CN110490913A - Feature based on angle point and the marshalling of single line section describes operator and carries out image matching method - Google Patents
Feature based on angle point and the marshalling of single line section describes operator and carries out image matching method Download PDFInfo
- Publication number
- CN110490913A CN110490913A CN201910660833.3A CN201910660833A CN110490913A CN 110490913 A CN110490913 A CN 110490913A CN 201910660833 A CN201910660833 A CN 201910660833A CN 110490913 A CN110490913 A CN 110490913A
- Authority
- CN
- China
- Prior art keywords
- angle point
- line segment
- image
- angle
- matched
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
Operator, which is described, the invention discloses a kind of feature based on angle point and the marshalling of single line section carries out image matching method, first, to Extraction of Image line segment and Harris angle point, it is searched for and is organized into groups, the angle point that will test and line segment building have angle point-single line section texture descriptor of scale, rotation and illumination invariant, wherein Harris angle point has the advantages that rotational invariance, and line segment can be lifted at the reliability in parallax scene change using half-breadth description；And carry out the spatial weighting shortest distance and estimate, obtain local matching results；Finally, establishing the candidate matches of each line segment, matching matrix is established, solves global registration result using spectrum analysis；Image Matching of the present invention, which describes operator, has the characteristics that scale, rotation and illumination invariant；The present invention establishes image pyramid to stereopsis respectively, matches one by one on the pyramid of different layers, can eliminate the influence of scale；And it can overcome in multi-line section matching and organize into groups the shortcomings that computationally intensive, time-consuming.
Description
Technical field
The invention belongs to remote sensing images identification and technical field of computer vision, are related to a kind of image matching method, specifically
It is related to a kind of feature based on Harris angle point and the marshalling of single line section and describes operator, and geometrical constraint is combined to carry out pyramid transmitting
The image matching method of strategy, can match characteristic point and characteristic curve simultaneously.
Background technique
Image Matching technology three-dimensional reconstruction, video search, target following, in terms of answered extensively
With with application values such as important military affairs, medical treatment and ECOLOGICAL ENVIRONMENTAL MONITORINGs.Although current Feature Correspondence Algorithm is special based on point
Levy and obtain significant achievement in matched research, but based on the matching technique of linear feature due to illumination, noise, block etc. because
Element influences, and there are still some problems: 1) line segment feature does not have conspicuousness, so the SIFT operator of conspicuousness can deliberately avoid line
Point in section；2) there are also problems for the line match based on geometrical constraint, such as directlys adopt the matching technique that core line constrains and mention
The endpoint taken is not accurate enough, and requires visual angle change small or predict the geometrical relationship between image；3) matching based on line segment marshalling
Method typically includes the matching relationship between internal a plurality of line segment, to construct how possible Feature Grouping, causes to calculate complicated, consumption
Duration；4) common line segment feature describes son can make buffer area to line segment two sides, and count to the texture of whole region
Description, but due to the variation of camera perspective, line segment may only have side texture be it is stable, the texture of the other side can occur
Biggish variation leads to matched uncertainty.
Harris angle point is used in edge detection and Corner Detection, is usually located at the joint at edge, can resist shooting
The variation at visual angle has rotational invariance, and the first derivative due to only only used image, image grayscale is translated and is changed
It is constant.These good features make it may be used as matched characteristic point.Meanwhile Harris angle point and line segment are on plan-position
With propinquity, complete and the line segment of accurate positioning endpoint is most likely Harris angle point, therefore is constructed the angle Harris
The matching that point is organized into groups with single line section describes operator, on the basis of overcoming rotation, grey scale change, can effectively filter and not conform to
The line segment group of reason, than multi-line section marshalling matching algorithm time-consuming is shorter, effect is more preferable.
Summary of the invention
In view of the shortcomings of the prior art, it is high that the present invention provides a kind of time and storage efficiencies, to image rotation, translation and
The more stable angle point of change of scale and the Image Matching of single line section marshalling describe operator and carry out image matching method.
