CN106228138A - A kind of Road Detection algorithm of integration region and marginal information - Google Patents

A kind of Road Detection algorithm of integration region and marginal information Download PDF

Info

Publication number
CN106228138A
CN106228138A CN201610594100.0A CN201610594100A CN106228138A CN 106228138 A CN106228138 A CN 106228138A CN 201610594100 A CN201610594100 A CN 201610594100A CN 106228138 A CN106228138 A CN 106228138A
Authority
CN
China
Prior art keywords
pixel
road
image
threshold
point
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
Application number
CN201610594100.0A
Other languages
Chinese (zh)
Inventor
周小龙
伏进
吴彬
宫林
万欣
石为人
王成疆
王大洪
肖杰
李新平
刘垒
李�杰
林日
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
SMARTECH AVANCED RESEARCH SHENZHEN
Electric Power Research Institute of State Grid Chongqing Electric Power Co Ltd
State Grid Corp of China SGCC
State Grid Chongqing Electric Power Co Ltd
Original Assignee
SMARTECH AVANCED RESEARCH SHENZHEN
Electric Power Research Institute of State Grid Chongqing Electric Power Co Ltd
State Grid Corp of China SGCC
State Grid Chongqing Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by SMARTECH AVANCED RESEARCH SHENZHEN, Electric Power Research Institute of State Grid Chongqing Electric Power Co Ltd, State Grid Corp of China SGCC, State Grid Chongqing Electric Power Co Ltd filed Critical SMARTECH AVANCED RESEARCH SHENZHEN
Priority to CN201610594100.0A priority Critical patent/CN106228138A/en
Publication of CN106228138A publication Critical patent/CN106228138A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses the Road Detection algorithm of a kind of integration region and marginal information, first gather road image and use adaptive median filter denoising enhancing to obtain strengthening image；The R component of selective enhancement image in RGB color space, and the road area segmentation using maximum between-cluster variance Otsu method to realize image obtains binary segmentation image, and with serial mathematical morphology optimization binary segmentation image；Then use optimal edge detection Canny operator detection binary segmentation image to obtain road-edge detection information, and use maximum variance between clusters to calculate the dual threshold in optimal edge detection Canny operator；Finally utilize binary segmentation image and the lane boundary line in road-edge detection acquisition of information image.The algorithm that the present invention provides utilizes image processing techniques to detect and extracts road information, for the autonomous navigation system of cable tunnel robot, is a kind of self adaptation lane boundary line segmentation fusion method, and this blending algorithm can obtain smooth road edge accurately.

