CN103714544A - Optimization method based on SIFT feature point matching - Google Patents
Optimization method based on SIFT feature point matching Download PDFInfo
- Publication number
- CN103714544A CN103714544A CN201310733191.8A CN201310733191A CN103714544A CN 103714544 A CN103714544 A CN 103714544A CN 201310733191 A CN201310733191 A CN 201310733191A CN 103714544 A CN103714544 A CN 103714544A
- Authority
- CN
- China
- Prior art keywords
- point
- point set
- matching
- value
- coupling
- 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
Images
Abstract
The invention discloses an optimization method based on SIFT feature point matching. The optimization method includes the steps that image feature points and descriptors of the image feature points are extracted through the SIFT algorithm, cross matching is carried out on Euclidean distances of the descriptors through the K-nearest node algorithm, the test specific value of each point in a basic point set is calculated, the coordinates of vanishing points are fitted based on coordinate information, and checking is carried out through the vanishing points. By means of the optimization method, wrong matching points generated when only scale space is considered in some SIFT matching methods are avoided.
Description
Technical field:
The present invention the invention belongs to the technical field of image matching in image processing and area of pattern recognition, be specifically related to a kind of optimization method based on SIFT (Scale-invariant feature transform, i.e. yardstick invariant features conversion) Feature Points Matching.
Background technology:
Image matching technology is mainly crossed the corresponding relation to presentation content, feature, structure, relation, texture and gray scale etc., and similarity and consistency analysis, seek same image order calibration method.Image matching technology can be divided into: the matching technique based on gradation of image, the matching technique based on characteristics of image, the matching technique based on template matches and the matching technique based on transform domain.But without spin in the situation that, matching result has much room for improvement at image to be matched.
Summary of the invention:
The object of the invention is to overcome the deficiencies in the prior art, a kind of optimization method based on SIFT Feature Points Matching is provided.
In order to solve the existing problem of background technology, the present invention by the following technical solutions:
An optimization method based on SIFT Feature Points Matching, it comprises the steps:
(1) read image to be matched;
(2 use yardstick invariant features transfer algorithms extract image characteristic point and descriptor thereof, with 128 dimensional vectors, represent;
(3) the neighbouring node algorithm of K carries out cross-matched to the Euclidean distance of descriptor, first by the feature of the right figure of feature profile matching of left figure, describe, then with the feature description of the left figure of feature profile matching of right figure; Choose the identical point of coupling twice, as the basic point set of coupling;
(4) calculate each some ratio test value of basic point set, the point that on left figure, unique point is mated with right figure most, the point of right figure coupling, makes respectively difference and compares at work; In like manner the point of right figure is done to identical point and do identical operation, get the mean value of two ratios;
(5) ratio test value, as SIFT unique point can not reliability, ratio test value is more little more credible.
(6) get part with a high credibility, value is between 0 to 0.3, to be standard point set here, and value is between 0.3 to 0.8, to be point set undetermined;
(7) to standard point set in same coordinate system, by the mode of straight line, describe;
(8) successively every straight line is asked the intersection point of all the other straight lines and this straight line, in these intersection points, got distance two intersection points farthest as the eigenwert of this straight line;
(9) two of cut-off line eigenwert minimum as optimal straight line, usings the intersection point of these two straight lines as the shadow point that disappears.
(10) it is starting point that each point for the treatment of fixed-point set solves left figure point coordinate, the shadow point that disappears for terminal and left figure point coordinate be starting point, the cosine value that right figure point coordinate is terminal;
(11) get cosine value approach 1 for correct point, get here and be greater than 0.99, form correct point set;
(12) final coupling point set is comprised of standard point set and correct coupling point set.
The present invention contrasts prior art following beneficial effect: the present invention introduces coordinate information and carries out verification, has rejected some SIFT couplings due to the error matching points of only having considered that metric space is introduced.
Accompanying drawing explanation:
Fig. 1 is the inventive method process flow diagram.
Embodiment:
Below in conjunction with accompanying drawing, the present invention is described in further detail:
Fig. 1 is the inventive method process flow diagram.An optimization method based on SIFT Feature Points Matching, it comprises the steps:
(1) read image to be matched;
(2 use yardstick invariant features transfer algorithms extract image characteristic point and descriptor thereof, with 128 dimensional vectors, represent;
(3) the neighbouring node algorithm of K carries out cross-matched to the Euclidean distance of descriptor, first by the feature of the right figure of feature profile matching of left figure, describe, then with the feature description of the left figure of feature profile matching of right figure; Choose the identical point of coupling twice, as the basic point set of coupling;
(4) calculate each some ratio test value of basic point set, the point that on left figure, unique point is mated with right figure most, the point of right figure coupling, makes respectively difference and compares at work; In like manner the point of right figure is done to identical point and do identical operation, get the mean value of two ratios;
(5) ratio test value, as SIFT unique point can not reliability, ratio test value is more little more credible.
(6) get part with a high credibility, value is between 0 to 0.3, to be standard point set here, and value is between 0.3 to 0.8, to be point set undetermined;
(7) to standard point set in same coordinate system, by the mode of straight line, describe;
(8) successively every straight line is asked the intersection point of all the other straight lines and this straight line, in these intersection points, got distance two intersection points farthest as the eigenwert of this straight line;
(9) two of cut-off line eigenwert minimum as optimal straight line, usings the intersection point of these two straight lines as the shadow point that disappears.
(10) it is starting point that each point for the treatment of fixed-point set solves left figure point coordinate, the shadow point that disappears for terminal and left figure point coordinate be starting point, the cosine value that right figure point coordinate is terminal;
(11) get cosine value approach 1 for correct point, get here and be greater than 0.99, form correct point set;
(12) final coupling point set is comprised of standard point set and correct coupling point set.
The principle of the invention:
At SIFT, detect unique point, calculate feature descriptor, for the Euclidean distance of feature descriptor, mate.At this matching process, for guaranteeing the algorithm that the efficiency of coupling has used K to close on node most, carry out Euclidean distance coupling, for guaranteeing the correct cross-matched of coupling.So far the matching double points obtaining, the correct point being in SIFT algorithm is right.The limitation of hesitation SIFT feature and the systematic error of algorithm cause the matching double points obtaining to have part incorrect.Not enough herein for making up, done following processing:
We are referred to as basic point set obtaining the current match point obtaining, and we will screen on this basis and reject.
Due to two width picture point pixels all corresponding true scenery in the world, so any point in left figure has at the most a point corresponding with it in right figure.If any in left figure has the feature descriptor of two or more points very approaching in right figure, we just think that this matching double points is insincere, and vice versa.
So the every a pair of match point of basic point set is all found to a time match point, the value obtaining than the matching degree of inferior match point by the matching degree of match point, as can not reliability.This value is between 0 to 1.
So far, our matching algorithm is all based on feature descriptor, is further verification matching result, and we introduce the two-dimensional coordinate of pixel.
Due in space, not with the parallel plane parallel lines of camera imaging, in the projection of imaging plane, meet at a bit, i.e. the shadow point that disappears of image.If we regard camera as the particle of a translation, take camera in object of reference, the movement locus of the object in space is one group of parallel lines.Between two frames, the coordinate line of match point meets at the shadow point that disappears in image like this, that is, do not give this shadow point matching double points that disappears incorrect.
So we are according to the confidence level of coupling, find part match point with a high credibility, simulate the position of the shadow point that disappears, the left figure point coordinate of take is criterion with the left figure point coordinate of vector sum of the shadow point composition that disappears and the vectorial cosine value of right figure composition, larger at this cosine value, when approaching 1, think that this matching double points is correct, on the contrary mistake.
It is to be understood that: the above is only the preferred embodiment of the present invention; for those skilled in the art; under the premise without departing from the principles of the invention, can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.
Claims (1)
1. the optimization method based on SIFT Feature Points Matching, is characterized in that, it comprises the steps:
(1) read image to be matched;
(2 use yardstick invariant features transfer algorithms extract image characteristic point and descriptor thereof, with 128 dimensional vectors, represent;
(3) the neighbouring node algorithm of K carries out cross-matched to the Euclidean distance of descriptor, first by the feature of the right figure of feature profile matching of left figure, describe, then with the feature description of the left figure of feature profile matching of right figure; Choose the identical point of coupling twice, as the basic point set of coupling;
(4) calculate each some ratio test value of basic point set, the point that on left figure, unique point is mated with right figure most, the point of right figure coupling, makes respectively difference and compares at work; In like manner the point of right figure is done to identical point and do identical operation, get the mean value of two ratios;
(5) ratio test value, as SIFT unique point can not reliability, ratio test value is more little more credible.
(6) get part with a high credibility, value is between 0 to 0.3, to be standard point set here, and value is between 0.3 to 0.8, to be point set undetermined;
(7) to standard point set in same coordinate system, by the mode of straight line, describe;
(8) successively every straight line is asked the intersection point of all the other straight lines and this straight line, in these intersection points, got distance two intersection points farthest as the eigenwert of this straight line;
(9) two of cut-off line eigenwert minimum as optimal straight line, usings the intersection point of these two straight lines as the shadow point that disappears.
(10) it is starting point that each point for the treatment of fixed-point set solves left figure point coordinate, the shadow point that disappears for terminal and left figure point coordinate be starting point, the cosine value that right figure point coordinate is terminal;
(11) get cosine value approach 1 for correct point, get here and be greater than 0.99, form correct point set;
(12) final coupling point set is comprised of standard point set and correct coupling point set.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310733191.8A CN103714544B (en) | 2013-12-27 | 2013-12-27 | A kind of optimization method based on SIFT feature Point matching |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310733191.8A CN103714544B (en) | 2013-12-27 | 2013-12-27 | A kind of optimization method based on SIFT feature Point matching |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103714544A true CN103714544A (en) | 2014-04-09 |
CN103714544B CN103714544B (en) | 2016-08-17 |
Family
ID=50407487
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310733191.8A Active CN103714544B (en) | 2013-12-27 | 2013-12-27 | A kind of optimization method based on SIFT feature Point matching |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103714544B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105513075A (en) * | 2015-12-09 | 2016-04-20 | 深圳市华和瑞智科技有限公司 | Image matching method for realizing multi-point layered support description on the basis of SIFT features |
WO2017181892A1 (en) * | 2016-04-19 | 2017-10-26 | 中兴通讯股份有限公司 | Foreground segmentation method and device |
CN108376409A (en) * | 2018-02-24 | 2018-08-07 | 首都师范大学 | A kind of light field image method for registering and system |
CN109410255A (en) * | 2018-10-17 | 2019-03-01 | 中国矿业大学 | A kind of method for registering images and device based on improved SIFT and hash algorithm |
CN113063353A (en) * | 2021-03-31 | 2021-07-02 | 深圳中科飞测科技股份有限公司 | Coordinate system establishing method, detection device, detection equipment and storage medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102004911A (en) * | 2010-12-31 | 2011-04-06 | 上海全景数字技术有限公司 | Method for improving accuracy of face identification |
CN103020945A (en) * | 2011-09-21 | 2013-04-03 | 中国科学院电子学研究所 | Remote sensing image registration method of multi-source sensor |
-
2013
- 2013-12-27 CN CN201310733191.8A patent/CN103714544B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102004911A (en) * | 2010-12-31 | 2011-04-06 | 上海全景数字技术有限公司 | Method for improving accuracy of face identification |
CN103020945A (en) * | 2011-09-21 | 2013-04-03 | 中国科学院电子学研究所 | Remote sensing image registration method of multi-source sensor |
Non-Patent Citations (2)
Title |
---|
REN JIAYUAN ET AL.: "Study on Eliminating Wrong Match Pairs of SIFT", 《IEEE INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING》, 31 December 2010 (2010-12-31), pages 992 - 995 * |
刘焕敏 等: "一种改进的SIFT双向匹配算法", 《兵工自动化》, vol. 28, no. 6, 30 June 2009 (2009-06-30) * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105513075A (en) * | 2015-12-09 | 2016-04-20 | 深圳市华和瑞智科技有限公司 | Image matching method for realizing multi-point layered support description on the basis of SIFT features |
WO2017181892A1 (en) * | 2016-04-19 | 2017-10-26 | 中兴通讯股份有限公司 | Foreground segmentation method and device |
CN108376409A (en) * | 2018-02-24 | 2018-08-07 | 首都师范大学 | A kind of light field image method for registering and system |
CN109410255A (en) * | 2018-10-17 | 2019-03-01 | 中国矿业大学 | A kind of method for registering images and device based on improved SIFT and hash algorithm |
CN113063353A (en) * | 2021-03-31 | 2021-07-02 | 深圳中科飞测科技股份有限公司 | Coordinate system establishing method, detection device, detection equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN103714544B (en) | 2016-08-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2015149534A1 (en) | Gabor binary pattern-based face recognition method and device | |
Sahin et al. | Category-level 6d object pose recovery in depth images | |
Zhou et al. | Robust plane-based structure from motion | |
CN103714544A (en) | Optimization method based on SIFT feature point matching | |
CN104821011A (en) | Method of generating 3D house type model by 2D house type model based on camera shooting | |
CN104167003A (en) | Method for fast registering remote-sensing image | |
CN105654421B (en) | Based on the projective transformation image matching method for converting constant low-rank texture | |
CN104361314A (en) | Method and device for positioning power transformation equipment on basis of infrared and visible image fusion | |
EP3550467B1 (en) | Image matching method, device and system, and storage medium | |
CN107862713B (en) | Camera deflection real-time detection early warning method and module for polling meeting place | |
CN104240231A (en) | Multi-source image registration based on local structure binary pattern | |
Junejo et al. | Optimizing PTZ camera calibration from two images | |
CN103839253A (en) | Arbitrary point matching method based on partial affine transformation | |
Li et al. | 4FP-structure: A robust local region feature descriptor | |
US20150131853A1 (en) | Stereo matching system and method for generating disparity map using same | |
Yu et al. | Riga: Rotation-invariant and globally-aware descriptors for point cloud registration | |
CN105654479A (en) | Multispectral image registering method and multispectral image registering device | |
Zhu et al. | An improved RANSAC algorithm based on similar structure constraints | |
Shen et al. | Semi-dense feature matching with transformers and its applications in multiple-view geometry | |
Zhao et al. | Learning probabilistic coordinate fields for robust correspondences | |
CN107274382B (en) | State identification method and device of hard pressing plate and electronic equipment | |
KR20210088436A (en) | Image processing methods, devices and electronic devices | |
CN103530872B (en) | A kind of error hiding delet method based on angle restriction | |
CN102034235B (en) | Rotary model-based fisheye image quasi dense corresponding point matching diffusion method | |
Guo et al. | Image registration method based on improved SIFT algorithm and essential matrix estimation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20190211 Address after: Unit B101-2, Phase II, International Science and Technology Park, 1355 Jinjihu Avenue, Suzhou Industrial Park, Jiangsu Province Patentee after: SUZHOU SHENGJING INFORMATION TECHNOLOGY CO., LTD. Address before: 215021 111C unit of international science and Technology Park, 1355 Jinji Hu Avenue, Suzhou Industrial Park, Jiangsu Patentee before: Suzhou grand view Spatial Information Technology company limited |