The technical scheme adopted by the invention is that: it is a kind of based on angle point and single line section marshalling feature describe operator carry out shadow
As matching process, which comprises the following steps:
Step 1: input refers to image and image to be matched, constructs multi-level Gaussian image pyramid, successively drops and adopt to image
Step 2-4 is executed to each layer of reference image and image to be matched after sample, calculates best match scale；
Step 2: straightway and Harris angle point are extracted respectively to reference image and image to be matched；
Harris angle point is searched within the specified range to the endpoint for the straightway that each is extracted, by single straight line segment with
The associated marshalling of nearest angle point constitutes angle point-line segment in conjunction with the feature vector of angle point and spectral information description of straightway
Half-breadth textural characteristics description；
Step 3: according to the angle point of line segment-line segment half-breadth textural characteristics description to the point of reference image and image to be matched
Line is matched, and candidate matches set is obtained；
Step 4: according between image to be matched geometrical relationship calculate candidate matches similarity, including image angle point away from
From the similarity with arc description, to the marshalling screening candidate matches and establishment each line of reference image and image to be matched
The candidate matches matrix M of section i；Established matching matrix M is solved by spectrum analysis, judges that candidate matches are received or refused
Absolutely；
Step 5: the corresponding stereopsis of output Optimum Matching scale obtains the best ruler with reference to image and image to be matched
Degree and its corresponding angle point-matched result of single line section.
Image Matching technology provided by the invention has the beneficial effect that (1) includes line feature abundant for remote sensing image
The characteristics of, angle point and single line section are organized into groups to and constructed angle point-arc description, makes full use of the textural characteristics and line of line segment
Section screens out candidate matches with the geometrical relationship between corresponding angle point, effectively reduces the time complexity of algorithm, improves matching result
Reliability；(2) image pyramid is established.(3) eliminating image rotation and scaling in the matching process influences, and has entire algorithm
There are rotation and scale invariability, and be not related to pixel grey scale information in the matching process, so also having to brightness of image transformation
There is good invariance.
Detailed description of the invention
Fig. 1 is the flow chart of inventive embodiments；
Fig. 2 is that the angle point of the embodiment of the present invention is associated with schematic diagram with line segment；
Fig. 3 is the sub- schematic diagram of half-breadth arc description of the embodiment of the present invention；
The Harris angle point and the schematic diagram with rotational invariance after line segment marshalling that Fig. 4 is inventive embodiments.
Specific embodiment
Understand for the ease of those of ordinary skill in the art and implement the present invention, with reference to the accompanying drawings and embodiments to this hair
It is bright to be described in further detail, it should be understood that implementation example described herein is merely to illustrate and explain the present invention, not
For limiting the present invention.
Referring to Fig.1, a kind of feature based on angle point and the marshalling of single line section provided by the invention describes operator and carries out image
Method of completing the square comprising following steps:
Step 1: input refer to image and image to be matched, based on Gauss it is down-sampled (other algorithms can also be used, such as:
Wavelet decomposition) building multi-level Gaussian image pyramid, it is successively down-sampled to image after to each layer of reference image and to be matched
Image executes step 2-4, calculates best match scale；
Step 2: straightway and Harris angle point are extracted respectively to reference image and image to be matched；
See Fig. 2, Harris angle point is searched within the specified range to the endpoint for the straightway that each is extracted, by single
Straightway marshalling associated with nearest angle point constitutes angle in conjunction with the feature vector of angle point and spectral information description of straightway
Point-line segment half-breadth textural characteristics description；
For straightway, take the rectangular area of its two sides vertical direction equal length as texture description region, and by its
It is divided into m cross-talk region.Obtain the line segment spectral signature description vectors L that a dimension is 2m:
L=(M1,S1,M2,S2..., Mm,Sm)T
M in formulai、SiKnot after every cross-talk area pixel gradient mean value and standard deviation individually normalize respectively in region
Fruit.
See Fig. 3, it is contemplated that the problem of the texture stability of line segment two sides, for each line segment, establish texture region and retouch
Variance calculating is carried out to the pixel value of line segment two sides after stating and shows that the zone-texture stability of the side is poor, pole if variance is larger
It is likely to be in change in depth region.The line segment for being greater than certain threshold value to one-way takes unilateral matched strategy, i.e., only makes
Texture constraint is constructed with half stable in description vectors, leakage matching is reduced with this.At this point, the spectral signature of line segment describes
Vector L becomes:
L=(M1,S1,M2,S2..., Mm/2,Sm/2)T
When extracting Harris angle point, the x in angle point regional area, y direction gradient value X, Y are calculated first, and construct
Angle point receptance function Ex,y:
Wherein,W is the Gauss weight put in angle point regional area, and I is angle
Gray value in point regional area,For convolution algorithm symbol.