Description

A kind of Road Detection algorithm of integration region and marginal information
Technical field
The present invention relates to robot control field, the Road Detection algorithm of a kind of integration region and marginal information.
Background technology
Vision guided navigation is a kind of important technology in intelligent mobile robot research field and study hotspot.Detect and carry It is the premise of vision guided navigation and autonomous by way of road guide wire.Compared with structurized road, non-structured road has Certain complexity and multiformity, and road scene has randomness, and the detection to the type road at present does not also have one The algorithm that adaptability is the strongest.General Road Detection algorithm or based on region segmentation or based on rim detection.Region segmentation It is generally basede on the global characteristics of image, background area and target area are split, but the marginal position being partitioned into is not very Accurately.Rim detection is generally basede on the local feature of image, can obtain more accurate marginal position, but the most also detect except The edge of a lot of redundancies.In the environment of under based on cable tunnel robot navigation, cable tunnel is equivalent to need that extracts to lead Navigation channel road, needs also exist for this kind of road is carried out region segmentation and rim detection, accordingly, it would be desirable to one can obtain segmentation effect relatively Good road boundary method, Application way can carry out road boundary segmentation to the unstructured road environment that other are similar.
Summary of the invention
It is an object of the invention to need the Road Detection algorithm of a kind of energy integration region and marginal information to destructuring road Road environment carries out road boundary segmentation.
It is an object of the invention to be achieved through the following technical solutions:
The integration region of present invention offer and the Road Detection algorithm of marginal information, comprise the following steps:
S1: gather road image and use adaptive median filter denoising enhancing to obtain strengthening image；
S2: the R component of selective enhancement image in RGB color space, and use maximum between-cluster variance Otsu method realization figure The road area segmentation of picture obtains binary segmentation image, and optimizes binary segmentation image with serial mathematical morphology；
S3: use optimal edge detection Canny operator detection binary segmentation image to obtain road-edge detection information, and adopt The dual threshold in optimal edge detection Canny operator is calculated with maximum variance between clusters；
S4: utilize binary segmentation image and the lane boundary line in road-edge detection acquisition of information image.
Further, the adaptive median filter in described step S1 specifically comprises the following steps that
Step A: calculate according to below equation:
A1=zmed-zmin, A2=zmed-zmax；
Wherein, A1Represent the difference between pixel grey scale intermediate value and pixel grey scale minima in template；zmedRepresent pixel in template The intermediate value of gray scale；zminThe minima of pixel grey scale in expression template；A2Represent between intermediate value and the maximum of pixel grey scale in template Difference；zmaxThe maximum of pixel grey scale in expression template；
Judge whether to meet relationship below: A1＞ 0 and A2＜ 0, if it is satisfied, then enter step B；If be unsatisfactory for, then Increase SxyValue, it is judged that template size Sxy≤Smax, if it is satisfied, then return to perform A, if be unsatisfactory for, then output valve zxy；
Wherein, SxyCentered by pixel (x, template window corresponding time y)；
Minima z of pixel in template windowmin=min (Sxy)；
The maximum z of pixel in template windowmax=max (Sxy)；
Intermediate value z of pixel in template windowmed=med (Sxy)；
zxyFor pixel (x, y) gray-scale pixel values at place；
SmaxRepresent SxyThe gray-scale pixel values allowed；
Step B: calculate B according to below equation1:
B1=zxy-zmin, A2=zxy-zmax；
Wherein, B1Represent the difference between pixel grey scale minima in template center's coordinate grey scale pixel value and template；
Judge whether to meet relationship below: if B1＞ 0 and B2＜ 0, if it is satisfied, then output zxy；If be unsatisfactory for, then Output zmed；
Step C: judge the z in step A and BmedAnd zxyWhether it is an impulsive noise, if not being the most impulsive noise, then Output current pixel value；If it is not, then output neighborhood intermediate value.
Further, the maximum between-cluster variance Otsu method in described step S2 realize image road area segmentation obtain two-value Segmentation threshold in segmentation image, obtains according to following steps:
S21: divide an image into L gray level；
S22: calculate number of pixels n belonging to gray level ii, and each gray level occur probability be:
p i = n i N ;
Wherein,Pixel count for whole image；
S23: pixel threshold value T in image is divided into two classes C0And C1, gray level pixel in [0, T-1] belongs to C0, Gray level pixel in [T, L-1] belongs to C1,
S24: zoning C0And C1Probability be respectively as follows:
S25: region C0And C1Average gray value be respectively as follows:
The average gray of S26: entire image is:
Wherein, T is threshold value, and the span of T is [0, L-1]；
S27: calculating threshold value is inter-class variance during T:
σ2(T)=P0(μ0-μ)2+P1(μ1-μ)2=P0P1(μ0-μ1)2；
S28: obtain and work as σ2(T) T when taking maximum the segmentation threshold as maximum between-cluster variance Otsu method.
Further, the dual threshold in described step S3 includes high threshold and threshold ones；Specifically according to following steps Obtain:
Otsu algorithm is used to calculate acquisition optimal threshold, and using optimal threshold as the high threshold of Canny operator；Press According to the below equation described threshold ones of calculating:
Tl=(0.4～0.6) Th, wherein, TlRepresent threshold ones, ThRepresent high threshold.
Further, described road-edge detection information obtains according to following steps:
S31: obtain the gradient magnitude of pixel in binary segmentation image, and in the following manner processes:
S32: if the gradient magnitude at pixel is more than the high threshold set, then pixel is marginal point；
S33: if the gradient magnitude at pixel is less than the threshold ones set, then pixel is non-edge point；
S34: if the gradient magnitude at pixel is between high and low threshold value, then increase constraints:
S35: if the gradient at pixel is more than time in the pixel neighborhood of a point of high threshold, then pixel is edge Point；Otherwise discard pixel.
Further, the lane boundary line in described step S4 obtains according to following steps:
S41: set the border trusted area of cut zone；
S42: obtain the binary segmentation figure of cut zone；
S43: by row scanning in binary segmentation figure, when scanning the pixel of black and white change, then in pixel being M the pixel border trusted area [i-M, i+M] as road boundary point is extended about the heart；Until searching for complete acquisition border Trusted area；
S44: in binary segmentation image, scans the pixel in trusted area [i-M, i+M], one by one if can in k-th In letter region, first scan is to marginal point, is the most all regarded as boundary point；
S45: scan for marginal point in the K+1 trusted area, then judge as follows to edge point:
S451: be in one be marked as in road boundary neighborhood of a point if existed in marginal point, then will be labeled as Road boundary point；
S452: be marked as in road boundary neighborhood of a point, then by whole marginal points if marginal point is not located in one It is designated as boundary point；
S453: search for each border trusted area one by one, and mark all road boundaries point, obtain road boundary Line；
Wherein, i represents the abscissa of image array；M represents constant, and oneself is optionally arranged；K represents variable, and k-th can Letter region.
Owing to have employed technique scheme, present invention have the advantage that:
The integration region of present invention offer and the Road Detection algorithm of marginal information, for the road under unstructured moving grids I.e. cable tunnel road, utilizes image processing techniques to detect and extracts road information, autonomous for cable tunnel robot Navigation system.First, adaptive median filter denoising is used to strengthen the road image gathered；Secondly, select at RGB color sky Between R component figure in apply maximum between-cluster variance Otsu method to realize road area segmentation, and with serial mathematical morphology to segmentation Binary map optimizes further；Then use optimal edge detection Canny operator detection road edge, and use maximum between-cluster variance Method calculates the dual threshold in Canny operator automatically, it is achieved adaptivity；Finally utilize region segmentation and rim detection respective excellent Gesture,；Lane boundary line is detected by integration region and marginal information；Through experimental verification, this blending algorithm can obtain smooth Road edge accurately.
Road Detection algorithm is the pith in vision guided navigation, and the accuracy of road boundary line drawing decides whole system The performance of system.With cable tunnel road as experimental subject, research and analyse the design sketch of region segmentation and rim detection, utilized two The complementary characteristic of person, devises the border detection algorithm of a kind of integration region and marginal information.It is above-mentioned it is demonstrated experimentally that this fusion is calculated Method is capable of detecting when the lane boundary line that position is accurate, smooth.
Other advantages, target and the feature of the present invention will be illustrated to a certain extent in the following description, and And to a certain extent, will be apparent to those skilled in the art based on to investigating hereafter, or can To be instructed from the practice of the present invention.The target of the present invention and other advantages can be realized by description below and Obtain.
Accompanying drawing explanation
The accompanying drawing of the present invention is described as follows.
Fig. 1 a is adaptive median filter.
Fig. 1 b is 9X9 mean filter.
Fig. 1 c is 9X9 medium filtering.
Fig. 2 a is the OTSU segmentation figure of R component.
Fig. 2 b is the OTSU segmentation figure of gray-scale map.
Fig. 2 c is the OTSU segmentation figure of R component.
Fig. 2 d is the OTSU segmentation figure of gray-scale map.
Fig. 3 a is the OTSU segmentation figure of R component.
After Fig. 3 b is Fig. 3 a Morphological scale-space.
Fig. 3 c is the OTSU segmentation figure of R component.
After Fig. 3 d is Fig. 3 c Morphological scale-space.
Fig. 4 a is mileage chart.
Fig. 4 b is self adaptation Canny operator.
Fig. 4 c is Prewitt boundary operator.
Fig. 4 d is Canny operator in matlab.
Fig. 4 e is mileage chart 2.
Fig. 4 f is self adaptation Canny operator.
My Prewitt boundary operator of Fig. 4 g.
Fig. 4 h is Canny operator in matlab.
Fig. 5 is the flow chart of blending algorithm.
Fig. 6 is the lower 10 neighborhood templates in blending algorithm.
Fig. 7 a is that road area splits Fig. 1.
Fig. 7 b is self adaptation Canny rim detection.
Fig. 7 c is the road boundary after merging.
Fig. 7 d is that road area splits Fig. 2.
Fig. 7 e is self adaptation Canny rim detection.
Fig. 7 f is the road boundary after merging.
Detailed description of the invention
The invention will be further described with embodiment below in conjunction with the accompanying drawings.
Embodiment 1
As it can be seen, the integration region of the present embodiment offer and the Road Detection algorithm of marginal information, comprise the following steps:
S1: gather road image and use adaptive median filter denoising enhancing to obtain strengthening image；
S2: the R component of selective enhancement image in RGB color space, and use maximum between-cluster variance Otsu method realization figure The road area segmentation of picture obtains binary segmentation image, and optimizes binary segmentation image with serial mathematical morphology；
S3: use optimal edge detection Canny operator detection binary segmentation image to obtain road-edge detection information, and adopt The dual threshold in optimal edge detection Canny operator is calculated with maximum variance between clusters；
S4: utilize binary segmentation image and the lane boundary line in road-edge detection acquisition of information image.
Adaptive median filter in described step S1 specifically comprises the following steps that
Step A: calculate according to below equation:
A1=zmed-zmin, A2=zmed-zmax；
Wherein, A1Represent the difference between pixel grey scale intermediate value and pixel grey scale minima in template；zmedRepresent pixel in template The intermediate value of gray scale；zminThe minima of pixel grey scale in expression template；A2Represent between intermediate value and the maximum of pixel grey scale in template Difference；zmaxThe maximum of pixel grey scale in expression template；
Judge whether to meet relationship below: A1＞ 0 and A2＜ 0, if it is satisfied, then enter step B；If be unsatisfactory for, then Increase SxyValue, it is judged that template size Sxy≤Smax, if it is satisfied, then return to perform A, if be unsatisfactory for, then output valve zxy；
Wherein, SxyCentered by pixel (x, template window corresponding time y)；
Minima z of pixel in template windowmin=min (Sxy)；
The maximum z of pixel in template windowmax=max (Sxy)；
Intermediate value z of pixel in template windowmed=med (Sxy)；
zxyFor pixel (x, y) gray-scale pixel values at place；
SmaxRepresent SxyThe gray-scale pixel values allowed；
Step B: calculate B according to below equation1:
B1=zxy-zmin, A2=zxy-zmax；
Wherein, B1Represent the difference between pixel grey scale minima in template center's coordinate grey scale pixel value and template；
Judge whether to meet relationship below: if B1＞ 0 and B2＜ 0, if it is satisfied, then output zxy；If be unsatisfactory for, then Output zmed；
Step C: judge the z in step A and BmedAnd zxyWhether it is an impulsive noise, if not being the most impulsive noise, then Output current pixel value；If it is not, then output neighborhood intermediate value.
Maximum between-cluster variance Otsu method in described step S2 realizes the road area segmentation of image and obtains binary segmentation figure Segmentation threshold in Xiang, obtains according to following steps:
S21: divide an image into L gray level；
S22: calculate number of pixels n belonging to gray level ii, and each gray level occur probability be:
p i = n i N ;
Wherein,Pixel count for whole image；
S23: pixel threshold value T in image is divided into two classes C0And C1, gray level pixel in [0, T-1] belongs to C0, Gray level pixel in [T, L-1] belongs to C1,
S24: zoning C0And C1Probability be respectively as follows:
S25: region C0And C1Average gray value be respectively as follows:
The average gray of S26: entire image is:
Wherein, T is threshold value, and the span of T is [0, L-1]；
S27: calculating threshold value is inter-class variance during T:
σ2(T)=P0(μ0-μ)2+P1(μ1-μ)2=P0P1(μ0-μ1)2；
S28: obtain and work as σ2(T) T when taking maximum the segmentation threshold as maximum between-cluster variance Otsu method.
Dual threshold in described step S3 includes high threshold and threshold ones；Specifically obtain according to following steps:
Otsu algorithm is used to calculate acquisition optimal threshold, and using optimal threshold as the high threshold of Canny operator；Press According to the below equation described threshold ones of calculating:
Tl=(0.4～0.6) Th, wherein, TlRepresent threshold ones, ThRepresent high threshold.
Described road-edge detection information obtains according to following steps:
S31: obtain the gradient magnitude of pixel in binary segmentation image, and in the following manner processes:
S32: if the gradient magnitude at pixel is more than the high threshold set, then pixel is marginal point；
S33: if the gradient magnitude at pixel is less than the threshold ones set, then pixel is non-edge point；
S34: if the gradient magnitude at pixel is between high and low threshold value, then increase constraints:
S35: if the gradient at pixel is more than time in the pixel neighborhood of a point of high threshold, then pixel is edge Point；Otherwise discard pixel.
Lane boundary line in described step S4 obtains according to following steps:
S41: set the border trusted area of cut zone；
S42: obtain the binary segmentation figure of cut zone；
S43: by row scanning in binary segmentation figure, when scanning the pixel of black and white change, then in pixel being M the pixel border trusted area [i-M, i+M] as road boundary point is extended about the heart；Can until searching for complete acquisition border Letter region；
S44: in binary segmentation image, scans the pixel in trusted area [i-M, i+M], one by one if can in k-th In letter region, first scan is to marginal point, is the most all regarded as boundary point；
S45: scan for marginal point in the K+1 trusted area, then judge as follows to edge point:
S451: be in one be marked as in road boundary neighborhood of a point if existed in marginal point, then will be labeled as Road boundary point；
S452: be marked as in road boundary neighborhood of a point, then by whole marginal points if marginal point is not located in one It is designated as boundary point；
S453: search for each border trusted area one by one, and mark all road boundaries point, obtain road boundary Line；
Wherein, i represents the abscissa of image array；M represents constant, and oneself is optionally arranged；K represents variable, and k-th can Letter region.
Embodiment 2
The Image semantic classification that the present embodiment provides uses adaptive median filter, and usual medium filtering is a kind of nonlinear filtering Ripple device, it is better than mean filter to the treatment effect of random noise, can be while keeping the original clear profile of image, figure The noise of picture filters.But standard medium filtering there is also some problems, when filter template selection is improper or template is the biggest Time, the details of image also can be erased, and be there will be the phenomenons such as edge thinning.Therefore, in order to obtain preferable denoising effect And edge details, specifically comprising the following steps that of the adaptive median filter of the present embodiment
If SxyCentered by pixel (x, the template window of many correspondences time y).The minima of pixel in default template window zmin=min (Sxy), the maximum z of pixel in template windowmax=max (Sxy), intermediate value z of pixel in template windowmed=med (Sxy), pixel (x, y) gray-scale pixel values z at placexy, SmaxRepresent SxyThe gray-scale pixel values allowed.Adaptive median filter is held Row process is as follows:
Step A:A1=zmed-zmin, A2=zmed-zmaxIf, A1＞ 0 and A2＜ 0, is carried out step B, the most just increases Sxy Value, if template size Sxy≤Smax, then return to perform A, otherwise output valve zxy。
Wherein, A1Represent the difference between pixel grey scale intermediate value and pixel grey scale minima in template；zmedRepresent pixel in template The intermediate value of gray scale；zminThe minima of pixel grey scale in expression template；A2Represent between intermediate value and the maximum of pixel grey scale in template Difference；zmaxThe maximum of pixel grey scale in expression template；
Step B:B1=zxy-zmin, A2=zxy-zmaxIf, B1＞ 0 and B2＜ 0, exports zxy, otherwise export zmed。
Wherein, B1Represent the difference between pixel grey scale minima in template center's coordinate grey scale pixel value and template；
Step A and B are respectively intended to judge zmedAnd zxyWhether it is an impulsive noise, if not being the most impulsive noise, then Algorithm just exports current pixel value, to avoid unnecessary loss of detail, and otherwise output field intermediate value.
Such as Fig. 1 a, b, for the filter effect of algorithms of different；Wherein, Fig. 1 a adaptive median filter；Fig. 1 b is the filter of 9X9 average Ripple；Fig. 1 c is 9X9 medium filtering；Result shows, after mean filter, whole image becomes the fuzzyyest；Although after standard medium filtering Eliminate random noise, but edge thinning phenomenon occurs, erase edge details.Adaptive median filter is owing to automatically selecting Use original pixel value at template size, and sound, not remove only noise and remain edge details, well imitated Really.
The road area segmentation that the present embodiment provides, is carried out as a example by the Threshold segmentation of R component based on RGB color space Explanation；By comparing maximum variance between clusters (Otsu method), maximum entropy threshold method and the segmentation figure of region-growing method, determine this Embodiment uses Otsu method to carry out road area segmentation.The basic thought of maximum variance between clusters (Otsu method) is to seek most preferably Threshold value is so that the variance between target area and background area reaches maximum；Described optimal threshold calculates according to Otsu formula, Detailed process is as follows:
If image has L gray level i.e. gray level [0, L-1], belong to gray level i has niIndividual pixel, the most whole image Pixel count is:The probability that each gray level occurs is:For pi, have
Pixel threshold value T in image is divided into two classes C0And C1, gray level pixel in [0, T-1] belongs to C0, gray scale Level pixel in [T, L-1] belongs to C1, then region C0And C1Probability be respectively as follows:Region C0And C1Average gray value be respectively as follows:
Then the average gray of entire image is:
When threshold value is T, inter-class variance is: σ2(T)=P0(μ0-μ)2+P1(μ1-μ)2=P0P1(μ0-μ1)2
Wherein, the span of T is [0, L-1], works as σ2(T) T when taking maximum is Otsu segmentation threshold.
If directly the gray-scale map of road image being carried out OTSU Threshold segmentation, the segmentation effect at road boundary is not very Good, some redundant blocks can be produced, blending algorithm below is produced impact.By road color characteristic and Experimental comparison, select The R component figure of RGB color space carries out Threshold segmentation, obtains more satisfactory lane segmentation figure.
Fig. 2 is the region segmentation figure of algorithms of different,
2.2 morphology processing
Due to the existence of noise, lane segmentation binary map is the most especially desirable, is studded with some little in target area Hole or crack, and on background area, it is studded with some little burrs, it is impossible to it is directly used in extraction lane boundary line, it is therefore desirable to Lane segmentation binary map is made next step optimization.Mathematical morphology can be used to process for these interference.
Mathematical morphology is to utilize algebraic set, allows structural element travel through all pixels, and with the pixel of surrounding neighbors Carry out logical operations.Its elementary operation has: expands and corrosion, opens operation and closed operation.
Vacation lets a and b be Z2In set, B is structural element, then:
A is defined as by B corrosion:
A is expanded by B and is defined as:
The B opening operation to A, is defined as:
The B closed operation to A, is defined as:
Opening operation is exactly first to use structural element B that original image A uses erosion operation, then with structural element B to corrosion knot Fruit uses dilation operation, and effect is can to disconnect narrow connection and eliminate fine, soft fur thorn；Closed operation is exactly first to use structural element B to former Beginning image A uses dilation operation, then with structural element B, expansion results to be used erosion operation, effect be to make narrow interruption up With fill up tiny fracture and hole.
Analyze road image region segmentation figure it is found that road area stitches with thin containing some noise holes, background area The burr of some redundancies is contained in territory, and road boundary is not the most smooth.In order to obtain more satisfactory lane segmentation figure, must Little burr must be eliminated and make tiny hole and gap up.The Morphological scale-space that opening and closing operation is combined by the present embodiment is calculated Method, wherein the structural element in opening operation is less than the structural element in closed operation.Experiments verify that, lane segmentation figure is used Opening operation and a closed operation, can well remove little noise spot and burr, fills some pixel cracks and noise hole. Design sketch is as shown in figs. 2 a-d.
Fig. 3 a-d is the Morphological scale-space of segmentation figure；
The road-edge detection of the present embodiment uses Canny rim detection, and Canny operator is same preferably suppression noise Time be also accurately positioned out edge, compared with other edge detection algorithms, that reduce the sensitivity to noise, and improve opposite side The sensitivity of edge.Canny operator is based on three criterions:
(1) signal to noise ratio is high, and marginal error rate is low, and the edge detected is usually true edge.
(2) positioning precision is high, and the edge detected is with true edge center closely.
(3) single endpoint detections, to true edge point, only returns a value.
Canny operator to realize process as follows:
(1) Gauss smoothing filter denoising is utilized
G ( x , y ) = 1 2 πσ 2 exp ( - x 2 + y 2 2 σ 2 )
G (x, y)=f (x, y) * G (x, y)
Wherein, σ is Gauss filter parameter, and it determines the denoising effect of image.The value of σ is it is generally required to according to experiment really Fixed.
(2) gradient magnitude and the calculating in direction
Canny algorithm utilizes first difference to divide to calculate gradient magnitude and the direction of each pixel in image, thus The histogram of gradients of image can be obtained.In image, (x y) is in difference G on x and y direction to pixelx(x, y) and Gy(x y) divides It is not:
Gx(x, y)=(I (x, y+1)-I (x, y)+I (x+1, y+1)-I (x+1, y))/2
Gy(x, y)=(I (x+1, y)-I (x, y)+I (x+1, y+1)-I (x, y+1))/2
Then (x, y) gradient magnitude and the direction at place is respectively as follows: point
G ( x , y ) = G x 2 ( x , y ) + G y 2 ( x , y )
θ ( x , y ) = arctan ( G x ( x , y ) G y ( x , y ) )
(3) non-maxima suppression
After the gradient magnitude of all pixels and direction all calculate, the non-maximum gradient in neighborhood be put It is zero, retains the pixel of gradient maximum, thus get rid of non-edge pixels, obtain thinner edge.In same gradient side Upwards, if gradient M (x, value y) be less than consecutive points gradient, then make M (x, y)=0.Otherwise it is regarded as candidate marginal.
(4) detection and adjoining edge
Finally, Canny operator employs hysteresis threshold (the most high and low two threshold values).Start to scan all pixels: the most such as Fruit gradient magnitude at some pixel is more than the high threshold set, then just this point is considered as marginal point；If 2. existed Gradient magnitude at some pixel is less than the Low threshold set, then just this point is considered as non-edge point；If 3. at certain Gradient magnitude at one pixel between high and low threshold value, then needs to add a constraints again: if this pixel is one Individual gradient, more than time in 8 neighborhoods of the pixel of high threshold, is just regarded as marginal point, is otherwise discarded.
The present embodiment uses self adaptation Canny rim detection, and first parameter of Canny operator is exactly gaussian filtering parameter σ, it is related to the filter effect of image, and the least then image denoising effect of σ value is bad, and σ has then obscured too greatly the details of image, Edge is the fuzzyyest.According to experiment, we take σ is 0.2.Two other parameter of Canny operator is exactly the selection of high-low threshold value, It is usually authority experience artificially to pre-set.The Gradient distribution of different road images also differs, therefore the height of every two field picture The setting of threshold value also differs, if high threshold arranges the highest, then segment path edge can not be detected, and examines The edge measured is likely to occur desultory situation；If high threshold arranges the lowest, will detect that much interference or not phase The edge closed.Additionally, robot is in motor process, image is on a frame-by-frame basis changing, and is manually set two fixing height Threshold value carries out rim detection can not meet all road images, it is therefore desirable to design is a kind of according to present frame road image information certainly The dynamic method obtaining high-low threshold value.The form of the gradient magnitude histogram of gradients of image is showed, finds that it presents Double-hump characteristics；Additionally, the high-low threshold value in Canny operator is the most also a kind of Grads threshold.In order to make tradition Canny calculate Method realizes adaptivity, uses Otsu algorithm to calculate an optimal threshold when determining the high-low threshold value of Canny operator, And using this threshold value as the high threshold of Canny operator.Threshold ones is determined by high threshold, typically takes Tl=(0.4～0.6) Th, the present embodiment takes Tl=0.4Th。
Fig. 4 a-h is the road edge identification of the rim detection of road image, integration region and marginal information, specific as follows:
Threshold segmentation based on region, although target road and background area can be separated, but fixed to lane boundary line Position is the most accurate.Although road-edge detection can the most accurately detect the border of road, but also detected the most simultaneously Relevant edge.If directly rim detection binary map application Hough transform being extracted leading line, result have the biggest not Definitiveness, it is therefore necessary to eliminate the interference at uncorrelated edge.The effect detected according to above-mentioned Threshold segmentation and Canny, Ke Yifa Both existing, there is complementary characteristic.Therefore, according to the complementary characteristic of the two, a kind of integration region proposed in the present embodiment and edge The lane boundary line detection algorithm of information.Utilize region segmentation can split the outlet feature with region, non-road by edge binary map In Clutter edge filter, retain target and background intersection single edge, lane boundary line can be detected exactly.
Concrete blending algorithm is:
Border trusted area is set according to region segmentation figure.
In region segmentation binary map, road boundary is typically in the cohesive position in target and background region.Divide in region Cutting by row scanning in binary map, as the pixel I scanning black and white change, (j, time i), is then taken as possible boundary point, and will Extending M the pixel trusted area [i-M, i+M] as road boundary point about this pixel, in the present embodiment, M is set as 10. N number of trusted area can be produced after search.Wherein, i represents the abscissa (arranging) of image array；M represents constant, and oneself regards feelings Condition is arranged；K represents variable, k-th trusted area.
Lane boundary line is positioned in trusted area.
In edge binary map, scan the pixel in trusted area [i-10, i+10] one by one, if in k-th confidence region In territory, first scan is to marginal point, is the most all regarded as boundary point；Then proceed to search in the K+1 trusted area These marginal points if there is marginal point within the range, are then made the following judgment by rope: (1) is if in these marginal points Exist and be in a point being marked as in lower 10 neighborhoods (as shown in Figure 5) of road boundary point, then only this point is labeled as Road boundary point；(2) it is marked as in lower 10 neighborhoods of road boundary point, then by this model if these marginal points are not located in one Enclose interior whole marginal points and be designated as boundary point.Search for each trusted area one by one according to this algorithm, mark roadside, all roads Boundary's point, the road boundary of i.e. available required single edges.
Fig. 5 is the flow chart of blending algorithm, the lower 10 neighborhood templates in the blending algorithm of the present embodiment；As shown in Figure 6, figure Road boundary in Xiang is generally single edges at distant view, is likely to occur false edge, dual edge and burr etc. dry at close shot Disturbing, in above-mentioned algorithm, if do not made a decision, the marginal point in trusted area being all considered as border, then result is probably double The lane boundary line at edge.And blending algorithm takes full advantage of These characteristics and the seriality of road boundary, obtain comparing reason The single road boundary thought.This algorithm make use of the complementary characteristic between region segmentation and rim detection, can be relatively accurately Detect lane boundary line.
Fig. 7 a-f is the road edge identification at integration region and edge, shown in Fig. 7 a-f；Road Detection algorithm is that vision is led Pith in boat, the accuracy of road boundary line drawing decides the performance of whole system.With cable tunnel road as reality Test object, researched and analysed the design sketch of region segmentation and rim detection, utilize the complementary characteristic of the two, devise a kind of fusion Region and the border detection algorithm of marginal information.Above-mentioned it is demonstrated experimentally that this blending algorithm be capable of detecting when position accurately, smooth Lane boundary line.
Finally illustrating, above example is only in order to illustrate technical scheme and unrestricted, although with reference to relatively The present invention has been described in detail by good embodiment, it will be understood by those within the art that, can be to the skill of the present invention Art scheme is modified or equivalent, and without deviating from objective and the scope of the technical program, it all should be contained in the present invention Protection domain in the middle of.

Claims (6)

1. an integration region and the Road Detection algorithm of marginal information, it is characterised in that: comprise the following steps:
S1: gather road image and use adaptive median filter denoising enhancing to obtain strengthening image；
S2: the R component of selective enhancement image in RGB color space, and use maximum between-cluster variance Otsu method to realize image Road area segmentation obtains binary segmentation image, and optimizes binary segmentation image with serial mathematical morphology；
S3: use optimal edge detection Canny operator detection binary segmentation image to obtain road-edge detection information, and use Big Ostu method calculates the dual threshold in optimal edge detection Canny operator；
S4: utilize binary segmentation image and the lane boundary line in road-edge detection acquisition of information image.
2. integration region as claimed in claim 1 and the Road Detection algorithm of marginal information, it is characterised in that: described step S1 In adaptive median filter specifically comprise the following steps that
Step A: calculate according to below equation:
A1=zmed-zmin, A2=zmed-zmax；
Wherein, A1Represent the difference between pixel grey scale intermediate value and pixel grey scale minima in template；zmedRepresent pixel grey scale in template Intermediate value；zminThe minima of pixel grey scale in expression template；A2Represent between intermediate value and the maximum of pixel grey scale in template Difference；zmaxThe maximum of pixel grey scale in expression template；
Judge whether to meet relationship below: A1> 0 and A2< 0, if it is satisfied, then enter step B；If be unsatisfactory for, then increase Sxy Value, it is judged that template size Sxy≤Smax, if it is satisfied, then return to perform A, if be unsatisfactory for, then output valve zxy；
Wherein, SxyCentered by pixel (x, template window corresponding time y)；
Minima z of pixel in template windowmin=min (Sxy)；
The maximum z of pixel in template windowmax=max (Sxy)；
Intermediate value z of pixel in template windowmed=med (Sxy)；
zxyFor pixel (x, y) gray-scale pixel values at place；
SmaxRepresent SxyThe gray-scale pixel values allowed；
Step B: calculate B according to below equation1:
B1=zxy-zmin, A2=zxy-zmax；
Wherein, B1Represent the difference between pixel grey scale minima in template center's coordinate grey scale pixel value and template；
Judge whether to meet relationship below: if B1> 0 and B2< 0, if it is satisfied, then output zxy；If be unsatisfactory for, then export zmed；
Step C: judge the z in step A and BmedAnd zxyWhether being an impulsive noise, if not being the most impulsive noise, then exporting Current pixel value；If it is not, then output neighborhood intermediate value.
3. integration region as claimed in claim 1 and the Road Detection algorithm of marginal information, it is characterised in that: described step S2 In maximum between-cluster variance Otsu method realize image road area split obtain in binary segmentation image segmentation threshold, be by Obtain according to following steps:
S21: divide an image into L gray level；
S22: calculate number of pixels n belonging to gray level ii, and each gray level occur probability be:
p i = n i N ;
Wherein,Pixel count for whole image；
S23: pixel threshold value T in image is divided into two classes C0And C1, gray level pixel in [0, T-1] belongs to C0, gray scale Level pixel in [T, L-1] belongs to C1,
S24: zoning C0And C1Probability be respectively as follows:
S25: region C0And C1Average gray value be respectively as follows:
The average gray of S26: entire image is:
Wherein, T is threshold value, and the span of T is [0, L-1]；
S27: calculating threshold value is inter-class variance during T:
σ2(T)=P0(μ0-μ)2+P1(μ1-μ)2=P0P1(μ0-μ1)2；
S28: obtain and work as σ2(T) T when taking maximum the segmentation threshold as maximum between-cluster variance Otsu method.
4. integration region as claimed in claim 1 and the Road Detection algorithm of marginal information, it is characterised in that: described step S3 In dual threshold include high threshold and threshold ones；Specifically obtain according to following steps:
Otsu algorithm is used to calculate acquisition optimal threshold, and using optimal threshold as the high threshold of Canny operator；According to The lower formula described threshold ones of calculating:
Tl=(0.4～0.6) Th, wherein, TlRepresent threshold ones, ThRepresent high threshold.
5. integration region as claimed in claim 1 and the Road Detection algorithm of marginal information, it is characterised in that: roadside, described road Edge detection information obtains according to following steps:
S31: obtain the gradient magnitude of pixel in binary segmentation image, and in the following manner processes:
S32: if the gradient magnitude at pixel is more than the high threshold set, then pixel is marginal point；
S33: if the gradient magnitude at pixel is less than the threshold ones set, then pixel is non-edge point；
S34: if the gradient magnitude at pixel is between high and low threshold value, then increase constraints:
S35: if the gradient at pixel is more than time in the pixel neighborhood of a point of high threshold, then pixel is marginal point；No Then discard pixel.
6. integration region as claimed in claim 1 and the Road Detection algorithm of marginal information, it is characterised in that: described step S4 In lane boundary line obtain according to following steps:
S41: set the border trusted area of cut zone；
S42: obtain the binary segmentation figure of cut zone；
S43: by row scanning in binary segmentation figure, when scanning the pixel of black and white change, then left centered by pixel Right M pixel of extension is as the border trusted area [i-M, i+M] of road boundary point；Until searching for confidence region, complete acquisition border Territory；
S44: in binary segmentation image, scans the pixel in trusted area [i-M, i+M] one by one, if in k-th confidence region In territory, first scan is to marginal point, is the most all regarded as boundary point；
S45: scan for marginal point in the K+1 trusted area, then judge as follows to edge point:
S451: be in one be marked as in road boundary neighborhood of a point if existed in marginal point, then will be labeled as road Boundary point；
S452: be marked as in road boundary neighborhood of a point if marginal point is not located in one, then whole marginal points be designated as Boundary point；
S453: search for each border trusted area one by one, and mark all road boundaries point, obtain lane boundary line；
Wherein, i represents the abscissa of image array；M represents constant, and oneself is optionally arranged；K represents variable, k-th confidence region Territory.
CN201610594100.0A 2016-07-26 2016-07-26 A kind of Road Detection algorithm of integration region and marginal information Pending CN106228138A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610594100.0A CN106228138A (en) 2016-07-26 2016-07-26 A kind of Road Detection algorithm of integration region and marginal information

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610594100.0A CN106228138A (en) 2016-07-26 2016-07-26 A kind of Road Detection algorithm of integration region and marginal information

Publications (1)

Publication Number Publication Date
CN106228138A true CN106228138A (en) 2016-12-14

Family

ID=57533434

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610594100.0A Pending CN106228138A (en) 2016-07-26 2016-07-26 A kind of Road Detection algorithm of integration region and marginal information

Country Status (1)

Country Link
CN (1) CN106228138A (en)

Cited By (43)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106682787A (en) * 2017-01-09 2017-05-17 北京航空航天大学 Method for quickly generating generalized Voronoi diagram based on wavefront algorithm
CN106981055A (en) * 2017-03-30 2017-07-25 西安交通大学 A kind of ICCD image de-noising methods
CN107274403A (en) * 2017-06-30 2017-10-20 长安大学 A kind of evaluation method of flotation surface quality
CN107369159A (en) * 2017-06-29 2017-11-21 大连理工大学 Threshold segmentation method based on multifactor two-dimensional gray histogram
CN107590501A (en) * 2017-07-30 2018-01-16 华南理工大学 A kind of improved Canny adaptive edge extracting methods based on Otsu
CN108109155A (en) * 2017-11-28 2018-06-01 东北林业大学 A kind of automatic threshold edge detection method based on improvement Canny
CN108242060A (en) * 2016-12-23 2018-07-03 重庆邮电大学 A kind of method for detecting image edge based on Sobel operators
CN108470340A (en) * 2017-02-23 2018-08-31 南宁市富久信息技术有限公司 A kind of improved Sobel edge detection algorithms
CN108573492A (en) * 2018-04-02 2018-09-25 电子科技大学 A kind of real time radar search coverage detection method
CN108710883A (en) * 2018-06-04 2018-10-26 国网辽宁省电力有限公司信息通信分公司 A kind of complete conspicuousness object detecting method using contour detecting
CN108734139A (en) * 2018-05-24 2018-11-02 辽宁工程技术大学 Feature based merges and the newer correlation filtering tracking of SVD adaptive models
CN109345586A (en) * 2018-11-02 2019-02-15 国网湖南省电力有限公司 Electrical equipment discharge characteristic extracting method based on ultraviolet imagery technology
CN109341524A (en) * 2018-10-17 2019-02-15 上海理工大学 A kind of optical fiber geometric parameter detection method based on machine vision
CN109543487A (en) * 2018-11-22 2019-03-29 福州图腾易讯信息技术有限公司 A kind of automatic sensing triggering method and its system based on bar edges segmentation
CN109559306A (en) * 2018-11-27 2019-04-02 广州供电局有限公司 Crosslinked polyetylene insulated layer surface planarization detection method based on edge detection
CN109993750A (en) * 2017-12-29 2019-07-09 中国科学院深圳先进技术研究院 A kind of segmentation recognition method and system, terminal and readable storage medium storing program for executing of hand jnjuries
CN110334692A (en) * 2019-07-17 2019-10-15 河南科技大学 A kind of blind way recognition methods based on image procossing
CN110415222A (en) * 2019-07-16 2019-11-05 东华大学 A kind of spinning cake side face defects recognition methods based on textural characteristics
CN110443821A (en) * 2019-07-09 2019-11-12 成都理工大学 Water body detection method and water body detection device based on image
CN110516610A (en) * 2019-08-28 2019-11-29 上海眼控科技股份有限公司 A kind of method and apparatus for road feature extraction
CN111191534A (en) * 2019-12-18 2020-05-22 长安大学 Road extraction method in fuzzy aerial image
CN111429372A (en) * 2020-03-24 2020-07-17 杭州趣维科技有限公司 Method for enhancing edge detection effect of low-contrast image
CN111539429A (en) * 2020-06-19 2020-08-14 天津施格机器人科技有限公司 Automatic circulation box positioning method based on image geometric features
CN111553868A (en) * 2020-04-29 2020-08-18 南京航空航天大学 Tunnel reflectivity image filtering method based on improved sobel operator
CN111739044A (en) * 2020-06-12 2020-10-02 鞍钢集团矿业有限公司 Mine road network extraction method based on remote sensing image
CN111815580A (en) * 2020-06-25 2020-10-23 北京航天新立科技有限公司 Image edge identification method and small module gear module detection method
CN112132848A (en) * 2020-09-01 2020-12-25 成都运达科技股份有限公司 Preprocessing method based on image layer segmentation and extraction
CN112365539A (en) * 2020-11-11 2021-02-12 中南大学 Ingot casting mold metal solution boundary positioning method integrating template matching and symmetrical straight line segmentation
CN112396581A (en) * 2020-11-13 2021-02-23 国网天津市电力公司营销服务中心 Method for detecting and repairing scratches on upper cover of disassembled and reused electrical information acquisition terminal
CN112508913A (en) * 2020-12-10 2021-03-16 国网江西省电力有限公司电力科学研究院 Cable section edge detection method based on image detection
CN112634269A (en) * 2021-01-14 2021-04-09 华东交通大学 Rail vehicle body detection method
CN112785571A (en) * 2021-01-20 2021-05-11 浙江理工大学 Famous tea tender leaf recognition and segmentation method based on improved watershed
CN112820109A (en) * 2021-01-15 2021-05-18 林安齐 Road inspection safety early warning method and system
CN113139975A (en) * 2021-04-19 2021-07-20 国交空间信息技术（北京）有限公司 Road characteristic-based road surface segmentation method and device
CN113554646A (en) * 2021-09-17 2021-10-26 江苏正金建筑装饰工程有限公司 Intelligent urban road pavement detection method and system based on computer vision
CN113837198A (en) * 2021-05-18 2021-12-24 中国计量大学 Improved self-adaptive threshold Canny edge detection method based on three-dimensional block matching
CN114494318A (en) * 2022-02-17 2022-05-13 天津市眼科医院 Method for extracting cornea contour from cornea dynamic deformation video based on Otsu algorithm
CN115035105A (en) * 2022-08-10 2022-09-09 山东西曼克技术有限公司 Multidimensional data fusion and decision-making method for AGV trolley steering control
CN115147414A (en) * 2022-09-01 2022-10-04 南通三信塑胶装备科技股份有限公司 Method for detecting surface breakdown defect of bipolar power transistor
CN115330798A (en) * 2022-10-13 2022-11-11 博格达智能装备(南通)有限公司 Defect assessment method and system for door plate wood
CN116630321A (en) * 2023-07-24 2023-08-22 铁正检测科技有限公司 Intelligent bridge health monitoring system based on artificial intelligence
CN116778263A (en) * 2023-08-22 2023-09-19 四川坤鸿电子科技有限公司 Sorting apparatus control method, electronic apparatus, and computer-readable medium
CN118097581A (en) * 2024-04-28 2024-05-28 山东领军智能交通科技有限公司 Road edge recognition control method and device

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1885317A (en) * 2006-07-06 2006-12-27 上海交通大学 Adaptive edge detection method based on morphology and information entropy
CN104636722A (en) * 2015-01-26 2015-05-20 江苏大学 Fast tracking recognition method for overlapped fruits by picking robot
US9305360B2 (en) * 2013-02-14 2016-04-05 Avago Technologies General Ip (Singapore) Pte. Ltd. Method and apparatus for image enhancement and edge verification using at least one additional image
CN105550665A (en) * 2016-01-15 2016-05-04 北京理工大学 Method for detecting pilotless automobile through area based on binocular vision

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1885317A (en) * 2006-07-06 2006-12-27 上海交通大学 Adaptive edge detection method based on morphology and information entropy
US9305360B2 (en) * 2013-02-14 2016-04-05 Avago Technologies General Ip (Singapore) Pte. Ltd. Method and apparatus for image enhancement and edge verification using at least one additional image
CN104636722A (en) * 2015-01-26 2015-05-20 江苏大学 Fast tracking recognition method for overlapped fruits by picking robot
CN105550665A (en) * 2016-01-15 2016-05-04 北京理工大学 Method for detecting pilotless automobile through area based on binocular vision

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
戴天虹 等: "基于OTSU算法与数学形态学的木材缺陷图像分割", 《森林工程》 *
李婷婷: "用于视觉导航的道路检测与跟踪技术的研究", 《万方数据知识平台》 *
魏武 等: "基于机器视觉的非结构化道路检测算法研究", 《传感器与微系统》 *

Cited By (64)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108242060A (en) * 2016-12-23 2018-07-03 重庆邮电大学 A kind of method for detecting image edge based on Sobel operators
CN106682787A (en) * 2017-01-09 2017-05-17 北京航空航天大学 Method for quickly generating generalized Voronoi diagram based on wavefront algorithm
CN108470340A (en) * 2017-02-23 2018-08-31 南宁市富久信息技术有限公司 A kind of improved Sobel edge detection algorithms
CN106981055A (en) * 2017-03-30 2017-07-25 西安交通大学 A kind of ICCD image de-noising methods
CN107369159A (en) * 2017-06-29 2017-11-21 大连理工大学 Threshold segmentation method based on multifactor two-dimensional gray histogram
CN107369159B (en) * 2017-06-29 2020-04-24 大连理工大学 Threshold segmentation method based on multi-factor two-dimensional gray level histogram
CN107274403A (en) * 2017-06-30 2017-10-20 长安大学 A kind of evaluation method of flotation surface quality
CN107274403B (en) * 2017-06-30 2020-12-18 长安大学 Evaluation method for flotation surface quality
CN107590501A (en) * 2017-07-30 2018-01-16 华南理工大学 A kind of improved Canny adaptive edge extracting methods based on Otsu
CN108109155A (en) * 2017-11-28 2018-06-01 东北林业大学 A kind of automatic threshold edge detection method based on improvement Canny
CN109993750A (en) * 2017-12-29 2019-07-09 中国科学院深圳先进技术研究院 A kind of segmentation recognition method and system, terminal and readable storage medium storing program for executing of hand jnjuries
CN109993750B (en) * 2017-12-29 2020-12-25 中国科学院深圳先进技术研究院 Segmentation identification method and system for wrist bones, terminal and readable storage medium
CN108573492A (en) * 2018-04-02 2018-09-25 电子科技大学 A kind of real time radar search coverage detection method
CN108734139A (en) * 2018-05-24 2018-11-02 辽宁工程技术大学 Feature based merges and the newer correlation filtering tracking of SVD adaptive models
CN108734139B (en) * 2018-05-24 2021-12-14 辽宁工程技术大学 Correlation filtering tracking method based on feature fusion and SVD self-adaptive model updating
CN108710883B (en) * 2018-06-04 2021-08-24 国网辽宁省电力有限公司信息通信分公司 Complete salient object detection method adopting contour detection
CN108710883A (en) * 2018-06-04 2018-10-26 国网辽宁省电力有限公司信息通信分公司 A kind of complete conspicuousness object detecting method using contour detecting
CN109341524A (en) * 2018-10-17 2019-02-15 上海理工大学 A kind of optical fiber geometric parameter detection method based on machine vision
CN109345586A (en) * 2018-11-02 2019-02-15 国网湖南省电力有限公司 Electrical equipment discharge characteristic extracting method based on ultraviolet imagery technology
CN109543487A (en) * 2018-11-22 2019-03-29 福州图腾易讯信息技术有限公司 A kind of automatic sensing triggering method and its system based on bar edges segmentation
CN109559306B (en) * 2018-11-27 2021-03-12 广东电网有限责任公司广州供电局 Crosslinked polyethylene insulating layer surface smoothness detection method based on edge detection
CN109559306A (en) * 2018-11-27 2019-04-02 广州供电局有限公司 Crosslinked polyetylene insulated layer surface planarization detection method based on edge detection
CN110443821B (en) * 2019-07-09 2023-05-05 成都理工大学 Image-based water body detection method and device
CN110443821A (en) * 2019-07-09 2019-11-12 成都理工大学 Water body detection method and water body detection device based on image
CN110415222A (en) * 2019-07-16 2019-11-05 东华大学 A kind of spinning cake side face defects recognition methods based on textural characteristics
CN110334692A (en) * 2019-07-17 2019-10-15 河南科技大学 A kind of blind way recognition methods based on image procossing
CN110334692B (en) * 2019-07-17 2021-07-23 河南科技大学 Blind road identification method based on image processing
CN110516610A (en) * 2019-08-28 2019-11-29 上海眼控科技股份有限公司 A kind of method and apparatus for road feature extraction
CN111191534B (en) * 2019-12-18 2023-05-23 长安大学 Road extraction method in fuzzy aviation image
CN111191534A (en) * 2019-12-18 2020-05-22 长安大学 Road extraction method in fuzzy aerial image
CN111429372A (en) * 2020-03-24 2020-07-17 杭州趣维科技有限公司 Method for enhancing edge detection effect of low-contrast image
CN111553868A (en) * 2020-04-29 2020-08-18 南京航空航天大学 Tunnel reflectivity image filtering method based on improved sobel operator
CN111553868B (en) * 2020-04-29 2022-03-01 南京航空航天大学 Tunnel reflectivity image filtering method based on improved sobel operator
CN111739044A (en) * 2020-06-12 2020-10-02 鞍钢集团矿业有限公司 Mine road network extraction method based on remote sensing image
CN111539429A (en) * 2020-06-19 2020-08-14 天津施格机器人科技有限公司 Automatic circulation box positioning method based on image geometric features
CN111815580B (en) * 2020-06-25 2024-04-30 北京航天新立科技有限公司 Image edge recognition method and small module gear module detection method
CN111815580A (en) * 2020-06-25 2020-10-23 北京航天新立科技有限公司 Image edge identification method and small module gear module detection method
CN112132848A (en) * 2020-09-01 2020-12-25 成都运达科技股份有限公司 Preprocessing method based on image layer segmentation and extraction
CN112365539B (en) * 2020-11-11 2022-12-27 中南大学 Method for positioning boundary of metal solution of ingot casting mold
CN112365539A (en) * 2020-11-11 2021-02-12 中南大学 Ingot casting mold metal solution boundary positioning method integrating template matching and symmetrical straight line segmentation
CN112396581A (en) * 2020-11-13 2021-02-23 国网天津市电力公司营销服务中心 Method for detecting and repairing scratches on upper cover of disassembled and reused electrical information acquisition terminal
CN112508913B (en) * 2020-12-10 2024-07-05 国网江西省电力有限公司电力科学研究院 Cable section edge detection method based on image detection
CN112508913A (en) * 2020-12-10 2021-03-16 国网江西省电力有限公司电力科学研究院 Cable section edge detection method based on image detection
CN112634269B (en) * 2021-01-14 2023-12-26 华东交通大学 Railway vehicle body detection method
CN112634269A (en) * 2021-01-14 2021-04-09 华东交通大学 Rail vehicle body detection method
CN112820109A (en) * 2021-01-15 2021-05-18 林安齐 Road inspection safety early warning method and system
CN112785571B (en) * 2021-01-20 2024-04-12 浙江理工大学 Famous tea tender leaf identification and segmentation method based on improved watershed
CN112785571A (en) * 2021-01-20 2021-05-11 浙江理工大学 Famous tea tender leaf recognition and segmentation method based on improved watershed
CN113139975A (en) * 2021-04-19 2021-07-20 国交空间信息技术（北京）有限公司 Road characteristic-based road surface segmentation method and device
CN113139975B (en) * 2021-04-19 2023-11-17 国交空间信息技术（北京）有限公司 Road feature-based pavement segmentation method and device
CN113837198A (en) * 2021-05-18 2021-12-24 中国计量大学 Improved self-adaptive threshold Canny edge detection method based on three-dimensional block matching
CN113837198B (en) * 2021-05-18 2023-11-28 中国计量大学 Improved self-adaptive threshold Canny edge detection method based on three-dimensional block matching
CN113554646A (en) * 2021-09-17 2021-10-26 江苏正金建筑装饰工程有限公司 Intelligent urban road pavement detection method and system based on computer vision
CN114494318A (en) * 2022-02-17 2022-05-13 天津市眼科医院 Method for extracting cornea contour from cornea dynamic deformation video based on Otsu algorithm
CN115035105B (en) * 2022-08-10 2022-10-25 山东西曼克技术有限公司 Multidimensional data fusion and decision-making method for AGV trolley steering control
CN115035105A (en) * 2022-08-10 2022-09-09 山东西曼克技术有限公司 Multidimensional data fusion and decision-making method for AGV trolley steering control
CN115147414A (en) * 2022-09-01 2022-10-04 南通三信塑胶装备科技股份有限公司 Method for detecting surface breakdown defect of bipolar power transistor
CN115147414B (en) * 2022-09-01 2023-08-29 南通三信塑胶装备科技股份有限公司 Surface breakdown defect detection method for bipolar power transistor
CN115330798A (en) * 2022-10-13 2022-11-11 博格达智能装备(南通)有限公司 Defect assessment method and system for door plate wood
CN116630321B (en) * 2023-07-24 2023-10-03 铁正检测科技有限公司 Intelligent bridge health monitoring system based on artificial intelligence
CN116630321A (en) * 2023-07-24 2023-08-22 铁正检测科技有限公司 Intelligent bridge health monitoring system based on artificial intelligence
CN116778263B (en) * 2023-08-22 2023-11-14 四川坤鸿电子科技有限公司 Sorting apparatus control method, electronic apparatus, and computer-readable medium
CN116778263A (en) * 2023-08-22 2023-09-19 四川坤鸿电子科技有限公司 Sorting apparatus control method, electronic apparatus, and computer-readable medium
CN118097581A (en) * 2024-04-28 2024-05-28 山东领军智能交通科技有限公司 Road edge recognition control method and device

Similar Documents

Publication Publication Date Title
CN106228138A (en) A kind of Road Detection algorithm of integration region and marginal information
CN105261017B (en) The method that image segmentation based on road surface constraint extracts pedestrian&#39;s area-of-interest
CN105550665B (en) A kind of pilotless automobile based on binocular vision can lead to method for detecting area
CN103364410B (en) Crack detection method of hydraulic concrete structure underwater surface based on template search
CN111985329A (en) Remote sensing image information extraction method based on FCN-8s and improved Canny edge detection
CN106650553A (en) License plate recognition method and system
CN106022243B (en) A kind of retrograde recognition methods of the car lane vehicle based on image procossing
CN104021574A (en) Method for automatically identifying pavement diseases
CN104966066A (en) Traffic block port monitoring oriented in-car human face detection method and system
CN107895151A (en) Method for detecting lane lines based on machine vision under a kind of high light conditions
CN106096491B (en) Automatic identification method for microaneurysms in fundus color photographic image
CN104809433A (en) Zebra stripe detection method based on maximum stable region and random sampling
CN103413303A (en) Infrared target segmentation method based on joint obviousness
Liu et al. Multi-type road marking recognition using adaboost detection and extreme learning machine classification
CN107704833A (en) A kind of front vehicles detection and tracking based on machine learning
CN111582004A (en) Target area segmentation method and device in ground image
CN108734172A (en) Target identification method, system based on linear edge feature
CN105184802A (en) Image processing method and device
CN109492647A (en) A kind of power grid robot barrier object recognition methods
CN105447489A (en) Character and background adhesion noise elimination method for image OCR system
CN111091071B (en) Underground target detection method and system based on ground penetrating radar hyperbolic wave fitting
CN101369312A (en) Method and equipment for detecting intersection in image
CN107239761B (en) Fruit tree branch pulling effect evaluation method based on skeleton angular point detection
KR102010086B1 (en) Method and apparatus for phase segmentation of microstructure
Xiaolin et al. Unstructured road detection based on region growing

Legal Events

Date Code Title Description
C06 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: 20161214

RJ01 Rejection of invention patent application after publication