CN108090455A - Parking stall line vertex localization method, system, terminal and medium based on cascade mechanism - Google Patents
Parking stall line vertex localization method, system, terminal and medium based on cascade mechanism Download PDFInfo
- Publication number
- CN108090455A CN108090455A CN201711439249.2A CN201711439249A CN108090455A CN 108090455 A CN108090455 A CN 108090455A CN 201711439249 A CN201711439249 A CN 201711439249A CN 108090455 A CN108090455 A CN 108090455A
- Authority
- CN
- China
- Prior art keywords
- pixel
- feature descriptor
- parking stall
- stall line
- image
- 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.)
- Granted
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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The present invention provides a kind of parking stall line vertex localization method, system, terminal and medium based on cascade mechanism, comprises the following steps:Input semantic segmentation bianry image;The pixel on semantic segmentation bianry image is traveled through, finds out the pixel that position is in image border;The Feature Descriptor of the pixel is calculated using at least level-one circular shuttering;Image is filtered one by one using at least level-one Feature Descriptor calculated;Pixel is clustered, and the pixel for meeting cluster condition is ranged to the candidate point on parking stall line vertex, terminates flow.The present invention is more more descriptive than Feature Descriptors such as traditional SIFT/SURF using the high-dimensional Feature Descriptor that two-stage Feature Descriptor joint is formed, and the advantage that classification calculates causes overwhelming majority point only to calculate the feature of first order Feature Descriptor, calculation amount substantially reduces.
Description
Technical field
The present invention relates to vehicle electronics technical fields, are positioned more particularly to a kind of parking stall line vertex based on cascade mechanism
Method, system, terminal and medium.
Background technology
ADAS, that is, advanced driving assistance system is also known as active safety system, mainly include body electronics systems stabilisation ESC, from
Adapt to cruise system ACC, lane shift alarm system LDW, Lane Keeping System LKA, forward direction collision warning systems FCW, enabling
Early warning DOW, automatic emergency brake system AEB, Traffic Sign Recognition TSR, blind spot detection BSD, night vision system NV, automatic parking system
Unite APS etc..
ADAS active safety systems will not only identify static object, also identify dynamic object.Depth convolution at present
Neutral net has been achieved for greatly success on image recognition tasks.But still remain problems with:
The calculation amount for traveling through the image of known features in the image of magnanimity by several known features is very big, and
And in the hardware of same treatment ability, the calculating cycle that known features image is traveled through from the image of magnanimity is long.So shorten
The traversal time for extracting feature is current urgent problem to be solved.
The content of the invention
In order to solve the technical issues of above-mentioned and other are potential, the present invention provides a kind of vehicles based on cascade mechanism
Bit line vertex localization method, system, terminal and medium, first, with two-stage parking stall line vertex candidate point strobe utility, because first
The preliminary filtering of grade can filter away more than 95% non-parking stall line vertex, and the first order Feature Descriptor used first
It is to be filtered with small template, although the not high computation complexity of precision is low, substantial amounts of computing cost is saved compared to large form.The
Two, the filtering of second level parking stall line vertex is on the basis of the result that is filtered in the first order, further increase to parking stall line vertex
Be accurately positioned, used large form to calculate the second level Feature Descriptor on parking stall line vertex.Using two-stage Feature Descriptor
This high-dimensional Feature Descriptor of Feature Descriptor that joint is formed all has more than Feature Descriptors such as traditional SIFT/SURF
It is descriptive, and be classified the advantage calculated and overwhelming majority point is caused only to calculate the feature of first order Feature Descriptor, calculation amount is shown
It writes and reduces.
A kind of parking stall line vertex localization method based on cascade mechanism, comprises the following steps:
S01:Input semantic segmentation bianry image;
S02:The pixel on semantic segmentation bianry image is traveled through, finds out the pixel that position is in image border;
S03:The Feature Descriptor of the pixel is calculated using at least level-one circular shuttering;
S04:Image is filtered one by one using at least level-one Feature Descriptor calculated;
S05:Pixel is clustered, and the pixel for meeting cluster condition is ranged to the candidate point on parking stall line vertex, is terminated
Flow.
Further, the Feature Descriptor includes first order Feature Descriptor and second level Feature Descriptor, first by the
Level-one Feature Descriptor filters image, and the result drawn filters image by second level Feature Descriptor again.
Further, the first order Feature Descriptor and the second level Feature Descriptor mould including feature vector are about
Beam condition and angle restriction condition.
Further, the first order Feature Descriptor and second level Feature Descriptor further include position constraint condition.
Further, the specific steps of the first order Feature Descriptor mould length constraint:The first order Feature Descriptor
The specific steps of the long constraints of mould:
S031:The small template that pixel matrix is set to arrange travels through bianry image with small template;
S032:Small template is divided into n dimensions by the center of circle of its central point, the pixel in small template is distributed to n dimension
Under, the number of pixel under each dimension is calculated, one is set by the upper limit of a numerical value of its pixel to each dimension
Dimension pixel threshold value if white pixel point number is more than the dimension pixel threshold value, retains the vector, if less than dimension
During pixel threshold value, then the vector is filtered;
S033:Filter out the pixel after the length constraint of first order Feature Descriptor mould..
Further, the specific steps of the first order Feature Descriptor angle restriction condition:
S034:It calculates mould and grows the angle constrained between the vector for filtering out pixel, if angle meets angle threshold value scope,
Then retain two vectors, be unsatisfactory for angle threshold value scope, then filter out two vectors;
S035:Filter out the pixel after first order Feature Descriptor angle restriction.
Further, angle threshold value scope is 75-105 degree in angle restriction condition in the step S034.
Further, the specific steps of the first order Feature Descriptor distance restraint:
S036:It filters out the pixel that first order Feature Descriptor angle restriction obtains and these pixels is formed into pixel
Point set judges the position of each pixel and the bianry image parking stall line edge where the pixel in the pixel point set
Distance,
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are more than two pixels
When, then retain the pixel;
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are less than two pixels,
It then filters this pixel and jumps to next pixel in pixel point set.
Further, the mould length constraint specific steps of the second level Feature Descriptor:
S041:The large form that pixel matrix is set to arrange traverses through first order Feature Descriptor with large form and screens
Acquired results;
S042:Large form is divided into m dimensions by the center of circle of its central point, the pixel in large form is distributed to m dimension
Under, the number of pixel under each dimension is calculated, one is set by the upper limit of a numerical value of its pixel to each dimension
Dimension pixel threshold value if white pixel point number is more than the dimension pixel threshold value, retains the vector, if less than dimension
During pixel threshold value, then the vector is filtered;
S043:Filter out the pixel after the Feature Descriptor mould length constraint of the second level;.
Further, the ranks number of the pixel matrix of the small template is less than the ranks of the pixel matrix of large form
Number.
Further, the angle restriction specific steps of the second level Feature Descriptor:
S044:It calculates mould and grows the angle constrained between the vector for filtering out pixel, if angle meets angle threshold value scope,
Then retain two vectors, be unsatisfactory for angle threshold value scope, then filter out two vectors;
S045:Filter out the pixel after the Feature Descriptor angle restriction of the second level.
Further, angle threshold value scope is 75-105 degree in angle restriction condition in the step S034.
Further, the specific steps of the second level Feature Descriptor distance restraint:
S046:It filters out the pixel that second level Feature Descriptor angle restriction obtains and these pixels is formed into pixel
Point set judges the position of each pixel and the bianry image parking stall line edge where the pixel in the pixel point set
Distance,
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are more than two pixels
When, then retain the pixel;
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are less than two pixels,
It then filters this pixel and jumps to next pixel in pixel point set.
Further, the pixel point set root that the remaining pixel after the second level Feature Descriptor is screened forms
It is clustered according to the distance between each pixel position, parking stall line Corner clustering condition will be met in cluster process
Pixel region recognition is the candidate point on parking stall line vertex.
A kind of parking stall line vertex alignment system based on cascade mechanism, including semantic segmentation module, image binaryzation module,
First order feature describing module, second level feature describing module and cluster module;
The semantic segmentation module is used for the pixel in segmentation figure picture, and the in vivo all pixels point of same object is made to be in one
In a semantic segmentation module;
Described image binarization block obtains binary image for handling original image;;
The small template filter binary image of the first order feature describing module, for tentatively filtering in binary image
It is not the pixel on parking stall line vertex;;
The pixel that the second level feature describing module obtains after screening small template filtering with large form, for further
It is not the pixel on parking stall line vertex in filtering binary image;
The cluster module is used for the pixel that the remaining pixel after the second level Feature Descriptor is screened forms
Point set is clustered according to the distance between each pixel position, and parking stall line angle point will be met in cluster process and is gathered
The pixel region recognition of class condition is parking stall line vertex.
The first order feature increase describing module and the second level feature describing module include the long constraints module of mould and
Angle restriction module.The first order feature describing module and second level feature describing module further include distance restraint module.
A kind of target following car-mounted terminal based on depth characteristic stream, which is characterized in that including processor and memory, institute
Stating memory storage has program instruction, and the step in above-mentioned method is realized in the processor operation program instruction.
A kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that:The program is by processor
The step in above-mentioned method is realized during execution.
As described above, the present invention's has the advantages that:
First, with two-stage parking stall line vertex candidate point strobe utility because the preliminary filtering of the first order can by 95% with
On non-parking stall line vertex filter away, and the first order Feature Descriptor used first is filtered with small template, although smart
It is low to spend not high but computation complexity, substantial amounts of computing cost is saved compared to large form.
Second, the filtering of second level parking stall line vertex is on the basis of the result that is filtered in the first order, further increase pair
Parking stall line vertex is accurately positioned, and has been used large form to calculate the second level Feature Descriptor on parking stall line vertex, has equally been 24
Dimension.Then, the dimension for the Feature Descriptor that two-stage Feature Descriptor joint is formed is exactly 24*24=576, so high-dimensional
Feature Descriptor is all more more descriptive than Feature Descriptors such as traditional SIFT/SURF.And the advantage that classification calculates causes absolutely greatly
Partial dot only calculates the feature of first order Feature Descriptor, i.e. 24 dimensions.
Description of the drawings
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, the accompanying drawings in the following description is only some embodiments of the present invention, for
For those of ordinary skill in the art, without creative efforts, other are can also be obtained according to these attached drawings
Attached drawing.
Fig. 1 is shown as the particular flow sheet of the present invention.
Fig. 2 is shown as the flow chart of one embodiment of the invention.
Fig. 3 is shown as the schematic diagram of the small template that the constraint of first order Feature Descriptor mould length uses in an embodiment.
Fig. 4 is shown as the schematic diagram of the large form that the constraint of second level Feature Descriptor mould length uses in an embodiment.
Specific embodiment
Illustrate embodiments of the present invention below by way of specific specific example, those skilled in the art can be by this specification
Disclosed content understands other advantages and effect of the present invention easily.The present invention can also pass through in addition different specific realities
The mode of applying is embodied or practiced, the various details in this specification can also be based on different viewpoints with application, without departing from
Various modifications or alterations are carried out under the spirit of the present invention.It should be noted that in the case where there is no conflict, following embodiment and implementation
Feature in example can be mutually combined.
It should be clear that structure, ratio, size depicted in this specification institute accompanying drawings etc., only specification to be coordinated to be taken off
The content shown so that those skilled in the art understands and reads, is not limited to the enforceable qualifications of the present invention, therefore
Do not have technical essential meaning, the modification of any structure, the change of proportionate relationship or the adjustment of size are not influencing the present invention
Under the effect of can be generated and the purpose that can reach, it should all still fall and obtain the model that can cover in disclosed technology contents
In enclosing.Meanwhile cited such as " on ", " under " in this specification, "left", "right", the term of " centre " and " one ", also it is only
Convenient for the clear of narration rather than to limit the enforceable scope of the present invention, relativeness is altered or modified, in no essence
It changes under technology contents, when being also considered as the enforceable scope of the present invention.
Referring to Fig. 1~Fig. 4, a kind of parking stall line vertex localization method based on cascade mechanism comprises the following steps:
A kind of parking stall line vertex localization method based on cascade mechanism, comprises the following steps:
S01:Input semantic segmentation bianry image;
S02:The pixel on semantic segmentation bianry image is traveled through, finds out the pixel that position is in image border;
S03:The Feature Descriptor of the pixel is calculated using at least level-one circular shuttering;
S04:Image is filtered one by one using at least level-one Feature Descriptor calculated;
S05:Pixel is clustered, and the pixel for meeting cluster condition is ranged to the candidate point on parking stall line vertex, is terminated
Flow.
As the presently preferred embodiments, the Feature Descriptor includes first order Feature Descriptor and second level Feature Descriptor,
Image is first filtered by first order Feature Descriptor, the result drawn filters image by second level Feature Descriptor again.
As the presently preferred embodiments, the first order Feature Descriptor and second level Feature Descriptor include feature vector
The long constraints of mould and angle restriction condition.
As the presently preferred embodiments, the first order Feature Descriptor and second level Feature Descriptor further include position constraint item
Part.
As the presently preferred embodiments, the specific steps of the first order Feature Descriptor mould length constraint:
The specific steps of the long constraints of first order Feature Descriptor mould:
S031:The small template that pixel matrix is set to arrange travels through bianry image with small template;
S032:The small template of 19*19 pixels is divided into 24 dimensions by the center of circle of its central point, by the pixel in small template
Under mean allocation to 24 dimensions, calculate the number of pixel under each dimension as 10 pixels, to each dimension with
A numerical value of its pixel sets a dimension pixel threshold value for the upper limit, it is assumed that dimension pixel threshold value is 8, if white pixel
Point number be more than the dimension pixel threshold value 8 when, then retain the vector, if be less than dimension pixel threshold value 8, filter this to
Amount;The size of dimension pixel threshold value is determined according to the quality of image.
S033:Filter out the pixel after the length constraint of first order Feature Descriptor mould..
As the presently preferred embodiments, the specific steps of the first order Feature Descriptor angle restriction condition:
S034:It calculates mould and grows the angle constrained between the vector for filtering out pixel, if angle meets angle threshold value scope,
Then retain two vectors, be unsatisfactory for angle threshold value scope, then filter out two vectors;
S035:Filter out the pixel after first order Feature Descriptor angle restriction.
As the presently preferred embodiments, angle threshold value scope is 75-105 degree in angle restriction condition in the step S034.
As the presently preferred embodiments, the specific steps of the first order Feature Descriptor distance restraint:
S036:It filters out the pixel that first order Feature Descriptor angle restriction obtains and these pixels is formed into pixel
Point set judges the position of each pixel and the bianry image parking stall line edge where the pixel in the pixel point set
Distance,
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are more than distance threshold,
Then retain the pixel;
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are less than distance threshold,
It filters this pixel and jumps to next pixel in pixel point set.
Further, the distance threshold upper limit of the first order Feature Descriptor distance restraint is the one of parking stall line width
Half, the distance threshold lower limit of the first order Feature Descriptor distance restraint is a pixel.
As the presently preferred embodiments, the mould length constraint specific steps of the second level Feature Descriptor:
S041:The large form of 29*29 pixel matrix arrangement is set, first order Feature Descriptor is traversed through with large form
Screen acquired results;
S042:The large form of 29*29 pixels is divided into 24 dimensions by the center of circle of its central point, by the pixel in large form
Under mean allocation to 24 dimensions, calculate the number of pixel under each dimension as 15 pixels, to each dimension with
A numerical value of its pixel sets a dimension pixel threshold value for the upper limit, it is assumed that dimension pixel threshold value is 15, if white picture
When vegetarian refreshments number is more than the dimension pixel threshold value 15, then retain the vector, if be less than dimension pixel threshold value 15, filter
The vector;The size of dimension pixel threshold value is determined according to the quality of image.
S043:Filter out the pixel after the Feature Descriptor mould length constraint of the second level.
Further, the ranks number of the pixel matrix of the small template is less than the ranks of the pixel matrix of large form
Number.
Further, the angle restriction specific steps of the second level Feature Descriptor:
S044:It calculates mould and grows the angle constrained between the vector for filtering out pixel, if angle meets angle threshold value scope,
Then retain two vectors, be unsatisfactory for angle threshold value scope, then filter out two vectors;
S045:Filter out the pixel after the Feature Descriptor angle restriction of the second level.
As the presently preferred embodiments, angle threshold value scope is 75-105 degree in angle restriction condition in the step S034.
As the presently preferred embodiments, the specific steps of the second level Feature Descriptor distance restraint:
S046:It filters out the pixel that second level Feature Descriptor angle restriction obtains and these pixels is formed into pixel
Point set judges the position of each pixel and the bianry image parking stall line edge where the pixel in the pixel point set
Distance,
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are more than distance threshold,
Then retain the pixel;
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are less than distance threshold,
It filters this pixel and jumps to next pixel in pixel point set.
Further, the distance threshold upper limit of the second level Feature Descriptor distance restraint is the one of parking stall line width
Half, the distance threshold lower limit of the second level Feature Descriptor distance restraint is a pixel.
As the presently preferred embodiments, the pixel that the remaining pixel after the second level Feature Descriptor is screened forms
Set is clustered according to the distance between each pixel position, and parking stall line Corner clustering will be met in cluster process
The pixel region recognition of condition is the candidate point on parking stall line vertex.
A kind of parking stall line vertex alignment system based on cascade mechanism, including semantic segmentation module, image binaryzation module,
First order feature describing module, second level feature describing module and cluster module;
The semantic segmentation module is used for the pixel in segmentation figure picture, and the in vivo all pixels point of same object is made to be in one
In a semantic segmentation module;
Described image binarization block obtains binary image for handling original image;;
The small template filter binary image of the first order feature describing module, for tentatively filtering in binary image
It is not the pixel on parking stall line vertex;;
The pixel that the second level feature describing module obtains after screening small template filtering with large form, for further
It is not the pixel on parking stall line vertex in filtering binary image;
The cluster module is used for the pixel that the remaining pixel after the second level Feature Descriptor is screened forms
Point set is clustered according to the distance between each pixel position, and parking stall line angle point will be met in cluster process and is gathered
The pixel region recognition of class condition is parking stall line vertex.
The first order feature describing module and the second level feature describing module include the long constraints module of mould and angle
Spend constraints module.The first order feature describing module and second level feature describing module further include distance restraint module.
A kind of target following car-mounted terminal based on depth characteristic stream, which is characterized in that including processor and memory, institute
Stating memory storage has program instruction, and the step in above-mentioned method is realized in the processor operation program instruction.
A kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that:The program is by processor
The step in above-mentioned method is realized during execution.
A kind of target following car-mounted terminal based on depth characteristic stream, which is characterized in that including processor and memory, institute
Stating memory storage has program instruction, and the step in above-mentioned method is realized in the processor operation program instruction.
A kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that:The program is by processor
The step in above-mentioned method is realized during execution.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.It is any ripe
Know the personage of this technology all can carry out modifications and changes under the spirit and scope without prejudice to the present invention to above-described embodiment.Cause
This, technical field includes usual skill complete without departing from disclosed spirit and institute under technological thought such as
Into all equivalent modifications or change, should by the present invention claim be covered.
Claims (10)
1. the parking stall line vertex localization method based on cascade mechanism, which is characterized in that comprise the following steps:
S01:Input semantic segmentation bianry image;
S02:The pixel on semantic segmentation bianry image is traveled through, finds out the pixel that position is in image border;
S03:The Feature Descriptor of the pixel is calculated using at least level-one circular shuttering;
S04:Image is filtered one by one using at least level-one Feature Descriptor calculated;
S05:Pixel is clustered, and the pixel for meeting cluster condition is ranged to the candidate point on parking stall line vertex, terminates flow.
2. the parking stall line vertex localization method according to claim 1 based on cascade mechanism, which is characterized in that the feature
Description attached bag includes first order Feature Descriptor and second level Feature Descriptor, first filters image by first order Feature Descriptor, obtains
The result gone out filters image by second level Feature Descriptor again.
3. the parking stall line vertex localization method according to claim 2 based on cascade mechanism, which is characterized in that described first
Grade Feature Descriptor and second level Feature Descriptor include the long constraints of mould of feature vector and angle restriction condition, appoint
The long constraints of beam mould constrains angle restriction condition again.
4. the parking stall line vertex localization method according to claim 3 based on cascade mechanism, which is characterized in that described first
Grade Feature Descriptor and second level Feature Descriptor further include position constraint condition, and the position constraint condition completes mould in operation
Position constraint is carried out again after long constraint and angle restriction.
5. the parking stall line vertex localization method according to claim 3 based on cascade mechanism, which is characterized in that
The specific steps of the long constraints of first order Feature Descriptor mould:
S031:The small template that pixel matrix is set to arrange travels through bianry image with small template;
S032:Small template is divided into n dimensions by the center of circle of its central point, the pixel in small template is distributed to n dimension, is counted
The number of pixel under each dimension is calculated, one dimension picture is set by the upper limit of a numerical value of its pixel to each dimension
Vegetarian refreshments threshold value if white pixel point number is more than the dimension pixel threshold value, retains the vector, if less than dimension pixel
During threshold value, then the vector is filtered;;
S033:Filter out the pixel after the length constraint of first order Feature Descriptor mould;
The specific steps of the long constraints of mould of the second level Feature Descriptor:
S041:The large form that pixel matrix is set to arrange traverses through first order Feature Descriptor screening gained with large form
As a result;
S042:Large form is divided into m dimensions by the center of circle of its central point, the pixel in large form is distributed to m dimension, is counted
The number of pixel under each dimension is calculated, one dimension picture is set by the upper limit of a numerical value of its pixel to each dimension
Vegetarian refreshments threshold value if white pixel point number is more than the dimension pixel threshold value, retains the vector, if less than dimension pixel
During threshold value, then the vector is filtered;;
S043:Filter out the pixel after the Feature Descriptor mould length constraint of the second level;
The ranks number of the pixel matrix of the small template is less than the ranks number of the pixel matrix of large form.
6. the parking stall line vertex localization method according to claim 3 based on cascade mechanism, which is characterized in that described first
The specific steps of grade Feature Descriptor angle restriction condition:
S034:It calculates mould and grows the angle constrained between the vector for filtering out pixel, if angle meets angle threshold value scope, protect
Two vectors are stayed, are unsatisfactory for angle threshold value scope, then filter out two vectors;
S035:Filter out the pixel after first order Feature Descriptor angle restriction;
The angle restriction specific steps of the second level Feature Descriptor:
S044:It calculates mould and grows the angle constrained between the vector for filtering out pixel, if angle meets angle threshold value scope, protect
Two vectors are stayed, are unsatisfactory for angle threshold value scope, then filter out two vectors;
S045:Filter out the pixel after the Feature Descriptor angle restriction of the second level.
7. according to parking stall line vertex localization method of claim 4 any one of them based on cascade mechanism, which is characterized in that
The specific steps of the first order Feature Descriptor distance restraint:
S036:It filters out the pixel that first order Feature Descriptor angle restriction obtains and these pixels is formed into pixel point set
Close, judge bianry image parking stall line edge in the pixel point set where the position of each pixel and the pixel away from
From,
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are more than distance threshold scope,
Then retain the pixel;
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are less than distance threshold scope,
It filters this pixel and jumps to next pixel in pixel point set;
The specific steps of the second level Feature Descriptor distance restraint:
S046:It filters out the pixel that second level Feature Descriptor angle restriction obtains and these pixels is formed into pixel point set
Close, judge bianry image parking stall line edge in the pixel point set where the position of each pixel and the pixel away from
From,
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are more than distance threshold scope,
Then retain the pixel;
If the position of pixel and the distance at the bianry image parking stall line edge where the pixel are less than distance threshold scope,
It then filters this pixel and jumps to next pixel in pixel point set.
8. a kind of parking stall line vertex alignment system based on cascade mechanism, which is characterized in that including semantic segmentation module, image two
Value module, first order feature describing module, second level feature describing module and cluster module;
The semantic segmentation module is used for the pixel in segmentation figure picture, and the in vivo all pixels point of same object is made to be in a language
In justice segmentation module;
Described image binarization block obtains binary image for handling original image;;
The small template filter binary image of the first order feature describing module is not for tentatively filtering in binary image
The pixel on parking stall line vertex;;
The pixel that the second level feature describing module obtains after screening small template filtering with large form, for further filtering
It is not the pixel on parking stall line vertex in binary image;
The cluster module is used for the pixel point set that the remaining pixel after the second level Feature Descriptor is screened forms
Conjunction is clustered according to the distance between each pixel position, and parking stall line Corner clustering item will be met in cluster process
The pixel region recognition of part is parking stall line vertex;
The first order feature, which increases describing module and the second level feature describing module, includes the long constraints module of mould and angle
Constraints module and distance restraint module.
9. a kind of target following car-mounted terminal based on depth characteristic stream, which is characterized in that described including processor and memory
Memory storage has program instruction, and the processor operation program instruction is realized as described in claim 1 to 7 any claim
Method in step.
10. a kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that:The program is by processor
The step in the method as described in claim 1 to 7 any claim is realized during execution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711439249.2A CN108090455B (en) | 2017-12-27 | 2017-12-27 | Cascade mechanism-based parking space line vertex positioning method, system, terminal and medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711439249.2A CN108090455B (en) | 2017-12-27 | 2017-12-27 | Cascade mechanism-based parking space line vertex positioning method, system, terminal and medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108090455A true CN108090455A (en) | 2018-05-29 |
CN108090455B CN108090455B (en) | 2023-08-22 |
Family
ID=62179562
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711439249.2A Active CN108090455B (en) | 2017-12-27 | 2017-12-27 | Cascade mechanism-based parking space line vertex positioning method, system, terminal and medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108090455B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109543609A (en) * | 2018-11-22 | 2019-03-29 | 四川长虹电器股份有限公司 | The method for detecting backing distance |
CN110796063A (en) * | 2019-10-24 | 2020-02-14 | 百度在线网络技术(北京)有限公司 | Method, device, equipment, storage medium and vehicle for detecting parking space |
CN110969655A (en) * | 2019-10-24 | 2020-04-07 | 百度在线网络技术(北京)有限公司 | Method, device, equipment, storage medium and vehicle for detecting parking space |
CN111274974A (en) * | 2020-01-21 | 2020-06-12 | 北京百度网讯科技有限公司 | Positioning element detection method, device, equipment and medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090146842A1 (en) * | 2007-10-26 | 2009-06-11 | Mando Corporation | Method and system for recognizing target parking location of vehicle |
CN103366602A (en) * | 2012-03-29 | 2013-10-23 | 施乐公司 | Method of determining parking lot occupancy from digital camera images |
CN104916163A (en) * | 2015-06-29 | 2015-09-16 | 惠州华阳通用电子有限公司 | Parking space detection method |
CN104933409A (en) * | 2015-06-12 | 2015-09-23 | 北京理工大学 | Parking space identification method based on point and line features of panoramic image |
CN107153823A (en) * | 2017-05-22 | 2017-09-12 | 北京北昂科技有限公司 | A kind of view-based access control model associates the lane line feature extracting method of double space |
CN107424116A (en) * | 2017-07-03 | 2017-12-01 | 浙江零跑科技有限公司 | Position detecting method of parking based on side ring depending on camera |
-
2017
- 2017-12-27 CN CN201711439249.2A patent/CN108090455B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090146842A1 (en) * | 2007-10-26 | 2009-06-11 | Mando Corporation | Method and system for recognizing target parking location of vehicle |
CN103366602A (en) * | 2012-03-29 | 2013-10-23 | 施乐公司 | Method of determining parking lot occupancy from digital camera images |
CN104933409A (en) * | 2015-06-12 | 2015-09-23 | 北京理工大学 | Parking space identification method based on point and line features of panoramic image |
CN104916163A (en) * | 2015-06-29 | 2015-09-16 | 惠州华阳通用电子有限公司 | Parking space detection method |
CN107153823A (en) * | 2017-05-22 | 2017-09-12 | 北京北昂科技有限公司 | A kind of view-based access control model associates the lane line feature extracting method of double space |
CN107424116A (en) * | 2017-07-03 | 2017-12-01 | 浙江零跑科技有限公司 | Position detecting method of parking based on side ring depending on camera |
Non-Patent Citations (1)
Title |
---|
JAE KYU SUHR等: "Fully-automatic Recognition of Various Parking Slot Markings in Around View Monitor (AVM) Image Sequences" * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109543609A (en) * | 2018-11-22 | 2019-03-29 | 四川长虹电器股份有限公司 | The method for detecting backing distance |
CN109543609B (en) * | 2018-11-22 | 2022-04-12 | 四川长虹电器股份有限公司 | Method for detecting reversing distance |
CN110796063A (en) * | 2019-10-24 | 2020-02-14 | 百度在线网络技术(北京)有限公司 | Method, device, equipment, storage medium and vehicle for detecting parking space |
CN110969655A (en) * | 2019-10-24 | 2020-04-07 | 百度在线网络技术(北京)有限公司 | Method, device, equipment, storage medium and vehicle for detecting parking space |
CN110969655B (en) * | 2019-10-24 | 2023-08-18 | 百度在线网络技术(北京)有限公司 | Method, device, equipment, storage medium and vehicle for detecting parking space |
CN111274974A (en) * | 2020-01-21 | 2020-06-12 | 北京百度网讯科技有限公司 | Positioning element detection method, device, equipment and medium |
CN111274974B (en) * | 2020-01-21 | 2023-09-01 | 阿波罗智能技术(北京)有限公司 | Positioning element detection method, device, equipment and medium |
Also Published As
Publication number | Publication date |
---|---|
CN108090455B (en) | 2023-08-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112528878B (en) | Method and device for detecting lane line, terminal equipment and readable storage medium | |
Gilroy et al. | Overcoming occlusion in the automotive environment—A review | |
JP5690688B2 (en) | Outside world recognition method, apparatus, and vehicle system | |
CN108090455A (en) | Parking stall line vertex localization method, system, terminal and medium based on cascade mechanism | |
US10380434B2 (en) | Vehicle detection system and method | |
CN114359851A (en) | Unmanned target detection method, device, equipment and medium | |
US8836812B2 (en) | Image processing device, image processing method, and image processing program | |
CN112793576B (en) | Lane change decision method and system based on rule and machine learning fusion | |
EP3979196A1 (en) | Image processing method and apparatus for target detection | |
JP2019021317A (en) | Driver support system and guard rail detection method | |
Em et al. | Vision-based lane departure warning framework | |
CN115376089A (en) | Deep learning-based lane line detection method | |
Heo et al. | Extraction of car license plate regions using line grouping and edge density methods | |
US11458892B2 (en) | Image generation device and image generation method for generating a composite image | |
KR102039723B1 (en) | Vehicle's behavior analyzing system using aerial photo and analyzing method using the same | |
CN111144361A (en) | Road lane detection method based on binaryzation CGAN network | |
Jakob et al. | Camera-based on-road detections for the visually impaired | |
CN114067186A (en) | Pedestrian detection method and device, electronic equipment and storage medium | |
CN113989753A (en) | Multi-target detection processing method and device | |
CN113408413A (en) | Emergency lane identification method, system and device | |
CN112989956A (en) | Traffic light identification method and system based on region of interest and storage medium | |
CN115588188A (en) | Locomotive, vehicle-mounted terminal and driver behavior identification method | |
KR20220040530A (en) | System and method for deep learning based semantic segmentation with low light images | |
CN114627439A (en) | Moving object detection method based on 360-degree look-around camera | |
Chingting et al. | Traffic Lane Line Classification System by Real-time Image Processing |
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 |