CN106558074A - Coarse-fine combination matching algorithm in assemble of the satellite based on rotational transformation matrix - Google Patents
Coarse-fine combination matching algorithm in assemble of the satellite based on rotational transformation matrix Download PDFInfo
- Publication number
- CN106558074A CN106558074A CN201510601197.9A CN201510601197A CN106558074A CN 106558074 A CN106558074 A CN 106558074A CN 201510601197 A CN201510601197 A CN 201510601197A CN 106558074 A CN106558074 A CN 106558074A
- Authority
- CN
- China
- Prior art keywords
- point
- coarse
- matrix
- matching
- assemble
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to the coarse-fine combination matching algorithm in a kind of assemble of the satellite based on rotational transformation matrix, its technical characterstic is to comprise the following steps：(1) detection of multiple dimensioned FAST characteristic points；(2) construction of new sampling model；(3) selection of sampled point pair；(4) establishment of feature description matrix；(5) the Description Matrix matching of coarse-fine combination；(6) using threshold value and RANSAC algorithms screening feature point pairs.The interference of various conversion in image, particularly rotation transformation can be preferably overcome in assemble of the satellite of the present invention based on the coarse-fine combination matching algorithm of rotational transformation matrix; Rapid matching is to proper characteristics point, it is ensured that robot system smoothly completes the work such as image registration, target recognition and region labeling.
Description
Technical field
The present invention relates to the object matching technology in assemble of the satellite image, the coarse-fine combination matching algorithm in specially a kind of assemble of the satellite based on rotational transformation matrix.
Background technology
Various countries pay much attention to government utilities such as national security, social progress, and in aerospace field, the accurate real-time ground mounting technology of all kinds of satellite equipments has become the key that field of aerospace can be fast-developing.The relevant regulations of demand are assembled according to satellite ground, it is necessary to assure optical camera, experimental provision and instrument, communicate integrated with the stand-alone device fast and stable such as detecting devices, space telescope drive mechanism, support and antenna.Current assemble of the satellite technology development is slower, whole assembling process still relies primarily on the experience of engineering staff and is qualitatively judged, artificial operation certainly exist cannot be between digitized measurement accessory relative position, pose adjustment can not quantify, the visual not strong situation in crucial docking site.The accurate real-time ground assembling of satellite can be carried out at present using the mobile robot of assemble of the satellite, but for the Practical Project environment that there are various rotation transformation interference, still describing son using floating type carries out matching judgment, and the matching algorithm of son is described generally with the operational capability that the matching algorithm than son is described based on floating type is easier based on binary type, therefore, it is possible to significantly improve the real-time of matching algorithm.How to overcome binary descriptor not enough defect of correct matching rate in rotation transformation image, and take into account the real-time of matching, there are still some key issues needs to solve.Robot system during assemble of the satellite itself has higher practical value, if the improved correlation matching algorithm for describing son based on binary system is applied to mobile-robot system, the matching and identification work of workpiece target are completed in rotation transformation picture fast and accurately, by with higher real value and theory significance.
At present, the classical matching algorithm based on binary descriptor mainly has following several：
1) using the BRIEF algorithms that the Gauss distribution point for randomly selecting is sub to building description；
2) feature and the ORB algorithms with directional characteristic description structure are detected using Accelerated fractionation；
3) the short distance set for up-sampling point composition using concentric circular builds the BRISK algorithms of description；
4) using the FREAK algorithms that there is on annulus the sampled point for overlapping acceptance region to build description；
Characteristic matching is the basis of the concrete applications such as image registration, target recognition, region labeling and robot navigation.Binary descriptor generally has excellent speed characteristics, but as invariable rotary performance is not enough, causes the error hiding phenomenon in assemble of the satellite image very serious.Therefore find and a kind of can effectively overcome being particularly important based on the matching algorithm of binary descriptor for various rotation transformations interference in image.
The content of the invention
According to the deficiencies in the prior art, the technical problem that the present invention is intended to solve is to propose a kind of coarse-fine combination matching algorithm based on rotational transformation matrix for being applied to assemble of the satellite.The method creates the new sampling model with invariable rotary feature and sampled point to selection mechanism；With 10 ° as interval, the thick rotation transformation Description Matrix of characteristics of image description vectors to be matched is obtained, and the Description Matrix with reference to benchmark image searches for optimal rotation angle when slightly matching；In ± 10 ° of positions of the thick anglec of rotation, repeat above step with 1 ° as interval, calculate the accurate anglec of rotation, while obtaining accurate match point.Method is easy, is easy to practical application.
The present invention solves the technical scheme of the technical problem, designs a kind of coarse-fine combination matching algorithm based on rotational transformation matrix for being applied to assemble of the satellite, the method institute using the step of be：
(1) detection of multiple dimensioned FAST characteristic points：Set up metric space pyramid and FAST characteristic points are extracted in per layer of pyramid；The current FAST characteristic points of comparison and the score of 26 pixels of surrounding, determine whether characteristic point；The location and yardstick of characteristic point are optimized.
(2) construction of new sampling model：In the local neighborhood of characteristic point, equal proportion builds four layers of donut.At interval of 1 ° of determination, one sampled point on annulus.
(3) selection of sampled point pair：Using new sampling model, 10 groups of sampled points pair are formed per 1 ° of direction, travel through all directions of sampling model.
(4) establishment of feature description matrix：All sampled points pair of all characteristic points are traveled through, using the gray value comparative result construction Description Matrix of two sampled points of sampled point centering；
(5) the Description Matrix matching of coarse-fine combination：With 10 ° as interval, the thick rotation transformation Description Matrix of characteristics of image description vectors to be matched is obtained, and the Description Matrix with reference to benchmark image searches for optimal rotation angle when slightly matching；In ± 10 ° of positions of the thick anglec of rotation, repeat above step with 1 ° as interval, calculate accurate match point and angle call number.
(6) using threshold value and RANSAC algorithms screening feature point pairs：Whether the matching result of judging characteristic point pair meets threshold condition, and the feature point pairs set to meeting threshold condition is screened using stochastical sampling concordance (RANSAC) algorithm；
Compared with prior art; the interference of various conversion in image, particularly rotation transformation can be preferably overcome in assemble of the satellite of the present invention based on the coarse-fine combination matching algorithm of rotational transformation matrix; Rapid matching is to proper characteristics point, it is ensured that robot system smoothly completes the work such as image registration, target recognition and region labeling.
Description of the drawings
A kind of flow chart of embodiment of coarse-fine combination matching algorithm in Fig. 1 assemble of the satellite of the present invention based on rotational transformation matrix；
A kind of overall structure of the invariable rotary sampling model of embodiment of coarse-fine combination matching algorithm in Fig. 2 assemble of the satellite of the present invention based on rotational transformation matrix；
A kind of sampled point FINE DISTRIBUTION of the invariable rotary sampling model of embodiment of coarse-fine combination matching algorithm in Fig. 3 assemble of the satellite of the present invention based on rotational transformation matrix；
A kind of built-up sequence of the sampled point pair of embodiment of coarse-fine combination matching algorithm in Fig. 4 assemble of the satellite of the present invention based on rotational transformation matrix.
Specific embodiment
Now with a kind of workpiece video image that there is rotation transformation interference in assemble of the satellite as embodiment, and its accompanying drawing is combined, the present invention is further described based on the coarse-fine combination matching algorithm of rotational transformation matrix.
