CN105551058B - A kind of combination SURF feature extractions and the cylindrical picture matching process of curve matching - Google Patents
A kind of combination SURF feature extractions and the cylindrical picture matching process of curve matching Download PDFInfo
- Publication number
- CN105551058B CN105551058B CN201610074411.4A CN201610074411A CN105551058B CN 105551058 B CN105551058 B CN 105551058B CN 201610074411 A CN201610074411 A CN 201610074411A CN 105551058 B CN105551058 B CN 105551058B
- Authority
- CN
- China
- Prior art keywords
- matching
- curve
- euclidean distance
- image
- angle
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
The present invention relates to a kind of combination SURF feature extractions and the cylindrical picture matching process of curve matching, step are as follows：By two images A and B to be matched in the way of left-justify more than the next placement location；Feature detection is carried out to two images using SURF feature detection algorithms respectively；Matching is found out into set；Feature Points Matching is calculated respectively to angle and Euclidean distance the angle of the straight line determined and horizontal direction is each matched in set and each matches two characteristic points of centering, establishes the angle set K of images match pair and the Euclidean distance set D of image；The angle set K of matching pair is carried out curve fitting, the independent variable of curve matching is the abscissa X of matching pair1i, the ordinate of curve matching is angle, θi, reject error hiding.The Euclidean distance set D of matching pair is carried out curve fitting, the independent variable of curve matching is abscissa X of the matching to the characteristic point in image A1i, reject error hiding.The present invention can more accurately reject the error hiding of cylinder object.
Description
Technical field
The invention belongs to digital image processing techniques field, is related to a kind of cylindrical picture matching process.
Background technology
With the progress of society, requirement of the people to industrial production efficiency also more and more higher, manually completed in the past
Production due to efficiency is low, cost is high, can no longer meet the needs of today's society.In order to improve production efficiency,
Production line of today is more and more intelligent, and the intelligentized important foundation of production line is exactly image matching technology, is mainly used in
The registration of product, identification, error detection.Image matching technology using it is relatively more be feature based image matching technology, tool
There is the advantages of robustness is high, calculating speed is fast.But the image matching algorithm of feature based due to characteristic point by noise, extraction and
The influence of matching way, certain error hiding can occur, it is all to linear change occurs mostly that traditional error hiding, which rejects algorithm,
The image changed carries out error hiding rejecting, and it is less that the error hiding for nonlinear transformation object occurs rejects algorithm.
The content of the invention
It is an object of the invention to provide a kind of matching of the cylindrical picture of error hiding that can more accurately reject cylinder object
Method.Because error hiding easily occur in actual use in SURF Feature Correspondence Algorithms, method provided by the invention can be realized
Error hiding after cylindrical picture SURF slightly matching is rejected, while correct matching of the and can to nonlinear transformation region is to carrying out
Retain to greatest extent, so as to realize the raising to the cylindrical picture matching precision of nonlinear transformation.Technical scheme
It is as follows：
A kind of combination SURF feature extractions and the cylindrical picture matching process of curve matching, step are as follows：
1) by two images A and B to be matched in the way of left-justify more than the next placement location；
2) feature detection is carried out to two images using SURF feature detection algorithms respectively；
3) characteristic point that all characteristic points of other in image A are matched in image B is found out, forms matching to set H1；
4) according to matching to set H1Ascending suitable of included characteristic point abscissa location of pixels in image A
Sequence, to matching to resequencing, constitutive characteristic Point matching is to set H2If matching is to the pixel in image A and image B
Position is respectively (x1i,y1i) and (x2i,y2i), then Feature Points Matching is to set H2For 4*N array, wherein N is matching pair
Quantity, the matching pair of array each column element representation one.
Feature Points Matching is calculated respectively to set H2In the straight line that determines of each matching pair and an angle, θ for horizontal directioniAnd
Angle, θ between each matching two characteristic points of centeringiWith Euclidean distance di, establish the angle set K and image of images match pair
Euclidean distance set D, preserves angle, θ respectivelyiWith Euclidean distance diAbscissa with corresponding matching to the characteristic point in image A
X1iBetween relation；
5) the angle set K by matching pair carries out curve fitting, and the independent variable of curve matching is the abscissa X of matching pair1i,
The ordinate of curve matching is angle, θi, the number of curve matching is set to three times, to draw angle matched curve f1(X), then will
The abscissa X of each characteristic point1iSubstitute into f1(X) in, the angle for meeting this abscissa in curve is obtained, and calculates this angle and phase
The angle, θ answerediDifference absolute value, when more than a certain particular value TθWhen, then it is assumed that this is paired into error hiding, is rejected.
6) the Euclidean distance set D by matching pair carries out curve fitting, and the independent variable of curve matching is to match in image A
The abscissa X of middle characteristic point1i, the ordinate of curve matching is matches the Euclidean distance of two characteristic points of centering, by curve matching
Number be set to three times, draw Euclidean distance matched curve f2(X), then, for each characteristic point, by its abscissa X1iSubstitute into
f2(X) in, the Euclidean distance for meeting this matched curve is calculated, and ask for this Euclidean distance and corresponding Euclidean distance diDifference
Absolute value, when more than a certain particular value TdWhen, then it is assumed that this Feature Points Matching is to for error hiding, being rejected.
Wherein, the method for step 3) can be as follows：To each characteristic point A in image A1i, calculate and all spies in image B
Levy the Euclidean distance of the characteristic vector of point and find out nearest characteristic point B1iWith secondary near characteristic point B2i, and obtain European recently
Distance and time nearly Euclidean distance, set threshold value T of some value between 0.6~0.8, when nearest Euclidean distance and time near European
When the ratio of distance is less than this threshold value T, it is believed that characteristic point A1iWith characteristic point B1iMatching, otherwise it is assumed that in image B be not present with
A1iThe characteristic point of matching.
The present invention by respectively by match to the straight angle of institute's shape and matching two characteristic points it is European away from
From carrying out curve fitting, images match is limited to position relationship according to trend, can not only be realized to all areas of cylindrical picture
Domain does not meet the rejecting of the error hiding of such a trend, and for traditional matching algorithm, prior improvement is to protect
Demonstrate,prove reservation of the cylindrical picture in the correct matching pair that nonlinear transformation region occurs.It is as can be seen that proposed by the present invention based on song
The error hiding of line fitting rejects algorithm and the error hiding of cylindrical picture can be rejected well, so as to improve cylindrical picture
The precision matched somebody with somebody.
Brief description of the drawings
The cylindrical picture of Fig. 1 combination SURF feature extractions and curve matching matches flow chart
The mode that Fig. 2 images to be matched are placed up and down
The thick matching result of Fig. 3 arest neighbors
Fig. 4 angle curve fitting results
Fig. 5 Euclidean distance curve-fitting results
Fig. 6 combination SURF feature extractions and the cylindrical picture matching result of curve matching
Embodiment
During cylindrical picture registration, it is necessary first to gather two width cylindrical pictures, the cylindrical picture of collection due to regarding
The reasons such as angle, position, the transformation relation between two images is a kind of nonlinear transformation relation.The first step of two images matching
Be first using SURF feature detection algorithms to two images carry out SURF feature extractions, then using characteristic vector recently with it is secondary
The ratio between nearly Euclidean distance determines the thick matching pair between two images.Due to SURF in actual use, it may appear that certain mistake
Matching, the plane picture of linear transformation occurs for tradition, the single linear transformation model between two images can be calculated, looked for
Go out to be unsatisfactory for the matching pair of this model, i.e. error hiding, realize the rejecting to Linear Transformations Image error hiding.It is but non-for occurring
The cylindrical picture of linear transformation, the transformation model between two images is not single, is become with the difference of region
Change, now using traditional linear error hiding reject algorithm will cause substantial amounts of correct matching to be taken as error hiding and by
Reject, cause matching precision very low.Present invention proposition is a kind of to use curve matching on the basis of SURF arest neighbors slightly matching
Cylindrical picture error hiding rejects algorithm, comprises the following steps that：
1) two images A and B to be matched is placed according to upper and lower mode, as shown in Figure 2.
2) feature detection is carried out to two images using SURF feature detection algorithms respectively, to each feature in image A
Point A1i, calculate the characteristic vector Euclidean distance with all characteristic points in B and find out nearest characteristic point B1iWith secondary near characteristic point
B2i, as L (A1i,B1i)/L(A1i,B2i) ＜ T when, it is believed that characteristic point A1iWith characteristic point B1iMatching, otherwise it is assumed that in B be not present with
A1iThe characteristic point of matching, wherein L (A, B) to calculate the Euclidean distance of two characteristic point characteristic vectors, threshold value T often takes 0.6~
0.8。
3) characteristic point that all characteristic points of other in image A are matched in image B is found out successively, forms matching to set
H1, the result that two width cylindrical pictures are slightly matched using arest neighbors is as shown in Figure 3.
4) according to set H1Included characteristic point ascending order of abscissa location of pixels in image A, to
Pairing is resequenced, and constitutive characteristic Point matching is to set H2If matching is to the location of pixels in image A and image B point
It is not：(x1i,y1i) and (x2i,y2i), x represents the columns where pixel, and y represents the line number where pixel.Then image characteristic point
Match set H2For 4*N array, as shown in Equation 1, wherein N is the quantity of matching pair, array each column element representation one
Matching pair.
5) difference set of computations H2In the straight line that determines of each matching pair and horizontal direction angle and each match pair
In Euclidean distance between two characteristic points.The present invention uses the method for placing two images up and down, if image A sizes are m1×n1,
Image B sizes are m2×n2, then matching is to the location of pixels difference (X in figure1i,Y1j) and (X2i,Y2j) determined respectively by formula 2：
Match the slope k between centering any two characteristic pointiWith Euclidean distance diCalculation as shown in formula 3 and 4, in order to
The convenience of calculating and solves the problems, such as vertical curve slope infinity, the present invention with squared difference and respectively generation using the angle of straight line
For slope and Euclidean distance, as shown in formula 5 and 6.Then the angle set K of images match pair and the Euclidean distance collection of image are established
D is closed, its relation between image abscissa is preserved respectively, as shown in formula 7 and 8.
θi=arctan (ki) (5)
d′i=(Y2i-Y1i)2+(X2i-X1i)2 (6)
6) the straight line angle set K by matching pair carries out curve fitting, and the independent variable of curve matching is to match in image A
The abscissa X of middle characteristic point1i, the ordinate of curve matching is straight line angle of the matching to formation, and the number of curve matching is set
Three times, to draw angle matched curve f1(X), as shown in figure 4, then by the abscissa X of each characteristic point1iSubstitute into f1(X) in,
The angle for meeting this abscissa in curve is obtained, and calculates the absolute value with the difference of respective angles, when more than a certain particular value Tθ
When, then it is assumed that this is paired into error hiding, is rejected.
7) the Euclidean distance set D by matching pair carries out curve fitting, and the independent variable of curve matching is to match in image A
The abscissa X of middle characteristic point1i, the ordinate of curve matching is matches the Euclidean distance of two characteristic points of centering, by curve matching
Number be set to three times, draw Euclidean distance matched curve f2(X), as shown in figure 5, then by the abscissa X of each characteristic point1i
Substitute into f2(X) in, the Euclidean distance for this feature Point matching pair for meeting this matched curve, and corresponding Euclidean distance of summing are calculated
Difference absolute value, when more than a certain particular value TdWhen, then it is assumed that this Feature Points Matching is to for error hiding, being rejected.
The straight angle of institute's shape and matching are carried out the Euclidean distance two characteristic points by will match respectively
Curve matching, images match is limited to position relationship according to trend, can not only realize and cylindrical picture all areas are not inconsistent
The rejecting of the error hiding of such a trend is closed, for traditional matching algorithm, prior improvement is can to ensure cylinder
Reservation of the image in the correct matching pair that nonlinear transformation region occurs.As can be seen that proposed by the present invention be based on curve matching
Error hiding reject algorithm the error hiding of cylindrical picture can be rejected well, so as to improve cylindrical picture match essence
Degree.
Claims (2)
1. a kind of combination SURF feature extractions and the cylindrical picture matching process of curve matching, step are as follows：
1) by two images A and B to be matched in the way of left-justify more than the next placement location；
2) feature detection is carried out to two images using SURF feature detection algorithms respectively；
3) characteristic point that all characteristic points are matched in image B in image A is found out, forms matching to set H1；
4) according to matching to set H1Included characteristic point ascending order of abscissa location of pixels in image A, to
Pairing is resequenced, and constitutive characteristic Point matching is to set H2If matching is to the location of pixels in image A and image B point
Wei not (x1i,y1i) and (x2i,y2i), then Feature Points Matching is to set H2For 4*N array, wherein N is the number of matching pair
Amount, one matching pair of array each column element representation；
Feature Points Matching is calculated respectively to set H2In the straight line that determines of each matching pair and an angle, θ for horizontal directioniIt is and each
Euclidean distance d between matching two characteristic points of centeringi, establish the angle set K and image of images match pair Euclidean distance set
D, angle, θ is preserved respectivelyiWith Euclidean distance diAbscissa X with corresponding matching to the characteristic point in image A1iBetween pass
System；
5) the angle set K by matching pair carries out curve fitting, and the independent variable of curve matching is the abscissa X of matching pair1i, curve
The ordinate of fitting is angle, θi, the number of curve matching is set to three times, to draw angle matched curve f1(X), then will be each
The abscissa X of characteristic point1iSubstitute into f1(X) in, the angle for meeting this abscissa in curve is obtained, and calculates this angle and step 4)
The angle, θ preservediDifference absolute value, when more than a certain particular value TθWhen, then it is assumed that this is paired into error hiding, is picked
Remove；
6) the Euclidean distance set D by matching pair carries out curve fitting, and the independent variable of curve matching is to match to special in image A
Levy the abscissa X of point1i, the ordinate of curve matching is matches the Euclidean distance of two characteristic points of centering, by time of curve matching
Number is set to three times, draw Euclidean distance matched curve f2(X), then, for each characteristic point, by its abscissa X1iSubstitute into f2
(X) in, calculate and meet the Euclidean distance of this matched curve, and ask for this Euclidean distance and Euclidean distance d that step 4) is preservedi
Difference absolute value, when more than a certain particular value TdWhen, then it is assumed that this Feature Points Matching is to for error hiding, being rejected.
2. matching process according to claim 1, the method for step 3) can be as follows：To each characteristic point in image A
A1i, calculate with image B in all characteristic points characteristic vector Euclidean distance and find out nearest characteristic point B1iWith secondary near spy
Levy point B2i, and nearest Euclidean distance and time nearly Euclidean distance are obtained, threshold value T of some value between 0.6~0.8 is set, when
When the ratio of nearest Euclidean distance and time nearly Euclidean distance is less than this threshold value T, it is believed that characteristic point A1iWith characteristic point B1iMatching, it is no
Then, it is believed that be not present in image B and A1iThe characteristic point of matching.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610074411.4A CN105551058B (en) | 2016-02-02 | 2016-02-02 | A kind of combination SURF feature extractions and the cylindrical picture matching process of curve matching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610074411.4A CN105551058B (en) | 2016-02-02 | 2016-02-02 | A kind of combination SURF feature extractions and the cylindrical picture matching process of curve matching |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105551058A CN105551058A (en) | 2016-05-04 |
| CN105551058B true CN105551058B (en) | 2018-04-03 |
Family
ID=55830234
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610074411.4A Expired - Fee Related CN105551058B (en) | 2016-02-02 | 2016-02-02 | A kind of combination SURF feature extractions and the cylindrical picture matching process of curve matching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105551058B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106023187B (en) * | 2016-05-17 | 2019-04-19 | 西北工业大学 | A kind of method for registering images based on SIFT feature and angle relative distance |
| CN106778510B (en) * | 2016-11-25 | 2020-03-20 | 江西师范大学 | Method for matching high-rise building characteristic points in ultrahigh-resolution remote sensing image |
| CN109117854B (en) * | 2018-07-25 | 2021-01-29 | 北京达佳互联信息技术有限公司 | Key point matching method and device, electronic equipment and storage medium |
| CN110020646B (en) * | 2019-04-16 | 2021-07-27 | 恒生电子股份有限公司 | File archiving method and device, electronic equipment and storage medium |
| CN110443128B (en) * | 2019-06-28 | 2022-12-27 | 广州中国科学院先进技术研究所 | Finger vein identification method based on SURF feature point accurate matching |
| CN114061501A (en) * | 2020-08-06 | 2022-02-18 | 广东博智林机器人有限公司 | Frame upright column parameter measuring method and device, electronic equipment and storage medium |
| CN113658080B (en) * | 2021-08-23 | 2023-12-22 | 宁波棱镜空间智能科技有限公司 | Linear scanning cylinder geometric correction method and device based on characteristic point matching |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7463773B2 (en) * | 2003-11-26 | 2008-12-09 | Drvision Technologies Llc | Fast high precision matching method |
| CN103217438A (en) * | 2013-04-02 | 2013-07-24 | 天津大学 | Accurate circuit board element location and detection method based on image feature |
| CN103400388A (en) * | 2013-08-06 | 2013-11-20 | 中国科学院光电技术研究所 | Method for eliminating Brisk (binary robust invariant scale keypoint) error matching point pair by utilizing RANSAC (random sampling consensus) |
| CN103530872A (en) * | 2013-09-18 | 2014-01-22 | 北京理工大学 | Mismatching deleting method based on angle constraint |
| CN104574421A (en) * | 2015-01-29 | 2015-04-29 | 北方工业大学 | Large-breadth small-overlapping-area high-precision multispectral image registration method and device |
-
2016
- 2016-02-02 CN CN201610074411.4A patent/CN105551058B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7463773B2 (en) * | 2003-11-26 | 2008-12-09 | Drvision Technologies Llc | Fast high precision matching method |
| CN103217438A (en) * | 2013-04-02 | 2013-07-24 | 天津大学 | Accurate circuit board element location and detection method based on image feature |
| CN103400388A (en) * | 2013-08-06 | 2013-11-20 | 中国科学院光电技术研究所 | Method for eliminating Brisk (binary robust invariant scale keypoint) error matching point pair by utilizing RANSAC (random sampling consensus) |
| CN103530872A (en) * | 2013-09-18 | 2014-01-22 | 北京理工大学 | Mismatching deleting method based on angle constraint |
| CN104574421A (en) * | 2015-01-29 | 2015-04-29 | 北方工业大学 | Large-breadth small-overlapping-area high-precision multispectral image registration method and device |
Non-Patent Citations (7)
| Title |
|---|
| Action Recognition Algorithm Based on Optical Flow and RANSAC in Frequency Domain;Hafiz Imtiaz 等;《SICE Annual Conference 2011》;20110918;1627-1631 * |
| Face Detection using SURF Cascade;Jianguo Li 等;《2011 IEEE International Conference on Computer Vision Workshops》;20111231;2183-2190 * |
| Fast Image Matching Based-on Improved SURF Algorithm;Zhang Huijuan 等;《Electronics, Communications and Control (ICECC), 2011 International Conference on》;20111103;1460-1463 * |
| Fully affine invariant SURF for image matching;Yanwei Pang 等;《Neurocomputing》;20121231;第85卷;6-10 * |
| Influence of Cylindrical Inclusion on the Elastic Field of Soft Material Loaded by Wedge Indenter;XiaoHua Tan 等;《Applied Mechanics and Materials》;20141008;第670-671卷;664-667 * |
| 一种基于曲线拟合的柱面图像匹配算法;田健飞;《电子测量技术》;20160215;第39卷(第2期);61-68 * |
| 基于SURF算法的电路板元件定位及检测;肖夏 等;《天津大学学报(自然科学与工程技术版)》;20140930;第47卷(第9期);765-770 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105551058A (en) | 2016-05-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105551058B (en) | A kind of combination SURF feature extractions and the cylindrical picture matching process of curve matching | |
| CN111080529A (en) | Unmanned aerial vehicle aerial image splicing method for enhancing robustness | |
| CN104036480B (en) | Quick elimination Mismatching point method based on surf algorithm | |
| CN105654421B (en) | Based on the projective transformation image matching method for converting constant low-rank texture | |
| CN109671110B (en) | Local geometric structure constrained urban wide baseline image feature point matching method | |
| CN104167003A (en) | Method for fast registering remote-sensing image | |
| CN101833767B (en) | Gradient and color characteristics-based automatic straight line matching method in digital image | |
| CN108550166B (en) | Spatial target image matching method | |
| EP3550467B1 (en) | Image matching method, device and system, and storage medium | |
| CN105550994B (en) | A kind of quick outline joining method of unmanned plane image based on satellite image | |
| CN102324134A (en) | Valuable document identification method and device | |
| CN106709870B (en) | Close-range image straight-line segment matching method | |
| CN108229500A (en) | A kind of SIFT Mismatching point scalping methods based on Function Fitting | |
| CN107862319A (en) | A kind of heterologous high score optical image matching error elimination method based on neighborhood ballot | |
| CN103914690A (en) | Shape matching method based on projective invariant | |
| CN107016414A (en) | A kind of recognition methods of footprint | |
| CN103886338B (en) | A kind of multispectral image matching process and system | |
| CN107038401A (en) | Lip contour segmentation and feature extraction method | |
| CN110288533B (en) | Rapid splicing method of non-rotating images | |
| CN110991501B (en) | Improved ORB feature point matching method based on Hessian matrix | |
| CN113255449A (en) | Real-time matching method of binocular video images | |
| CN106408023B (en) | Image characteristic point two valued description and matching process based on group comparison strategy | |
| CN113095385A (en) | Multimode image matching method based on global and local feature description | |
| CN103208003B (en) | Geometric graphic feature point-based method for establishing shape descriptor | |
| CN104036494A (en) | Fast matching computation method used for fruit picture |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180403 Termination date: 20220202 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |