surface gray level image of Fig. 1 using existing Sauvola algorithms；
Fig. 4 is the flow chart according to a kind of method for detecting lane lines shown in an exemplary embodiment；
Fig. 5 is the dividing mode that pavement image is divided into several submodules signal shown in a kind of exemplary embodiment
Figure；
Fig. 6 is the details flow chart of the step 420 of Fig. 4 corresponding embodiments；
Fig. 7 is the schematic diagram of the division foreground area and background area shown in a kind of exemplary embodiment；
Fig. 8 is the result schematic diagram carried out using method provided by the invention after binary conversion treatment；
Fig. 9 is the details flow chart of the step 430 of Fig. 4 corresponding embodiments；
Figure 10 is the schematic diagram that cluster areas is formed by alternate modular shown in a kind of exemplary embodiment；
Figure 11 is the stream of the method for detecting lane lines that another exemplary embodiment provides on the basis of Fig. 9 corresponding embodiments
Cheng Tu；
Figure 12 is that an exemplary embodiment shows the principle schematic clustered to the alternate modular in target area；
Figure 13 is the details flow chart of the step 450 of Fig. 4 corresponding embodiments；
Figure 14 is the principle schematic of submodule quantity occupied by a kind of determining candidate straight line shown in exemplary embodiment；
Figure 15 is the details flow chart of the step 453 of Figure 13 corresponding embodiments；
Figure 16 is that the result of the progress of road surface gray level image shown in Fig. 1 lane detection is shown using method provided by the invention
It is intended to；
Figure 17 is the block diagram according to a kind of lane detection device shown in an exemplary embodiment；
Figure 18 is the details block diagram of the prospect screening module of Figure 17 corresponding embodiments；
Figure 19 is a kind of structure diagram for lane detection terminal that exemplary embodiment of the present provides.
Specific embodiment
Here explanation will be performed to exemplary embodiment in detail, example is illustrated in the accompanying drawings.Following description is related to
During attached drawing, unless otherwise indicated, the same numbers in different attached drawings represent the same or similar element.Following exemplary embodiment
Described in embodiment do not represent and the consistent all embodiments of the present invention.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects being described in detail in claims, of the invention.
Fig. 4 is the flow chart according to a kind of method for detecting lane lines shown in an exemplary embodiment.The lane detection
The scope of application and executive agent of method, for example, this method is for the lane detection terminal in following embodiments.Such as Fig. 4 institutes
Show, which may comprise steps of.
In step 410, the pavement image for including lane line is obtained, the pavement image is divided into multiple submodule；
Wherein, pavement image can be the two dimensional image that CCD camera assembly acquires in lane detection terminal, in the image
Include lane line to be detected.Wherein, it can be coloured image that the image of CCD camera assembly acquisition, which can be two dimensional image, according to
It needs coloured image are converted to gray level image, using the gray level image as pavement image.
It optionally, can be according to the coordinate of end point in the gray level image point of distant place (parallel lines disappear to), from gray scale
ROI (interested) region, i.e. road area are intercepted in image, using the ROI region of interception as pavement image.
As shown in figure 5, pavement image can be divided into multiple submodule 501, the shape of each submodule 501 can be
Arbitrary shape, such as rectangle, trapezoidal, the form similar with track wire shaped such as parallelogram can also be other irregular shapes
Shape.The size of each submodule 501 can be the same or different, specifically can be according to warp of the lane line in image distribution position
Test the shape and size that value (is obtained) setting submodule by tracking data or statistical data.For example, submodule 501 can be arranged to
Rectangle, the subblock size in pavement image centre position can more greatly, the submodule in pavement image centre position both sides
Size can be a little bit smaller.In one embodiment, pavement image can be divided into N number of submodule that size is identical and shape is identical
Block 501.
At step 420, according to the pixel value of feature pixel in the submodule, the feature of the submodule is calculated
Value；
Wherein, feature pixel refers to the big pixel of gradient is high in current sub-block pixel or gray scale, that is,
Strong feature pixel.Each submodule is scanned one by one, each submodule feature pixel in the block can be marked, Ran Hougen
According to the pixel value of feature pixels all in current sub-block, the average value of the pixel value by calculating all feature pixels,
Using the average value as the characteristic value of current sub-block.As needed, all character pixels in current sub-block can also be calculated
Intermediate value, maximum value or the mode of the pixel value of point, as the characteristic value of current sub-block.
In a kind of exemplary embodiment, as shown in fig. 6, above-mentioned steps 420 specifically include：
In step 421, according to the pixel value of pixel in the submodule, submodule character pixel in the block is determined
Point；
The method of feature pixel in determination sub-module can be according to the gradient or gray scale of the submodule, with highest gradient
Or greatest gradient is as feature pixel；Either ask for submodule average gradient value or gray value or request submodule block gradient
The variance of value or gray value, and retain one according to the pixel value and the magnitude relationship of mean value or variance of pixel each in submodule
The pixel of certainty ratio is as feature pixel.For example, according to the pixel value of pixel each in submodule, retain pixel
Pixel of the value more than in the pixel of mean value 30% is as feature pixel or preceding 20% pixel of reservation variance maximum
Point is as feature pixel, so as to which each submodule has a certain number of feature pixels.
Assuming that the pixel value of pixel is 230,180,90,215,80,150,175 in some submodule, then can calculate
It is 160 to go out mean value, and the pixel more than mean value is 230,180,215,175, then takes in these pixels preceding 30% pixel
As feature pixel, 230 can be taken in this example.With respect to mean value 160, preceding 20% larger pixel of variance is 230.Due to
Pixel existing for practical each submodule is more, can find out multiple character pixels in each submodule in this way
Point.These pixels are the pixels that pixel value is larger in the submodule.
By taking submodule block gradient as an example, each pixel is ranked up from big to small according to Grad, it is assumed that k-th submodule
Pixel sum be Spixel, choose the forward R that sortsgrad%*SpixelFeature pixel of a pixel as the submodule,
Rgrad% is preset feature pixel ratio.Assuming that the pixel sum of k-th submodule is 2000, Rgrad% values are
10%, that is to say, that each pixel value that will be sorted according to Grad chooses 200 forward (i.e. 10% × 2000) a pictures that sort
Vegetarian refreshments is as feature pixel.
In step 422, it by calculating the pixel average or pixel maximum of feature pixel in the submodule, obtains
To the characteristic value of the submodule.
It, can be by calculating all feature pixels in current sub-block after the feature pixel of each submodule is obtained
Pixel value mean value or maximum value, using the mean value or maximum value as the characteristic value of current sub-block.
In one embodiment, after the feature pixel for determining each submodule, the characteristic value of each submodule, feature are calculated
Value is asked for by taking the pixel mean value of feature pixel as an example, for example, the feature pixel of k-th submodule has SeigenIt is a, the submodule
Block eigenvalueGred (p) represents the pixel value of p-th of feature pixel.
In step 430, the submodule that characteristic value is higher than threshold value, the son filtered out are filtered out from multiple submodules
Module is as foreground area, other regions in the pavement image in addition to foreground area are as background area；
Specifically, all submodules can from high to low be carried out according to characteristic value according to the characteristic value of each submodule
Sequence, using the forward a certain proportion of submodule that sorts (i.e. characteristic value is higher than the submodule of threshold value) as foreground area.Due to
The pixel value of lane line position be higher than other regions pixel value, the submodule higher by filtering out characteristic value, also
It is to say the submodule that can be filtered out where lane line, therefore the foreground area being made of these submodules includes lane line
Region.And other regions in pavement image in addition to foreground area may be considered background area.Background area refers to not include
The region of lane line.
As shown in fig. 7, the submodule that characteristic value is higher than threshold value is filtered out from all submodules, as foreground area 701,
Using the region in pavement image in addition to foreground area 701 as background area 702.
In step 440, binary conversion treatment is carried out to the foreground area, the pixel value zero setting of background area obtains institute
State the binary image of pavement image；
It should be noted that according to theory, lane line region with respect to the gray value in other regions of road surface should differ compared with
Greatly, region of the road surface in addition to lane line answers gray value approximate, so as to be conducive to accurately extract lane line, but due to practical road
Road scene is complicated, and road surface is even by uneven illumination, larger so as to cause road grey scale change itself, the pixel ladder of normal lane line
It spends relatively low.Assuming that the pixel value of lane line region should be close to 210 under normal circumstances, and the pixel value in other regions of road surface
Should be close to 90, lane line pixel gradient is higher at this time, is distinguished conducive to other regions with road surface；When road surface, uneven illumination is even
When, the variation of 60-190 is presented in other area pixel values of road surface, then part road surface region is closer to lane line pixel value, vehicle
Diatom pixel gradient reduces, i.e., there are more background sideways pixels to interfere lane line pixel.Therefore using existing
The binary image that the global binarization method and local binarization method having obtain, it is serious by background sideways pixel interference,
So as to accurately can not therefrom extract lane line.
The present invention eliminates the relatively low submodule of pixel value (i.e. background area) first, and two are no longer participate in these submodules
Value, but directly avoid interfering the binaryzation of other submodules labeled as 0 by the pixel value of its pixel.For
The higher submodule of pixel value (i.e. foreground area) only carries out binary conversion treatment to foreground area, so as to the two-value of foreground area
It is small by the interference of background sideways pixel to change handling result.
Wherein, binary conversion treatment refers to that the gray value by the pixel in foreground area is set as 0 or 255, that is, will
Foreground area, which shows, significantly only has black and white visual effect.It specifically, can be according to submodule each in foreground area
Characteristic value, by pixel value be more than current sub-block characteristic value pixel gray value be set as 255, other pixel gray values
It is set as 0.The gray value of pixel in background area is set as 0, i.e. the pixel value zero setting of background area.Before after binaryzation
Background area after scene area and zero setting forms the binary image of pavement image.As shown in figure 8, for using provided by the present invention
Scheme carry out binary conversion treatment after binary image schematic diagram.
In step 450, the straight line for representing lane line is extracted in the binary image.
Specifically, the straight line for representing lane line can be extracted from binary image by existing line detection algorithm.
For example, can the non-zero pixel in the binary image be projected to by polar coordinates by Hough transformation line detection algorithm
In space, Hough matrix is formed, and local maximum is found according to the votes of coordinate each in Hough matrix, maximum is sat
The straight line in corresponding binary image is marked as lane line, detailed process can refer to the prior art.
Technical solution provided by the invention by the way that pavement image is divided into multiple submodule, and calculates each submodule
Characteristic value, filter out the higher submodule of characteristic value as foreground area carry out binary conversion treatment, so as to characteristic value it is relatively low
Submodule will not interfere the binaryzation of foreground area, and then can obtain the apparent binary image of foreground pixel, base
It is small due to being interfered by background pixel in the binary image into the detection of driveway line, lane line can be accurately extracted, into
And the accuracy that lane line deviates early warning can be improved.
Further, as shown in figure 9, above-mentioned steps 430 specifically include：
In step 431, selected characteristic value is higher than the submodule of first threshold, the son of selection from the multiple submodule
Module alternately module；
Wherein, alternate modular refers to that characteristic value is higher than the submodule of first threshold.Specifically, all submodules can be pressed
It is ranked up according to its characteristic value.To reduce calculation amount, can only keeping characteristics value it is sub higher than M of threshold value T (i.e. first threshold)
Module, M submodule of reservation alternately module.With reference to shown in Figure 10, each small rectangle represents a submodule, filters out
Characteristic value is higher than submodule alternately module 1004 of first threshold, i.e., shaded rectangle shown in Fig. 10.
In step 432, different alternate modulars are clustered according to characteristic value, it will will be different alternative according to cluster result
Module is divided into multiple cluster areas；
It needs what is explained, carries out clustering the characteristic value referred to according to each alternate modular according to characteristic value, characteristic value is connect
Near alternate modular is as a kind of.Wherein, cluster result refers to which alternate modular belongs to the close same class of characteristic value.Cluster
Region refers to the close alternate modular of characteristic value being labeled as same region.By the way that the close alternate modular of characteristic value is marked
For same cluster areas, and then all alternate modulars can be divided into multiple cluster areas.Wherein, characteristic value is close can be with
Judged using modes such as mean value or variances.Such as variance it is smaller when represent characteristic value approach.
As shown in Figure 10, the close alternate modular 1004 of characteristic value is classified as same class, the first cluster areas 101 includes spy
Multiple alternate modulars 1004 of the value indicative in the range of first, it is more in the range of second that the second cluster areas 102 includes characteristic value
A alternate modular 1004.Such as by the alternate modular 1004 of characteristic value 80-100 labeled as second cluster areas 102, by feature
The alternate modular 1004 of value 100-120 is labeled as first cluster areas 101.
In step 433, according to the characteristic value of the cluster areas Neutron module, determine that the combination of the cluster areas is special
Value indicative；
It, can be by calculating the feature of each submodule specifically, according to the characteristic value of each submodule in a certain cluster areas
The average value or maximum value of value, the assemblage characteristic value as the cluster areas.As shown in Figure 10, the first cluster areas 101 is calculated
The mean value of the characteristic value of interior all alternate modulars 1004, using the mean value as the assemblage characteristic value of the first cluster areas 101.It calculates
The mean value of the characteristic value of all alternate modulars 1004 in second cluster areas 102, using the mean value as the second cluster areas 102
Assemblage characteristic value.
In step 434, the cluster areas that assemblage characteristic value is higher than second threshold is chosen from multiple cluster areas,
The cluster areas chosen is as foreground area, other regions in the pavement image in addition to foreground area are as background area
Domain, the first threshold are less than second threshold.
It should be noted that first threshold is less than second threshold, it is therefore an objective to further submodule be screened, before reduction
Scape character pixel.Pass through empirically determined second threshold according to the image area characteristics of lane line.According to the group of each cluster areas
Characteristic value is closed, the cluster areas that assemblage characteristic value is higher than second threshold is filtered out, using these cluster areas filtered out as before
Scene area carries out binaryzation to foreground area.It other regions in pavement image in addition to foreground area, will as background area
The pixel gray value of background area is set as 0.
This mode generated excessive character pixel after can not only having prevented binary conversion treatment, but also when can solve uneven illumination
Normal lane line gradient is low, it is difficult to the problem of detecting.
As shown in Figure 10, it is assumed that the assemblage characteristic value of the first cluster areas 101 be 110, more than second threshold (such as
100), and the assemblage characteristic value (such as 90) of the second cluster areas 102 is less than second threshold, then the first cluster areas 101 belongs to
Foreground area, all areas in addition to the first cluster areas 101 belong to background area (including the second cluster areas 102).
Binaryzation is carried out to foreground area, background area gray scale is set to 0, and then can detect lane line 1003 in foreground area.
In a kind of exemplary embodiment, as shown in figure 11, step 432, different alternate modulars are carried out according to characteristic value
Different alternate modulars will be divided into multiple cluster areas according to cluster result and included by cluster：
In step 1101, by the pavement image it is longitudinally divided be at least two target areas；
Since the gray scale at left and right sides of pavement image is uneven, as shown in figure 12, by area limit line by pavement image
Longitudinally divided is two target areas in left and right.Submodule is divided into two parts by region segmentation line.It at this time can be respectively from a left side
It is extracted in right two target areas into driveway line.
In step 1102, the alternate modular is drawn in the position according to the alternate modular in the pavement image
Divide to affiliated target area；
By taking two target areas as an example, according to position of the alternate modular in pavement image, all alternate modulars are belonged to
To two target areas.Specifically, the alternate modular in current goal region is belonged to current goal region, if some is alternative
Module occupies two target areas, then can determine that this is standby according to its area distributions ratio or internal feature pixel distribution situation
The target area of modeling block ownership.Such as the 60% of some alternate modular is all in some target area, then the alternate modular category
In the target area.Such as 60% feature pixel of some alternate modular is all in some target area, then the alternative mould
Block belongs to the target area.
In step 1103, the alternate modular in the target area is clustered according to characteristic value, by the target
Alternate modular in region is divided into multiple cluster areas according to cluster result.
The alternate modular belonged in same target area is clustered according to characteristic value, that is to say, that by a mesh
It marks the alternate modular that characteristic value is close in region and is labeled as same cluster areas, so as to which each target area can mark off
Multiple cluster areas.As shown in figure 12, it is assumed that each rectangle represents an alternate modular, the close alternate modular of characteristic value all by
Labeled as same cluster areas, there are multiple cluster areas on the left of region segmentation line, there are multiple cluster areas on the right side of region segmentation line
Domain.
As shown in figure 12, pavement image can be divided by area limit line by two target areas in left and right, left side mesh
Alternate modular in mark region is marked as 4 cluster areas (i.e. cluster areas 1,2,3,4).Wherein, cluster areas 1 includes
3 alternate modulars.Alternate modular in the target area of right side is merged into 3 cluster areas (i.e. merging module 1,2,3).Its
In, cluster areas 1 includes 4 alternate modulars.The cluster areas in two target areas can be screened respectively later,
Each target area retains higher preceding 3 cluster areas of assemblage characteristic value, and the cluster areas of reservation is carried out as foreground area
Binary conversion treatment.
In a kind of exemplary embodiment, as shown in figure 13, above-mentioned steps 450 specifically include：
In step 451, straight-line detection is carried out to the binary image by Hough transformation, is obtained a plurality of candidate straight
Line；
It should be noted that due to that may have the straight line of a plurality of non-lane line in pavement image, such as road boundary, protective fence
Deng.These straight lines cause lane detection strong jamming, and detected straight line could possibly be higher than correct lane line quantity.By
This, the candidate straight line of lane line is may be considered by the straight line that Hough transformation is directly detected from binary image, is needed
These candidate's straight lines are screened, remove road boundary and the corresponding straight line of protective fence.
In step 452, according to the candidate straight line in the pavement image number of occupied submodule, occupied
The straight line votes of the characteristic value of submodule and the candidate straight line determine the weight of the candidate straight line；
Where it is assumed that the length of candidate straight line is L, it is distributed in 2 submodules, the length included in each submodule is
L/2, at this time it is considered that candidate's straight line occupies 2 submodules.But if one of submodule in the Y-axis direction
Width be more than or equal to L, it is believed that candidate's straight line only occupies 1 submodule.It specifically, can be by putting down up and down
Submodule is moved, all pixels point allowed as far as possible on candidate straight line includes less submodule.Submodule number at this time is exactly
Submodule number occupied by candidate straight line, is denoted as Lsub.The submodule number that i-th candidate straight line occupies can be denoted as Lsub
(i)。
As shown in figure 14, it is assumed that candidate straight length is h, and each submodule width is H, and a candidate straight line part exists at this time
First submodule, a part is in second submodule.Assuming that h is less than or equal to H, directly say that candidate straight line needs to occupy two sons
Module is not just right, because a sub- module width H just has been above h.Therefore second submodule can be translated up, it is believed that
Candidate straight line need to only be in a submodule or first submodule translates downwards, it is believed that candidate straight line need to only be in one
Submodule, all pixels point allowed as far as possible on candidate straight line include less submodule.
Wherein, the characteristic value of submodule occupied by candidate straight line refers to the characteristic value for the submodule that candidate straight line is distributed,
It is denoted as Teigen.The corresponding characteristic value of t-th of submodule occupied by i-th candidate straight line can be denoted as
Wherein, what the straight line votes of candidate straight line referred in Hough matrix candidate line correspondences parameter coordinate goes out occurrence
Number, can be denoted as count.
Specifically, the weight of i-th candidate straight line can be according to weightWcountIt is comprehensive to determine, such as
It willWcountThree weights are multiplied or are added.The weight of submodule number occupied by expression,The weight of occupied submodule block eigenvalue, WcountRepresent the weight of straight line votes.For example, a submodule is occupied
When, the weight of occupied submodule numberA can be denoted as.When straight line votes are b, the weight W of straight line votescount
B can be denoted as.When the characteristic value of occupied submodule is 50~70, the weight of occupied submodule block eigenvalueIt is 1,
When the characteristic value of occupied submodule is 70~90, the weight of occupied submodule block eigenvalueBe 2, and so on progress
Assignment.
It should be noted that when i-th straight line occupies more than one submodule, the weight of occupied submodule block eigenvalueIt can be the weight of occupied each submodule block eigenvalueThe sum of.
In a kind of exemplary embodiment, the submodule quantity that i-th candidate straight line occupies is Lsub(i) it is a, t-th of son
The corresponding characteristic value of module isThe weight W of i-th candidate straight lineiAccording to weight And WcountReally
Fixed, calculation is by taking formula below as an example：
In step 453, according to the weight of the candidate straight line, weight selection meets pre- from a plurality of candidate straight line
If the candidate straight line of condition obtains representing the straight line of lane line in the binary image as target line.
Wherein, preset condition can be weight selection be more than preset value candidate straight line, the larger certain ratio of weight selection
The candidate straight line of example, the preceding x straight line of weight selection maximum.It in one embodiment, can be according to every candidate straight line
Weight is ranked up candidate straight line according to weight size, it is assumed that the straight line quantity exported is needed then to export sequence for x items and lean on
Preceding preceding x items candidate's straight line, obtains the lane line in binary image.In another embodiment, power can directly be exported
The great candidate straight line in preset value.
In a kind of exemplary embodiment, as shown in figure 15, step 453 specifically includes：
In step 4531, according to position of a plurality of candidate straight line in the pavement image, to a plurality of time
Straight line is selected to be grouped according to present position, forms several straight line set；
For needing to export two tracks, pavement image can be longitudinally divided into left and right two by an area limit line
A target area, according to candidate straight line in pavement image position, can using the candidate straight line in same target area as
Candidate straight line is divided into two groups as second group by one group, candidate straight line in another target area, obtain two it is straight
Line set.Candidate straight line quantity is denoted as z in each straight line set,
As needed, can by pavement image by two area limit lines by pavement image it is longitudinally divided be 3 targets
Region, using the candidate straight line in same target area as a group.
And so on, it can be longitudinally divided for n+1 target by pavement image by n region segmentation line by pavement image
Region using the candidate straight line in same target area as a group, that is, forms a straight line set.
In step 4532, candidate straight line quantity and track line number to be extracted according to included in straight line set
Amount, chooses the candidate straight line of identical quantity as target line, the target line is in residing straight line from different straight line set
Weight highest in set, the target line is as the lane line in the pavement image.
Assuming that being divided into the candidate straight line of c groups, each (i.e. each straight line set) corresponding candidate straight line number that is grouped is number
(c).Assuming that it is L to need the straight line quantity exportedneed, then the mode for choosing straight line isNamely
It says, every group of candidate's straight line is ranked up according to weight height, straight line quantity as needed is chosen in every group of candidate's straight line
Several candidate straight lines for sorting forward are as lane line.
For example, it is divided into 2 groups, every group there are 3 candidate straight lines, when to need the straight line quantity exported be 2, from first group
The highest candidate straight line of 1 weight is selected, the highest candidate straight line of a weight is selected from second group.If necessary to output
Straight line quantity is 4, and highest first 2 candidate straight lines of weight are selected from first group, select weight highest from first group
First 2 candidate straight lines.That is, the candidate straight line quantity that each grouping is chosen is identical, and group during the candidate straight line of selection
The highest candidate straight line of interior weight.
If the straight line number number (j) of jth group is less than the straight line number that the group needs, then from group output number (j) item
Straight line, other groupings synchronous with the group can be reduced, and can also be exported according to the output straight line number set before.That is,
It, can be based on the candidate straight line quantity in the grouping, out of each grouping when the candidate straight line negligible amounts in some grouping
Choose the straight line of current quantity.
For example, when being divided into 2 groups, first group has 4 candidate straight lines, and second group is only had 2 candidate straight lines, it is assumed that is needed
The candidate straight line of output 6, the 3 candidate straight lines that can select weight sequencing forward from first group under normal circumstances, from second
The 3 candidate straight lines that can select weight sequencing forward in group.But second group is only had 2 candidate straight lines, less than 3.At this point,
2 that select weight sequencing forward from first group candidate straight lines can be selected, 2 candidate straight lines are also selected in second group.It needs
It is noted that can be configured as needed, selection and candidate straight line the same number of in second group from first group,
3 straight lines can be exported according to preset need.
As shown in figure 16,3 candidate straight lines are detected by Hough transformation, it is assumed that only choose the corresponding left and right of current lane
Two lane lines, the votes straight line 1 and straight line 2 of hough transformation statistics are approximate, it is difficult to it distinguishes, according toIt is 2 that straight line 1, which occupies submodule quantity, and it is 3 that straight line 2, which occupies submodule quantity, and the submodule of straight line 2
Characteristic value is higher than submodule 1, therefore weight sequencing is forward for straight line 2, then straight line 2 and straight line 3 this two straight lines are as final
Output result.
Aforesaid way can preferentially export the highest candidate straight line of weight, solution by calculating the weight of every candidate straight line
The certainly straight line interference problem of the non-lane line such as road boundary, protective fence.
Following is apparatus of the present invention embodiment, can be used for performing the above-mentioned method for detecting lane lines embodiment of the present invention.It is right
The details not disclosed in apparatus of the present invention embodiment please refers to method for detecting lane lines embodiment of the present invention.
Figure 17 is according to a kind of block diagram of lane detection device shown in an exemplary embodiment, lane detection dress
Put the whole that can be used for performing any shown method for detecting lane lines of Fig. 4, Fig. 6, Fig. 9, Figure 11, Figure 13, Figure 15 or portion
Step by step.As shown in figure 17, which includes but not limited to：Image division module 1510, characteristic value calculating module 1520, prospect
Screening module 1530, binarization block 1540, lines detection module 1550.
Image division module 1510 for obtaining the pavement image for including lane line, the pavement image is divided into more
A submodule；
Characteristic value calculating module 1520 for the pixel value according to feature pixel in the submodule, calculates the son
The characteristic value of module；
Prospect screening module 1530, for filtering out the submodule that characteristic value is higher than threshold value from multiple submodules,
The submodule filtered out is as foreground area, other regions in the pavement image in addition to foreground area are as background area；
Binarization block 1540, for the foreground area carry out binary conversion treatment, the pixel value zero setting of background area,
Obtain the binary image of the pavement image；
Lines detection module 1550, for extracting the straight line for representing lane line in the binary image.
The function of modules and the realization process of effect specifically refer to above-mentioned method for detecting lane lines side in above device
The realization process of step is corresponded in method, details are not described herein.
Optionally, as shown in figure 18, prospect screening module 1530 includes but not limited to：
Alternate modular selection unit 1531, for from the multiple submodule selected characteristic value be higher than first threshold son
Module, the submodule of selection alternately module；
Region clustering unit 1532, will according to cluster result for being clustered to different alternate modulars according to characteristic value
Different alternate modulars are divided into multiple cluster areas；
Regional characteristic value determination unit 1533 for the characteristic value according to the cluster areas Neutron module, determines described
The assemblage characteristic value of cluster areas；
Region selection unit 1534, for choosing assemblage characteristic value from multiple cluster areas higher than second threshold
Cluster areas, the cluster areas of selection is as foreground area, other regions in the pavement image in addition to foreground area
As background area, the first threshold is less than second threshold.
Optionally, the present invention also provides a kind of lane detection terminal, which can perform the present invention
Any shown method for detecting lane lines of Fig. 4, Fig. 6, Fig. 9, Figure 11, Figure 13, Figure 15 is all or part of in embodiment of the method
Step.The lane detection terminal includes：
Processor；
For storing the memory of processor-executable instruction；
Wherein, the processor is configured as performing the method for detecting lane lines described in the above exemplary embodiments.
The processor of lane detection terminal performs the concrete mode of operation in the related lane line in the embodiment
Detailed description is performed in the embodiment of detection method, explanation will be not set forth in detail herein.
As needed, which can also include CCD camera assembly, pass through between CCD camera assembly and processor
Wired or wireless connection.Specifically, CCD camera assembly may be mounted at the headstock of vehicle, towards vehicle heading, adopt in real time
Collect the two dimensional image of vehicle heading.The tailstock of vehicle, the two dimensional image at real-time collection vehicle rear can also be mounted on.It takes the photograph
As the two dimensional image of acquisition is sent to the processor by head assembly.Scheme provided by the invention may be used to camera in processor
The two dimensional image of component acquisition carries out lane detection.
As needed, the two dimensional image of acquisition can also be sent to by CCD camera assembly or processor has data processing energy
The server of power.Server may be used scheme provided by the invention and lane detection carried out to the two dimensional image of reception, and will
Lane detection terminal is back to during lane detection fructufy to be shown.
Figure 19 is the structure diagram of a kind of lane detection terminal 200 that exemplary embodiment of the present provides.Such as figure
Shown in 19, lane detection terminal 200 can include following one or more components：Processing component 202, memory 204, power supply
Component 206, multimedia component 208, audio component 210, sensor module 214 and communication component 216.
The integrated operation of the usually control lane detection terminal 200 of processing component 202, such as with display, data communication, phase
Machine operates and record operates associated operation etc..Processing component 202 can be performed including one or more processors 218
Instruction, to complete all or part of step of the above method.Memory 204 is configured as storing various types of data to support
In the operation of lane detection terminal 200.The example of these data includes appointing for what is operated in lane detection terminal 200
The instruction of what application program or method.
Power supply module 206 provides electric power for the various assemblies of lane detection terminal 200.Multimedia component 208 is included in
The screen of one output interface of offer between the lane detection terminal 200 and user.Audio component 210 is configured as defeated
Go out and/or input audio signal.In some embodiments, audio component 210 further includes a loud speaker, for exporting audio letter
Number.Sensor module 214 includes one or more sensors, for providing the shape of various aspects for lane detection terminal 200
State is assessed.Communication component 216 is configured to facilitate wired or wireless way between lane detection terminal 200 and other equipment
Communication.
In the exemplary embodiment, a kind of storage medium is additionally provided, which is computer readable storage medium,
Such as can be the provisional and non-transitorycomputer readable storage medium for including instruction.The storage medium is stored with computer
Program, the computer program can be performed to complete above-mentioned method for detecting lane lines by the processor of lane detection terminal.
It should be understood that the invention is not limited in the precision architecture for being described above and being shown in the drawings, and
And various modifications and change can be being performed without departing from the scope.The scope of the present invention is only limited by appended claim.
Claims (10)
1. a kind of method for detecting lane lines, which is characterized in that the method includes：
The pavement image for including lane line is obtained, the pavement image is divided into multiple submodule；
According to the pixel value of feature pixel in the submodule, the characteristic value of the submodule is calculated；
The submodule that characteristic value is higher than threshold value is filtered out from multiple submodules, the submodule filtered out is as foreground zone
Domain, other regions in the pavement image in addition to foreground area are as background area；
Binary conversion treatment is carried out to the foreground area, the pixel value zero setting of background area obtains the two-value of the pavement image
Change image；
Extraction represents the straight line of the lane line in the binary image.
2. the according to the method described in claim 1, it is characterized in that, pixel according to feature pixel in the submodule
Value calculates the characteristic value of the submodule, including：
According to the pixel value of pixel in the submodule, submodule feature pixel in the block is determined；
By calculating the pixel average or pixel maximum of feature pixel in the submodule, the spy of the submodule is obtained
Value indicative.
3. according to the method described in claim 1, it is characterized in that, characteristic value is filtered out from multiple submodules higher than threshold
The submodule of value, the submodule filtered out as foreground area, make by other regions in the pavement image in addition to foreground area
For background area, including：
Selected characteristic value is higher than the submodule of first threshold from the multiple submodule, the submodule of selection alternately mould
Block；
Different alternate modulars according to characteristic value are clustered, different alternate modulars are divided into multiple clusters according to cluster result
Region；
According to the characteristic value of the cluster areas Neutron module, the assemblage characteristic value of the cluster areas is determined；
The cluster areas that assemblage characteristic value is higher than second threshold, the cluster area of selection are chosen from multiple cluster areas
Domain is as foreground area, other regions in the pavement image in addition to foreground area are as background area, the first threshold
Less than second threshold.
4. according to the method described in claim 3, it is characterized in that, described gather different alternate modulars according to characteristic value
Different alternate modulars are divided into multiple cluster areas according to cluster result and included by class：
By the pavement image it is longitudinally divided be at least two target areas；
According to position of the alternate modular in the pavement image, the alternate modular is divided to affiliated target area
Domain；
Alternate modular in the target area according to characteristic value is clustered, the alternate modular in the target area is pressed
Multiple cluster areas are divided into according to cluster result.
5. according to the method described in claim 1, represent the lane line it is characterized in that, being extracted in the binary image
Straight line, including：
Straight-line detection is carried out to the binary image by Hough transformation, obtains a plurality of candidate straight line；
According to the candidate straight line in the pavement image number of occupied submodule, the characteristic value of occupied submodule with
And the straight line votes of the candidate straight line, determine the weight of the candidate straight line；
According to the weight of the candidate straight line, weight selection meets the candidate straight line of preset condition from a plurality of candidate straight line
As target line, obtain representing the straight line of lane line in the binary image.
6. according to the method described in claim 5, it is characterized in that, according to the weight of the candidate straight line, from a plurality of time
The candidate straight line that weight selection meets preset condition in straight line is selected to obtain representing vehicle in the binary image as target line
The straight line of diatom, including：
According to position of a plurality of candidate straight line in the pavement image, to a plurality of candidate straight line according to present position
It is grouped, forms several straight line set；
Candidate straight line quantity and lane line quantity to be extracted according to included in straight line set, from different straight line set
The middle candidate straight line for choosing identical quantity is as target line, weight highest of the target line in residing straight line set,
The target line is as the lane line in the pavement image.
7. a kind of lane detection device, which is characterized in that described device includes：
The pavement image for obtaining the pavement image for including lane line, is divided into multiple submodule by image division module；
Characteristic value calculating module for the pixel value according to feature pixel in the submodule, calculates the spy of the submodule
Value indicative；
Prospect screening module for filtering out the submodule that characteristic value is higher than threshold value from multiple submodules, filters out
Submodule is as foreground area, other regions in the pavement image in addition to foreground area are as background area；
Binarization block, for carrying out binary conversion treatment to the foreground area, the pixel value zero setting of background area obtains described
The binary image of pavement image；
Lines detection module, for extracting the straight line for representing lane line in the binary image.
8. device according to claim 7, which is characterized in that the prospect screening module includes：
Alternate modular selection unit, for from the multiple submodule selected characteristic value be higher than first threshold submodule, choosing
The submodule taken alternately module；
Region clustering unit, will be different alternative according to cluster result for being clustered to different alternate modulars according to characteristic value
Module is divided into multiple cluster areas；
Regional characteristic value determination unit for the characteristic value according to the cluster areas Neutron module, determines the cluster areas
Assemblage characteristic value；
Region selection unit, for choosing the cluster area that assemblage characteristic value is higher than second threshold from multiple cluster areas
Domain, the cluster areas of selection is as foreground area, other regions in the pavement image in addition to foreground area are as the back of the body
Scene area, the first threshold are less than second threshold.
9. a kind of lane detection terminal, which is characterized in that the terminal includes：
Processor；
For storing the memory of processor-executable instruction；
Wherein, the processor is configured as the method for detecting lane lines described in perform claim requirement 1-6 any one.
10. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer journey
Sequence, the computer program can be performed the method for detecting lane lines described in completing claim 1-6 any one as processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810208562.3A CN108205667A (en) | 2018-03-14 | 2018-03-14 | Method for detecting lane lines and device, lane detection terminal, storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810208562.3A CN108205667A (en) | 2018-03-14 | 2018-03-14 | Method for detecting lane lines and device, lane detection terminal, storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108205667A true CN108205667A (en) | 2018-06-26 |
Family
ID=62606927
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810208562.3A Pending CN108205667A (en) | 2018-03-14 | 2018-03-14 | Method for detecting lane lines and device, lane detection terminal, storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108205667A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109034047A (en) * | 2018-07-20 | 2018-12-18 | 京东方科技集团股份有限公司 | A kind of method for detecting lane lines and device |
| CN109635737A (en) * | 2018-12-12 | 2019-04-16 | 中国地质大学（武汉） | Automobile navigation localization method is assisted based on pavement marker line visual identity |
| CN109886122A (en) * | 2019-01-23 | 2019-06-14 | 珠海市杰理科技股份有限公司 | Method for detecting lane lines, device, computer equipment and storage medium |
| CN109886081A (en) * | 2018-12-31 | 2019-06-14 | 武汉中海庭数据技术有限公司 | A kind of lane line form point string extracting method and device |
| CN109948504A (en) * | 2019-03-13 | 2019-06-28 | 东软睿驰汽车技术（沈阳）有限公司 | A kind of Lane detection method and device |
| CN110163930A (en) * | 2019-05-27 | 2019-08-23 | 北京百度网讯科技有限公司 | Lane line generation method, device, equipment, system and readable storage medium storing program for executing |
| CN111126209A (en) * | 2019-12-09 | 2020-05-08 | 上海博泰悦臻电子设备制造有限公司 | Lane line detection method and related equipment |
| CN111460072A (en) * | 2020-04-01 | 2020-07-28 | 北京百度网讯科技有限公司 | Lane line detection method, apparatus, device, and storage medium |
| CN111738034A (en) * | 2019-03-25 | 2020-10-02 | 杭州海康威视数字技术股份有限公司 | Method and device for detecting lane line |
| CN114724118A (en) * | 2022-06-08 | 2022-07-08 | 苏州魔视智能科技有限公司 | Zebra crossing detection method, device, equipment and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102592114A (en) * | 2011-12-26 | 2012-07-18 | 河南工业大学 | Method for extracting and recognizing lane line features of complex road conditions |
| CN102930279A (en) * | 2012-09-29 | 2013-02-13 | 广西工学院 | Image identification method for detecting product quantity |
| CN103136754A (en) * | 2012-10-09 | 2013-06-05 | 四川师范大学 | Image blurring direction identification method based on characteristic block direction differential |
| CN103559695A (en) * | 2013-09-29 | 2014-02-05 | 哈尔滨工程大学 | Adaptive block compressive sensing reconstruction method |
| CN103914155A (en) * | 2013-01-09 | 2014-07-09 | 奇高电子股份有限公司 | Optical navigation method and related device |
| CN104463097A (en) * | 2014-10-31 | 2015-03-25 | 武汉工程大学 | High-voltage wire image detection method based on local self-adaptation threshold value partitioning algorithm |
| CN105101878A (en) * | 2013-04-05 | 2015-11-25 | 东芝医疗系统株式会社 | Medical image processing apparatus and medical image processing method |
| CN105631880A (en) * | 2015-12-31 | 2016-06-01 | 百度在线网络技术（北京）有限公司 | Lane line segmentation method and apparatus |
-
2018
- 2018-03-14 CN CN201810208562.3A patent/CN108205667A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102592114A (en) * | 2011-12-26 | 2012-07-18 | 河南工业大学 | Method for extracting and recognizing lane line features of complex road conditions |
| CN102930279A (en) * | 2012-09-29 | 2013-02-13 | 广西工学院 | Image identification method for detecting product quantity |
| CN103136754A (en) * | 2012-10-09 | 2013-06-05 | 四川师范大学 | Image blurring direction identification method based on characteristic block direction differential |
| CN103914155A (en) * | 2013-01-09 | 2014-07-09 | 奇高电子股份有限公司 | Optical navigation method and related device |
| CN105101878A (en) * | 2013-04-05 | 2015-11-25 | 东芝医疗系统株式会社 | Medical image processing apparatus and medical image processing method |
| CN103559695A (en) * | 2013-09-29 | 2014-02-05 | 哈尔滨工程大学 | Adaptive block compressive sensing reconstruction method |
| CN104463097A (en) * | 2014-10-31 | 2015-03-25 | 武汉工程大学 | High-voltage wire image detection method based on local self-adaptation threshold value partitioning algorithm |
| CN105631880A (en) * | 2015-12-31 | 2016-06-01 | 百度在线网络技术（北京）有限公司 | Lane line segmentation method and apparatus |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109034047B (en) * | 2018-07-20 | 2021-01-22 | 京东方科技集团股份有限公司 | Lane line detection method and device |
| CN109034047A (en) * | 2018-07-20 | 2018-12-18 | 京东方科技集团股份有限公司 | A kind of method for detecting lane lines and device |
| US10970566B2 (en) | 2018-07-20 | 2021-04-06 | Boe Technology Group Co., Ltd. | Lane line detection method and apparatus |
| CN109635737A (en) * | 2018-12-12 | 2019-04-16 | 中国地质大学（武汉） | Automobile navigation localization method is assisted based on pavement marker line visual identity |
| CN109635737B (en) * | 2018-12-12 | 2021-03-26 | 中国地质大学（武汉） | Auxiliary vehicle navigation positioning method based on road marking line visual identification |
| CN109886081A (en) * | 2018-12-31 | 2019-06-14 | 武汉中海庭数据技术有限公司 | A kind of lane line form point string extracting method and device |
| CN109886122A (en) * | 2019-01-23 | 2019-06-14 | 珠海市杰理科技股份有限公司 | Method for detecting lane lines, device, computer equipment and storage medium |
| CN109886122B (en) * | 2019-01-23 | 2021-01-29 | 珠海市杰理科技股份有限公司 | Lane line detection method, lane line detection device, computer device, and storage medium |
| CN109948504A (en) * | 2019-03-13 | 2019-06-28 | 东软睿驰汽车技术（沈阳）有限公司 | A kind of Lane detection method and device |
| CN109948504B (en) * | 2019-03-13 | 2022-02-18 | 东软睿驰汽车技术（沈阳）有限公司 | Lane line identification method and device |
| CN111738034A (en) * | 2019-03-25 | 2020-10-02 | 杭州海康威视数字技术股份有限公司 | Method and device for detecting lane line |
| CN111738034B (en) * | 2019-03-25 | 2024-02-23 | 杭州海康威视数字技术股份有限公司 | Lane line detection method and device |
| CN110163930A (en) * | 2019-05-27 | 2019-08-23 | 北京百度网讯科技有限公司 | Lane line generation method, device, equipment, system and readable storage medium storing program for executing |
| CN111126209A (en) * | 2019-12-09 | 2020-05-08 | 上海博泰悦臻电子设备制造有限公司 | Lane line detection method and related equipment |
| CN111460072A (en) * | 2020-04-01 | 2020-07-28 | 北京百度网讯科技有限公司 | Lane line detection method, apparatus, device, and storage medium |
| CN111460072B (en) * | 2020-04-01 | 2023-10-03 | 北京百度网讯科技有限公司 | Lane line detection method, device, equipment and storage medium |
| CN114724118A (en) * | 2022-06-08 | 2022-07-08 | 苏州魔视智能科技有限公司 | Zebra crossing detection method, device, equipment and storage medium |
| CN114724118B (en) * | 2022-06-08 | 2022-09-27 | 苏州魔视智能科技有限公司 | Zebra crossing detection method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108205667A (en) | Method for detecting lane lines and device, lane detection terminal, storage medium | |
| CN106683076B (en) | The method of train wheel tread damage detection based on textural characteristics cluster | |
| CN102663760B (en) | Location and segmentation method for windshield area of vehicle in images | |
| CN106127137A (en) | A kind of target detection recognizer based on 3D trajectory analysis | |
| CN105844621A (en) | Method for detecting quality of printed matter | |
| CN105261017A (en) | Method for extracting regions of interest of pedestrian by using image segmentation method on the basis of road restriction | |
| CN106682586A (en) | Method for real-time lane line detection based on vision under complex lighting conditions | |
| CN111915583B (en) | Vehicle and pedestrian detection method based on vehicle-mounted thermal infrared imager in complex scene | |
| CN110119726A (en) | A kind of vehicle brand multi-angle recognition methods based on YOLOv3 model | |
| CN109726717A (en) | A kind of vehicle comprehensive information detection system | |
| CN107832674B (en) | Lane line detection method | |
| CN105956632A (en) | Target detection method and device | |
| CN106127145A (en) | Pupil diameter and tracking | |
| CN110866430A (en) | License plate recognition method and device | |
| CN111476804A (en) | Method, device and equipment for efficiently segmenting carrier roller image and storage medium | |
| CN104484680A (en) | Multi-model multi-threshold combined pedestrian detection method | |
| CN106529441B (en) | Depth motion figure Human bodys' response method based on smeared out boundary fragment | |
| CN104182728A (en) | Vehicle logo automatic location and recognition method based on pattern recognition | |
| CN108830274A (en) | Characters on license plate detects positioning and recognition methods automatically under a kind of complex background environment | |
| CN101369312B (en) | Method and equipment for detecting intersection in image | |
| CN110163109A (en) | A kind of lane line mask method and device | |
| CN112085675A (en) | Depth image denoising method, foreground segmentation method and human motion monitoring method | |
| CN111079675A (en) | Driving behavior analysis method based on target detection and target tracking | |
| CN112801021B (en) | Method and system for detecting lane line based on multi-level semantic information | |
| CN104966064A (en) | Pedestrian ahead distance measurement method based on visual sense |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180626 |