The coarse-fine combination matching process (abbreviation method, referring to Fig. 1-4) based on rotational transformation matrix of present invention design, creates the new sampling model with invariable rotary feature and sampled point first to selection mechanism；Then with 10 ° as interval, the thick rotation transformation Description Matrix of characteristics of image description vectors to be matched is obtained, and the Description Matrix with reference to benchmark image searches for optimal rotation angle when slightly matching；Repeat above step with 1 ° as interval in ± 10 ° of positions of the thick anglec of rotation, while calculating the accurate anglec of rotation, obtain accurate match point.Methods described is concretely comprised the following steps：
1st, the detection of multiple dimensioned FAST characteristic points
Metric space pyramid is initially set up, FAST characteristic points are extracted in per layer of pyramid, while calculating the FAST scores of all positions；Then compare the fraction of current FAST characteristic points and 26 pixels of surrounding, if projecting pixel, retain current signature point, otherwise delete this characteristic point；Quadratic function fitting is carried out to the pixel in each characteristic point current layer and adjacent two layers 3*3*3 neighborhoods finally, find the very big scoring position after each self-optimizing, this 3 maximum are carried out into a fitting of parabola along yardstick axle again, this parabolical extreme point is obtained and is counted corresponding FAST scores and dimensional information.Finally, interpolation is carried out to the coordinate of current extreme value score on the layer near this scale-value, obtains the final accurate coordinate position of characteristic point.
2nd, the construction of new sampling model
In the local neighborhood of characteristic point, with 2s, 4s, 6s, 8s are that radius equal proportion builds 4 layers of donut.At interval of 1 ° of determination, one sampled point on annulus, the sampling dot spacing on identical annulus gradually increases from the increase with annular radii.Meanwhile, it is to reduce the noise jamming in image, and the position deviation of coordinate points can be overcome, all sampled points are through mean filter and linear interpolation processing.Fig. 2,3 be new sampling model schematic diagram, sampled point is evenly distributed on each position of whole local neighborhood, can effectively describe most information of local neighborhood.
3rd, the selection of sampled point pair
Using new sampling model, combine successively and can form 10 groups of sampled points pair, travel through all directions and can obtain all sampled points pair, it is concrete as shown in Figure 4.Special instruction, it is excessive to introduce the multiplicity that analog information increased Description Matrix, make the unique reduction between Description Matrix；The desin speed of Description Matrix can be reduced simultaneously, increase the match time of Description Matrix.Therefore this paper only considers all combining forms of equidirectional sampled point, adds the sampled point of surrounding angle.In addition, the sequence number of Fig. 4 1.～10. give sampled point per 1 ° of direction to selection order.
4th, the establishment of feature description matrix
All sampled points pair of all characteristic points are traveled through, if the gray value of first sampled point of the sampled point pair of current signature point is more than second sampled point, the value of Description Matrix relevant position puts 1；If less than the gray value of second sampled point, the value of Description Matrix relevant position puts -1, and so improved advantage is：If two feature description matrixes carry out multiplying, the value of correspondence position is identical, then increase the judged result of similarity；If the value of correspondence position is different, the similarity between two characteristic points is cut down.In addition, in the matrix of all feature point description vectors of benchmark image is represented, columns represents the dimension of feature point description vector, and line number represents the number of characteristic point.In the rotational transformation matrix of all feature point description vectors of image to be matched is represented.Columns equally represents the dimension of feature point description vector, and line number represents the sum of all characteristic point rotation transformation forms.
5th, the Description Matrix matching of coarse-fine combination
Using the matrix of all feature point description vectors of benchmark image, thick matching stage, represents that the matrix of the 36 group thick rotation transformation forms vectorial with all feature point descriptions of image to be matched is represented and is multiplied, calculate Matrix C 1.The maximum position that C1 often goes need to be counted only, you can obtain L characteristic point of benchmark image corresponding thick matching characteristic point and angle call number a in image to be matched.Smart matching stage, ± 10 ° of positions on the basis of thick matching rotation angle, with 1 ° of matrix for all feature point description vectors of benchmark image being reused as interval, represent that the matrix of 21 groups of smart rotation transformation forms for being multiplied by all feature point description vectors of image to be matched is represented, obtain Matrix C 2.The maximum that C2 often goes is calculated, is both the accurate match point and angle call number β of L Feature point correspondence in benchmark image.The accurate result of the final anglec of rotation should be γ=a+ β, calculate the anglec of rotation using the mode of this coarse-fine combination, and the mean error of estimation can be less than 1 °.
6th, using threshold value and RANSAC algorithms screening feature point pairs
Whether the matching result of judging characteristic point pair meets threshold condition, if being more than threshold value, retains current signature point pair；Otherwise delete.The feature point pairs set of threshold condition is met to more than, is screened using stochastical sampling concordance (RANSAC) algorithm.The optimized parameter M and N of transformation matrix are calculated, all feature point pairs of the parameter will be met as final correct matching double points.
Although there is being described in detail as example for the workpiece video image in assemble of the satellite engineering for rotation transformation, the object matching of its principle and process suitable for the rotation transformation video image of other occasions with a kind of in inventive algorithm.
The present invention does not address part and is applied to prior art.
Claims (1)
1. a kind of coarse-fine combination matching algorithm in assemble of the satellite based on rotational transformation matrix, it is characterised in that bag
Include following steps：
(1) detection of multiple dimensioned FAST characteristic points：Set up metric space pyramid and in per layer of pyramid
Extract FAST characteristic points；Compare the score of current FAST characteristic points and 26 pixels of surrounding, judgement is
It is no to be characterized a little；The location and yardstick of characteristic point are optimized；
(2) construction of new sampling model：In the local neighborhood of characteristic point, equal proportion builds four layers of donut；
At interval of 1 ° of determination, one sampled point on annulus；
(3) selection of sampled point pair：Using new sampling model, 10 groups of sampled points pair are formed per 1 ° of direction,
All directions of traversal sampling model；
(4) establishment of feature description matrix：All sampled points pair of all characteristic points are traveled through, using sampled point
The gray value comparative result construction Description Matrix of two sampled points of centering；
(5) the Description Matrix matching of coarse-fine combination：With 10 ° as interval, characteristics of image description to be matched is obtained
The thick rotation transformation Description Matrix of vector, and the Description Matrix with reference to benchmark image searches for optimal when slightly matching
The anglec of rotation；In ± 10 ° of positions of the thick anglec of rotation, repeat above step with 1 ° as interval, calculate accurate
Match point and angle call number；
(6) using threshold value and RANSAC algorithms screening feature point pairs：The matching result of judging characteristic point pair is
No to meet threshold condition, the feature point pairs set to meeting threshold condition is entered using stochastical sampling consistency algorithm
Row screening.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510601197.9A CN106558074B (en) | 2015-09-18 | 2015-09-18 | Coarse-fine combination matching algorithm in assemble of the satellite based on rotational transformation matrix |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510601197.9A CN106558074B (en) | 2015-09-18 | 2015-09-18 | Coarse-fine combination matching algorithm in assemble of the satellite based on rotational transformation matrix |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106558074A true CN106558074A (en) | 2017-04-05 |
| CN106558074B CN106558074B (en) | 2019-04-12 |
Family
ID=58414875
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510601197.9A Expired - Fee Related CN106558074B (en) | 2015-09-18 | 2015-09-18 | Coarse-fine combination matching algorithm in assemble of the satellite based on rotational transformation matrix |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106558074B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110189368A (en) * | 2019-05-31 | 2019-08-30 | 努比亚技术有限公司 | Method for registering images, mobile terminal and computer readable storage medium |
| CN110207666A (en) * | 2019-04-11 | 2019-09-06 | 南京航空航天大学 | The vision pose measuring method and device of analog satellite on a kind of air floating platform |
| CN110751223A (en) * | 2019-10-25 | 2020-02-04 | 北京达佳互联信息技术有限公司 | Image matching method and device, electronic equipment and storage medium |
| CN112164107A (en) * | 2020-10-14 | 2021-01-01 | 上海汽车集团股份有限公司 | End-to-end camera modeling method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103136751A (en) * | 2013-02-05 | 2013-06-05 | 电子科技大学 | Improved scale invariant feature transform (SIFT) image feature matching algorithm |
| 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) |
| US20140226913A1 (en) * | 2009-01-14 | 2014-08-14 | A9.Com, Inc. | Method and system for matching an image using image patches |
-
2015
- 2015-09-18 CN CN201510601197.9A patent/CN106558074B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140226913A1 (en) * | 2009-01-14 | 2014-08-14 | A9.Com, Inc. | Method and system for matching an image using image patches |
| CN103136751A (en) * | 2013-02-05 | 2013-06-05 | 电子科技大学 | Improved scale invariant feature transform (SIFT) image feature matching algorithm |
| 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) |
Non-Patent Citations (1)
| Title |
|---|
| 白丰 等: "卫星装配中基于强区分性描述子的识别方法", 《仪器仪表学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110207666A (en) * | 2019-04-11 | 2019-09-06 | 南京航空航天大学 | The vision pose measuring method and device of analog satellite on a kind of air floating platform |
| CN110189368A (en) * | 2019-05-31 | 2019-08-30 | 努比亚技术有限公司 | Method for registering images, mobile terminal and computer readable storage medium |
| CN110189368B (en) * | 2019-05-31 | 2023-09-19 | 努比亚技术有限公司 | Image registration method, mobile terminal and computer readable storage medium |
| CN110751223A (en) * | 2019-10-25 | 2020-02-04 | 北京达佳互联信息技术有限公司 | Image matching method and device, electronic equipment and storage medium |
| CN110751223B (en) * | 2019-10-25 | 2022-09-30 | 北京达佳互联信息技术有限公司 | Image matching method and device, electronic equipment and storage medium |
| CN112164107A (en) * | 2020-10-14 | 2021-01-01 | 上海汽车集团股份有限公司 | End-to-end camera modeling method and device |
| CN112164107B (en) * | 2020-10-14 | 2023-06-09 | 上海汽车集团股份有限公司 | End-to-end camera modeling method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106558074B (en) | 2019-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107886528B (en) | Distribution line operation scene three-dimensional reconstruction method based on point cloud | |
| CN111540005B (en) | Loop detection method based on two-dimensional grid map | |
| CN106558074A (en) | Coarse-fine combination matching algorithm in assemble of the satellite based on rotational transformation matrix | |
| CN101398907B (en) | Two-dimension code structure and decoding method for movable robot | |
| CN103632366A (en) | Parameter identification method for elliptical target | |
| CN112665575B (en) | SLAM loop detection method based on mobile robot | |
| CN103778433B (en) | Generalized-point-set matching method based on distances from points to lines | |
| CN104536009A (en) | Laser infrared composite ground building recognition and navigation method | |
| CN106096497B (en) | A kind of house vectorization method for polynary remotely-sensed data | |
| US9697438B2 (en) | Image description and image recognizable method | |
| CN101604166A (en) | A kind of method for planning path for mobile robot based on particle swarm optimization algorithm | |
| CN109829853A (en) | A kind of unmanned plane image split-joint method | |
| CN101359400A (en) | Process for positioning spatial position of pipe mouth based on vision | |
| CN109712071B (en) | Unmanned aerial vehicle image splicing and positioning method based on track constraint | |
| CN102800097A (en) | Multi-feature multi-level visible light and infrared image high-precision registering method | |
| CN102496181A (en) | True-orthophotomap making method oriented to large-scale production | |
| CN108122255A (en) | It is a kind of based on trapezoidal with circular combination terrestrial reference UAV position and orientation method of estimation | |
| CN103761739A (en) | Image registration method based on half energy optimization | |
| CN112785724B (en) | LiDAR point cloud and two-dimensional image-based visual color matching method for ancient architecture | |
| CN113361499B (en) | Local object extraction method and device based on two-dimensional texture and three-dimensional attitude fusion | |
| CN106023183A (en) | Real-time line segment matching method | |
| CN110433467A (en) | Picking up table tennis ball robot operation method and equipment based on binocular vision and ant group algorithm | |
| CN113239953A (en) | SAR image rotating ship detection method based on directed Gaussian function | |
| Zhang et al. | Distributed coverage optimization for deployment of directional sensor networks | |
| CN105678791A (en) | Lane line detection and tracking method based on parameter non-uniqueness property |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190412 Termination date: 20190918 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |