CN106767506A - Method, device and vehicle for detecting the bend curvature in road - Google Patents
Method, device and vehicle for detecting the bend curvature in road Download PDFInfo
- Publication number
- CN106767506A CN106767506A CN201611143467.7A CN201611143467A CN106767506A CN 106767506 A CN106767506 A CN 106767506A CN 201611143467 A CN201611143467 A CN 201611143467A CN 106767506 A CN106767506 A CN 106767506A
- Authority
- CN
- China
- Prior art keywords
- curve
- road
- pixel
- edge
- growth
- 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
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
- G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses the method for detecting the bend curvature in road, device and vehicle, method therein mainly includes：The road area in image is determined according to algorithm of region growing；Edge pixel point is extracted from the border of the road area；The edge pixel point of the extraction is supplied to the curve model for forming curves, and road edge curve is determined according to the curve that the curve model is exported；Determine the bend curvature of the road edge curve.The technical scheme that the present invention is provided can make the vehicle bend curvature detected in road in the process of moving accurately and timely, and can preferably be applied in structured road and the different unstructured road of road conditions, so as to improve the intelligent driving performance of vehicle.
Description
Technical field
The present invention relates to the intelligent vehicles technology, and in particular to a kind of method, use for detecting the bend curvature in road
The vehicle of the device of the device of the bend curvature in road is detected and the bend curvature in being provided with for detecting road.
Background technology
Various traffic problems can effectively be alleviated or even be solved to intelligent transportation system due to it, and pass is enjoyed so as to have become
The research topic of note, and intelligent vehicle is a very important part in intelligent transportation system.
Intelligent vehicle can perceive the information such as vehicle attitude and external environment condition, and intelligence by install sensor on vehicle
The information that energy vehicle can be perceived according to it carries out path planning, so as to realize the autonomous driving of vehicle.
In intelligent vehicle perceives external environmental information, be to the perception of road information it is very important, only accurately
Road information (such as bend curvature information) has been perceived to be possible to obtain position and side of the car body relative to road exactly
To etc. information, so as to realize the functions such as follow-up adaptive cruise control and detection of obstacles.
Inventor has found in process of the present invention is realized：Real road is generally divided into structured road and destructuring
Road.Structured road (such as highway) due to its lane line factor more more obvious etc. than more visible and road boundary, easily
The bend curvature in road is detected；And unstructured road (as depicted in figs. 1 and 2) due to its do not have obvious lane line and
Road boundary does not have the factors such as obvious mark, generally more difficult accurately to detect bend curvature in road.How vehicle is made
In the process of moving accurately and timely detect bend curvature in road, and bend curvature measuring technology is preferably applicable
In the unstructured road that structured road and road conditions are different, so as to improve the intelligent driving performance of vehicle, being one is worth
The technical problem of concern.
The content of the invention
In view of the above problems, it is proposed that the present invention so as to provide one kind overcome above mentioned problem or at least in part solve on
State the method for detecting the bend curvature in road, device and the vehicle of problem.
According to one aspect of the invention, there is provided a kind of method for detecting the bend curvature in road, should
Method mainly includes：The road area in image is determined according to algorithm of region growing；Extracted from the border of the road area
Edge pixel point；The edge pixel point of the extraction is supplied to curve model as the control point of forming curves, and according to institute
The curve for stating curve model output determines road edge curve；Determine the bend curvature of the road edge curve.
Optionally, the above-mentioned method for detecting the bend curvature in road, wherein, it is described true according to algorithm of region growing
The step of determining the road area in image includes：From described image downside, zone line chooses initial growth element blocks, and by institute
State initial growth element blocks and add growth district；Left growth elements block, the right growth elements adjacent with growth district are calculated respectively
The characteristic value of block and upper growth elements block, the growth elements block that will meet predetermined condition is added in the growth district, weight
Duplicate step, until growth district left growth elements block, right growth elements block and upper growth elements block characteristic value not
When meeting predetermined condition, using current growth district as the road area.
Optionally, the above-mentioned method for detecting the bend curvature in road, wherein, the side from the road area
The step of edge pixel point is extracted in boundary includes：Each zone boundary in described image is determined using edge detection algorithm；One
During the borderline pixel of pixel and the road area that individual pixel is respectively on the zone boundary, by the picture
Vegetarian refreshments is used as the edge pixel point in the border of road area；To be selected from the edge pixel point in the border of the road area
The edge pixel point for taking is used as the edge pixel point for extracting.
Optionally, the above-mentioned method for detecting the bend curvature in road, wherein, the curve model includes：Three times
Bezier model.
Optionally, the above-mentioned method for detecting the bend curvature in road, wherein, the marginal point by the extraction
The curve model for forming curves is supplied to, and the step of road edge curve is determined according to the curve that the curve model is exported
Suddenly include：The multigroup edge pixel point that will be extracted is respectively supplied to curve model as the control point of forming curves；Calculate institute
State the characteristic value of the curve that curve model is exported respectively for each group of marginal point；Using the maximum curve of eigenvalue of curve as road
Road Edge curve.
Optionally, the above-mentioned method for detecting the bend curvature in road, wherein, it is described to calculate the curve model pin
The step of characteristic value of the curve exported respectively to each group of marginal point, includes：For any bar curve, extension width is calculated
The pixel value sum of all pixels point in curve, and the coefficient of the curve is calculated according to length of curve and embroidery；
The characteristic value of the curve is determined according to the coefficient and pixel value sum.
Optionally, the above-mentioned method for detecting the bend curvature in road, wherein, it is described to determine that the road edge is bent
The step of bend curvature of line, includes：The curvature of the multiple pixels on the road edge curve is calculated, and will be the multiple
The average value of the curvature of pixel as the road edge curve bend curvature.
It is described according to another aspect of the present invention, there is provided a kind of device for detecting the bend curvature in road
Device includes：Region growing module, for determining the road area in image according to algorithm of region growing；Extract edge pixel point
Module, for extracting edge pixel point from the border of the road area；Curve module is determined, for by the side of the extraction
Edge pixel is supplied to curve model as the control point of forming curves, and is determined according to the curve that the curve model is exported
Road Edge curve；Determine curvature module, the bend curvature for determining the road edge curve.
Optionally, the above-mentioned device for detecting the bend curvature in road, wherein, the region growing module is specifically used
In：From described image downside, zone line chooses initial growth element blocks, and the initial growth element blocks are added into vitellarium
Domain；The characteristic value of the left growth elements block, right growth elements block and upper growth elements block adjacent with growth district is calculated respectively,
The growth elements block that predetermined condition will be met is added in the growth district, this step is repeated, until the left life of growth district
When the characteristic value of element blocks long, right growth elements block and upper growth elements block is unsatisfactory for predetermined condition, by current vitellarium
Domain is used as the road area.
Optionally, the above-mentioned device for detecting the bend curvature in road, wherein, the extraction edge pixel point module
Including：Rim detection submodule, for determining each zone boundary in described image using edge detection algorithm；Filtering pixel
Submodule, for the pixel and the borderline picture of the road area that are respectively in a pixel on the zone boundary
During vegetarian refreshments, using the pixel as the edge pixel point in the border of road area；Extracting sub-module, for will be from the road
The edge pixel point chosen in edge pixel point in the border in road region is used as the edge pixel point for extracting.
Optionally, the above-mentioned device for detecting the bend curvature in road, wherein, the curve model includes：Three times
Bezier model.
Optionally, the above-mentioned device for detecting the bend curvature in road, wherein, the determination curve module is specifically wrapped
Include：Characteristic value submodule is calculated, the multigroup edge pixel point for that will extract is provided respectively as the control point of forming curves
To curve model, and calculate the characteristic value of the curve that the curve model is exported respectively for each group of marginal point；Compare submodule
Block, for using the maximum curve of eigenvalue of curve as road edge curve.
Optionally, the above-mentioned device for detecting the bend curvature in road, wherein, the calculating characteristic value submodule tool
Body is used for：For any bar curve, the pixel value sum of all pixels point in the curve of extension width is calculated, and according to curve
Length and embroidery calculate the coefficient of the curve；The spy of the curve is determined according to the coefficient and pixel value sum
Value indicative.
Optionally, the above-mentioned device for detecting the bend curvature in road, wherein, the determination curvature module is specifically used
In：Calculate the curvature of the multiple pixels on the road edge curve, and by the average value of the curvature of the multiple pixel
As the bend curvature of the road edge curve.
According to another aspect of the invention, there is provided a kind of vehicle, the vehicle includes：It is above-mentioned for detecting road in
Bend curvature device.
Method, device and vehicle provided by the present invention for the bend curvature in detection road at least have following advantages
And beneficial effect：The present invention not only can be quickly true for the structured road in image by using algorithm of region growing
Road area is made, quickly road area can also be determined for the unstructured road in image；By that will belong to
The pixel on the border in road region is directed to road roadside as edge pixel point (i.e. road edge pixel point) using curve model
Edge pixel forming curves, so as to can not only realize fast and accurately determining road edge curve for structured road
(i.e. the road boundary of structured road), can also realize fast and accurately determining that road edge is bent for unstructured road
Line (i.e. the road boundary of unstructured road), therefore the present invention can determine structured road and non-knot with accurate quick
The bend curvature of structure road；It follows that the technical scheme that the present invention is provided vehicle can be made accurate in the process of moving and
When the bend curvature detected in road, and can preferably be applied to structured road and the different destructuring of road conditions
In road, so as to improve the intelligent driving performance of vehicle.
Described above is only the general introduction of technical solution of the present invention, in order to better understand technological means of the invention,
And can be practiced according to the content of specification, and in order to allow the above and other objects of the present invention, feature and advantage can
Become apparent, below especially exemplified by specific embodiment of the invention.
Brief description of the drawings
By reading the detailed description of hereafter preferred embodiment, various other advantages and benefit is common for this area
Technical staff will be clear understanding.The accompanying drawing of the present embodiment is only used for showing the purpose of preferred embodiment, and is not regarded as
It is limitation of the present invention.And in whole accompanying drawing, identical part is denoted by the same reference numerals.In the accompanying drawings：
Fig. 1 is a schematic diagram for specific example of unstructured road；
Fig. 2 is the schematic diagram of another specific example of unstructured road；
Fig. 3 is the flow chart for detecting the method for the bend curvature in road of the invention；
Fig. 4 is a flow chart for specific example of the road area in determination image of the invention；
Fig. 5 is a signal for example of the road area in the image that utilization algorithm of region growing of the invention is determined
Figure；
Fig. 6 is an example at the edge of the road area in the image that utilization algorithm of region growing of the invention is determined
Schematic diagram；
Fig. 7 is the schematic diagram for waiting inspection region of the invention；
Fig. 8 is present invention utilization Cubic kolmogorov's differential system model formation based on four signals of control point curve
Figure；
Fig. 9 is the structural representation for detecting the device of the bend curvature in road of the invention；
Figure 10 is the structured flowchart of the intelligent vehicle of one embodiment of the invention.
Specific embodiment
The exemplary embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although showing the present invention in accompanying drawing
Disclosed exemplary embodiment, it being understood, however, that may be realized in various forms the present invention and disclose without that should be illustrated here
Embodiment limited.Conversely, there is provided these embodiments are able to be best understood from technical scheme, and
Can by scope disclosed by the invention it is complete convey to those skilled in the art.
In a specific embodiment, the invention provides a kind of method for detecting the bend curvature in road.Should
The flow of method as shown in figure 3, and the method in Fig. 3 mainly include：Step S300, step S310, step S320 and step
S330.Each step included by the method for the present embodiment is described in detail respectively below.
S300, the road area determined according to algorithm of region growing in image.
As an example, the image in the present embodiment is typically the image of the camera real time shooting installed in vehicle.For
For any image that camera is absorbed, the present embodiment be using the growth elements block with predefined size in image progressively
Road area is formed, and the size of growth elements block can be set according to the actual requirements, it is such as more powerful and more in computing resource
In the case of many concern bend curvature degrees of accuracy, what can be set the size of growth elements block is smaller, is calculating for another example
Resource is weaker and more concern bend curvature measuring speed in the case of, what can be set the size of growth elements block is big by one
A bit.In a specific example, growth elements block is the growth elements block of 10 × 10 pixel sizes.
Under normal conditions, it is empty growth district that the present embodiment can first set one, and is determined in the picture at the beginning of one
Beginning growth elements block；Then, the initial growth element blocks are added in growth district, and the initial growth element can be as
Current growth elements；Afterwards, detection of the growth is carried out to surrounding based on the growth district, so that gradually expand growth district,
Finally, when growth district stops growing, using dormant growth district as the road area in image.
As an example, because the central region of image generally falls into road area, therefore, the present embodiment will can initially give birth to
Element blocks long are arranged in the central region of the downside of image, and such as the present embodiment can be with the intermediary image vegetarian refreshments of the lower side of image
On the basis of put to the left, right side and upper side extend the pixel of respective numbers respectively, so as to form 10 × 10 pixels
The initial growth element blocks of point size.
As an example, the present embodiment be provided with initial growth element blocks, and by the initial growth element blocks add grow
After region is as current growth elements block, the feature of the left growth elements block adjacent with current growth district can be respectively calculated
It is worth the characteristic value and the upper growth elements with current growth district vector of the right growth elements block adjacent with current growth district
The characteristic value of block, then, judges whether the characteristic value of left growth elements block meets predetermined condition, the characteristic value of right growth elements block
Whether whether the characteristic value for meeting predetermined condition and upper growth elements block meets predetermined condition, and will meet the life of predetermined condition
Element blocks long are added in current growth district, the above-mentioned process for growing current growth district are repeated, until current vitellarium
When the characteristic value of the left growth elements block, right growth elements block and upper growth elements block in domain is unsatisfactory for predetermined condition, currently
Growth district stops growing, and the present embodiment can be using current dormant growth district as the road detected from image
Region.
As an example, the characteristic value of the growth elements block in the present embodiment can be specially the gray value side of growth elements block
Difference, certainly, the characteristic value that the present embodiment is also not excluded for growth elements block is the possibility of other forms.Specifically, the present embodiment
A gray value variance for growth elements block can be calculated using following formula (1)：
In above-mentioned formula (1), var is the gray value variance for growing element blocks, and M is to grow the pixel that element blocks are included
Line number, N is to grow the columns of pixel that element blocks are included, and I (i, j) represents the pixel of the i-th row jth row in growth elements block
The gray value of point, mean represents the average gray of all pixels of the growth elements block.
The utilization algorithm of region growing of the present embodiment determines a specific example such as Fig. 4 institutes of the road area in image
Show.
In Fig. 4, S400, initial growth element blocks are determined in the central region of the lower side of image, and by the initial life
Element blocks long are added in currently empty growth district, and the initial growth element blocks are by as current growth elements block.Arrive
Step S410 and step S420.
S410, be expert in current growth elements block, calculate the left growth elements block adjacent with current growth district (such as with
The growth elements block of left 10 × 10 adjacent pixel sizes of current growth district) characteristic value (such as gray value variance), arrive
S411。
S411, judge whether the characteristic value of left growth elements block meets predetermined condition, such as judge the ash of left growth elements block
Whether angle value variance is less than predetermined value (such as 40)；If it is judged that to meet predetermined condition (such as gray value variance is less than 40),
Then arrive step S412；If it is judged that to be unsatisfactory for predetermined condition (such as gray value variance is not less than 40), then should stop current
The left side growth course that growth elements block is expert at, to step S430.
S412, the left growth elements block is added in current growth district, to step S413.
S413, judge whether the left growth elements block has reached the leftmost side of image, that is, judge the left growth elements block
Whether the left margin of image is had arrived at, if it is judged that the leftmost side to have reached image, then to step S430；If
Judged result is also to be not up to the leftmost side of image, then return to step S410.
S420, be expert in current growth elements block, calculate the right growth elements block adjacent with current growth district (such as with
The growth elements block of right 10 × 10 adjacent pixel sizes of current growth district) characteristic value (such as gray value variance), to step
Rapid S421.
S421, judge whether the characteristic value of right growth elements block meets predetermined condition, such as judge the ash of right growth elements block
Whether angle value variance is less than predetermined value (such as 40), if it is judged that to meet predetermined condition (such as gray value variance is less than 40),
Then arrive step S422；If it is judged that to be unsatisfactory for predetermined condition (such as gray value variance is not less than 40), then should stop current
The left side growth course that growth elements block is expert at, to step S430.
S422, the right growth elements block is added in current growth district, to step S423.
S423, judge whether the right growth elements block has reached the rightmost side of image, that is, judge the right growth elements block
Whether the right margin of image is had arrived at, if it is judged that the rightmost side to have reached image, then to step S430；If
Judged result is also to be not up to the rightmost side of image, then return to step S420.
S430, judge whether current growth elements block has reached the top side of image, that is, growth elements before judging to deserve
Whether block has arrived at the coboundary of image, if it is judged that the top side to have reached image, then to step S440；Such as
Fruit judged result is also to be not up to the top side of image, then to step S431.
The situation that S431, the left side growth course and right side growth course be expert in current growth elements block stop
Under, using the growth elements block (such as growth elements blocks of 10 × 10 pixel sizes) on the upside of current growth elements block as working as previous existence
Element blocks long, to step S432.
S432, the characteristic value (such as gray value variance) for calculating current growth elements block, to step S433.
S433, judge whether the characteristic value of current growth elements block meets predetermined condition, such as judge current growth elements block
Gray value variance whether be less than predetermined value (such as 40), if it is judged that (such as gray value variance is less than to meet predetermined condition
40), then to step S434；Return to step S410 and step S420；If it is judged that to be unsatisfactory for predetermined condition (such as gray value
Variance is not less than 40), then to step S440.
S440, this determine that the process of the road area in image terminates using algorithm of region growing, can will work as previous existence
Region long is used as the road area in image.
It should be strongly noted that the flow shown in Fig. 4 is to first carry out the left side and right side of current growth elements block
Growth course, then, then performs the growth course of the upside of current growth elements block, so repeats and ultimately form growth
Region；However, the present embodiment first can also constantly perform the growth course of the upside of current growth elements block, and in upside
Growth course stop after, according still further to row putting in order or putting in order from top to bottom from top to bottom, for every a line
The growth course on left side and right side is performed respectively, until the left side of all rows and the growth course on right side stop, it is determined that this
Secondary utilization algorithm of region growing determines that the process of the road area in image terminates, such that it is able to using current growth district as figure
Road area as in.The detailed process of the method is no longer described in detail herein.In addition, the leftmost side growth unit in the present embodiment
The size of plain block, rightmost side growth elements block and top side growth elements block is possible to that the big of other growth elements blocks can be less than
It is small.
As an example, for the image shown in Fig. 1, in the image that the present embodiment is determined using algorithm of region growing
One example of road area is as shown in Figure 5.
The present embodiment determines the road area in image by using algorithm of region growing, either for structuring road
Road, is also directed to unstructured road, and the present embodiment can fast and accurately obtain the road area in image, so that in fact
Now determine that the bend curvature in structured road and unstructured road provides the foundation.
S310, the extraction edge pixel point from the border of the above-mentioned road area determined.
As an example, the present embodiment is after the road area in determining image, directly can be carried from the road area
Take out road area border, a specific example, due to growth elements block be shaped as it is square, therefore, road area
Border has generally comprised substantial amounts of right angle, and the present embodiment can be obtained by detecting the right angle being mutually linked in road area
The border of road area, the white line in Fig. 6 is the border extracted from the road area shown in Fig. 5.
As an example, the present embodiment is behind the border for extracting road area from road area, in order to be further ensured that
The accuracy on the border of road area, the present embodiment can be using edge detection algorithm (such as Canny edge detection algorithms) really
Each zone boundary in the image made verifies come the border to above-mentioned road area.The present embodiment is calculated using rim detection
Method can detect the border in all regions including the border including road area in image.The present embodiment is determining one
Individual pixel is the pixel on zone boundary, be again road area borderline pixel in the case of, just understand will should
Pixel is used as the edge pixel point in the border of road area；And if pixel is not the pixel on zone boundary
Point, and be only the borderline pixel of road area, then the present embodiment can carry out filtration treatment to the pixel, i.e., from road
By the pixel point deletion in the borderline pixel point set in region.
As an example, the zone boundary in the image determined using edge detection algorithm of the present embodiment is to road area
The specific example that border is verified is：After setting is processed image using edge detection algorithm, if one
Pixel belongs to some zone boundary in image, then the pixel value of the pixel is set to 1, and if a pixel
The zone boundary in image is not belonging to, then the pixel value of the pixel is set to 0；Setting is using algorithm of region growing to image
After being processed, if pixel belongs to the border of the road area in image, the pixel value of the pixel is set
It is 1, and if pixel is not belonging to the border of road area, then the pixel value of the pixel is set to 0；Set above-mentioned
Under fixed condition, the present embodiment edge detection algorithm can be processed after image in the treatment of pixel and algorithm of region growing after
Image in pixel carry out and computing, so as to using with computing after pixel value be still 1 all pixels point as road area
Border in edge pixel point, i.e., with computing after all pixels value still for 1 pixel formed road area border.Separately
Outward, under above-mentioned imposing a condition, in order to reduce amount of calculation, the present embodiment can be from the image after algorithm of region growing treatment really
The time inspection region on the border for including road area is made, the size in time inspection region can be by a lateral boundaries of road area
All edge pixel points in top side edge pixel point, leftmost side edge pixel point, rightmost side edge pixel point and most under
Lateral edges pixel is determined；The Hou Jian areas of road area left border and right side boundary are such as the square frame of left and right two in Fig. 7
Domain, so that the present embodiment is when carrying out with computing, can examine region just for the time in the image after edge detection algorithm treatment
In the treatment of pixel and algorithm of region growing after image in time inspection region in pixel carry out and computing.The present embodiment
Carry out with after computing, it will usually obtain the edge pixel point of discrete distribution, two in such as Fig. 7 wait in inspection regions it is discrete
The white edge pixel point of distribution.
As an example, the present embodiment can be random from borderline all edge pixel points of the road area after verification
Extract the edge pixel point for determining road edge curve.One specific example, is carried out according to pre-determined number (such as 15 times)
(such as randomly selecting) repeatedly is chosen, so as to form multigroup edge pixel point；The present embodiment chooses the edge of predetermined quantity every time
Include the edge pixel point of predetermined quantity in pixel, i.e., one group.Predetermined quantity in the present embodiment can be according to curvilinear mold
Type sets to the demand of number of control points.
It is each in the image that the present embodiment is determined by using edge detection algorithm (such as Canny edge detection algorithms)
Zone boundary is verified come the border to the road area determined based on algorithm of region growing, can make what is finally determined
The border of road area is more accurate, such that it is able to avoid not having obvious lane line and road boundary due to unstructured road
Do not have to determine the more difficult problem in the border of road area caused by obvious mark.
S320, curve model is supplied to using the marginal point of extraction as the control point of forming curves, and according to curve model
The curve of output determines road edge curve.
As an example, the curve model in the present embodiment can be specially Cubic kolmogorov's differential system model.Certainly, this implementation
Example can also use other curve models.In the case of using Cubic kolmogorov's differential system model, due to Cubic kolmogorov's differential system
Model needs 4 control points, therefore, the present embodiment can every time randomly select 4 edge pixel points and be supplied to curve model.This
Embodiment using Cubic kolmogorov's differential system model can be expressed as formula (2) in the form of：
In above-mentioned formula (2), P0、P1、P2、P34 control points of Cubic kolmogorov's differential system model are represented, t represents sampling
Width, and the span of t is usually between 0-1, the value of such as t can be set as 0.01, P (t) and represent three Bezier songs
The curve that line model is exported, the curve should be by P0And P3The two control points, not through P1And P2The two control points,
As shown in Figure 8.
As an example, the present embodiment can generate a curve for each group of edge pixel point using curve model,
The present embodiment should choose a curve as road edge curve from a plurality of curve, and a curve is being chosen from a plurality of curve
When, length of a curve and flexibility are considered as, such as can determine curve using the flexibility of length of a curve and curve
Characteristic value, so that an optimal curve can be selected by the eigenvalue of curve of relatively more all curves, such as by eigenvalue of curve
Maximum curve is used as road edge curve.
The present embodiment, can be from numerous curves by determining road edge curve based on length of a curve and flexibility
In accurately pick out the curve for more meeting road edge curvilinear characteristic, for unstructured road, can effectively improve
The accuracy on the border of the road determined, so that the present embodiment can improve the accuracy of the final bend curvature determined.
As an example, the present embodiment chooses curve specific example from a plurality of curve being：
Step a, each bar curve for curve model output, be using what following formula (3) calculated each bar curve respectively
Number：
K=l/d+ (θ+1)/2 formula (3)
In above-mentioned formula (3), k is the coefficient of curve, and l is P0And P3Line distance between the two control points, d is
The catercorner length (waiting the description in inspection region such as above-mentioned embodiment) in the time inspection region where curve, as d can be in Fig. 7
The catercorner length of left/right square frame；θ for curve flexibility, and curve flexibility can by following formula (4) calculate obtain
：
θ=(cos α+cos β)/2 formula (4)
In above-mentioned formula (4), α is by control point P0、P1The straight line of formation with by control point P1、P2The straight line of formation it
Between angle, β is by control point P1、P2The straight line of formation with by control point P2、P3Angle between the straight line of formation is (such as Fig. 8 institutes
Show).
Because the catercorner length for waiting inspection region can to a certain extent reflect road edge length of a curve, because
This, the present embodiment is by using P0And P3Line distance between two control points examines the diagonal in region with the time where curve
The ratio of length can to a certain extent reflect the relation between the length of a curve and road edge length of a curve, P0
And P3Line distance between two control points examines the catercorner length in region closer to the time where curve, then the spy of the curve
Value indicative is bigger.
Step b, each curve for curve model output, calculate all pixels in the curve of extension width respectively
The pixel value sum of point.The curve of the curve that the curve model in the present embodiment is exported usually single pixel point width, this reality
The curve for applying the extension width in example refers to the curve with many pixel width.If the both sides of curve are referred to as into inner side
If (such as the left side of the curve in Fig. 8) and outside (such as the right side of the curve in Fig. 8), then can be for each in curve
Side extends N number of pixel to pixel respectively inwards, and extends M pixel laterally respectively, so that forming one extends M+N
The curve of+1 pixel width a, specific example, the curve set in Fig. 8 is generated based on sampling width as 0.01
Curve, so that the curve is the curve formed by 100 pixels, the present embodiment can will be each in this 100 pixels
Individual pixel extends 2 pixels to the left respectively, 2 pixels is extended to the right, so as to form 5 curves of pixel width.
Extend each pixel in the curve of width and be respectively provided with pixel value, and the pixel value of pixel can represent the pixel
Whether point is edge pixel point, if for example, a pixel is edge pixel point, its pixel value could be arranged to 255, and
If a pixel is not edge pixel point, its pixel value could be arranged to 0；If again for example, a pixel is side
Edge pixel, then its pixel value could be arranged to 1, and if a pixel is not edge pixel point, then its pixel value can be with
It is set to 0.
It should be strongly noted that the specific value of the pixel value of each pixel in the present embodiment can be passed through
What above-mentioned and computing was determined.Above-mentioned steps a and step b can be performed simultaneously, i.e. the present embodiment not conditioning step a and step b
Execution sequence.
Because the peripheral image vegetarian refreshments of a pixel on curve falls within edge pixel point, then the pixel is roadside
A possibility for pixel on edge curve can be larger, and if the peripheral image vegetarian refreshments of a pixel on curve does not belong to
In edge pixel point, then the pixel is that a possibility for pixel on road edge curve can be smaller, therefore, one
It is road to extend the pixel value sum of the point of all pixels on the curve of width and can to a certain extent represent virgin curve
The possibility of boundary curve.
Step c, the eigenvalue of curve that each bar curve is calculated using following formula (5)：
V=k × sum formula (5)
In above-mentioned formula (5), v is the eigenvalue of curve of curve, and k is the coefficient of the curve that above-mentioned formula (3) is calculated,
Sum is the pixel value sum for extending all pixels point on the curve after width.
The eigenvalue of curve of step d, each curve of comparing, is defined as road edge bent by the maximum curve of eigenvalue of curve
Line.
S330, the bend curvature for determining road edge curve.
As an example, in the case where road edge curve is determined, the present embodiment can be determined using various methods
The bend curvature of the road edge curve, can such as calculate the curvature of any one pixel on road edge curve, and will
The curvature of the pixel can calculate the multiple pixels on road edge curve for another example as the bend curvature of road edge curve
Point curvature, and using the average value of the curvature of multiple pixels as the road edge curve bend curvature.By using many
The average value of the curvature of individual pixel determines the bend curvature of road edge curve, is conducive to improving the present embodiment and finally determines
The accuracy of the bend curvature for going out.
In the case of being based on width is sampled for 0.01 curve for generating in road edge curve, the present embodiment can lead to
Cross 100 average values of the radius of curvature of pixel that following formula (6) are calculated on road edge curve：
In above-mentioned formula (6), ρ is the bend radius of curvature of road edge curve, Δ αiFor on road edge curve
The angle between the tangent line at the i+1 pixel on tangent line and road edge curve at ith pixel point, Δ SiIt is road
The arc length between the i+1 pixel on ith pixel point and road edge curve on Road Edge curve, and the arc length can
Approximately to take the air line distance between two pixels.
In the case where 100 average values of the radius of curvature of pixel are obtained, 100 curvature of pixel it is average
Value is 100 inverses of the average value of the radius of curvature of pixel.
In a specific embodiment, the invention provides a kind of device for detecting the bend curvature in road.Should
The structure of device as shown in figure 9, and the device in Fig. 9 mainly include：Region growing module 900, extraction edge pixel point module
910th, determine curve module 920 and determine curvature module 930.
Each module included by the control device of the present embodiment is described in detail respectively below.
Region growing module 900 is mainly used in determining the road area in image according to algorithm of region growing.
As an example, the camera real time shooting installed in the image typically vehicle that is used of region growing module 900
Image.For any image that camera is absorbed, region growing module 900 is using big with making a reservation in image
Small growth elements block gradually forms road area, and the big I of growth elements block that region growing module 900 is used
To set according to the actual requirements, such as computing resource is more powerful and the more concern bend curvature degrees of accuracy in the case of, can be with
It is smaller that the size of the growth elements block that region growing module 900 is used is set, weaker and more in computing resource for another example
In the case of many concern bend curvature measuring speed, the growth elements block that region growing module 900 can be used it is big
Small setting it is larger.In a specific example, the growth elements block that region growing module 900 is used is 10 × 10 pictures
The growth elements block of vegetarian refreshments size.
Under normal conditions, it is empty growth district that region growing module 900 can first set one, and is determined in the picture
One initial growth element blocks；Then, be added to the initial growth element blocks in growth district by region growing module 900；It
Afterwards, region growing module 900 carries out detection of the growth based on the growth district to surrounding, so that growth district gradually expands
Greatly, finally, when determining that growth district stops growing, region growing module 900 is using dormant growth district as figure
Road area as in.
As an example, because the central region of image generally falls into road area, therefore, region growing module 900 can be with
Initial growth element blocks are arranged in the central region of the downside of image, such as region growing module 900 can be with image most
Put on the basis of the intermediary image vegetarian refreshments of downside to the left, right side and upper side extend the pixel of respective numbers respectively so that shape
Into an initial growth element blocks for 10 × 10 pixel sizes.
As an example, region growing module 900 is being provided with initial growth element blocks, and the initial growth element blocks are added
Enter after growth district, region growing module 900 can respectively calculate the left growth elements block adjacent with current growth district
The characteristic value of the characteristic value right growth elements block adjacent with current growth district and the upper growth with current growth district vector
The characteristic value of element blocks, then, region growing module 900 judge the characteristic value of left growth elements block whether meet predetermined condition,
Whether the characteristic value of right growth elements block meets predetermined condition and whether the characteristic value of upper growth elements block meets predetermined condition,
And the growth elements block of predetermined condition will be met be added in current growth district, region growing module 900 repeats above-mentioned to make to work as
The process of preceding growth region growing, left growth elements block, right growth elements block and upper the growth unit until current growth district
When the characteristic value of plain block is unsatisfactory for predetermined condition, region growing module 900 determines that current growth district stops growing, region
Pop-in upgrades 900 can be using current dormant growth district as the road area detected from image.
As an example, the characteristic value of growth elements block that region growing module 900 is calculated can be specially growth elements
The gray value variance of block, certainly, the present embodiment is also not excluded for the characteristic value of the growth elements block that region growing module 900 is calculated
It is the possibility of other forms.Specifically, region growing module 900 can calculate a growth elements using above-mentioned formula (1)
The gray value variance of block, no longer carries out repeat specification for formula (1) herein.In addition, region growing module 900 is given birth to using region
Algorithm long determines the description for Fig. 4 in a specific example such as above-mentioned embodiment of the road area in image, herein no longer
Repeat specification.
As an example, for the image shown in Fig. 1, region growing module 900 is using being somebody's turn to do that algorithm of region growing is determined
One example of the road area in image is as shown in Figure 5.
Extraction edge pixel point module 910 is mainly used in the border of the road area determined from region growing module 900
Middle extraction edge pixel point；And in one embodiment, the extraction edge pixel point module 910 can include：Rim detection
Module, filtering pixel submodule and extracting sub-module；Rim detection submodule therein is mainly used in utilizing rim detection
Algorithm determines each zone boundary in described image；Filtering pixel submodule therein is mainly used in a pixel difference
During pixel borderline for pixel and road area on the zone boundary, using pixel as the side of road area
Edge pixel point in boundary；Extracting sub-module is mainly used in the side that will be chosen from the edge pixel point in the border of road area
Edge pixel is used as the edge pixel point for extracting.
As an example, extracting edge pixel 910 road area in region growing module 900 determines image of point module
Afterwards, the border of road area, a specific example, due to growth elements block can be directly extracted from the road area
Be shaped as it is square, therefore, the border of road area has generally comprised substantial amounts of right angle, and extracting edge pixel point module 910 can be with
The border for obtaining road area by detecting the right angle being mutually linked in road area, the white line in Fig. 6 is extraction side
The border that edge pixel point module 910 is extracted from the road area shown in Fig. 5.
As an example, extracting edge pixel point module 910 behind the border for extracting road area from road area, it is
Be further ensured that the accuracy on the border of road area, extract edge pixel point module 910 (such as rim detection submodule and
Filtering pixel submodule) can utilize the image that edge detection algorithm (such as Canny edge detection algorithms) be determined
Each zone boundary verifies come the border to above-mentioned road area.Extract (such as rim detection submodule of edge pixel point module 910
Block) border in all regions including the border including road area in image can be detected using edge detection algorithm.
It is the picture on zone boundary that edge pixel point module 910 (such as filtering pixel submodule) is extracted pixel is determined
Vegetarian refreshments, be again road area borderline pixel in the case of, just can be using the pixel as in the border of road area
Edge pixel point；And if pixel is not the pixel on zone boundary, and only it is the borderline of road area
Pixel, then extracting edge pixel point module 910 (such as filtering pixel submodule) can carry out filtration treatment to the pixel, i.e.,
Extracting edge pixel point module 910 (such as filtering pixel submodule) should from the borderline pixel point set of road area
Pixel point deletion.
As an example, the regional edge in extracting the image that edge pixel point module 910 is determined using edge detection algorithm
The specific example that the border of bound pair road area is verified is：Setting rim detection submodule is calculated using rim detection
After method is processed image, if a pixel belongs to some zone boundary in image, the pixel of the pixel
Value is set as 1 by rim detection submodule, and if pixel is not belonging to the zone boundary in image, then the pixel
Pixel value is set as 0 by rim detection submodule；Setting regions pop-in upgrades 900 using algorithm of region growing to image at
After reason, if pixel belongs to the border of the road area in image, the pixel value of the pixel is by region growing mould
Block 900 is set as 1, and if pixel is not belonging to the border of road area, then the pixel value of the pixel is given birth to by region
Module long 900 is set as 0；Under above-mentioned imposing a condition, extracting edge pixel point module 910 (such as filtering pixel submodule) can
The pixel in image after being processed with the pixel in the image after edge detection algorithm is processed and algorithm of region growing is clicked through
Row and computing, thus filter pixel submodule using with computing after pixel value be still 1 all pixels point as road area
Edge pixel point in border, i.e., with computing after all pixels value still for 1 pixel formed road area border.
In addition, under above-mentioned imposing a condition, in order to reduce amount of calculation, filtering pixel submodule can be calculated from region growing
The time inspection region on the border for including road area is determined in image after method treatment, the size in time inspection region can be by road
Top side edge pixel point, leftmost side edge pixel point, the rightmost side in all edge pixel points in one lateral boundaries in road region
Edge pixel point and lower side edge pixel point are determined；As the square frame of left and right two in Fig. 7 is road area left border
Time inspection region and right side boundary time inspection region, so as to filter pixel submodule when carrying out with computing, can be just for
In pixel and the image after algorithm of region growing treatment in time inspection region in image after edge detection algorithm treatment
The pixel waited in inspection region is carried out and computing.Filtering pixel submodule carry out with after computing, it will usually obtain discrete point
The edge pixel point of cloth, two in such as Fig. 7 white edge pixel points for waiting the discrete distribution in inspection region.
As an example, extract edge pixel point module 910 (such as extracting sub-module) can be from filtering pixel submodule school
(such as random extraction) is extracted in borderline all edge pixel points of the road area after testing for determining road edge curve
Edge pixel point.One specific example, extracting sub-module is repeatedly randomly selected according to pre-determined number (such as 15 times), from
And form multigroup edge pixel point；Extracting sub-module chooses the edge pixel point of predetermined quantity every time, i.e., include in one group
The edge pixel point of predetermined quantity.The predetermined quantity that extracting sub-module is used can be according to curve model to number of control points
Demand sets.
Determine that curve module 920 is mainly used in for the edge pixel point of extraction being supplied to song as the control point of forming curves
Line model, and road edge curve is determined according to the curve that curve model is exported；And in one embodiment, the determination curvilinear mold
Block 920 can include：Calculate characteristic value submodule and comparison sub-module；It is therein calculating characteristic value submodule be mainly used in by
The multigroup edge pixel point for extracting is respectively supplied to curve model as the control point of forming curves, and calculates the curvilinear mold
The characteristic value of the curve that type is exported respectively for each group of marginal point；Comparison sub-module therein is mainly used in eigenvalue of curve
Maximum curve is used as road edge curve.
As an example, determining that the curve model that curve module 920 is used can be specially Cubic kolmogorov's differential system model.
Certainly, determine that curve module 920 can also use other curve models.It is determined that curve module 920 uses three Bezier songs
In the case of line model, because Cubic kolmogorov's differential system model needs 4 control points, therefore, extract edge pixel point module 910
4 edge pixel points can be every time chosen (such as randomly select), determines that curve module 920 (as calculated characteristic value submodule) will take
4 edge pixel points that edge pixel point module 910 is randomly selected every time are supplied to curve model.Determine the institute of curve module 920
The Cubic kolmogorov's differential system model for using can be expressed as the form of above-mentioned formula (2), not be repeated.
As an example, determining that curve module 920 can be using curve model generation one for each group of edge pixel point
Bar curve, determines that curve module 920 should choose a curve as road edge curve from a plurality of curve, from a plurality of curve
During one curve of middle selection, each bar length of a curve and flexibility etc. are considered as, such as determine that curve module 920 is (as special in calculated
Value indicative submodule) eigenvalue of curve can be determined using the flexibility of length of a curve and curve, so that it is determined that curvilinear mold
Block 920 (such as comparison sub-module) can select an optimal curve by the eigenvalue of curve of relatively more all curves, such as compare
Submodule is using the maximum curve of eigenvalue of curve as road edge curve.Determine curve module 920 by the length based on curve
Degree and flexibility determine road edge curve, can accurately be picked out from numerous curves and more meet road edge curve
The curve of feature, the accuracy of the final bend curvature determined of curvature module 930 is determined such that it is able to improve.
As an example, determining that curve module 920 chooses curve specific example from a plurality of curve and is：For
Each bar curve of curve model output, calculates characteristic value submodule and calculates each bar song respectively using above-mentioned formula (3) and formula (4)
The coefficient of line.For each curve of curve model output, the curve that characteristic value submodule calculates extension width respectively is calculated
In all pixels point pixel value sum.The curve that curve model in the present embodiment is exported usually single pixel point width
Curve, the curve of the extension width in the present embodiment refers to the curve with many pixel width.Extend the curve of width
In each pixel be respectively provided with pixel value, and the pixel value of pixel can represent whether the pixel is edge pixel
Point, if for example, a pixel is edge pixel point, its pixel value can be configured so that 255, and if a pixel
It is not edge pixel point, then its pixel value can be configured so that 0；If again for example, a pixel is edge pixel point, its
Pixel value could be arranged to 1, and if a pixel is not edge pixel point, then its pixel value could be arranged to 0.Need spy
The specific value for not mentionleting alone the pixel value of bright each pixel in the present embodiment can be by determining curve module
Being determined with computing performed by 920.Calculate characteristic value submodule special using the curve that above-mentioned formula (5) calculates each bar curve
Value indicative.The eigenvalue of curve of more each curve of comparison sub-module, and the maximum curve of eigenvalue of curve is defined as road edge
Curve.
Determine that curvature module 930 is mainly used in determining the bend curvature of road edge curve.
As an example, in the case of it is determined that curve module 920 determines road edge curve, determining curvature module 930
The bend curvature of the road edge curve can be determined using various methods, such as determines that curvature module 930 can calculate road
The curvature of any one pixel on boundary curve, and the curvature of the pixel is bent as the bend of road edge curve
Rate, determines that curvature module 930 can calculate the curvature of the multiple pixels on road edge curve for another example, and by multiple pixels
Curvature average value as the road edge curve bend curvature.Determine curvature module 930 by using multiple pixels
The average value of curvature determine the bend curvature of road edge curve, be conducive to improving the accuracy of bend curvature.
In the case of being based on width is sampled for 0.01 curve for generating in road edge curve, curvature module is determined
930 can calculate 100 average values of the radius of curvature of pixel on road edge curve by above-mentioned formula (6).It is determined that
Curvature module 930 in the case where 100 average values of the radius of curvature of pixel are obtained, 100 curvature of pixel
Average value is 100 inverses of the average value of the radius of curvature of pixel.
In one embodiment, the present invention provides a kind of car for being provided with the device for detecting the bend curvature in road
(such as intelligent vehicle), uses the above-mentioned method for detecting the bend curvature in road to detect structured road and non-structural
Change the bend curvature of road.
In one embodiment, the present invention provides a kind of intelligent vehicle, and Figure 10 shows the intelligence of one embodiment of the invention
The structured flowchart of vehicle, as shown in Figure 10, the vehicle 1000 includes：It is middle control module, instrument board 1010, drive recorder 1011,
HUD (Head Up Display, head-up display) HUD 1012, intelligent vehicle-carried information entertainment 1013, intelligence are driven
Sail module 1013.
Instrument board 1010 has 12.3 cun of LCD display devices, and the instrument board can be using the J6CPU of TI；The behaviour of instrument board
Making system can be displayed for vehicle-state, map, vehicle navigation information, vehicle based on QNX embedded systems, instrument board
Music etc. is played, the car status information includes：Speed, rotating speed, electricity, tire pressure, vehicle parking, gear etc..HUD comes back aobvious
Show that device 1012 can show GPS navigation information, navigation route information, temporal information etc..
In one embodiment, intelligent driving module 1013 can be used for the treatment operation related to intelligent driving, intelligence
Drive module 1013 can include senior DAS (Driver Assistant System) (Advanced Driver Assistance Systems,
ADAS), active safety system, notice accessory system (Attention Assist System, AAS), tired warning system
(Fatigue Warning System, FWS), Vehicular intelligent acoustical alarm system (Acoustic Vehicle Alerting
System, AVAS) etc..Vehicle can combine ADAS systems etc. and carry out intelligent driving, the intelligent driving can be completely nobody
Drive, or driver carries out the senior auxiliary such as auxiliary doubling, the lane shift of Driving control and drives function.
Control device can be made up of multiple modules, can mainly include：Mainboard 1001；SATA(Serial Advanced
Technology Attachment, Serial Advanced Technology Attachment) module 1002, the storage device such as SSD1003 is connected to, can
For data storage information；(Frequency Modulation are adjusted AM (Amplitude Modulation, amplitude modulation)/FM
Frequently module 1004), the function of radio is provided for vehicle；Power amplifier module 1005, for acoustic processing；WIFI(Wireless-
Fidelity, Wireless Fidelity)/Bluetooth modules 1006, the service of WIFI/Bluetooth is provided for vehicle；LTE(Long
Term Evolution, Long Term Evolution) communication module 1007, for vehicle provides the communication function with telecom operators；Power supply mould
Block 1008, power module 1008 provides power supply for the control device；Switch interconnecting modules 1009, the Switch interconnecting modules
1009 can connect multiple sensors as a kind of expansible interface, for example if necessary to addition night vision function sensor,
PM2.5 function sensors, can be connected to the mainboard of control device, so as to control device by the Switch interconnecting modules 1009
Processor carry out data processing, and transfer data to middle control display.
In one embodiment, the vehicle also includes looking around camera, ADAS cameras, night vision cam, millimeter wave thunder
Up to, the sensor such as ultrasonic radar, ESR radars.
Provided herein algorithm and display not with any certain computer, virtual system or the intrinsic phase of miscellaneous equipment
Close.Various general-purpose systems can also be used together with based on teaching in this.As described above, this kind of system is constructed to want
The structure asked is obvious.Additionally, the present invention is not also directed to any certain programmed language.It is understood that, it is possible to use it is each
The content that programming language realizes invention described herein is planted, and the description done to language-specific above is to disclose this
The preferred forms of invention.
In specification mentioned herein, numerous specific details are set forth.It is to be appreciated, however, that implementation of the invention
Example can be put into practice in the case of without these details.In some instances, known method, structure is not been shown in detail
And technology, so as not to obscure the understanding of this description.
Similarly, it will be appreciated that in order to simplify one or more that the disclosure and helping understands in each inventive aspect, exist
Above to the description of exemplary embodiment of the invention in, each feature of the invention is grouped together into single implementation sometimes
In example, figure or descriptions thereof.However, the method for the disclosure should be construed to reflect following intention：I.e. required guarantor
The application claims of shield features more more than the feature being expressly recited in each claim.More precisely, such as this hair
As bright claims reflect, inventive aspect is all features less than single embodiment disclosed above.Cause
This, it then follows thus claims of specific embodiment are expressly incorporated in the specific embodiment, wherein each claim
Itself is all as separate embodiments of the invention.
Those skilled in the art are appreciated that can be carried out adaptively to the module in the equipment in embodiment
Change and they are arranged in one or more equipment different from the embodiment.Can be the module or list in embodiment
Unit or component be combined into a module or unit or component, and can be divided into addition multiple submodule or subelement or
Sub-component.In addition at least some in such feature and/or process or unit exclude each other, can use any
Combine to all features disclosed in this specification (including adjoint claim, summary and accompanying drawing) and so disclosed appoint
Where all processes or unit of method or equipment are combined.Unless expressly stated otherwise, this specification (including adjoint power
Profit is required, summary and accompanying drawing) disclosed in each feature can the alternative features of or similar purpose identical, equivalent by offer carry out generation
Replace.
Although additionally, it will be appreciated by those of skill in the art that embodiment described herein includes institute in other embodiments
Including some features rather than further feature, but the combination of the feature of different embodiments means in the scope of the present invention
Within and form different embodiments.For example, in claims of the present invention, embodiment required for protection it is any
One of mode can use in any combination.
All parts embodiment of the invention can be realized with hardware, or be run with one or more processor
Software module realize, or with combinations thereof realize.It will be understood by those of skill in the art that can use in practice
Microprocessor or digital signal processor (DSP) are according to embodiments of the present invention for detecting that the bend in road is bent to realize
The method of rate and for some or all functions in the device for detecting the bend curvature in road.It is of the invention acceptable real
Be now for perform some or all equipment or system program of method as described herein (such as computer program and
Computer program product).It is such to realize that program of the invention be stored on a computer-readable medium, or can have
The form of one or more signal.Such signal can be downloaded from internet website and obtained, it is also possible in carrier signal
Upper offer, or provided in any other form.
It should be noted that above-described embodiment is that the present invention will be described rather than limiting the invention, and
Those skilled in the art can design alternative embodiment without departing from the scope of the appended claims.In claim
In, any reference symbol being located between bracket should not be configured to limitations on claims.Word "comprising" is not excluded for depositing
In element or step not listed in the claims etc..Word "a" or "an" before element does not exclude the presence of many
Individual such element.The present invention can be by means of the hardware for including some different elements and by means of properly programmed calculating
Machine is realized.If in the unit claim for listing dry systems, several in these systems can be by same
Hardware branch is embodied.Word first, second and third use do not indicate that any order, can explain these words
It is title.
Claims (15)
1. a kind of method for detecting the bend curvature in road, it is characterised in that methods described includes：
The road area in image is determined according to algorithm of region growing；
Edge pixel point is extracted from the border of the road area；
The edge pixel point of the extraction is supplied to curve model as the control point of forming curves, and according to the curvilinear mold
The curve of type output determines road edge curve；
Determine the bend curvature of the road edge curve.
2. the method for claim 1, it is characterised in that the roadway area determined according to algorithm of region growing in image
The step of domain, includes：
From described image downside, zone line chooses initial growth element blocks, and the initial growth element blocks are added into vitellarium
Domain；
The feature of the left growth elements block, right growth elements block and upper growth elements block adjacent with growth district is calculated respectively
Value, the growth elements block that will meet predetermined condition is added in the growth district, this step is repeated, until a left side for growth district
When the characteristic value of growth elements block, right growth elements block and upper growth elements block is unsatisfactory for predetermined condition, will currently grow
Region is used as the road area.
3. the method for claim 1, it is characterised in that described to extract edge pixel from the border of the road area
The step of point, includes：
Each zone boundary in described image is determined using edge detection algorithm；
In the borderline pixel of pixel and the road area that a pixel is respectively on the zone boundary,
Using the pixel as the edge pixel point in the border of road area；
The edge pixel point that will be chosen from the edge pixel point in the border of the road area is used as the side for extracting
Edge pixel.
4. the method for claim 1, it is characterised in that the curve model includes：Cubic kolmogorov's differential system model.
5. the method as described in any claim in Claims 1-4, it is characterised in that the edge by the extraction
Point is supplied to the curve model for forming curves, and determines road edge curve according to the curve that the curve model is exported
Step includes：
The multigroup edge pixel point that will be extracted is respectively supplied to curve model as the control point of forming curves；
Calculate the characteristic value of the curve that the curve model is exported respectively for each group of marginal point；
Using the maximum curve of eigenvalue of curve as road edge curve.
6. method as claimed in claim 5, it is characterised in that the calculating curve model is for each group of marginal point point
The step of characteristic value of the curve not exported, includes：
For any bar curve, the pixel value sum of all pixels point in the curve of extension width is calculated, and it is long according to curve
Degree and embroidery calculate the coefficient of the curve；
The characteristic value of the curve is determined according to the coefficient and pixel value sum.
7. the method as described in any claim in claim 1 to 6, it is characterised in that the determination road edge
The step of bend curvature of curve, includes：
Calculate the curvature of the multiple pixels on the road edge curve, and by the average value of the curvature of the multiple pixel
As the bend curvature of the road edge curve.
8. a kind of device for detecting the bend curvature in road, it is characterised in that described device includes：
Region growing module, for determining the road area in image according to algorithm of region growing；
Edge pixel point module is extracted, for extracting edge pixel point from the border of the road area；
Curve module is determined, for the edge pixel point of the extraction to be supplied into curvilinear mold as the control point of forming curves
Type, and road edge curve is determined according to the curve that the curve model is exported；
Determine curvature module, the bend curvature for determining the road edge curve.
9. device as claimed in claim 8, it is characterised in that the region growing module specifically for：
From described image downside, zone line chooses initial growth element blocks, and the initial growth element blocks are added into vitellarium
Domain；
The feature of the left growth elements block, right growth elements block and upper growth elements block adjacent with growth district is calculated respectively
Value, the growth elements block that will meet predetermined condition is added in the growth district, this step is repeated, until a left side for growth district
When the characteristic value of growth elements block, right growth elements block and upper growth elements block is unsatisfactory for predetermined condition, will currently grow
Region is used as the road area.
10. device as claimed in claim 8, it is characterised in that the extraction edge pixel point module includes：
Rim detection submodule, for determining each zone boundary in described image using edge detection algorithm；
Filtering pixel submodule, for the pixel and the road that are respectively in a pixel on the zone boundary
During pixel on zone boundary, using the pixel as the edge pixel point in the border of road area；
Extracting sub-module, for will from the edge pixel point in the border of the road area choose edge pixel point as
The edge pixel point for extracting.
11. devices as claimed in claim 8, it is characterised in that the curve model includes：Cubic kolmogorov's differential system model.
Device in 12. such as claim 8 to 11 as described in any claim, it is characterised in that the determination curve module tool
Body includes：
Characteristic value submodule is calculated, the multigroup edge pixel point for that will extract is provided respectively as the control point of forming curves
To curve model, and calculate the characteristic value of the curve that the curve model is exported respectively for each group of marginal point；
Comparison sub-module, for using the maximum curve of eigenvalue of curve as road edge curve.
13. devices as claimed in claim 12, it is characterised in that the calculating characteristic value submodule specifically for：
For any bar curve, the pixel value sum of all pixels point in the curve of extension width is calculated, and it is long according to curve
Degree and embroidery calculate the coefficient of the curve；
The characteristic value of the curve is determined according to the coefficient and pixel value sum.
Device in 14. such as claim 8 to 13 as described in any claim, it is characterised in that the determination curvature module tool
Body is used for：
Calculate the curvature of the multiple pixels on the road edge curve, and by the average value of the curvature of the multiple pixel
As the bend curvature of the road edge curve.
15. a kind of vehicles, it is characterised in that the vehicle includes：
The device for detecting the bend curvature in road in the claims 8 to 14 described in any claim.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611143467.7A CN106767506A (en) | 2016-12-13 | 2016-12-13 | Method, device and vehicle for detecting the bend curvature in road |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611143467.7A CN106767506A (en) | 2016-12-13 | 2016-12-13 | Method, device and vehicle for detecting the bend curvature in road |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106767506A true CN106767506A (en) | 2017-05-31 |
Family
ID=58880592
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611143467.7A Pending CN106767506A (en) | 2016-12-13 | 2016-12-13 | Method, device and vehicle for detecting the bend curvature in road |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106767506A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109109597A (en) * | 2018-08-29 | 2019-01-01 | 北京汽车股份有限公司 | Vehicle Suspension Control System and automobile |
| CN109115127A (en) * | 2018-08-24 | 2019-01-01 | 合肥工业大学 | A kind of sub-pix peak point extraction algorithm based on Bezier |
| CN110057316A (en) * | 2019-04-25 | 2019-07-26 | 华南理工大学 | A method of the tower crane rod piece buckling monitoring based on unmanned plane image recognition |
| CN110307806A (en) * | 2018-03-27 | 2019-10-08 | 广东电网有限责任公司 | A kind of cable curvature estimation method and detection device based on binocular stereo vision |
| CN112686904A (en) * | 2020-12-14 | 2021-04-20 | 深兰人工智能(深圳)有限公司 | Lane division method, lane division device, electronic equipment and storage medium |
| CN113470102A (en) * | 2021-06-23 | 2021-10-01 | 依未科技(北京)有限公司 | Method, device, medium and equipment for measuring fundus blood vessel curvature with high precision |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1804928A (en) * | 2005-11-24 | 2006-07-19 | 上海交通大学 | Machine vision based estimation method for local geometry of driveway and vehicle location |
| CN101041355A (en) * | 2006-01-19 | 2007-09-26 | 通用汽车环球科技运作公司 | Lane departure warning and avoidance system with warning modification criteria |
| CN101296833A (en) * | 2005-09-26 | 2008-10-29 | 通用汽车环球科技运作公司 | Selectable lane-departure warning system and method |
| CN101630412A (en) * | 2008-07-18 | 2010-01-20 | 通用汽车环球科技运作公司 | Camera-based lane marker detection |
| CN101970273A (en) * | 2008-03-12 | 2011-02-09 | 本田技研工业株式会社 | Vehicle travel support device, vehicle, and vehicle travel support program |
| CN102706291A (en) * | 2012-05-18 | 2012-10-03 | 长安大学 | Method for automatically measuring road curvature radius |
| CN104166400A (en) * | 2014-07-11 | 2014-11-26 | 杭州精久科技有限公司 | Multi-sensor fusion-based visual navigation AGV system |
-
2016
- 2016-12-13 CN CN201611143467.7A patent/CN106767506A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101296833A (en) * | 2005-09-26 | 2008-10-29 | 通用汽车环球科技运作公司 | Selectable lane-departure warning system and method |
| CN1804928A (en) * | 2005-11-24 | 2006-07-19 | 上海交通大学 | Machine vision based estimation method for local geometry of driveway and vehicle location |
| CN101041355A (en) * | 2006-01-19 | 2007-09-26 | 通用汽车环球科技运作公司 | Lane departure warning and avoidance system with warning modification criteria |
| CN101970273A (en) * | 2008-03-12 | 2011-02-09 | 本田技研工业株式会社 | Vehicle travel support device, vehicle, and vehicle travel support program |
| CN101630412A (en) * | 2008-07-18 | 2010-01-20 | 通用汽车环球科技运作公司 | Camera-based lane marker detection |
| CN102706291A (en) * | 2012-05-18 | 2012-10-03 | 长安大学 | Method for automatically measuring road curvature radius |
| CN104166400A (en) * | 2014-07-11 | 2014-11-26 | 杭州精久科技有限公司 | Multi-sensor fusion-based visual navigation AGV system |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110307806A (en) * | 2018-03-27 | 2019-10-08 | 广东电网有限责任公司 | A kind of cable curvature estimation method and detection device based on binocular stereo vision |
| CN109115127A (en) * | 2018-08-24 | 2019-01-01 | 合肥工业大学 | A kind of sub-pix peak point extraction algorithm based on Bezier |
| CN109109597A (en) * | 2018-08-29 | 2019-01-01 | 北京汽车股份有限公司 | Vehicle Suspension Control System and automobile |
| CN110057316A (en) * | 2019-04-25 | 2019-07-26 | 华南理工大学 | A method of the tower crane rod piece buckling monitoring based on unmanned plane image recognition |
| CN112686904A (en) * | 2020-12-14 | 2021-04-20 | 深兰人工智能(深圳)有限公司 | Lane division method, lane division device, electronic equipment and storage medium |
| CN113470102A (en) * | 2021-06-23 | 2021-10-01 | 依未科技(北京)有限公司 | Method, device, medium and equipment for measuring fundus blood vessel curvature with high precision |
| CN113470102B (en) * | 2021-06-23 | 2024-06-11 | 依未科技(北京)有限公司 | Method, device, medium and equipment for measuring fundus blood vessel curvature with high precision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106767506A (en) | Method, device and vehicle for detecting the bend curvature in road | |
| US11801861B2 (en) | Using image augmentation with simulated objects for training machine learning models in autonomous driving applications | |
| CN111095291B (en) | Real-time detection of lanes and boundaries by autonomous vehicles | |
| US11087186B2 (en) | Fixation generation for machine learning | |
| US11900629B2 (en) | Surface profile estimation and bump detection for autonomous machine | |
| US20230012645A1 (en) | Deep neural network processing for sensor blindness detection in autonomous machine applications | |
| US10451425B2 (en) | Autonomous navigation system | |
| JP2022531092A (en) | Simulating realistic test data from transformed real-world sensor data for autonomous machine applications | |
| CN114902295A (en) | Three-dimensional intersection structure prediction for autonomous driving applications | |
| US11961243B2 (en) | Object detection using image alignment for autonomous machine applications | |
| CN110618678A (en) | Behavioral guided path planning in autonomous machine applications | |
| US12086208B2 (en) | Measuring the effects of augmentation artifacts on a machine learning network | |
| US20140159886A1 (en) | Driving assist device for vehicle | |
| US11313696B2 (en) | Method and apparatus for a context-aware crowd-sourced sparse high definition map | |
| US20230110713A1 (en) | Training configuration-agnostic machine learning models using synthetic data for autonomous machine applications | |
| US20240001957A1 (en) | Using image augmentation with simulated objects for training machine learning models in autonomous driving applications | |
| CN115909237A (en) | Visibility distance estimation using deep learning in autonomous machine applications | |
| CN115344117A (en) | Adaptive eye tracking machine learning model engine | |
| CN114841336A (en) | Patching a deployed deep neural network for autonomous machine applications | |
| WO2022061289A1 (en) | Simulating viewpoint transformations for sensor independent scene understanding in autonomous systems | |
| JP2022104773A (en) | Restraint device localization | |
| US20230351638A1 (en) | Detecting hazards based on disparity maps using computer vision for autonomous machine systems and applications | |
| US20230288223A1 (en) | Keyframe-based compression for world model representation in autonomous systems and applications | |
| CN108428356B (en) | Road condition map display and driving assistance application method based on fluid density field | |
| EP3850539B1 (en) | Deep neural network processing for sensor blindness detection in autonomous machine applications |
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 |
Application publication date: 20170531 |
|
| RJ01 | Rejection of invention patent application after publication |