CN108596165B - Road traffic marking detection method and system based on unmanned plane low latitude Aerial Images - Google Patents
Road traffic marking detection method and system based on unmanned plane low latitude Aerial Images Download PDFInfo
- Publication number
- CN108596165B CN108596165B CN201810951206.0A CN201810951206A CN108596165B CN 108596165 B CN108596165 B CN 108596165B CN 201810951206 A CN201810951206 A CN 201810951206A CN 108596165 B CN108596165 B CN 108596165B
- Authority
- CN
- China
- Prior art keywords
- road traffic
- traffic marking
- image
- connected domain
- unmanned plane
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
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 present invention provides a kind of road traffic marking detection method and system based on unmanned plane low latitude Aerial Images, the carriageway image taken photo by plane including obtaining unmanned plane low latitude；The carriageway image of acquisition is pre-processed；Calculate after pretreatment the stroke width of pixel in each connected domain in carriageway image；Connected domain screening is carried out according to the stroke width of pixel with the sorting for carrying out road traffic marking and road；The position of road traffic marking is accurately determined with the B-spline Curve fitting algorithm based on random sampling consistency in the road traffic marking connected domain filtered out.The influence that exclusion unmanned plane low latitude shooting angle, height change, intensity of illumination variation, water mark, lane line breakage and greenbelt and shade detect road traffic marking, detect lane line, arrow, pavement, stop line etc. simultaneously, the unmanned aerial vehicle onboard for realizing road traffic marking detection is handled in real time.The present invention is applied to traffic information technology and computer vision technique traffic application field.
Description
Technical field
The present invention relates to traffic information technologies and computer vision technique traffic application field, more particularly to one kind to be based on nothing
The road traffic marking detection method and system of man-machine low latitude Aerial Images.
Background technique
Clearly road traffic marking can help driver to carry out driving judgement, safe driving, when road administration is patrolled, need
Damaged road traffic marking is detected in time, and is repaired.It, can also be by detecting road traffic in traffic administration inspection
Graticule has monitored whether rule-breaking vehicle traveling.In the intelligent driving of rapid development, the technologies such as unmanned, also it is widely used
The method of lane line is detected to assist vehicle to carry out lane holding.In addition when making high-precision map, it is also desirable to accurate to extract
The position of road traffic marking and attribute out.
The carriageway image that unmanned plane low latitude is taken photo by plane, shooting angle are different（Have just take the photograph, the oblique photograph of multiple and different angles,
It is proactive）, graticule type multiplicity（Such as straight line, shaped form lane line, pavement, stop line, arrow）, outside lane with road traffic mark
The similar chaff interferent of wire shaped is more, width, the length of road traffic marking also due to shooting angle, far and near difference and send out
Changing, the shade of greenbelt, building etc. can obstruct the road traffic marking, and in some lanes certain positions of driving vehicle
The chaff interferent of road traffic marking detection can be become.Therefore, detect that road is handed over from the carriageway image that unmanned plane low latitude is taken photo by plane
Logical graticule, is related to two Important Problems, first is that the inside and outside interference with road traffic marking similar object in lane how is excluded, second is that
The accuracy of road traffic marking detection how is improved in lane.
Existing road traffic marking detection technique is generally be directed to driver visual angle, the visual angle of such method single-frame images
It is limited in scope, the road traffic marking type in single-frame images is less, is not suitable for the processing of Aerial Images.And it is directed to unmanned plane
The road traffic marking of Aerial Images detects, then based on positive photograph picture, the detection of Lane marking line.Using artificial method,
Although the road traffic marking in the Aerial Images of unmanned plane low latitude, time-consuming and laborious, inefficiency can be detected accurately.
Existing road traffic marking detection algorithm can substantially be divided into the detection method based on feature and be based on model
Two class of detection method matched.
Detection method based on feature, general color, texture, geometry, the change of gradient for utilizing road traffic marking
Etc. features road traffic marking is split from image, such as " a kind of vehicle-mounted lane detection system and method, application number/
The patent No.：201710086674.1 invention designer：It is super that the yellow just Cai Hao poplar Zhe Mumeng of political affairs is encouraged recklessly " face based on lane line
Color characteristic detects lane line；" Automatic lane line identification method based on low latitude Aerial Images, the application number/patent No.：
201310409740.6 invention designer：Your positive Zhang Yipeng " of treasure's Zheng Xin stamen Xu Yong more than Wang Yunpeng, " it is based on unmanned plane
The lane line of image automatically extracts and recognition methods, the application number/patent No.：201710450069.8 invention designer：Ge Song ",
" a kind of highway graticule detection method based on line-structured light three-dimensional measurement, the application number/patent No.：201610827906.X invention
Designer：Li Qingquan Zhang Dejin Cao Min woods is red ", a kind of " highway graticule damage testing based on unmanned plane highway image
Method, the application number/patent No.：201710908434.5 invention designer：The week super Wang Lichun Wang Yong " of quick Liu Ning clock Zhu Zhi
Lane line is extracted by the geometric characteristic of lane line, " video highway lane detection technical research of taking photo by plane, the southeast are big
It learns, master thesis, author：Tang Tao " then extracts lane line in conjunction with the color characteristic of lane line and geometric characteristic.Wherein
Method for detecting lane lines based on color characteristic and change of gradient is easy to be illuminated by the light Strength Changes, road surface shade, water mark, lane
The influence of the factors such as line breakage.Detection method based on geometry, has higher requirements to shooting angle, such as " is navigated based on low latitude
Clap the Automatic lane line identification method of image, the application number/patent No.：201310409740.6 invention designer：More than Wang Yunpeng
Your positive Zhang Yipeng " of treasure Zheng Xin stamen Xu Yong require shooting angle be horizontal direction and road area be located at image middle position,
" a kind of roadmarking detection method, the application number/patent No.：201310193755.3 invention designer：The refined bright Sun Tao " of soup is needed
Imaging sensor is demarcated, while such method is easy by driving vehicle, abrasion vehicle in roadside thread-shaped body, road
The influence of diatom, and Lane marking line and shaped form lane need to detect respectively.
Detection method based on Model Matching describes lane using specific parameter of curve using road priori knowledge
The process simplification of lane detection is the process of computation model parameter by line, to extract lane line, common model has directly
Line, parabola, hyperbola, spline curve.Such method is not suitable for the graticule detection of unstructured road, and such method
Calculation amount it is larger, real-time is not good enough.
Summary of the invention
In view of the deficienciess of the prior art, the object of the present invention is to provide a kind of based on unmanned plane low latitude Aerial Images
Road traffic marking automatic testing method effectively excludes the variation of unmanned plane low latitude shooting angle, shooting height variation to road
The influence of traffic marking detection is reducing intensity of illumination variation, water mark, lane line breakage and greenbelt and building effects to road
Traffic marking detection in road simplifies calculating process while influence, calculation amount is reduced, to effectively realize road traffic
The airborne real-time processing of graticule detection, and at the same time detecting lane line（Including Lane marking line, shaped form lane line）With
And other road traffic markings such as arrow, pavement, stop line, improve the position precision of lane detection.
The technical solution adopted by the present invention is that：
Based on the road traffic marking detection method of unmanned plane low latitude Aerial Images, following steps are specifically included：
S1, the carriageway image that unmanned plane low latitude is taken photo by plane is obtained；
S2, the carriageway image of acquisition is pre-processed with the bianry image for obtaining carriageway image；
S3, the stroke width for calculating after pretreatment pixel in each connected domain in carriageway image；
S4, according to the stroke width of pixel carry out connected domain screening with for carry out the lane line in road traffic marking with
The first sorting of road；
S5, Fuzzy Linear discriminatory analysis is carried out as sample using the lane line and road that just sort out, calculates road traffic mark
Line grad enhancement parameter, specifically includes following steps：
Lane figure in position coordinates the collection C1, read step S1 of filtered out lane line connected domain in S51, acquisition step S4
As the rgb value at coordinate set C1, as the first kind sample of Fuzzy Linear discriminatory analysis, by lane filtered out in step S4
Line connected domain is expanded, and the connected domain before the connected domain and expansion after expansion is subtracted each other, and acquisition is subtracted each other rear connected domain position and sat
Rgb value of the carriageway image at coordinate set C2 in mark collection C2, read step S1, the second class sample as Fuzzy Linear discriminatory analysis
This, the intersection of two class samples is population sample；
S52, the mean value for calculating first kind sample, the second class sample mean valueWith the mean value of population sample, class
Interior distance, between class distance：
In formula,Indicate theThe number of class sample,Indicate sample,Indicate theThe set of class sample,It indicates
The transposition of matrix；
S53, find out rgb space to gray space optimal transformation ratio：
The calculation criterion of optimal projection coefficient is --- the within-cluster variance of similar sampleIt is as small as possible, it is discrete between class
DegreeIt is as big as possible, it introduces Fisher and identifies criterion expression formula,
In formula,wIndicate rgb space to gray space converting vector,It is corresponding when maximumwAs RGB
Space is to the optimum translation coefficient of gray space, i.e. road traffic marking grad enhancement parameter；
S6, step S3 is repeated after repeating step S2 using road traffic marking grad enhancement parameter, and according to the pen of pixel
Width progress connected domain screening is drawn to pick with the subdivision for carrying out all road traffic markings and road；
S7, in the road traffic marking connected domain that step S6 is filtered out with the B sample three times based on random sampling consistency
Curve fitting algorithm accurately determines the position of road traffic marking.
As a further improvement of the above technical scheme, in step S4, the lane line and road in road traffic marking are carried out
The first sorting on road specifically includes following steps：
S41, the stroke width according to lane line, and screening range is determined according to height of taking photo by plane, it filters out stroke width and exists
Screen the connected domain in range；
S42, microdilatancy, filling processing are carried out to the result of step S41, the lane line being broken after step S41 screening is connected
It picks up and；
S43, step S42 treated connected domain is carried out according to the geometrical characteristic parameter of connected domain area, long axis length
Further screening.
As a further improvement of the above technical scheme, in step S6, the thin of all road traffic markings and road is carried out
Sorting specifically includes following steps：
S61, according to the stroke width of all road traffic markings, filter out the connection of stroke width in a certain range
Domain, wherein road traffic marking includes lane line, arrow, pavement and/or stop line；
S62, microdilatancy, filling processing are carried out to the result of step S61, the lane line being broken after step S61 screening is connected
It picks up and；
S63, according to the parameter of connected domain area, long axis length, to step S62, treated that connected domain is further sieved
Choosing.
As a further improvement of the above technical scheme, in step S3, each connected domain in carriageway image after calculating pretreatment
The stroke width of middle pixel specifically includes following steps：
S31, the bianry image image of a size obtained after pre-processing is established in a width and step S2, it is wide becomes stroke
Image is spent, initializing each pixel value in stroke width image is infinity；
S32, edge extracting is carried out to the bianry image obtained after pre-processing in step S2, from any edge point, edge
The gradient direction of the point searches for other marginal points, if find other marginal points, and its gradient direction just with searcher
To consistent, then all pixels point connected on this two o'clock line segment in stroke width image is assigned a value of to the Euclidean distance of the two o'clock；
S33, when the arbitrary point of stroke width image is by multiple assignment, each assignment always with last institute's assignment carry out
Compare, if smaller than last institute's assignment, the pixel value of the point is replaced with into this assignment, i.e., each pixel always records institute
The most thin stroke width belonged to.
As a further improvement of the above technical scheme, in step S7, in the road traffic marking connected domain filtered out
Accurately determine that the position of road traffic marking is specifically included with the B-spline Curve fitting algorithm based on random sampling consistency
Following steps：
S71, setting fitting number；
S72, in analyzed connected domain, randomly select 4 points, seek the parameter of B-spline Curve, and according to parameter
Determine curve；
S73, calculate in analyzed connected domain in step S72 selected by other than 4 points other to curve distance, if away from
From being less than threshold value, then it is assumed that the point is intra-office point, and the number put statistics bureau in, if the number of intra-office point greater than preset value,
Think that optimal curve has been found, stops fitting, corresponding intra-office point is road traffic marking pixel；If of intra-office point
Number is less than preset value, then return step S72, finds optimal curve again, if finding number reaches fitting number, optimal curve is also
It does not find, then in found curve, the most curve of intra-office point is optimal curve, corresponding intra-office point is road traffic marking
Pixel.
As a further improvement of the above technical scheme, in step S1, when unmanned plane shoots carriageway image, primary shooting is made
In the industry the variation of shooting angle within 20 degree, the variation of shooting height within 15 meters, shooting height within 60 meters, shooting
The road traffic marking be subject in macroscopic image of weather condition.
As a further improvement of the above technical scheme, in step S2, carrying out pretreatment to the carriageway image of acquisition includes
In image grayscale conversion, the variation of image top cap, contrast stretching, carrying out image threshold segmentation, image expansion or Image erosion at least
It is a kind of.
The present invention also provides a kind of road traffic marking detection systems based on unmanned plane low latitude Aerial Images, specifically include
Memory and processor, the memory are stored with computer program, and the processor is realized when executing the computer program
The step of above method.
Advantageous effects of the invention：
When the present invention carries out road traffic marking detection, road traffic marking ladder is sought by Fuzzy Linear discriminatory analysis
Degree enhancing parameter, and then to road traffic marking carry out grad enhancement, reduce intensity of illumination variation, water mark, lane line breakage and
Greenbelt and building effects, which detect road traffic marking, to be influenced, and improves the accuracy of detection, while reflecting in Fuzzy Linear
Not Fen Xi during, simplify the finding process of first kind sample and the second class sample, reduce calculation amount, thus effectively
Realize the airborne real-time processing of road traffic marking detection；Using the method in the present invention, linear type can not only be detected
Lane line and shaped form lane line, can also detect the various roads traffic marking such as arrow, stop line, pavement simultaneously, have
Effect improves the efficiency of single detection, while not requiring the shooting angle of unmanned plane, improves when unmanned plane is shot
Safety expands road shooting area.
Detailed description of the invention
Fig. 1 is flow diagram of the invention；
Fig. 2 is stroke width algorithm schematic diagram.
Specific embodiment
For the ease of implementation of the invention, it is further described below with reference to specific example.
Road traffic marking detection method based on unmanned plane low latitude Aerial Images as shown in Figure 1, specific steps include：
S1, the carriageway image that unmanned plane low latitude is taken photo by plane is obtained
It takes photo by plane carriageway image to acquired unmanned plane low latitude, no shooting angle requirement, but shooting in a shooting operation
The variation of angle, height requires in a certain range, and variation of the variation no more than 20 degree, shooting height of shooting angle is no more than
15 meters, shooting height be no more than 60 meters.The weather condition of shooting, the road traffic marking being subject in human eye energy recognisable image.
The carriageway image obtained in this step is known as original image in the present embodiment.
S2, the original image of acquisition is pre-processed with the bianry image for obtaining carriageway image
In order to improve the effect of carrying out image threshold segmentation, needs before Threshold segmentation, carry out the variation of image top cap and contrast
It stretches.Road traffic marking, color are white and yellow, and white and yellow object in picture are extracted by Threshold segmentation.Threshold
Road traffic marking after value segmentation, especially lane line, the case where being easy to appear fracture, by image expansion by breaking part into
Row connection.In addition to the later period is easier to reject interference, the image after expansion is corroded, so as to by part chaff interferent
Connected domain is isolated.
Pretreated carriageway image is bianry image, and pixel value only has 1 and 0, only to picture when calculating stroke width images
The region that plain value is 1 is calculated, and is generally less than 20 the percent of whole image in bianry image for 1 region, step
S1 unmanned plane requires shooting height when shooting, and belongs to low latitude shooting, the ratio that road area accounts for whole picture is more than hundred
/ eight ten, it is 0 after road area binaryzation, does not need to carry out stroke width calculating, therefore pre-process to by original image as two-value
Image can effectively reduce subsequent calculation amount, so that the present embodiment can be realized nobody of road traffic marking detection
The airborne real-time processing of machine.
S3, the stroke width for calculating after pretreatment pixel in each connected domain in image
In a unmanned plane operation, the angle and height of shooting change in a certain range, then captured image
In, either straight way or bend, the width of road traffic marking can also change in a certain range, have preferable width one
Cause property, is based on this feature, and the present embodiment carries out the detection of road traffic marking, tool by calculating the stroke width of pixel in picture
Body implementation method is as follows：
S31, the bianry image image of a size obtained after pre-processing is established in a width and step S2, it is wide becomes stroke
Image is spent, initializing each pixel value in stroke width image is infinity；
S32, edge extracting is carried out to the bianry image obtained after pre-processing in step S2, from any edge point, edge
The gradient direction of the point searches for other marginal points, if find other marginal points, and its gradient direction just with searcher
To consistent, then all pixels point connected on this two o'clock line segment in stroke width image is assigned a value of to the Euclidean distance of the two o'clock；
S33, when the arbitrary point of stroke width image is by multiple assignment, each assignment always with last institute's assignment carry out
Compare, if smaller than last institute's assignment, the pixel value of the point is replaced with into this assignment, i.e., each pixel always records institute
The most thin stroke width belonged to.
In conjunction with Fig. 2, above-mentioned steps S32 is further described, from edge a bitIt sets out, edgeThe gradient side of point
To other marginal points are searched for, other marginal points are found,Point gradient direction just withPoint search direction is consistent,
Then in width imagesWithAll the points between line are assignedExtremelyEuclidean distance,WithRelationship it is sameWith。
S4, according to the stroke width of pixel carry out connected domain screening with for carry out the lane line in road traffic marking with
The first sorting of road, includes the following steps：
S41, the stroke width according to lane line, and screening range is determined according to height of taking photo by plane, it filters out stroke width and exists
Screen the connected domain in range；
S42, microdilatancy, filling processing are carried out to the result of step S41, the lane line being broken after step S41 screening is connected
It picks up and；
S43, step S42 treated connected domain is carried out according to the geometrical characteristic parameter of connected domain area, long axis length
Further screening.
S5, Fuzzy Linear discriminatory analysis is carried out as sample using the lane line and road that just sort out, calculates road traffic mark
Line grad enhancement parameter, includes the following steps：
Lane figure in position coordinates the collection C1, read step S1 of filtered out lane line connected domain in S51, acquisition step S4
As the rgb value at coordinate set C1, as the first kind sample of Fuzzy Linear discriminatory analysis, by lane filtered out in step S4
Line connected domain is expanded, and the connected domain before the connected domain and expansion after expansion is subtracted each other, and acquisition is subtracted each other rear connected domain position and sat
Rgb value of the carriageway image at coordinate set C2 in mark collection C2, read step S1, the second class sample as Fuzzy Linear discriminatory analysis
This, the intersection of two class samples is population sample, simplifies the finding process of first kind sample and the second class sample, reduces calculating
Amount, so that the unmanned aerial vehicle onboard for effectively realizing road traffic marking detection is handled in real time；
S52, the mean value for calculating first kind sample, the second class sample mean valueWith the mean value of population sample, class
Interior distance, between class distance：
In formula,Indicate theThe number of class sample,Indicate sample,Indicate theThe set of class sample,It indicates
The transposition of matrix；
S53, find out rgb space to gray space optimal transformation ratio：
The calculation criterion of optimal projection coefficient is --- the within-cluster variance of similar sampleIt is as small as possible, it is discrete between class
DegreeIt is as big as possible, it introduces Fisher and identifies criterion expression formula,
In formula,wIndicate rgb space to gray space converting vector,It is corresponding when maximumwAs RGB
Space is to the optimum translation coefficient of gray space, i.e. road traffic marking grad enhancement parameter.
S6, step S3 is repeated after repeating step S2 using road traffic marking grad enhancement parameter, and according to the pen of pixel
It draws width progress connected domain screening to pick with the subdivision for carrying out all road traffic markings and road, include the following steps：
S61, according to the stroke width of all road traffic markings, filter out the connection of stroke width in a certain range
Domain, wherein road traffic marking includes lane line, arrow, pavement and/or stop line；
S62, microdilatancy, filling processing are carried out to the result of step S61, the lane line being broken after step S61 screening is connected
It picks up and；
S63, according to the parameter of connected domain area, long axis length, to step S62, treated that connected domain is further sieved
Choosing.
S7, in the road traffic marking connected domain filtered out with the B-spline Curve based on random sampling consistency
Fitting algorithm accurately determines the position of road traffic marking, specifically includes following steps：
S71, setting fitting number；
S72, in analyzed connected domain, randomly select 4 points, seek the parameter of B-spline Curve, and according to parameter
Determine curve；
S73, calculate in analyzed connected domain in step S72 selected by other than 4 points other to curve distance, if away from
From being less than threshold value, then it is assumed that the point is intra-office point, and the number put statistics bureau in, if the number of intra-office point greater than preset value,
Think that optimal curve has been found, stops fitting, corresponding intra-office point is road traffic marking pixel；If of intra-office point
Number is less than preset value, then return step S72, finds optimal curve again, if finding number reaches fitting number, optimal curve is also
It does not find, then in found curve, the most curve of intra-office point is optimal curve, corresponding intra-office point is road traffic marking
Pixel.
The present embodiment also provides a kind of road traffic marking detection system based on unmanned plane low latitude Aerial Images,
Specifically include memory and processor, the memory is stored with computer program, described in the processor executes
The step of above method is realized when computer program.
Generally speaking, it takes photo by plane the detection of road traffic marking in carriageway image for unmanned plane low latitude, the present embodiment is abundant
Using road traffic marking have the characteristics that metastable stroke width this（On condition that the angle and height of unmanned plane shooting exist
Variation in prescribed limit）, the stroke width value that each pixel is most possible in bianry image is calculated, according to the stroke of lane line
Width preliminary screening goes out lane line, so that it is determined that the region in lane, in the road traffic marking domain having determined, by fuzzy
Linear discriminant analysis enhances the gradient of road traffic marking, carries out stroke width calculating again to the picture for completing grad enhancement,
Filter out the road traffic markings such as lane line and arrow, pavement, stop line, finally in the lane line connected domain filtered out into
Row curve matching, the accurate position for determining lane line.And during Fuzzy Linear discriminatory analysis, first kind sample is simplified
This finding process with the second class sample reduces calculation amount, to effectively realize nobody of road traffic marking detection
The airborne real-time processing of machine.The present embodiment can detect a plurality of types of road traffic markings, including Lane marking line, song simultaneously
Line style lane line, arrow, stop line, pavement etc.；Strong environmental adaptability wants the angle requirement and height of unmanned plane shooting
Ask loose, the sundries such as water mark, shade, road surface stone influence it smaller；The march in the lane line region filtered out
Line fitting, improves the detection accuracy of lane line；The stroke width for only calculating bianry image pixel, is only examined in the region of lane
The road traffic markings such as arrow, pavement are surveyed, only in extracted limited connected domain（It is usually no more than 20）Into
Row curve matching, reduces calculation amount, improves detection efficiency.
Contain the explanation of the preferred embodiment of the present invention above, this be for the technical characteristic that the present invention will be described in detail, and
Be not intended to for summary of the invention being limited in concrete form described in embodiment, according to the present invention content purport carry out other
Modifications and variations are also protected by this patent.The purport of the content of present invention is to be defined by the claims, rather than by embodiment
Specific descriptions are defined.
Claims (8)
1. the road traffic marking detection method based on unmanned plane low latitude Aerial Images, which is characterized in that specifically include following step
Suddenly：
S1, the carriageway image that unmanned plane low latitude is taken photo by plane is obtained；
S2, the carriageway image of acquisition is pre-processed with the bianry image for obtaining carriageway image；
S3, the stroke width for calculating after pretreatment pixel in each connected domain in carriageway image；
S4, connected domain screening is carried out according to the stroke width of pixel for carrying out lane line and road in road traffic marking
First sorting；
S5, Fuzzy Linear discriminatory analysis is carried out as sample using the lane line and road that just sort out, calculates road traffic marking ladder
Degree enhancing parameter, specifically includes following steps：
Carriageway image exists in position coordinates the collection C1, read step S1 of filtered out lane line connected domain in S51, acquisition step S4
Rgb value at coordinate set C1 connects lane line filtered out in step S4 as the first kind sample of Fuzzy Linear discriminatory analysis
Logical domain is expanded, and the connected domain before the connected domain and expansion after expansion is subtracted each other, and rear connected domain position coordinates collection is subtracted each other in acquisition
Rgb value of the carriageway image at coordinate set C2 in C2, read step S1, as the second class sample of Fuzzy Linear discriminatory analysis,
The intersection of two class samples is population sample；
S52, the mean value for calculating first kind sample, the second class sample mean valueWith the mean value of population sample, inter- object distance, between class distance：
In formula,Indicate theThe number of class sample,Indicate sample,Indicate theThe set of class sample,Representing matrix
Transposition；
S53, find out rgb space to gray space optimal transformation ratio：
The calculation criterion of optimal projection coefficient is --- the within-cluster variance of similar sampleIt is as small as possible, inter _ class relationshipTo the greatest extent
May be big, it introduces Fisher and identifies criterion expression formula,
In formula, the converting vector of w expression rgb space to gray space,Corresponding w is rgb space when maximum
To the optimum translation coefficient of gray space, i.e. road traffic marking grad enhancement parameter；
S6, step S3 is repeated after repeating step S2 using road traffic marking grad enhancement parameter, and wide according to the stroke of pixel
Degree is carried out connected domain screening and is picked with the subdivision for carrying out all road traffic markings and road；
It is S7, bent with the cubic B-spline based on random sampling consistency in the road traffic marking connected domain that step S6 is filtered out
Line fitting algorithm accurately determines the position of road traffic marking.
2. the road traffic marking detection method according to claim 1 based on unmanned plane low latitude Aerial Images, feature exist
In in step S4, the first sorting for carrying out lane line and road in road traffic marking specifically includes following steps：
S41, the stroke width according to lane line, and screening range is determined according to height of taking photo by plane, it filters out stroke width and is screening
Connected domain in range；
S42, microdilatancy, filling processing are carried out to the result of step S41, the lane line being broken after step S41 screening is connected
Come；
S43, according to the geometrical characteristic parameter of connected domain area, long axis length, to step S42, treated that connected domain is carried out into one
Step screening.
3. the road traffic marking detection method according to claim 1 based on unmanned plane low latitude Aerial Images, feature exist
In in step S6, the subdivision for carrying out all road traffic markings and road, which picks, specifically includes following steps：
S61, according to the stroke width of all road traffic markings, filter out the connected domain of stroke width in a certain range,
Middle road traffic marking includes lane line, arrow, pavement and/or stop line；
S62, microdilatancy, filling processing are carried out to the result of step S61, the lane line being broken after step S61 screening is connected
Come；
S63, according to the parameter of connected domain area, long axis length, to step S62, treated that connected domain is further screened.
4. the road traffic marking detection method according to any one of the claim 1 to 3 based on unmanned plane low latitude Aerial Images,
It is characterized in that, in step S3, calculate after pretreatment in carriageway image the stroke width of pixel in each connected domain specifically include with
Lower step：
S31, the bianry image image of a size obtained after pre-processing is established in a width and step S2, is stroke width figure
Picture, initializing each pixel value in stroke width image is infinity；
S32, edge extracting is carried out to the bianry image obtained after pre-processing in step S2, from any edge point, along the point
Gradient direction search for other marginal points, if find other marginal points, and its gradient direction just with the direction of search one
It causes, then all pixels point connected on this two o'clock line segment in stroke width image is assigned a value of to the Euclidean distance of the two o'clock；
S33, when the arbitrary point of stroke width image is by multiple assignment, each assignment is always compared with last institute's assignment
Compared with if smaller than last institute's assignment, the pixel value of the point being replaced with this assignment, i.e., belonging to each pixel always records
In most thin stroke width.
5. the road traffic marking detection method according to any one of the claim 1 to 3 based on unmanned plane low latitude Aerial Images,
It is characterized in that, in step S7, with the B three times based on random sampling consistency in the road traffic marking connected domain filtered out
Spline curve fitting algorithm accurately determines that the position of road traffic marking specifically includes following steps：
S71, setting fitting number；
S72, in analyzed connected domain, randomly select 4 points, seek the parameter of B-spline Curve, and determined according to parameter
Curve；
S73, calculate in analyzed connected domain in step S72 selected by other than 4 points other to curve distance, if apart from small
In threshold value, then it is assumed that the point is intra-office point, and the number put in statistics bureau, if the number of intra-office point is greater than preset value, then it is assumed that
Optimal curve has been found, and stops fitting, corresponding intra-office point is road traffic marking pixel；If the number of intra-office point is small
In preset value, then return step S72, finds optimal curve again, if finding number reaches fitting number, optimal curve is not looked for also
It arrives, then in found curve, the most curve of intra-office point is optimal curve, corresponding intra-office point is road traffic marking pixel
Point.
6. the road traffic marking detection method according to any one of the claim 1 to 3 based on unmanned plane low latitude Aerial Images,
It is characterized in that, in step S1, when unmanned plane shoots carriageway image, in a shooting operation variation of shooting angle 20 degree with
Interior, shooting height variation within 15 meters, shooting height within 60 meters, the weather condition of shooting is in macroscopic image
Road traffic marking subject to.
7. the road traffic marking detection method according to any one of the claim 1 to 3 based on unmanned plane low latitude Aerial Images,
It is characterized in that, in step S2, to the carriageway image of acquisition carry out pretreatment include image grayscale conversion, the variation of image top cap,
At least one of contrast stretching, carrying out image threshold segmentation, image expansion or Image erosion.
8. based on the road traffic marking detection system of unmanned plane low latitude Aerial Images, including memory and processor, it is described to deposit
Reservoir is stored with computer program, which is characterized in that the processor realized when executing the computer program claim 1 to
The step of any one of 7 the method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810951206.0A CN108596165B (en) | 2018-08-21 | 2018-08-21 | Road traffic marking detection method and system based on unmanned plane low latitude Aerial Images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810951206.0A CN108596165B (en) | 2018-08-21 | 2018-08-21 | Road traffic marking detection method and system based on unmanned plane low latitude Aerial Images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108596165A CN108596165A (en) | 2018-09-28 |
| CN108596165B true CN108596165B (en) | 2018-11-23 |
Family
ID=63619009
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810951206.0A Active CN108596165B (en) | 2018-08-21 | 2018-08-21 | Road traffic marking detection method and system based on unmanned plane low latitude Aerial Images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108596165B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109215053B (en) * | 2018-10-16 | 2021-04-27 | 西安建筑科技大学 | Method for detecting moving vehicle with pause state in aerial video shot by unmanned aerial vehicle |
| CN109684938A (en) * | 2018-12-06 | 2019-04-26 | 广西大学 | It is a kind of to be taken photo by plane the sugarcane strain number automatic identifying method of top view based on crop canopies |
| CN111488762A (en) * | 2019-01-25 | 2020-08-04 | 阿里巴巴集团控股有限公司 | Lane-level positioning method and device and positioning equipment |
| CN110070012B (en) * | 2019-04-11 | 2022-04-19 | 电子科技大学 | Refinement and global connection method applied to remote sensing image road network extraction |
| CN110245600B (en) * | 2019-06-11 | 2021-07-23 | 长安大学 | Unmanned aerial vehicle road detection method for self-adaptive initial quick stroke width |
| CN110516532B (en) * | 2019-07-11 | 2022-03-11 | 北京交通大学 | Unmanned aerial vehicle railway track line identification method based on computer vision |
| CN110989613A (en) * | 2019-12-18 | 2020-04-10 | 北京新能源汽车技术创新中心有限公司 | Vehicle positioning method and device, electronic equipment and storage medium |
| CN113371185B (en) * | 2021-07-19 | 2023-08-08 | 江苏中天吉奥信息技术股份有限公司 | Terrain aerial investigation method and aerial aircraft |
| CN113763342B (en) * | 2021-08-30 | 2024-04-30 | 东南大学 | Expressway marking detection method based on unmanned aerial vehicle remote sensing |
| CN115100696A (en) * | 2022-08-29 | 2022-09-23 | 山东圣点世纪科技有限公司 | Connected domain rapid marking and extracting method and system in palm vein recognition |
| CN115294293B (en) * | 2022-10-08 | 2023-03-24 | 速度时空信息科技股份有限公司 | Method for automatically compiling high-precision map road reference line based on low-altitude aerial photography result |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103500322B (en) * | 2013-09-10 | 2016-08-17 | 北京航空航天大学 | Automatic lane line identification method based on low latitude Aerial Images |
| CN105740809B (en) * | 2016-01-28 | 2019-03-12 | 东南大学 | A kind of highway method for detecting lane lines based on Airborne camera |
| CN107180228A (en) * | 2017-05-02 | 2017-09-19 | 开易（北京）科技有限公司 | A kind of grad enhancement conversion method and system for lane detection |
| CN107330380A (en) * | 2017-06-14 | 2017-11-07 | 千寻位置网络有限公司 | Lane line based on unmanned plane image is automatically extracted and recognition methods |
-
2018
- 2018-08-21 CN CN201810951206.0A patent/CN108596165B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108596165A (en) | 2018-09-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108596165B (en) | Road traffic marking detection method and system based on unmanned plane low latitude Aerial Images | |
| Wei et al. | Urban building extraction from high-resolution satellite panchromatic image using clustering and edge detection | |
| CN106022381B (en) | Automatic extraction method of street lamp pole based on vehicle-mounted laser scanning point cloud | |
| CN110647850A (en) | Automatic lane deviation measuring method based on inverse perspective principle | |
| Kong et al. | General road detection from a single image | |
| Kong et al. | Vanishing point detection for road detection | |
| CN111340797A (en) | Laser radar and binocular camera data fusion detection method and system | |
| US20160154999A1 (en) | Objection recognition in a 3d scene | |
| CN108875911A (en) | One kind is parked position detecting method | |
| Zhao et al. | Road network extraction from airborne LiDAR data using scene context | |
| CN110414355A (en) | The right bit sky parking stall of view-based access control model and parking stall line detecting method during parking | |
| CN107392929B (en) | Intelligent target detection and size measurement method based on human eye vision model | |
| CN112734761B (en) | Industrial product image boundary contour extraction method | |
| CN108764012A (en) | The urban road shaft recognizer of mobile lidar data based on multi-frame joint | |
| CN107729853A (en) | A kind of automatic identifying method suitable for the narrow tuning drive gear formula instrument of transformer station | |
| CN108052886A (en) | A kind of puccinia striiformis uredospore programming count method of counting | |
| CN108538052A (en) | Night traffic flow rate testing methods based on headlight track following and dynamic pairing | |
| CN111199245A (en) | Rape pest identification method | |
| CN113281782A (en) | Laser radar snow point filtering method based on unmanned vehicle | |
| CN110458019B (en) | Water surface target detection method for eliminating reflection interference under scarce cognitive sample condition | |
| CN112197705A (en) | Fruit positioning method based on vision and laser ranging | |
| Zou et al. | Path voting based pavement crack detection from laser range images | |
| CN116958837A (en) | Municipal facilities fault detection system based on unmanned aerial vehicle | |
| CN105260559A (en) | Paper pulp fiber morphology parameter calculation method based on contour area and contour refinement | |
| CN111160231A (en) | Automatic driving environment road extraction method based on Mask R-CNN |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 410000 1501, building 1, Xincheng Science Park, Lugu street, Yuelu District, Changsha City, Hunan Province Patentee after: Hunan Kunpeng Zhihui Technology Co.,Ltd. Address before: 410000 1501, building 1, Xincheng Science Park, Lugu street, Yuelu District, Changsha City, Hunan Province Patentee before: HUNAN KUNPENG ZHIHUI UNMANNED PLANE TECHNOLOGY CO.,LTD. |
|
| CP01 | Change in the name or title of a patent holder |