By Ex,yIt is written as matrix form, E (x, y)=(x, y) H (x, y)T, whereinCalculate two spies of H
Value indicative [α β] and the smallest characteristic value α correspond to the feature vector [α for the oval major semiaxis that Harris angle point determines1 α2], make
For one of the description vectors factor of angle point.Wherein characteristic value [α β] represents the size of angle point regional area inside gradient variation, and
Feature vector [α1 α2] represent the direction (oval major semiaxis direction) that population gradient changes in angle point regional area.
Step 2.1 is executed to the endpoint for the line segment that each is extracted, if searching Harris angle point within the specified range
Then follow the steps 2.2, emphasis eliminates the Rotation of image using the rotational invariance of Harris angle point, by single line with it is right
It answers angle point to be associated (marshalling), guarantees that angle point is exactly the endpoint of corresponding single line as far as possible.In conjunction with the feature vector and line of angle point
Spectral information description of section constitutes angle point-line segment texture descriptor.
Step 2.1, according to the propinquity between angle point and line segment endpoint, reference image and image to be matched will be extracted respectively
Two endpoints of each line segment angle point is searched in distance threshold t, one or more angle points then follow the steps if it exists
2.2。
In the present embodiment, to reference image and image to be matched it is with two endpoints of each line segment of extraction respectively
The heart opens the circular window that radius is 3 pixels, angle point is searched in window.
Step 2.2: see Fig. 4, taking to establish apart from nearest angle point and single line section and organize into groups, give up other angle points, an angle
Point can correspond to a plurality of line segment, and single line only corresponds to most two angle points.Utilize the rotational invariance of Harris angle point, i.e. feature
Elliptical long axis and the corresponding direction of short axle are constant, determine that " long axis direction " and " line segment direction " rotation (is fixed as side counterclockwise
To) corner, with eliminate two width images between rotational deformation.If the starting point coordinate of straightway is respectively with terminating point coordinate
For [sx sy]、[ex ey], then the angle theta counterclockwise of straight line and Harris angle point major semiaxis gradient direction can pass through following formula meter
It calculates:
Finally by the texture description of the angle theta of the description vectors of angle point [α β], straight line and angle point gradient direction and line segment to
Amount L combines, and forms angle point-line segment textural characteristics description of candidate marshalling, when matched angle point only one when, angle
Point-line segment textural characteristics description are as follows:
L Μ=(M1,S1, M2,S2..., Mm/2,Sm/2,α,β,θ)T
When matched angle point there are two when, angle point-line segment textural characteristics description are as follows:
L Μ=(M1,S1, M2,S2..., Mm/2,Sm/2,α1,β1,θ1,α2,β2,θ2)T
Wherein, α1、β1、θ1、α2、β2、θ2It respectively represents the corresponding feature vector of 2 Harris angle points and arrives the line segment inverse time
The corner of needle.
Step 3: according to the angle point of line segment-line segment half-breadth textural characteristics description to the point of reference image and image to be matched
Line is matched, and candidate matches set is obtained；
It is of different size due to line segment, also change therewith with the weight accounting of angle point.It is m's for a dimension
Half-breadth line segment spectral signature description vectors L, angle point-line segment textural characteristics describe each of son value, arc description
Weighing has with the weight more typically weighting of Harris corner feature:
LHD indicates angle point-corresponding weighted value of line segment textural characteristics description, and should meet:
Its corresponding weight is assigned according to the above rule:
(1) if two angle points of line match, arc description and the corresponding weight of corner description are as follows:
In formulaFor the weight of i-th of line segment neighborhood subregion statistic (including standard deviation and mean value), WHFor angle point spy
The weight of description vectors is levied,For the Gauss weight of i-th of line segment neighborhood subregion；Subregion is remoter apart from line segment, weight
It is lower, the calculation formula of Gauss weight are as follows:
D is distance of i-th of line segment neighborhood subregion to line segment in formula, and is had:
Weight matrix W is stated at this time are as follows:
(2) if line segment only has matched an angle point, the weight of arc description are as follows:
Weight matrix W is stated at this time are as follows:
Angle point-line segment textural characteristics description is L Μ W after considering the final weighting of above-mentioned Gauss powerT。
Step 4: according between image to be matched geometrical relationship calculate candidate matches similarity, including image angle point away from
From the similarity with arc description, to the marshalling screening candidate matches and establishment each line of reference image and image to be matched
The candidate matches matrix M of section i；Established matching matrix M is solved by spectrum analysis, judges that candidate matches are received or refused
Absolutely.
When matching, for each line segment in reference image, sub- weighted value is described into its angle point-line segment textural characteristics,
Then the Euclidean distance with line descriptor for a segment weighted values all in image to be matched for calculating it arranges these Euclidean distances
Sequence, it is assumed that shortest distance is s1, secondary short distance is s2, work as s1With s2, while when meeting the following conditions, take s1Corresponding two
Line segment group is combined into candidate matches pair:
T is indicated apart from threshold limit in formula, only when the shortest distance is less than this threshold value, can just be considered corresponding line
Duan Zuowei candidate matches line segment.tsMost short-secondary short distance rate threshold is indicated, when the shortest distance and time short distance are got too close to
When, illustrate that there is no correct matching line segment, s1With s2It may all be error hiding.
Candidate matches all in two width images are finally calculated, line segment candidate matches set is established:
In formulaRespectively with reference to the corresponding line segment of i-th group of candidate matches in image and image to be matched, n is candidate
Matching is to number.
The texture and Geometrical consistency score between two candidate matches are calculated, and constructs adjacency matrix M, M is considered as one
Non-directed graph solves non-directed graph by spectrum analysis, and screening candidate matches obtain final line match result.
For two pairs of candidate matches M (a, b), whereinUsing following constraint building, its is similar
Property.
(1) cross rate Ii, indicate the registration in two lines section horizontal direction.
(2) Throw ratio Pi, indicate the distance in two lines section vertical direction.
(3) line segment angle Θ, the i.e. angle of two lines section can be found out by line segment angle formulae.
(4) texture paging V, for line segment their texture descriptor in reference image and image to be matched
It respectively indicates are as follows:
Their texture and geometrical constraint can be respectively obtained with after geometrical constraint to (a, b) building texture to candidate matches,With
After obtaining texture and geometrical constraint, so that it may which the texture and Geometrical consistency calculated between matching candidate pair obtains
Divide Mij, by M if the Geometrical change of two groups of candidate matches or texture variations are greater than preset valueijValue be set as 0.MijFor neighbour
Connect the value of the i-th row jth column element in matrix M.
It is calculate by the following formula out texture and Geometrical consistency score Mij:
D in formula Chinese styleI, dp, dΘ,Respectively
Wherein tI,tP,tΘ, t is the threshold value for limiting two groups of candidate matches geometry and texture variations, only works as dI, dp, dΘ,When all less than 1, Γ is true, otherwise texture and Geometrical consistency score MijValue be set as zero.
After constructing adjacency matrix M, line match problem, which is converted to, finds out matching cluster C, so thatMost
Greatly；All candidate matches are represented with vector x, for i-th group of candidate matches, if it, which belongs to, matches cluster C, x (i)=1, instead
It, x (i)=0；Therefore best match solution x*Are as follows:
x*=argmax (xTMx)。
Step 5: the corresponding stereopsis of output Optimum Matching scale obtains the best ruler with reference to image and image to be matched
Degree and its corresponding angle point-matched result of single line section.
It should be understood that the part that this specification does not elaborate belongs to the prior art.
It should be understood that the above-mentioned description for preferred embodiment is more detailed, can not therefore be considered to this
The limitation of invention patent protection range, those skilled in the art under the inspiration of the present invention, are not departing from power of the present invention
Benefit requires to make replacement or deformation under protected ambit, fall within the scope of protection of the present invention, this hair
It is bright range is claimed to be determined by the appended claims.
Claims (7)
1. a kind of feature based on angle point and the marshalling of single line section describes operator and carries out image matching method, which is characterized in that including
Following steps:
Step 1: input refers to image and image to be matched, constructs multi-level Gaussian image pyramid, after successively down-sampled to image
Reference image and image to be matched to each layer execute step 2-4, calculate best match scale；
Step 2: straightway and Harris angle point are extracted respectively to reference image and image to be matched；
Harris angle point is searched within the specified range to the endpoint for the straightway that each is extracted, by single straight line segment and recently
The associated marshalling of angle point, constitute angle point-line segment half-breadth in conjunction with the feature vector of angle point and spectral information description of straightway
Textural characteristics description；
Step 3: according to the angle point of line segment-line segment half-breadth textural characteristics description to the dotted line of reference image and image to be matched into
Row matching, obtains candidate matches set；
Step 4: calculate the similarity of candidate matches according to the geometrical relationship between image to be matched, distance including image angle point and
The similarity of arc description, to the marshalling screening candidate matches and establishment each line segment i of reference image and image to be matched
Candidate matches matrix M；Established matching matrix M is solved by spectrum analysis, judges that candidate matches are received or refused；
Step 5: the corresponding stereopsis of output Optimum Matching scale obtains the best scale with reference to image and image to be matched,
And its corresponding angle point-matched result of single line section.
2. the feature according to claim 1 based on angle point and the marshalling of single line section describes operator and carries out image matching method,
It is down-sampled using Gauss or wavelet decomposition constructs multi-level Gaussian image pyramid it is characterized by: in step 1.
3. the feature according to claim 1 based on angle point and the marshalling of single line section describes operator and carries out image matching method,
It is characterized by:, for straightway, taking the rectangular area of its two sides vertical direction equal length as texture description in step 2
Region, and it is divided into m cross-talk region；Obtain the line segment spectral signature description vectors L that a dimension is 2m:
L=(M1,S1,M2,S2..., Mm,Sm)T；
M in formulai、SiResult after every cross-talk area pixel gradient mean value and standard deviation individually normalize respectively in region；
For each line segment, establishes after texture region description and variance calculating is carried out to the pixel value of line segment two sides, to unilateral side side
The line segment that difference is greater than certain threshold value takes unilateral matched strategy, i.e., constructs line using only half stable in description vectors
Reason constraint reduces leakage matching with this；At this point, the spectral signature description vectors L of line segment becomes:
L=(M1,S1,M2,S2..., Mm/2,Sm/2)T
When extracting Harris angle point, the x in angle point regional area, y direction gradient value X, Y are calculated first, and construct angle point
Receptance function Ex,y:
Ex,y=Ax2+2Cxy+By2
Wherein,W is the Gauss weight put in angle point regional area, and I is angle point office
Gray value in portion region,For convolution algorithm symbol；
By Ex,yIt is written as matrix form, E (x, y)=(x, y) H (x, y)T, whereinCalculate two characteristic values of H
[α β] and the smallest characteristic value α correspond to the feature vector [α for the oval major semiaxis that Harris angle point determines1 α2], as angle
One of the description vectors factor of point；Wherein characteristic value [α β] represents the size of angle point regional area inside gradient variation, and feature
Vector [α1 α2] represent the direction of population gradient variation in angle point regional area, i.e. oval major semiaxis direction.
4. the feature according to claim 3 based on angle point and the marshalling of single line section describes operator and carries out image matching method,
It is characterized in that, constituting angle point-line segment half in conjunction with the spectral information of the feature vector of angle point and line segment description described in step 2
Wide textural characteristics description, is the straightway extracted for each, executes following operating procedure:
Step 2.1: according to the propinquity between angle point and straightway endpoint, to reference image and image to be matched respectively with extraction
Centered on two endpoints of each line segment, angle point is searched in distance threshold t；One or more angle points then execute step if it exists
Rapid 2.2；
Step 2.2: according to being ranked up apart from angle steel joint for angle point to endpoint, taking to establish apart from nearest angle point and the line segment and compile
Group gives up other angle points, and an angle point can correspond to a plurality of line segment, and single line section only corresponds to an angle point；
If the starting point coordinate of straightway and terminating point coordinate are [sx sy]、[ex ey], then straightway and the angle Harris
The angle theta counterclockwise of point major semiaxis gradient direction is calculate by the following formula:
Finally by the texture description of the angle theta of the description vectors of angle point [α β], straightway and angle point gradient direction and straightway to
Amount L combines, and forms angle point-straightway textural characteristics description of candidate marshalling；
Wherein: when matched angle point only one when, angle point-line segment half-breadth textural characteristics description are as follows:
L Μ=(M1,S1, M2,S2..., Mm/2,Sm/2,α,β,θ)T；
When matched angle point there are two when, angle point-line segment half-breadth textural characteristics description are as follows:
L Μ=(M1,S1, M2,S2..., Mm/2,Sm/2,α1,β1,θ1,α2,β2,θ2)T；
Wherein, α1、β1、θ1、α2、β2、θ2Respectively represent the corresponding feature vector of 2 Harris angle points and counterclockwise to line segment
Corner.
5. the feature according to claim 4 based on angle point and the marshalling of single line section describes operator and carries out image matching method,
It is characterized by: in step 3, the half-breadth line segment spectral signature description vectors L for being m for a dimension, angle point-straightway line
Each of Feature Descriptor value is managed, the power of straight line segment description and the weight more typically weighting of Harris corner feature have:
LHD indicates angle point-corresponding weighted value of line segment textural characteristics description, and should meet:
Its corresponding weight is assigned according to the above rule:
(1) if two angle points of line match, arc description and the corresponding weight of corner description are as follows:
In formulaFor the weight of i-th of line segment neighborhood subregion statistic, WHFor the weight of corner feature description vectors；Due to son
Region distance line segment is remoter, and weight is lower, the calculation formula of Gauss weight are as follows:
D is distance of i-th of line segment neighborhood subregion to line segment in formula, and is had:
In formulaFor the Gauss weight of i-th of line segment neighborhood subregion；
Weight matrix W is stated at this time are as follows:
(2) if line segment only has matched an angle point, the weight of arc description are as follows:
Weight matrix W is stated at this time are as follows:
Angle point-line segment textural characteristics description is L Μ W after considering the final weighting of above-mentioned Gauss powerT。
6. the feature according to claim 5 based on angle point and the marshalling of single line section describes operator and carries out image matching method,
It is characterized by: the similarity of candidate matches is calculated described in step 4 according to the geometrical relationship between image to be matched, when matching,
For each straightway in reference image, its angle point-straightway textural characteristics are described into sub- weighted value, then calculate its
With the Euclidean distance of line descriptor for a segment weighted values all in image to be matched, sort to these Euclidean distances, it is assumed that shortest
Distance is s1, secondary short distance is s2, work as s1With s2, while when meeting the following conditions, take s1Corresponding two lines section group is combined into
Candidate matches pair:
T is indicated apart from threshold limit in formula, only when the shortest distance is less than this threshold value, can just consider to make on corresponding line segment
For candidate matches line segment；
Candidate matches all in two width images are finally calculated, line segment candidate matches set is established:
In formulaRespectively with reference to the corresponding line segment of i-th group of candidate matches in image and image to be matched, n is candidate matches
To number；
The texture and Geometrical consistency score between two candidate matches are calculated, and constructs adjacency matrix M, M is considered as a Zhang Wuxiang
Figure solves non-directed graph by spectrum analysis, and screening candidate matches obtain final line match result；
For two pairs of candidate matches M (a, b), whereinIts similitude is constructed using following constraint:
(1) cross rate Ii, indicate the registration in two lines section horizontal direction；
(2) Throw ratio Pi, indicate the distance in two lines section vertical direction；
(3) line segment angle Θ, the i.e. angle of two lines section can be found out by line segment angle formulae；
(4) texture paging V respectively indicates reference image and line segment their texture descriptor in image to be matched
Are as follows:
Their texture and geometrical constraint are respectively obtained with after geometrical constraint to (a, b) building texture to candidate matches,With
After obtaining texture and geometrical constraint, the texture and Geometrical consistency score M between matching candidate pair are calculatedijIf
The Geometrical change or texture variations of two groups of candidate matches are greater than preset value then by MijValue be set as 0；MijFor in adjacency matrix M
The value of i-th row jth column element；After constructing adjacency matrix M, line match problem, which is converted to, finds out matching cluster C, so thatIt is maximum；
All candidate matches are represented with vector x, for i-th group of candidate matches, if it, which belongs to, matches cluster C, x (i)=1,
Conversely, x (i)=0；Therefore best match solution x*Are as follows:
x*=argmax (xTMx)。
7. the feature according to claim 6 based on angle point and the marshalling of single line section describes operator and carries out image matching method,
It is characterized by: being calculate by the following formula out texture and Geometrical consistency score Mij:
D in formula Chinese styleI, dp, dΘ,Respectively
Wherein tI,tP,tΘ, t is the threshold value for limiting two groups of candidate matches geometry and texture variations, only works as dI, dp, dΘ,
When all less than 1, Γ is true, otherwise texture and Geometrical consistency score MijValue be set as zero.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910660833.3A CN110490913B (en) | 2019-07-22 | 2019-07-22 | Image matching method based on feature description operator of corner and single line segment grouping |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910660833.3A CN110490913B (en) | 2019-07-22 | 2019-07-22 | Image matching method based on feature description operator of corner and single line segment grouping |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110490913A true CN110490913A (en) | 2019-11-22 |
| CN110490913B CN110490913B (en) | 2022-11-22 |
Family
ID=68547833
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910660833.3A Active CN110490913B (en) | 2019-07-22 | 2019-07-22 | Image matching method based on feature description operator of corner and single line segment grouping |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110490913B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111898646A (en) * | 2020-07-06 | 2020-11-06 | 武汉大学 | Cross-view image straight line feature matching method based on point line graph optimization solution |
| CN112163622A (en) * | 2020-09-30 | 2021-01-01 | 山东建筑大学 | Overall situation and local fusion constrained line segment feature matching method for aviation wide-baseline stereopair |
| CN113095384A (en) * | 2021-03-31 | 2021-07-09 | 安徽工业大学 | Remote sensing image matching method based on context characteristics of straight line segments |
| CN113378507A (en) * | 2021-06-01 | 2021-09-10 | 中科晶源微电子技术（北京）有限公司 | Mask data cutting method and device, equipment and storage medium |
| CN113538503A (en) * | 2021-08-21 | 2021-10-22 | 西北工业大学 | Solar panel defect detection method based on infrared image |
| CN114012736A (en) * | 2021-12-08 | 2022-02-08 | 北京云迹科技有限公司 | Positioning object for assisting environment positioning and robot system |
| CN116309837A (en) * | 2023-03-16 | 2023-06-23 | 南京理工大学 | Method for identifying and positioning damaged element by combining characteristic points and contour points |
| CN117114971A (en) * | 2023-08-01 | 2023-11-24 | 北京城建设计发展集团股份有限公司 | Pixel map-to-vector map conversion method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184801A (en) * | 2015-09-28 | 2015-12-23 | 武汉大学 | Optical and SAR image high-precision registration method based on multilevel strategy |
| WO2017049994A1 (en) * | 2015-09-25 | 2017-03-30 | 深圳大学 | Hyperspectral image corner detection method and system |
| CN106709870A (en) * | 2017-01-11 | 2017-05-24 | 辽宁工程技术大学 | Close-range image straight-line segment matching method |
-
2019
- 2019-07-22 CN CN201910660833.3A patent/CN110490913B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017049994A1 (en) * | 2015-09-25 | 2017-03-30 | 深圳大学 | Hyperspectral image corner detection method and system |
| CN105184801A (en) * | 2015-09-28 | 2015-12-23 | 武汉大学 | Optical and SAR image high-precision registration method based on multilevel strategy |
| CN106709870A (en) * | 2017-01-11 | 2017-05-24 | 辽宁工程技术大学 | Close-range image straight-line segment matching method |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111898646B (en) * | 2020-07-06 | 2022-05-13 | 武汉大学 | Cross-view image straight line feature matching method based on point-line graph optimization solution |
| CN111898646A (en) * | 2020-07-06 | 2020-11-06 | 武汉大学 | Cross-view image straight line feature matching method based on point line graph optimization solution |
| CN112163622A (en) * | 2020-09-30 | 2021-01-01 | 山东建筑大学 | Overall situation and local fusion constrained line segment feature matching method for aviation wide-baseline stereopair |
| CN112163622B (en) * | 2020-09-30 | 2022-07-05 | 山东建筑大学 | Global and local fusion constrained aviation wide-baseline stereopair line segment matching method |
| CN113095384A (en) * | 2021-03-31 | 2021-07-09 | 安徽工业大学 | Remote sensing image matching method based on context characteristics of straight line segments |
| CN113095384B (en) * | 2021-03-31 | 2023-04-28 | 安徽工业大学 | Remote sensing image matching method based on linear segment context characteristics |
| CN113378507A (en) * | 2021-06-01 | 2021-09-10 | 中科晶源微电子技术（北京）有限公司 | Mask data cutting method and device, equipment and storage medium |
| CN113378507B (en) * | 2021-06-01 | 2023-12-05 | 中科晶源微电子技术（北京）有限公司 | Mask data cutting method and device, equipment and storage medium |
| CN113538503A (en) * | 2021-08-21 | 2021-10-22 | 西北工业大学 | Solar panel defect detection method based on infrared image |
| CN113538503B (en) * | 2021-08-21 | 2023-09-01 | 西北工业大学 | Solar panel defect detection method based on infrared image |
| CN114012736A (en) * | 2021-12-08 | 2022-02-08 | 北京云迹科技有限公司 | Positioning object for assisting environment positioning and robot system |
| CN116309837A (en) * | 2023-03-16 | 2023-06-23 | 南京理工大学 | Method for identifying and positioning damaged element by combining characteristic points and contour points |
| CN116309837B (en) * | 2023-03-16 | 2024-04-26 | 南京理工大学 | Method for identifying and positioning damaged element by combining characteristic points and contour points |
| CN117114971A (en) * | 2023-08-01 | 2023-11-24 | 北京城建设计发展集团股份有限公司 | Pixel map-to-vector map conversion method and system |
| CN117114971B (en) * | 2023-08-01 | 2024-03-08 | 北京城建设计发展集团股份有限公司 | Pixel map-to-vector map conversion method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110490913B (en) | 2022-11-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110490913A (en) | Feature based on angle point and the marshalling of single line section describes operator and carries out image matching method | |
| Liew et al. | Image segmentation based on adaptive cluster prototype estimation | |
| CN110210276A (en) | A kind of motion track acquisition methods and its equipment, storage medium, terminal | |
| CN108268859A (en) | A kind of facial expression recognizing method based on deep learning | |
| CN108491786B (en) | Face detection method based on hierarchical network and cluster merging | |
| CN110378997A (en) | A kind of dynamic scene based on ORB-SLAM2 builds figure and localization method | |
| CN109903313A (en) | Real-time pose tracking method based on target three-dimensional model | |
| CN113095263B (en) | Training method and device for pedestrian re-recognition model under shielding and pedestrian re-recognition method and device under shielding | |
| CN113963032A (en) | Twin network structure target tracking method fusing target re-identification | |
| CN109886267A (en) | A kind of soft image conspicuousness detection method based on optimal feature selection | |
| CN109447036A (en) | A kind of segmentation of image digitization and recognition methods and system | |
| CN109559273A (en) | A kind of quick joining method towards vehicle base map picture | |
| CN109754440A (en) | A kind of shadow region detection method based on full convolutional network and average drifting | |
| CN111009005A (en) | Scene classification point cloud rough registration method combining geometric information and photometric information | |
| CN107154017A (en) | A kind of image split-joint method based on SIFT feature Point matching | |
| CN105809678B (en) | A kind of line segment feature global registration method between two views under short base line condition | |
| CN108364300A (en) | Vegetables leaf portion disease geo-radar image dividing method, system and computer readable storage medium | |
| CN109101985A (en) | It is a kind of based on adaptive neighborhood test image mismatch point to elimination method | |
| CN106529441A (en) | Fuzzy boundary fragmentation-based depth motion map human body action recognition method | |
| CN108334870A (en) | The remote monitoring system of AR device data server states | |
| CN109344758B (en) | Face recognition method based on improved local binary pattern | |
| CN113011359B (en) | Method for simultaneously detecting plane structure and generating plane description based on image and application | |
| CN112329662B (en) | Multi-view saliency estimation method based on unsupervised learning | |
| CN111445540B (en) | Automatic registration method for RGB colored three-dimensional point cloud | |
| CN108446639A (en) | Low-power consumption augmented reality equipment |
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 |