CN103714538B - road edge detection method, device and vehicle - Google Patents
road edge detection method, device and vehicle Download PDFInfo
- Publication number
- CN103714538B CN103714538B CN201310711070.3A CN201310711070A CN103714538B CN 103714538 B CN103714538 B CN 103714538B CN 201310711070 A CN201310711070 A CN 201310711070A CN 103714538 B CN103714538 B CN 103714538B
- Authority
- CN
- China
- Prior art keywords
- line segment
- straight
- road
- pixel
- curb
- 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 Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of road edge detection method, including: obtain the picture frame of the road edge information comprising the present road that vehicle is travelled;Picture frame is carried out rim detection to obtain multiple marginal points;Multiple marginal point is utilized to extract multiple straight-line segment;Curb architectural characteristic according to present road extracts curb line segment from multiple straight-line segment.The invention also discloses a kind of road edge detection device, vehicle.By the way, the present invention is capable of the curb line segment of the present road that automatic detection vehicle is travelled, and the operation complexity and the accuracy of detection that reduce tractor driver are higher.
Description
Technical field
The present invention relates to field of information processing, particularly relate to a kind of road edge detection method, device and vehicle.
Background technology
The vehicles such as manned vehicle or automatic driving vehicle are expert at during sailing, it is often necessary to detection vehicle is travelled
The curb line segment of present road, in order to the actual range between subsequent calculations vehicle and curb line segment, it is ensured that the safety of vehicle
Travel.Prior art generally uses following two method carry out the detection of road edge: a kind of for tractor driver by vehicle
The curb line segment of illuminator detection present road;Another kind is installation photographic head in vehicle, is collecting road edge image
Rear this image that transmits in real time, in vehicle, carries out manual detection road edge for tractor driver.
Present inventor finds in long-term R & D, and two kinds of road edge detection method of prior art are for tractor driver's
Operation requires complex, and tractor driver's labor intensity is bigger;In the case of the light environments such as night are dark, tractor driver is difficult to see and cleans the street
Road Edge, accuracy of detection is relatively low.
Summary of the invention
The technical problem that present invention mainly solves is to provide a kind of road edge detection method, device and vehicle, it is possible to real
The curb line segment of the present road that existing automatic detection vehicle is travelled, the operation complexity and the accuracy of detection that reduce tractor driver are higher.
For solving above-mentioned technical problem, the technical scheme that the present invention uses is: provide a kind of road-edge detection side
Method, including: obtain the picture frame of the road edge information comprising the present road that vehicle is travelled;Picture frame is carried out edge inspection
Survey, to obtain multiple marginal point;Multiple marginal point is utilized to extract multiple straight-line segment;Curb architectural characteristic according to present road
Curb line segment is extracted from multiple straight-line segment.
Wherein, the step that picture frame carries out rim detection farther includes: obtain mark set in advance from picture frame
Topography in fixed point surrounding predetermined area;Rim detection is carried out in topography.
Wherein, the step carrying out rim detection in topography farther includes: calculate the pixel in topography
Gray average;Gray average according to the pixel in topography sets the Low threshold parameter of canny edge detection algorithm
With high threshold parameter, canny edge detection algorithm is utilized to carry out rim detection in topography.
Wherein, the curb architectural characteristic according to present road extracts the step bag of curb line segment from multiple straight-line segment
Include: according to the comparing result of the pixel distance between the actual range between the road edge of present road and straight-line segment and/
Or the contrast according to the actual color difference of the road edge both sides of present road with the pixel color difference of straight-line segment both sides
Result extracts curb line segment from multiple straight-line segment.
Wherein, right according to the pixel distance between the actual range between the road edge of present road and straight-line segment
Actual color difference than result and/or according to the road edge both sides of present road is poor with the pixel color of straight-line segment both sides
Different comparing result farther includes before extracting curb line segment from multiple straight-line segment: delete tiltedly from multiple straight-line segment
Rate is unsatisfactory for the straight-line segment of predetermined slope requirement.
Wherein, right according to the pixel distance between the actual range between the road edge of present road and straight-line segment
The step extracting curb line segment than result from multiple straight-line segment includes: utilize calibration coefficient will obtain under space coordinates
Present road road edge between actual range and under image coordinate system obtain straight-line segment between pixel away from
From being transformed into the same coordinate system, wherein calibration coefficient is by the fixed point set in advance actual coordinate under space coordinates and mark
Fixed point image coordinate under image coordinate system calculates and obtains;Under the same coordinate system to the road edge of present road between
Pixel distance between actual range and straight-line segment carries out difference operation, and therefrom selects the difference straight line less than redundant error
Line segment.
Wherein, according to the pixel color of the actual color difference of the road edge both sides of present road with straight-line segment both sides
The comparing result of difference extracts the step of curb line segment from multiple straight-line segment and includes: calculate each straight-line segment with adjacent
The gray average of the pixel between straight-line segment or in the predetermined lateral width range of each straight-line segment both sides, and according to ash
Degree average determines the pixel color difference of straight-line segment both sides;The pixel face of straight-line segment both sides is extracted from multiple straight-line segment
The actual color difference of the different road edge both sides with present road of aberration is consistent or straight-line segment in error allowed band.
Wherein, right according to the pixel distance between the actual range between the road edge of present road and straight-line segment
Than result with according to the pixel color difference of the actual color difference of the road edge both sides of present road with straight-line segment both sides
Comparing result from multiple straight-line segment, extract the step of curb line segment include: utilize the calibration coefficient will be under space coordinates
The picture between actual range and the straight-line segment obtained under image coordinate system between the road edge of the present road obtained
Element distance is transformed into the same coordinate system, and wherein calibration coefficient is by the fixed point set in advance actual coordinate under space coordinates
Calculate with fixed point image coordinate under image coordinate system and obtain;Under the same coordinate system to the road edge of present road it
Between actual range and straight-line segment between pixel distance carry out difference operation, and therefrom select difference less than redundant error
Multiple alternative straight line segments;Calculate between each alternative straight line segment and adjacent alternative straight line segment or each alternative straight line
The gray average of the pixel in the predetermined lateral width range of section both sides, and determine alternative straight line segment two according to gray average
The pixel color difference of side;The pixel color difference of alternative straight line segment both sides is extracted with current from multiple alternative straight line segments
The actual color difference of the road edge both sides of road alternative straight line segment unanimously or in error allowed band.
Wherein, method farther includes: utilize multiple to the subsequent image frames of follow-up acquisition of acquired curb line segment
Straight-line segment is tracked, and then extracts curb line segment from multiple straight-line segment of subsequent image frames.
Wherein, method farther includes: calculate curb line segment according to curb line segment pixel coordinate under image coordinate system
Relative to the actual range of vehicle under space coordinates.
For solving above-mentioned technical problem, what the present invention used another solution is that offer a kind of road-edge detection dress
Put, including: picture frame acquisition module, for obtaining the image of the road edge information comprising the present road that vehicle is travelled
Frame;Edge detection module, for carrying out rim detection to picture frame, to obtain multiple marginal point;Straight-line segment extraction module, uses
Multiple straight-line segment are extracted in utilizing multiple marginal point;Curb line segment extraction module, according to the curb architectural characteristic of present road
Curb line segment is extracted from multiple straight-line segment.
Wherein, edge detection module is further used for obtaining fixed point surrounding predetermined area set in advance from picture frame
Interior topography, and in topography, carry out rim detection.
Wherein, edge detection module is further used for calculating the gray average of the pixel in topography, and according to office
The gray average of the pixel in portion's image sets Low threshold parameter and the high threshold parameter of canny edge detection algorithm, and
Canny edge detection algorithm is utilized to carry out rim detection in topography.
Wherein, the actual range between curb line segment extraction module is further used for according to the road edge of present road with
The comparing result of the pixel distance between straight-line segment and/or the actual color difference of the road edge both sides according to present road
From multiple straight-line segment, curb line segment is extracted with the comparing result of the pixel color difference of straight-line segment both sides.
Wherein, straight-line segment extraction module is further used at curb line segment extraction module according to the roadside, road of present road
The comparing result of the pixel distance between actual range and straight-line segment between edge and/or the road edge according to present road
The comparing result of the actual color difference of both sides and the pixel color difference of straight-line segment both sides extracts from multiple straight-line segment
Before curb line segment, from multiple straight-line segment, delete slope be unsatisfactory for the straight-line segment of predetermined slope requirement.
Wherein, curb line segment extraction module is further used for utilizing calibration coefficient current by obtain under space coordinates
The pixel distance conversion between actual range and the straight-line segment obtained under image coordinate system between the road edge of road
To the same coordinate system, and between the actual range between the road edge of present road and straight-line segment under the same coordinate system
Pixel distance carry out difference operation, and therefrom select the difference straight-line segment less than redundant error, wherein calibration coefficient is by advance
The fixed point first set actual coordinate under space coordinates and the fixed point image coordinate under image coordinate system calculates and obtains
?.
Wherein, curb line segment extraction module be further used for calculating between each straight-line segment and adjacent straight-line segment or
The gray average of the pixel in the predetermined lateral width range of each straight-line segment both sides, and determine straight line according to gray average
The pixel color difference of line segment both sides, and then from multiple straight-line segment, extract the pixel color difference of straight-line segment both sides and work as
The actual color difference of the road edge both sides of front road straight-line segment unanimously or in error allowed band.
Wherein, curb line segment extraction module is further used for utilizing calibration coefficient current by obtain under space coordinates
The pixel distance conversion between actual range and the straight-line segment obtained under image coordinate system between the road edge of road
To the same coordinate system, and between the actual range between the road edge of present road and straight-line segment under the same coordinate system
Pixel distance carry out difference operation, and therefrom select the difference multiple alternative straight line segments less than redundant error, wherein demarcate
Coefficient by the fixed point set in advance actual coordinate under space coordinates and fixed point the image under image coordinate system sit
Mark calculates and obtains;Curb line segment extraction module is further used for calculating each alternative straight line segment and adjacent alternative straight line segment
Between or the predetermined lateral width range of each alternative straight line segment both sides in the gray average of pixel, and equal according to gray scale
Value determines the pixel color difference of alternative straight line segment both sides, and then extracts alternative straight line segment from multiple alternative straight line segments
The pixel color difference of both sides is consistent with the actual color difference of the road edge both sides of present road or in error allowed band
Interior alternative straight line segment.
Wherein, curb line segment extraction module is further used for the subsequent figure utilizing acquired curb line segment to follow-up acquisition
As multiple straight-line segment of frame are tracked, and then from multiple straight-line segment of subsequent image frames, extract curb line segment.
Wherein, device farther includes: actual distance calculation module, is used for according to curb line segment under image coordinate system
Pixel coordinate calculates curb line segment under space coordinates relative to the actual range of vehicle.
For solving above-mentioned technical problem, the another technical scheme that the present invention uses is: providing a kind of vehicle, this vehicle includes
The road edge detection device of a upper technical scheme.
The invention has the beneficial effects as follows: be different from the situation of prior art, the present invention comprises vehicle by acquisition and is travelled
The picture frame of road edge information of present road;Picture frame is carried out rim detection to obtain multiple marginal points;Further
Multiple marginal point is utilized to extract multiple straight-line segment;Finally according to the curb architectural characteristic of present road from multiple straight-line segment
Extract curb line segment;Being capable of the curb line segment of the present road that automatic detection vehicle is travelled, the operation reducing tractor driver is multiple
Miscellaneous degree and accuracy of detection are higher.
Accompanying drawing explanation
Fig. 1 is the flow chart of road edge detection method the first embodiment of the present invention;
Fig. 2 is the flow chart of road edge detection method the second embodiment of the present invention;
Fig. 3 is vehicle and the schematic diagram of road edge in road edge detection method the second embodiment of the present invention;
Fig. 4 is showing of the picture frame that comprises road edge information in road edge detection method the second embodiment of the present invention
It is intended to;
Fig. 5 is pixel distance and the schematic diagram of actual range in road edge detection method the second embodiment of the present invention;
Fig. 6 is the theory diagram of road edge detection device one embodiment of the present invention.
Detailed description of the invention
Below in conjunction with the accompanying drawing in embodiment of the present invention, the technical scheme in embodiment of the present invention is carried out clearly
Chu, it is fully described by, it is clear that described embodiment is only a part of embodiment of the present invention rather than whole realities
Execute mode.Based on the embodiment in the present invention, those of ordinary skill in the art are institute under not making creative work premise
The every other embodiment obtained, belongs to the scope of protection of the invention.
Referring to Fig. 1, road edge detection method the first embodiment of the present invention includes:
Step S101: obtain the picture frame of the road edge information comprising the present road that vehicle is travelled;
In this step, the picture frame of the road edge information comprising the present road that vehicle is travelled specifically can be by installing
The image capture devices such as video camera on vehicle or photographing unit are acquired, and video camera can be digital camera or simulation shooting
Machine.The photographic head of video camera can be infrared camera, so that all can carry out the collection of image by day with night.Real at other
Executing in mode, video camera and photographic head are alternatively other kinds, do not make too many restrictions.Vehicle can be to be in manned shape
State or the vehicle of automatic Pilot state.Above-mentioned picture frame, except comprising road edge information, may also include and plants in road side
Trees, the information being installed on other roads such as building in the other street lamp of road, roadside.
Step S102: picture frame is carried out rim detection, to obtain multiple marginal point;
In this step, rim detection is the common technique means in image procossing and computer vision, rim detection
Purpose is to identify the obvious point of brightness flop in digital picture, i.e. marginal point.The algorithm of rim detection can be the inspection of canny edge
Method of determining and calculating or other algorithms well known in the art, hereafter by as a example by canny edge detection algorithm to concrete side edge detection
Formula is described in detail.
Step S103: utilize multiple marginal point to extract multiple straight-line segment;
In this step, available Hough transform or other algorithms well known in the art obtain many from step S102
Individual marginal point extracts multiple straight-line segment.Contain in these straight-line segment including road edge information is all possible
Marginal information.
Step S104: extract curb line segment from multiple straight-line segment according to the curb architectural characteristic of present road.
In this step, curb structure include lane line (track white line or track yellow line etc.) and/or paving stone and/or
The structures such as kerbstone are in interior all curb structures.Curb line segment correspondence includes the left side line of lane line, right-hand line, and/or road
The left side line of face stone, right-hand line, and/or the left side line of kerbstone, right-hand line.Curb architectural characteristic can be by for various differences
The priori data that road design is obtained ahead of time carries out arranging acquisition, mainly can include the different width of lane line of road, road
The information such as the color Changing Pattern of face stone width, sidewalk width and above-mentioned zones of different.
It is appreciated that road edge detection method the first embodiment of the present invention is by working as that acquisition comprises that vehicle travelled
The picture frame of the road edge information of front road;Picture frame is carried out rim detection to obtain multiple marginal points;Further with
Multiple marginal points extract multiple straight-line segment;Curb architectural characteristic finally according to present road is extracted from multiple straight-line segment
Curb line segment, it is possible to realize the curb line segment of the present road that automatic detection vehicle is travelled, reduces the operation complexity of tractor driver
And accuracy of detection is higher.
Seeing also Fig. 2-5, road edge detection method the second embodiment of the present invention includes:
Step S201: obtain the picture frame of the road edge information comprising the present road that vehicle is travelled;
In this step, this picture frame is specifically entered by image capture devices such as the video camera being installed on vehicle or photographing units
Row gathers.Specifically, as it is shown on figure 3, the image capture device of present embodiment include preposition image capture device 3 and/or
Side image capture device 4, preposition image capture device 3 is specifically installed on driver's cabin 1 front of vehicle and comprises vehicle to gather
The picture frame of the relevant information of the road edge 5 in front, straight line 31 is the axis, visual angle of preposition image capture device 3, side
Image capture device 4 is specifically installed on vehicle body 2 side of vehicle to gather the relevant information of the road edge 5 comprising automobile side
Picture frame, straight line 41 is the axis, visual angle of side image capture device 4.Hereafter will be clapped with preposition image capture device 3
As a example by the picture frame taken the photograph, present invention is described, the concrete processing mode of the picture frame captured by side image capture device 4
Similar with the picture frame captured by preposition image capture device 3, do not repeat them here.
Step S202: obtain the topography in fixed point surrounding predetermined area set in advance from picture frame;
In this step, from picture frame, obtain the topography in fixed point P surrounding predetermined area set in advance, its
Process is described in detail below:
Picture frame is carried out greyscale transformation, and this picture frame is generally coloured image, by greyscale transformation by RGB color image
Be converted to gray level image, to promote the speed that picture frame is processed.
Obtain centered by fixed point P and area size as s × the topography R of tst, image-region RstThe choosing of size
Select lane line 51, paving stone 52 and kerbstone 53 in coverage diagram 4 to be preferred.Fixed point P can be according to actually detected needs
Carry out choosing setting, for preposition image capture device 3, as detected and vehicle distances road edge farther out, then choose from
The more longer-distance fixed point as front end of vehicle;For side image capture device 4, can be at side image capture device
The fixed point a little as side is chosen on the axis, visual angle 41 of 4.The fixed point of front end and the fixed point of side are setting
Afterwards, it is not necessary to repeat to set.Certainly, in other embodiments, above-mentioned topography R can chosenstCarry out the most again
Grey scale change, or also can not carry out gray scale change in the case of follow-up edge detection algorithm and Straight Line Extraction allow
Change.
Can be by set in advance additionally, farther include to obtain calibration coefficient λ, calibration coefficient λ after setting fixed point P
Fixed point P actual coordinate under space coordinates and the fixed point P image coordinate under image coordinate system calculates and obtains, in the past
As a example by putting the picture frame that image capture device 3 gathers, fixed point P abscissa under space coordinates is xActual, fixed point P exists
Abscissa under image coordinate system is xImage, the axis, visual angle 31 of preposition image capture device 3 with through fixed point P and vertical
It is respectively x in the intersection point of fixed point straight line 6 of the y-axis abscissa under space coordinates and image coordinate systemActual′、xImage',
Then calibration coefficientFor ensure road-edge detection precision, calibration coefficient λ should be less than 1, i.e. effective unit away from
From at least corresponding 1 image pixel distance.
Step S203: carry out rim detection in topography;
In this step, the step carrying out rim detection in topography includes:
The gray average of the pixel in calculating topography, shown in formula specific as follows (1):
Wherein, BL is the gray average of the pixel in topography, and (x y) is the gray scale of pixel in topography to f
Value.
Gray average BL according to the pixel in topography set canny edge detection algorithm Low threshold parameter and
High threshold parameter, utilizes canny edge detection algorithm to carry out rim detection in topography.Canny edge detection algorithm is
The multistage edge detection algorithm that John F.Canny developed in 1986, present embodiment utilizes canny edge to examine
Method of determining and calculating carries out rim detection in topography, and details are provided below:
Utilize Gaussian filter to smooth above-mentioned topography to remove picture noise, improve the precision of road-edge detection.
Obtain Grad and the direction value of each pixel in topography, shown in formula specific as follows (2), (3):
Wherein, (x, y) is the Grad of pixel to M, and (x y) is the direction value of pixel, g to αx、gyIt is respectively pixel to exist
Local derviation on the x-axis direction of image coordinate system and y-axis direction.gx、gyCan be tried to achieve by Sobel template correspondence, Sobel template uses
Sobel operator, this operator comprises two group of 3 × 3 matrix, respectively transverse direction and longitudinal direction, and two groups of matrixes are respectively as follows: Above-mentioned two groups of matrixes and image are made planar convolution, transverse direction and longitudinal direction can be drawn respectively
Brightness difference approximation, i.e. gx、gy。
Utilize pixel Grad M (x, y) and direction value α (x y) carries out non-maximum suppression to obtain candidate's picture
Vegetarian refreshments, contains all marginal points in topography and part non-edge point in candidate pixel point.
Gray average BL according to the pixel in topography set canny edge detection algorithm Low threshold parameter and
High threshold parameter, shown in formula specific as follows (4), (5):
TL=BL × γ (4)
TH=3 × TL(5)
Wherein, TLFor Low threshold parameter, THFor high threshold parameter, γ is light diversity factor coefficient, and γ can be determined by experiment
Or seek optimal value by EM algorithm.EM(Expectation-maximization algorithm) algorithm is greatest hope
Algorithm, EM algorithm is to find parameter maximal possibility estimation or the algorithm of MAP estimation, wherein probability in probabilistic model
Model depends on the hidden variable that cannot observe.
Utilize Low threshold parameter TL, high threshold parameter THMultiple marginal point is obtained further in candidate pixel point.Wherein,
Less than Low threshold parameter TLCandidate pixel point be non-edge point, more than high threshold parameter THCandidate pixel point be marginal point,
TL-THBetween candidate pixel point may be marginal point, present embodiment is at TL-THBetween choose neighbouring high threshold parameter THNecessarily
The candidate pixel point of scope and choosing more than high threshold parameter THCandidate pixel point be marginal point.
Step S204: utilize multiple marginal point to extract multiple straight-line segment;
In this step, multiple marginal point is utilized to extract multiple straight-line segment, Hough especially by Hough transform mode
Conversion is a kind of parameter estimation techniques using voting principle, and it utilizes the dotted line antithesis of image space and Hough parameter space
Property, the test problems in image space is transformed into parameter space.The first step of straight-line segment is extracted by Hough transform mode
For obtaining the polar equation of straight-line segment corresponding to multiple marginal points, shown in formula specific as follows (6):
ρ=xcos θ+ysin θ (6)
Wherein, formula (6) is the polar equation of straight-line segment, and x, y are respectively marginal point horizontal stroke in image coordinate system
Coordinate, vertical coordinate, ρ is the footpath, pole of marginal point, and θ is the polar angle of marginal point.
Second step is the rectangular equation that the polar equation of straight-line segment is converted to correspondence.
Step S205: delete slope from multiple straight-line segment and be unsatisfactory for the straight-line segment of predetermined slope requirement;
In this step, from image-forming principle, when vehicle linearly road driving, the road edge of its both sides is in institute
The picture frame of shooting becomes certain slope.Such as, it is negative for the right-side course Road Edge of the vehicle slope in picture frame,
And be positive for the left-side course Road Edge of the vehicle slope in picture frame, and the edge line between paving stone 52 is at picture frame
On slope be essentially close to zero.Therefore, the chosen position according to fixed point P is different, and according to the present road to obtain
The prior information of curb structure, can calculate different road edge theoretical slope in picture frame by image-forming principle, with
Time consider certain redundant error, and then can determine that the slope requirement of different road edge, and then get rid of and be clearly not roadside
The straight-line segment of the edge line between the straight-line segment of edge line, such as paving stone 52, it is possible to increase follow-up road-edge detection
Efficiency.
Further, in other embodiments, can be by detection vehicle row such as the GPS function of vehicle or angle induction apparatuss
Sail direction, thus change above-mentioned slope requirement, such as, when vehicle turns to the right, the right side of vehicle according to vehicle heading
Side line Road Edge slope in picture frame then levels off to zero, it is therefore desirable to adjust above-mentioned slope requirement according to practical situation.
Therefore also included extract the second step of straight-line segment above by Hough transform mode before: limit taking of polar angle θ
Value scope and calculate footpath, the pole ρ of correspondence, and by corresponding ρ, θ parameter matrix unit accumulated counts, select ρ, θ parameter matrix unit
In bigger summing elements and then determine polar equation formula specific as follows (7) meeting the straight-line segment that predetermined slope requires
Shown in:
ρi=xcos θi+ysinθi(7)
In the present embodiment, θiSpan be [-90,0].
Further the polar equation (7) meeting the straight-line segment of predetermined slope requirement is converted to rectangular equation,
Shown in formula specific as follows (8):
Y=fi(x) (8)
Wherein, formula (8) is the rectangular equation of the straight-line segment meeting predetermined slope requirement.
Step S206: extract curb line segment from multiple straight-line segment according to the curb architectural characteristic of present road;
In this step, can be according to the picture between the actual range between the road edge of present road and straight-line segment
The comparing result of element distance and/or according to the actual color difference of the road edge both sides of present road and straight-line segment both sides
The comparing result of pixel color difference extracts curb line segment from multiple straight-line segment.
When the contrast according to the pixel distance between the actual range between the road edge of present road and straight-line segment
When result extracts curb line segment from multiple straight-line segment, it specifically includes:
Utilize calibration coefficient λ by under space coordinates obtain present road road edge between actual range and
Pixel distance between the straight-line segment obtained under image coordinate system is transformed into the same coordinate system.The road edge of present road
Between the prior information that actual range is road edge, specifically include the left side line 531 of kerbstone 53 and the reality of right-hand line 532
Border distance Scurb, the actual range S of left side line 521 and right-hand line 522 of paving stone 52roundAnd the left side line of lane line 51
511 and the actual range S of right-hand line 512white, SwhiteNumerical range typically be respectively 12-15cm, 40cm, 10-12cm, on
The numerical range stating prior information is alternatively other sizes, does not makes too many restrictions.Obtain pixel between straight-line segment away from
From process described in detail below:
Obtain the intersection point meeting the straight-line segment of predetermined slope requirement and fixed point straight line 6 seat in image coordinate system
Mark, particularly as follows: when picture frame is gathered by preposition image capture device 3, fixed point straight line 6 is through fixed point P and vertically
In the straight line of y-axis, the intersection point of acquisition straight-line segment and fixed point straight line 6 abscissa in image coordinate system, public affairs specific as follows
Shown in formula (9):
xi=fi'(yP) (9)
Wherein, xiFor the intersection point of straight-line segment and fixed point straight line 6 abscissa in image coordinate system, yPFor fixed point P
Vertical coordinate in image coordinate system.
To the intersection point of straight-line segment and the fixed point straight line 6 abscissa x in image coordinate systemiIt is ranked up by size,
Abscissa x further with intersection pointiObtain the pixel distance between straight-line segment, shown in formula specific as follows (10):
dk=|xi-xj| (10)
Wherein, dkFor the pixel distance between straight-line segment, i < j.
Additionally, when picture frame is gathered by side image capture device 4, fixed point straight line is through fixed point P and to hang down
Directly in the straight line of x-axis, in like manner now by obtaining the intersection point of straight-line segment and the fixed point straight line vertical seat in image coordinate system
Pixel distance between mark and then acquisition straight-line segment.
After the pixel distance obtained between straight-line segment, utilize calibration coefficient λ by the actual range between road edge
With the pixel distance between straight-line segment is transformed into the same coordinate system, it pixel distance including actual range is converted to correspondence
So that actual range and pixel distance are in together under an image coordinate system, or the pixel distance between straight-line segment is changed
For corresponding actual range so that actual range and pixel distance are in together under space coordinates.For example, by actual range
It is transformed under image coordinate system, the most above-mentioned each actual range Scurb、Sround、SwhiteCorresponding pixel distance is dcurb、
dground、dwhite, It is appreciated that and pixel distance is converted to correspondence
Actual range be then: by pixel distance dkIt is multiplied by calibration coefficient λ and then can obtain the actual range of correspondence.
Further between the actual range between the road edge of present road and straight-line segment under the same coordinate system
Pixel distance carry out difference operation, and therefrom select the difference straight-line segment less than redundant error.With under image coordinate system
As a example by, difference is less than shown in pixel distance formula specific as follows (11) between the straight-line segment of redundant error, (12), (13):
Dcurb={dk:|dcurb-dk| < e} (11)
Dground={dk:|dground-dk| < e} (12)
Dwhite={dk:|dwhite-dk| < e} (13)
Wherein, ε is redundant error, 0 < ε < 0.5*min{dcurb,dground,dwhite}。Dcurb、Dground、DwhiteFor curb
The difference that stone, paving stone and lane line are corresponding is less than the pixel distance between the straight-line segment of redundant error ε, obtains further
Dcurb、Dground、DwhiteThe corresponding difference straight-line segment less than redundant error ε, the difference straight-line segment less than redundant error ε
Be curb line segment, curb line segment include the left side line 531 of kerbstone 53 and right-hand line 532, paving stone 52 left side line 521 with
Left side line 511 and the right-hand line 512 of right-hand line 522 and lane line 51.
When the actual color difference of the road edge both sides according to present road and the pixel color of straight-line segment both sides are poor
Different comparing result extracts curb line segment from multiple straight-line segment, and it specifically includes:
Calculate between each straight-line segment and adjacent straight-line segment or the predetermined lateral width of each straight-line segment both sides
In the range of the gray average of pixel.When picture frame is gathered by preposition image capture device 3, gray average is specific as follows
Shown in formula (14):
Wherein,For the pixel between each straight-line segment and adjacent straight-line segment or in predetermined lateral width range
The gray average of point, type ∈ curb, ground, white}, Represent Dt respectivelyypeIn the left and right sides of t candidate straight
Line line segment with through fixed point P and the abscissa of the intersection point of the fixed point straight line 6 being perpendicular to y-axis, or formed predetermined lateral width
The left and right abscissa of degree scope.Predetermined lateral width range can be set as 1/ of the pixel distance between adjacent straight-line segment
2,1/3 equal in width scope, predetermined lateral width range may be alternatively provided as the fixed values such as such as 2 pixel unit distances;The most right
In straight-line segment a, it is straight that the predetermined lateral width range of its both sides can be respectively set to straight-line segment a and adjacent left side, right side
1/4 of pixel distance between line line segment.It is appreciated that when picture frame is gathered by side image capture device 4, corresponding profit
With straight-line segment and through fixed point P and the vertical coordinate of the intersection point of the fixed point straight line being perpendicular to x-axis, and the horizontal stroke of fixed point P
Coordinate obtains gray average.
Further according to gray averageDetermining the pixel color difference of straight-line segment both sides, pixel color difference is
The difference of the gray average of straight-line segment both sides;From multiple straight-line segment extract straight-line segment both sides pixel color difference with
The actual color difference of the road edge both sides of present road straight-line segment unanimously or in error allowed band, is extracted
Straight-line segment is curb line segment.Curb line segment includes left side line 531 and right-hand line 532, a left side for paving stone 52 for kerbstone 53
Left side line 511 and the right-hand line 512 of side line 521 and right-hand line 522 and lane line 51.Wherein, the left and right sides of kerbstone 53 is straight
Actual grey average between line is Vcurb, the actual grey average between the left and right sides straight line of paving stone 52 is Vground, track
Actual grey average between the left and right sides straight line of line 51 is Vwhite, the magnitude relationship of three kinds of actual grey averages is: Vwhite>
Vcurb>Vground.The pixel color difference such as obtaining straight-line segment b both sides is c, and the left side of the kerbstone 53 of present road
The pixel color difference of line 531 both sides is also c, the most i.e. can determine that straight-line segment b corresponds to the left side line 531 of kerbstone 53.
Further, in the present embodiment, use distance to calculate and curb line segment is entered by the two ways of color contrast simultaneously
Row extracts, and i.e. ties according to the contrast of the pixel distance between the actual range between the road edge of present road and straight-line segment
Fruit and right according to the actual color difference of the road edge both sides of present road and the pixel color difference of straight-line segment both sides
Extracting curb line segment from multiple straight-line segment than result, it specifically includes:
Utilize calibration coefficient by under space coordinates obtain present road road edge between actual range and
Pixel distance between the straight-line segment obtained under image coordinate system is transformed into the same coordinate system;To working as under the same coordinate system
The pixel distance between actual range and straight-line segment between the road edge of front road carries out difference operation, and therefrom selects
Difference is less than multiple alternative straight line segments of redundant error.
Calculate between each alternative straight line segment and adjacent alternative straight line segment or each alternative straight line segment both sides
The gray average of the pixel in predetermined lateral width range, and the pixel of alternative straight line segment both sides is determined according to gray average
Color distortion;The pixel color difference of alternative straight line segment both sides and the road of present road is extracted from multiple alternative straight line segments
The actual color difference of Road Edge both sides alternative straight line segment unanimously or in error allowed band.
Step S207: utilize acquired curb line segment that multiple straight-line segment of the subsequent image frames of follow-up acquisition are carried out
Follow the tracks of, and then from multiple straight-line segment of subsequent image frames, extract curb line segment;
Utilize acquired curb line segment that multiple straight-line segment of the subsequent image frames of follow-up acquisition are tracked, and then
Curb line segment is extracted from multiple straight-line segment of subsequent image frames.Concrete available nearest neighbour method or Kalman filter method pair
Multiple straight-line segment of subsequent image frames are tracked.
The process utilizing nearest neighbour method to be tracked straight-line segment specifically includes: obtain multiple straight line lines of subsequent image frames
Section coordinate under image coordinate system, and then this coordinate is deducted a upper picture frame each curb line segment seat under image coordinate system
Mark, when two error of coordinates are less than redundant error ε, it is determined that the straight-line segment of subsequent image frames and the kerb line of a upper picture frame
Section is same straight line.Such as utilize above-mentioned formula (9) obtain the straight-line segment of subsequent image frames with through fixed point P and be perpendicular to
The intersection point of the fixed point straight line 6 of y-axis abscissa under image coordinate system;The abscissa of intersection point is deducted a upper each road of picture frame
The abscissa of the intersection point of edge line segment and fixed point straight line 6, if this straight-line segment and the difference of a certain curb line segment abscissa
Same straight line then it is defined as less than redundant error ε: | x "-x ' | < e, wherein x " straight-line segment and the fixed point for subsequent image frames
The abscissa of the intersection point of the abscissa of the intersection point of straight line 6, a certain curb line segment that x ' is a upper picture frame and fixed point straight line 6.
Kalman filter is a kind of recursion filter for time-varying linear systems proposed by Kalman (Kalman),
These time-varying linear systems can describe by the Differential Equation Model comprising quadrature variable, and Kalman filter was by the past
Measurement estimation difference is merged into new measurement error and estimates error in the future.Kalman filter method by by straight-line segment,
The coordinate points tables of data of curb line segment is shown as Kalman filter, utilizes the principle of Kalman filter, enters each straight-line segment
Line trace.
After the tracking of each straight-line segment completing a picture frame, repeat to update the coordinate information of current each curb line segment.
At water stain, shade, blocking etc. under complex working condition, the true curb of some picture frame is likely difficult to detection, by both the above with
Track method utilizes the history information data of road edge structure to be capable of under complex working condition and is tracked straight-line segment, enters
And extract curb line segment.
Step S208: calculate curb line segment in space coordinates according to curb line segment pixel coordinate under image coordinate system
Under relative to the actual range of vehicle.
According to curb line segment pixel coordinate under image coordinate system calculate curb line segment under space coordinates relative to
The actual range of vehicle.For utilizing the right-hand line 522 of the paving stone 52 in curb line segment at image coordinate system in present embodiment
Under pixel coordinate so that calculate actual range: when picture frame is gathered by preposition image capture device 3, as it is shown in figure 5, obtain
Take the right-hand line 522 of paving stone 52 in curb line segment and the abscissa x of the intersection point of fixed point straight line 61, and preposition image adopts
The axis, visual angle 31 of collection equipment 3 and the abscissa x of the intersection point of fixed point straight line 62, by two abscissa x1、x2Subtract each other and obtain
Pixel distance L between right-hand line 522 and axis, visual angle 311=|x1-x2|, further with calibration coefficient λ and pixel away from
From L1Calculate curb line segment under space coordinates relative to the actual range S of vehicle1=λ×L1;When picture frame is side image
When collecting device 4 is gathered, fixed point is on the axis, visual angle 41 of side image capture device 4, and now fixed point straight line is
Through fixed point P and the straight line that is perpendicular to x-axis, fixed point straight line is the axis, visual angle 41 of side image capture device 4, this
Time obtain the vertical coordinate L of right-hand line 522 and the intersection point of fixed point straight line2, and then utilize calibration coefficient λ and this vertical coordinate L2Meter
Calculate curb line segment under space coordinates relative to the actual range S of vehicle2=λ×L2。
In other embodiments, it is possible to utilize left side line 531 other curb line segments such as grade of such as kerbstone 53 at image
Pixel coordinate under coordinate system calculates curb line segment under space coordinates relative to the actual range of vehicle, does not the most make too much
Limit.
For utilize picture frame that side image capture device gathers and corresponding calculate current time curb line segment relative to
The actual range of vehicle, can judge the actual range of current time further, if the actual range of current time is less than
A certain predeterminable range threshold value, then point out current time vehicle and curb line segment by sending the warning mode such as sound or text importing
Actual range beyond safe distance scope, then tractor driver or automated driving system can adjust vehicle and road according to this information
The actual range of edge line segment.For utilize picture frame that preposition image capture device gathers and corresponding obtain curb line segment relative to
The actual range of vehicle, this actual range is pre-to vehicle actual range between following a certain moment and curb line segment
Sentence, this actual range can be judged equally, if it exceedes a certain predeterminable range threshold value, send information the most equally to carry
Show that vehicle will be with the actual range of curb line segment beyond safe distance scope, then tractor driver or automated driving system can adjust in advance
Vehicle and the actual range of curb line segment.
Additionally, be calculated curb line segment in space coordinates under relative to the actual range of vehicle after, exist further
Show the actual range of curb line segment and vehicle on vehicle in real time, specifically can carry out reality by the display screen being installed on vehicle
The real-time display of distance.Above-mentioned information can show equally on vehicle.
It is appreciated that road edge detection method the second embodiment of the present invention is by working as that acquisition comprises that vehicle travelled
The picture frame of the road edge information of front road, obtains the office in fixed point surrounding predetermined area set in advance from picture frame
Portion's image;To carrying out rim detection in topography to obtain multiple marginal points;Extract multiple further with multiple marginal points
Straight-line segment;From multiple straight-line segment, delete slope be unsatisfactory for the straight-line segment of predetermined slope requirement;According to present road
The comparing result of the pixel distance between actual range and straight-line segment between road edge and/or the road according to present road
The comparing result of the actual color difference of Road Edge both sides and the pixel color difference of straight-line segment both sides is from multiple straight-line segment
Middle extraction curb line segment;Acquired curb line segment is utilized to be tracked and then multiple straight-line segment of subsequent image frames from rear
Multiple straight-line segment of continuous picture frame extract curb line segment;Finally according to curb line segment pixel coordinate under image coordinate system
Calculate curb line segment under space coordinates relative to the actual range of vehicle.
By the way, it is possible to realize the curb line segment of the present road that automatic detection vehicle is travelled, tractor driver is reduced
Operation complexity and accuracy of detection higher.In addition pass through in topography, to carry out rim detection and delete slope being unsatisfactory for
The straight-line segment that predetermined slope requires, it is possible to increase the efficiency of road-edge detection;By in subsequent image frames to straight line line
Section is tracked being capable of rapid extraction curb line segment equally;Sit finally according to curb line segment pixel under image coordinate system
Mark calculates curb line segment under space coordinates relative to the actual range of vehicle, it is possible to realize automatically measuring current time vehicle
And the actual range between actual range and the following a certain moment vehicle of anticipation and curb line segment between curb line segment, reduces
The operation complexity of tractor driver and the range accuracy of acquisition are higher, it is achieved the safe driving of vehicle.
Referring to Fig. 6, road edge detection device one embodiment of the present invention includes:
Picture frame acquisition module 71, edge detection module 72, straight-line segment extraction module 73 and curb line segments extraction mould
Block 74.
Picture frame acquisition module 71 is for obtaining the image of the road edge information comprising the present road that vehicle is travelled
Frame.Preposition image capture device or side image acquisition described in picture frame acquisition module 71 specially the respective embodiments described above set
Standby.
Edge detection module 72 is for carrying out rim detection to picture frame, to obtain multiple marginal point.
Edge detection module 72 is further used for obtaining from picture frame in fixed point surrounding predetermined area set in advance
Topography, and in topography, carry out rim detection.
During carrying out rim detection in topography, edge detection module 72 is further used for calculating topography
The gray average of interior pixel, and set canny edge detection algorithm according to the gray average of the pixel in topography
Low threshold parameter and high threshold parameter, and utilize canny edge detection algorithm to carry out rim detection in topography.
Straight-line segment extraction module 73 is used for utilizing multiple marginal point to extract multiple straight-line segment.
Curb line segment extraction module 74 is for extracting from multiple straight-line segment according to the curb architectural characteristic of present road
Curb line segment.Curb line segment extraction module 74 be further used for according to the road edge of present road between actual range with straight
The comparing result of the pixel distance between line line segment and/or according to the actual color difference of the road edge both sides of present road with
The comparing result of the pixel color difference of straight-line segment both sides extracts curb line segment from multiple straight-line segment.
Straight-line segment extraction module 73 is further used at curb line segment extraction module 74 according to the roadside, road of present road
The comparing result of the pixel distance between actual range and straight-line segment between edge and/or the road edge according to present road
The comparing result of the actual color difference of both sides and the pixel color difference of straight-line segment both sides extracts from multiple straight-line segment
Before curb line segment, from multiple straight-line segment, delete slope be unsatisfactory for the straight-line segment of predetermined slope requirement.
When curb line segment extraction module 74 is for according to the actual range between the road edge of present road and straight line line
When the comparing result of the pixel distance between Duan extracts curb line segment from multiple straight-line segment, it is further used for:
Utilize calibration coefficient by under space coordinates obtain present road road edge between actual range and
Pixel distance between the straight-line segment obtained under image coordinate system is transformed into the same coordinate system;And it is right under the same coordinate system
The pixel distance between actual range and straight-line segment between the road edge of present road carries out difference operation, and therefrom selects
Select the difference straight-line segment less than redundant error.Wherein calibration coefficient is by the fixed point set in advance reality under space coordinates
Border coordinate and the fixed point image coordinate under image coordinate system calculates and obtains.
When the curb line segment extraction module 74 actual color difference for the road edge both sides according to present road is with straight
When the comparing result of the pixel color difference of line line segment both sides extracts curb line segment from multiple straight-line segment, it is used further
In:
Calculate between each straight-line segment and adjacent straight-line segment or the predetermined lateral width of each straight-line segment both sides
In the range of the gray average of pixel, and determine the pixel color difference of straight-line segment both sides according to gray average;And then from
Multiple straight-line segment extract the actual face of the pixel color difference of straight-line segment both sides and the road edge both sides of present road
The different straight-line segment unanimously or in error allowed band of aberration.
When curb line segment extraction module 74 is for according to the actual range between the road edge of present road and straight line line
The comparing result of the pixel distance between Duan and according to the actual color difference of the road edge both sides of present road and straight line line
When the comparing result of the pixel color difference of section both sides extracts curb line segment from multiple straight-line segment, it is further used for:
Utilize calibration coefficient by under space coordinates obtain present road road edge between actual range and
Pixel distance between the straight-line segment obtained under image coordinate system is transformed into the same coordinate system;And it is right under the same coordinate system
The pixel distance between actual range and straight-line segment between the road edge of present road carries out difference operation, and therefrom selects
Select the difference multiple alternative straight line segments less than redundant error.Wherein calibration coefficient by fixed point set in advance in space coordinates
Actual coordinate and fixed point image coordinate under image coordinate system under Xi calculate and obtain.
Curb line segment extraction module 74 is further used for calculating each alternative straight line segment and adjacent alternative straight line segment
Between or the predetermined lateral width range of each alternative straight line segment both sides in the gray average of pixel, and equal according to gray scale
Value determines the pixel color difference of alternative straight line segment both sides;And then from multiple alternative straight line segments, extract alternative straight line segment
The pixel color difference of both sides is consistent with the actual color difference of the road edge both sides of present road or in error allowed band
Interior alternative straight line segment.
After extracting curb line segment, curb line segment extraction module 74 is further used for utilizing acquired curb line segment to rear
Multiple straight-line segment of the continuous subsequent image frames obtained are tracked, and then extract from multiple straight-line segment of subsequent image frames
Curb line segment.
Additionally, road edge detection device farther includes: actual distance calculation module, for scheming according to curb line segment
As the pixel coordinate under coordinate system calculates curb line segment under space coordinates relative to the actual range of vehicle.
Present invention also offers a kind of vehicle, this vehicle includes the road edge detection device described in above-mentioned embodiment,
The road edge of vehicle real time automatic detection present road in the process of moving is realized by this road edge detection device.
The foregoing is only embodiments of the present invention, not thereby limit the scope of the claims of the present invention, every utilization is originally
Equivalent structure or equivalence flow process that description of the invention and accompanying drawing content are made convert, or are directly or indirectly used in what other were correlated with
Technical field, is the most in like manner included in the scope of patent protection of the present invention.
Claims (21)
1. a road edge detection method, it is characterised in that including:
Obtain the picture frame of the road edge information comprising the present road that vehicle is travelled;
Described picture frame is carried out rim detection, to obtain multiple marginal point;
The plurality of marginal point is utilized to extract multiple straight-line segment;
Curb architectural characteristic according to present road extracts curb line segment from multiple straight-line segment, and this step includes:
The actual color difference of the road edge both sides according to described present road and the pixel color of described straight-line segment both sides
The comparing result of difference extracts described curb line segment from the plurality of straight-line segment.
Method the most according to claim 1, it is characterised in that the described step that described picture frame carries out rim detection is entered
One step includes:
The topography in fixed point surrounding predetermined area set in advance is obtained from described picture frame;
Rim detection is carried out in described topography.
Method the most according to claim 2, it is characterised in that the described step carrying out rim detection in described topography
Suddenly farther include:
Calculate the gray average of pixel in described topography;
Gray average according to the pixel in described topography sets Low threshold parameter and the height of canny edge detection algorithm
Threshold parameter, utilizes described canny edge detection algorithm to carry out rim detection in described topography.
Method the most according to claim 1, it is characterised in that the described curb architectural characteristic according to described present road from
The step extracting curb line segment in the plurality of straight-line segment also includes:
The actual range between road edge according to described present road and the pixel distance between described straight-line segment right
From the plurality of straight-line segment, described curb line segment is extracted than result.
Method the most according to claim 4, it is characterised in that between the described road edge according to described present road
The comparing result of the pixel distance between actual range and described straight-line segment and the road edge two according to described present road
The comparing result of the pixel color difference of the actual color difference of side and described straight-line segment both sides is from the plurality of straight-line segment
Farther include before middle extraction described curb line segment:
From the plurality of straight-line segment, delete slope be unsatisfactory for the described straight-line segment of predetermined slope requirement.
Method the most according to claim 4, it is characterised in that between the described road edge according to described present road
The comparing result of the pixel distance between actual range and described straight-line segment extracts described road from the plurality of straight-line segment
The step of edge line segment includes:
Utilize calibration coefficient by under space coordinates obtain described present road road edge between actual range and
Pixel distance between the described straight-line segment obtained under image coordinate system is transformed into the same coordinate system, wherein said demarcation system
Number by the fixed point set in advance actual coordinate under described space coordinates and described fixed point at described image coordinate system
Under image coordinate calculate obtain;
Under described the same coordinate system to the actual range between the road edge of described present road and described straight-line segment it
Between pixel distance carry out difference operation, and therefrom select the difference described straight-line segment less than redundant error.
Method the most according to claim 1, it is characterised in that the described road edge both sides according to described present road
Actual color difference carries from the plurality of straight-line segment with the comparing result of the pixel color difference of described straight-line segment both sides
The step taking described curb line segment includes:
Calculate between each described straight-line segment and adjacent described straight-line segment or each described straight-line segment both sides predetermined
The gray average of the pixel in the range of lateral width, and the pixel of described straight-line segment both sides is determined according to described gray average
Color distortion;
The pixel color difference of described straight-line segment both sides and the road of described present road is extracted from the plurality of straight-line segment
The actual color difference of Road Edge both sides straight-line segment unanimously or in error allowed band.
Method the most according to claim 4, it is characterised in that between the described road edge according to described present road
The comparing result of the pixel distance between actual range and described straight-line segment and the road edge two according to described present road
The comparing result of the pixel color difference of the actual color difference of side and described straight-line segment both sides is from the plurality of straight-line segment
The step of middle extraction described curb line segment includes:
Utilize calibration coefficient by under space coordinates obtain described present road road edge between actual range and
Pixel distance between the described straight-line segment obtained under image coordinate system is transformed into the same coordinate system, wherein said demarcation system
Number by the fixed point set in advance actual coordinate under described space coordinates and described fixed point at described image coordinate system
Under image coordinate calculate obtain;
Under described the same coordinate system to the actual range between the road edge of described present road and described straight-line segment it
Between pixel distance carry out difference operation, and therefrom select the difference multiple alternative straight line segments less than redundant error;
Calculate between each described alternative straight line segment and adjacent described alternative straight line segment or each described alternative straight line
The gray average of pixel in the predetermined lateral width range of section both sides, and according to described gray average determine described alternative directly
The pixel color difference of line line segment both sides;
Extract from the plurality of alternative straight line segment the pixel color difference of described alternative straight line segment both sides with described currently
The actual color difference of the road edge both sides of road alternative straight line segment unanimously or in error allowed band.
Method the most according to claim 1, it is characterised in that described method farther includes:
Utilize acquired described curb line segment that the multiple described straight-line segment of the subsequent image frames of follow-up acquisition is tracked,
And then from the multiple described straight-line segment of subsequent image frames, extract described curb line segment.
10. according to the method described in claim 1-9 any one, it is characterised in that described method farther includes:
Described curb line segment phase under space coordinates is calculated according to described curb line segment pixel coordinate under image coordinate system
Actual range for described vehicle.
11. 1 kinds of road edge detection device, it is characterised in that including:
Picture frame acquisition module (71), for obtaining the image of the road edge information comprising the present road that vehicle is travelled
Frame;
Edge detection module (72), for carrying out rim detection to described picture frame, to obtain multiple marginal point;
Straight-line segment extraction module (73), is used for utilizing the plurality of marginal point to extract multiple straight-line segment;
Curb line segment extraction module (74), for actual color difference and the institute of the road edge both sides according to described present road
The comparing result of the pixel color difference stating straight-line segment both sides extracts curb line segment from multiple straight-line segment.
12. road edge detection device according to claim 11, it is characterised in that
Described edge detection module (72) is further used for obtaining from described picture frame around fixed point set in advance predetermined
Topography in region, and in described topography, carry out rim detection.
13. road edge detection device according to claim 12, it is characterised in that
Described edge detection module (72) is further used for calculating the gray average of the pixel in described topography, and according to
The gray average of the pixel in described topography sets Low threshold parameter and the high threshold ginseng of canny edge detection algorithm
Number, and utilize described canny edge detection algorithm to carry out rim detection in described topography.
14. road edge detection device according to claim 11, it is characterised in that
The reality that described curb line segment extraction module (74) is further used between the road edge according to described present road away from
The comparing result of the pixel distance between described straight-line segment extracts described curb line segment from the plurality of straight-line segment.
15. road edge detection device according to claim 14, it is characterised in that
Described straight-line segment extraction module (73) be further used for described curb line segment extraction module (74) according to described currently
The comparing result of the pixel distance between actual range and described straight-line segment between the road edge of road and according to described
The contrast knot of the pixel color difference of the actual color difference of the road edge both sides of present road and described straight-line segment both sides
Before fruit extracts described curb line segment from the plurality of straight-line segment, from the plurality of straight-line segment, delete slope be unsatisfactory for
The described straight-line segment that predetermined slope requires.
16. road edge detection device according to claim 14, it is characterised in that
Described curb line segment extraction module (74) is further used for utilizing described in calibration coefficient will obtain under space coordinates
The pixel between actual range and the described straight-line segment obtained under image coordinate system between the road edge of present road
Distance is transformed into the same coordinate system, and under described the same coordinate system to the reality between the road edge of described present road away from
From and described straight-line segment between pixel distance carry out difference operation, and therefrom select difference less than redundant error described directly
Line line segment, wherein said calibration coefficient is by the fixed point set in advance actual coordinate under described space coordinates and described mark
Fixed point image coordinate under described image coordinate system calculates and obtains.
17. road edge detection device according to claim 11, it is characterised in that
Described curb line segment extraction module (74) is further used for calculating each described straight-line segment and adjacent described straight line line
The gray average of the pixel between Duan or in the predetermined lateral width range of each described straight-line segment both sides, and according to described
Gray average determines the pixel color difference of described straight-line segment both sides, so extract from the plurality of straight-line segment described directly
The pixel color difference of line line segment both sides is consistent with the actual color difference of the road edge both sides of described present road or is missing
Straight-line segment in difference allowed band.
18. road edge detection device according to claim 14, it is characterised in that
Described curb line segment extraction module (74) is further used for utilizing described in calibration coefficient will obtain under space coordinates
The pixel between actual range and the described straight-line segment obtained under image coordinate system between the road edge of present road
Distance is transformed into the same coordinate system, and under described the same coordinate system to the reality between the road edge of described present road away from
From and described straight-line segment between pixel distance carry out difference operation, and therefrom select multiple standby less than redundant error of difference
Selecting straight-line segment, wherein said calibration coefficient is by the fixed point set in advance actual coordinate under described space coordinates and institute
State fixed point image coordinate under described image coordinate system and calculate acquisition;
It is described standby with adjacent that described curb line segment extraction module (74) is further used for calculating each described alternative straight line segment
The gray scale selecting pixel between straight-line segment or in the predetermined lateral width range of each described alternative straight line segment both sides is equal
Value, and the pixel color difference of described alternative straight line segment both sides is determined according to described gray average, and then from the plurality of standby
Select the road edge two of pixel color difference and the described present road extracting described alternative straight line segment both sides in straight-line segment
The actual color difference of side alternative straight line segment unanimously or in error allowed band.
19. road edge detection device according to claim 11, it is characterised in that
After described curb line segment extraction module (74) is further used for utilizing acquired described curb line segment to follow-up acquisition
The multiple described straight-line segment of continuous picture frame is tracked, and then extracts institute from the multiple described straight-line segment of subsequent image frames
State curb line segment.
20. according to the road edge detection device described in claim 11-19 any one, it is characterised in that described device enters
One step includes:
Actual distance calculation module, for calculating described curb according to described curb line segment pixel coordinate under image coordinate system
Line segment under space coordinates relative to the actual range of described vehicle.
21. 1 kinds of vehicles, it is characterised in that described vehicle includes the road edge as described in claim 11-20 any one
Detection device.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310711070.3A CN103714538B (en) | 2013-12-20 | 2013-12-20 | road edge detection method, device and vehicle |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310711070.3A CN103714538B (en) | 2013-12-20 | 2013-12-20 | road edge detection method, device and vehicle |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103714538A CN103714538A (en) | 2014-04-09 |
CN103714538B true CN103714538B (en) | 2016-12-28 |
Family
ID=50407481
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310711070.3A Active CN103714538B (en) | 2013-12-20 | 2013-12-20 | road edge detection method, device and vehicle |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103714538B (en) |
Families Citing this family (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103927526B (en) * | 2014-04-30 | 2017-02-15 | 长安大学 | Vehicle detecting method based on Gauss difference multi-scale edge fusion |
WO2015177648A1 (en) | 2014-05-14 | 2015-11-26 | Ofer Springer | Systems and methods for curb detection and pedestrian hazard assessment |
CN105224908A (en) * | 2014-07-01 | 2016-01-06 | 北京四维图新科技股份有限公司 | A kind of roadmarking acquisition method based on orthogonal projection and device |
JP6576664B2 (en) * | 2015-03-31 | 2019-09-18 | 株式会社ミツトヨ | Edge detection bias correction value calculation method, edge detection bias correction method, and program |
TWI609807B (en) * | 2016-05-17 | 2018-01-01 | 緯創資通股份有限公司 | Image evaluation method and electronic apparatus thereof |
CN106546185B (en) * | 2016-10-18 | 2019-09-20 | 福州觉感视觉软件科技有限公司 | A kind of profile quality determining method based on Machine Vision Detection |
CN106491040B (en) * | 2016-11-14 | 2019-04-09 | 浙江工业大学 | A kind of swing type window wiping robot glass edge touching detection processing method |
CN106491042B (en) * | 2016-11-14 | 2019-01-08 | 浙江工业大学 | Processing method is moved down at a kind of swing type window wiping robot glass edge |
CN106491039B (en) * | 2016-11-14 | 2019-01-08 | 浙江工业大学 | A kind of swing type window wiping robot glass edge detection method |
CN106650726A (en) * | 2016-12-05 | 2017-05-10 | 渤海大学 | License plate recognition method |
CN107133625B (en) * | 2017-03-31 | 2020-05-22 | 纵目科技(上海)股份有限公司 | Yellow lane line extraction method and system, and vehicle-mounted terminal |
CN107194940A (en) * | 2017-05-23 | 2017-09-22 | 北京计算机技术及应用研究所 | A kind of coloured image contour extraction method based on color space and line segment |
CN107862699B (en) * | 2017-09-22 | 2020-11-10 | 中国科学院深圳先进技术研究院 | Bone edge extraction method, device, equipment and storage medium of bone CT image |
CN108416306B (en) * | 2018-03-12 | 2020-12-25 | 海信集团有限公司 | Continuous obstacle detection method, device, equipment and storage medium |
CN109740469B (en) * | 2018-12-24 | 2021-01-22 | 百度在线网络技术(北京)有限公司 | Lane line detection method, lane line detection device, computer device, and storage medium |
CN114359871A (en) * | 2021-12-31 | 2022-04-15 | 中国第一汽车股份有限公司 | Method and device for detecting area edge of vehicle driving area and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102314599A (en) * | 2011-10-11 | 2012-01-11 | 东华大学 | Identification and deviation-detection method for lane |
CN102521589A (en) * | 2011-11-18 | 2012-06-27 | 深圳市宝捷信科技有限公司 | Method and system for detecting lane marked lines |
CN102682292A (en) * | 2012-05-10 | 2012-09-19 | 清华大学 | Method based on monocular vision for detecting and roughly positioning edge of road |
CN103400150A (en) * | 2013-08-14 | 2013-11-20 | 浙江大学 | Method and device for road edge recognition based on mobile platform |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3651387B2 (en) * | 2000-11-22 | 2005-05-25 | 日産自動車株式会社 | White line detector |
CN101567086B (en) * | 2009-06-03 | 2014-01-08 | 北京中星微电子有限公司 | Method of lane line detection and equipment thereof |
CN101620732A (en) * | 2009-07-17 | 2010-01-06 | 南京航空航天大学 | Visual detection method of road driving line |
CN102436644B (en) * | 2011-11-02 | 2013-10-16 | 南京物联网研究院发展有限公司 | Unstructured road detection method based on adaptive edge registration |
CN102609938B (en) * | 2012-01-16 | 2014-04-16 | 浙江大学 | Method for detecting vanishing points of road based on single image |
-
2013
- 2013-12-20 CN CN201310711070.3A patent/CN103714538B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102314599A (en) * | 2011-10-11 | 2012-01-11 | 东华大学 | Identification and deviation-detection method for lane |
CN102521589A (en) * | 2011-11-18 | 2012-06-27 | 深圳市宝捷信科技有限公司 | Method and system for detecting lane marked lines |
CN102682292A (en) * | 2012-05-10 | 2012-09-19 | 清华大学 | Method based on monocular vision for detecting and roughly positioning edge of road |
CN103400150A (en) * | 2013-08-14 | 2013-11-20 | 浙江大学 | Method and device for road edge recognition based on mobile platform |
Also Published As
Publication number | Publication date |
---|---|
CN103714538A (en) | 2014-04-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103714538B (en) | road edge detection method, device and vehicle | |
US10846874B2 (en) | Method and apparatus for processing point cloud data and storage medium | |
CN111551958B (en) | Mining area unmanned high-precision map manufacturing method | |
Son et al. | Real-time illumination invariant lane detection for lane departure warning system | |
CN107045629B (en) | Multi-lane line detection method | |
CN102682292B (en) | Method based on monocular vision for detecting and roughly positioning edge of road | |
CN105426864B (en) | One kind being based on the matched Multi-lane Lines Detection method of equidistant marginal point | |
CN105260699B (en) | A kind of processing method and processing device of lane line data | |
CN105678285B (en) | A kind of adaptive road birds-eye view transform method and road track detection method | |
CN103413308B (en) | A kind of obstacle detection method and device | |
CN101382497B (en) | Visibility detecting method based on monitoring video of traffic condition | |
CN105654073B (en) | A kind of speed automatic control method of view-based access control model detection | |
CN111209780A (en) | Lane line attribute detection method and device, electronic device and readable storage medium | |
CN105987684A (en) | Monocular vision-based agricultural vehicle navigation line detection system and method | |
CN101807352A (en) | Method for detecting parking stalls on basis of fuzzy pattern recognition | |
CN202134079U (en) | Unmanned vehicle lane marker line identification and alarm device | |
CN102592454A (en) | Intersection vehicle movement parameter measuring method based on detection of vehicle side face and road intersection line | |
CN110379168A (en) | A kind of vehicular traffic information acquisition method based on Mask R-CNN | |
CN103206957B (en) | The lane detection and tracking method of vehicular autonomous navigation | |
CN106887004A (en) | A kind of method for detecting lane lines based on Block- matching | |
CN106529443A (en) | Method for improving detection of lane based on Hough transform | |
CN107644528A (en) | A kind of vehicle queue length detection method based on vehicle tracking | |
CN109241855B (en) | Intelligent vehicle travelable area detection method based on stereoscopic vision | |
KR20110001427A (en) | High speed road lane detection method based on extraction of roi-lb | |
CN103544489A (en) | Device and method for locating automobile logo |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |