CN106289188A - A kind of measuring method based on multi-vision aviation image and system - Google Patents
A kind of measuring method based on multi-vision aviation image and system Download PDFInfo
- Publication number
- CN106289188A CN106289188A CN201610639829.5A CN201610639829A CN106289188A CN 106289188 A CN106289188 A CN 106289188A CN 201610639829 A CN201610639829 A CN 201610639829A CN 106289188 A CN106289188 A CN 106289188A
- Authority
- CN
- China
- Prior art keywords
- image
- matrix
- junction point
- pose parameter
- point
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C11/00—Photogrammetry or videogrammetry, e.g. stereogrammetry; Photographic surveying
- G01C11/04—Interpretation of pictures
- G01C11/30—Interpretation of pictures by triangulation
- G01C11/34—Aerial triangulation
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Image Processing (AREA)
Abstract
A kind of measuring method based on multi-vision aviation image and system, the technical scheme provided by the embodiment of the present invention are provided, solve owing to tilting image inclination angle excessive, cause the problem that in traditional photography surveying, relative orientation iteration does not restrains.Use method computer vision and photogrammetry combined, recover true pose parameter and the junction point true coordinate of all images exactly.
Description
Technical field
The application relates to technical field of image processing, particularly relate to a kind of measuring method based on multi-vision aviation image and
System.
Background technology
Aerial triangulation (Aerial Triangulation) be photogrammetric in determine in photographing region and all photograph
Element outside photo or digitized video foreign side, and the effective ways of pass point geographical coordinates, be photogrammetric data processing procedure
In indispensable job step.Along with the development of the technology such as computer, artificial intelligence, image procossing, photogrammetric initially enter
Total digitalization, full-automation, the direction of multisensor imaging model are developed, photogrammetric more close with computer vision
Combine, and interpenetrate.Particularly rebuilding field in multiview three-dimensional, multiview three-dimensional is rebuild and is referred to by shooting
Certain scene and object are shot by machine in different angles, then utilize the multi-view image sequences photographed to recover true
Three-dimensional scenic, its critical process includes: feature extraction is mated with sparse, multi views aerial triangulation, dense Stereo Matching, surface weight
Build.Multi views aerial triangulation technology mainly calculates multi-view geometry restriction relation, and Exact recovery goes out the relative of image
Position orientation relation.It it is one of the important focus of computer vision and photogrammetry area research.
Multi-view geometry restriction relation calculates, and is by Epipolar geometry, geometrical-restriction relation is converted to basis matrix
Model parameter estimation.Once it is determined that the geometrical-restriction relation between two views, i.e. can be applicable to camera self-calibration and Three-dimensional Gravity
Build, be a basic research of computer vision field.The research work the earliest of multi-view geometry relation is by Longuet-
Higgins proposes, and the geometrical-restriction relation between multi views, in euclidean geometry, available essential matrix represents.The most for many years
In, it is all the research contents with essential matrix as multi-view geometry.The thesis for the doctorate of Luong proposes basis matrix for solution two
Geometrical relationship between width image, elaborates Epipolar geometry fundamental property under projective space.Hereafter, the robust of basis matrix
Property estimate to obtain studying in a large number and widely, classical algorithm is 7 methods and 8 methods, is retrained by Epipolar geometry, by 7 methods
Or 8 methods can solve basis matrix, obtain camera interior and exterior parameter.Owing to Epipolar geometry relation has clear and definite geometric meaning,
Thus scholars begin one's study and minimize the nonlinear algorithm such as back projection's error or Sampson error, are used for solving basis matrix.
China researchers have done substantial amounts of research work in terms of the estimation of basis matrix, achieve certain achievement.Solve base
During plinth matrix, under Epipolar geometry retrains, the input of various algorithms is the Corresponding matching point between multi views, i.e. represents same field
In scape, same object and the point of the matching characteristic in the same space position, can be obtained with mating by feature point detection.Often scheme
Exist between Xiang block, the factor such as illumination variation, the matching characteristic point detected can be caused not exclusively correct, mistake may be obtained
Match point (outliers), data inconsistent with hypothesized model the most on the whole, these data points are not followed and are assumed by mistake
Model.If the corresponding point containing Mismatching point are substituted into above-mentioned algorithm, then can cause the model parameter estimation of mistake.How
The match point of filter false and retain the problem that correct data are computer vision and other field is being studied always.People
Make substantial amounts of research work in this respect, but without finding a perfect robustness model parameter estimation technology.At present
There is many robustness method for parameter estimation, maximum likelihood estimate (M-estimator), minimum median algorithm (LMedS),
Three kinds of technology of RANSAC algorithm (RANSAC) are the most universal Robust estimation algorithms and are applied to wide in a large number
General field.But, when the outliers proportion in data is more than 50%, M-estimator Yu LMedS algorithm is the most not
Can normally work.RANSAC algorithm realizes the simple and robustness of method due to it, is widely used in solution model parameter
Estimation problem.In recent years, in order to improve efficiency and the performance of RANSAC algorithm, research worker proposes the improvement of many and calculates
Method.What some of which strategy was main aims at the process of Optimized model inspection, and the method for main employing Pre-testing is by model
A part of data are tested, on the portion by checking in total data the most again.And another kind of being mainly sends out
Now or amendment sampling process produces more effectively hypothesis.But, a most basic hypothesis of RANSAC algorithm is a model
Parameter be to calculate by not polluting from one in the sample of noise, and to meet mate all of with data
Point is consistent, in practice, owing to there being noise to occur, causes the increase of Riming time of algorithm, or the result of mistake occurs.From
In camera motion, restoration scenario structure is to input the most proven one group of image, represents with describing algorithm acquisition by feature detection
The matching characteristic point of Same Scene, then uses multi-view geometry restriction relation, carries out projective reconstruction, obtain the inside and outside of video camera
Parameter, finally carries out metric reconstruction, exports sparse three-dimensional scene structure point cloud.After the matching characteristic point obtained between view,
Projective reconstruction is studied by substantial amounts of scholar, solves how to be obtained in video camera by multi-view geometry restriction relation
The position of ginseng and video camera and the problem in direction.After the demarcation completing camera parameters, so that it may carry out European geometric space
Metric reconstruction, generate three dimensional structure point cloud in space.After having obtained initial projective reconstruction result, in order to make error equal
Even is distributed between each width image, and also to obtain more accurate as a result, it is often necessary ro enter initial reconstructed results
Line nonlinearity optimizes, and in SFM, measurement error being applied the most accurate nonlinear optimization method is bundle adjustment algorithm.Light beam
Method adjustment has become an important component part of SFM and three-dimensional reconstruction.
In recent years, numerous bundle adjustment algorithms is suggested, the problem that these algorithms mainly solve two aspects.One is
Solving bundle adjustment validity problem, LM (Levenberg-Marquardt) optimized algorithm proposes a classical light beam
Method adjustment solution, Lourakis utilizes trusted area concept, it is proposed that dog leg algorithm, performance is better than LM
(Levenberg-Marquardt) optimized algorithm.Ni rebuilds for large scale scene, and utilization figure cuts and optimizes bundle adjustment.Use the structure distribution of variable, use the pretreatment of conjugate gradient to optimize bundle adjustment.On the other hand it is to solve
Bundle adjustment size and calculating speed issue.Shum, by splitting the overall situation, with different levels carries out bundle adjustment
Carry out three-dimensional reconstruction.Steedly utilizes optical segmentation method, large-scale bundle adjustment is divided into little subproblem and retains former
The a small amount of error model of system.Mouragnon proposes when each new key frame adds to come in, and just runs local flux of light method and puts down
Difference, Engels, for uncertain noise model, proposes local bundle adjustment, and it comprises uncertain diffusion and maximum likelihood is estimated
Meter, reduces computing with this, improves speed.
But, excessive, so causing relative orientation iteration in traditional photography surveying not restrain owing to tilting image inclination angle
Problem.
Summary of the invention
Embodiments provide the generation method and device of a kind of virtual observation data, in order to solve in prior art
Excessive, so causing the problem that in traditional photography surveying, relative orientation iteration does not restrains owing to tilting image inclination angle.
Its concrete technical scheme is as follows:
A kind of measuring method based on multi-vision aviation image, described method includes:
S1, selects two images in all images, and according to described two images, obtains essential matrix, according to described
Essential matrix, obtains the relative pose parameter that described two images are corresponding；
S2, according to default optimized algorithm, is optimized computing to described relative pose parameter, rejects described relative pose ginseng
Rough error point in number and error, obtain stable relative pose parameter；
S3, the number of the point that the unknown image of statistics residue is overlapping with junction point, the number selecting overlapping point is more than
The image of predetermined threshold value, and according to described image, obtain camera matrix, by consistency algorithm, described camera matrix is commented
Valency, obtains stable camera matrix, according to stable camera matrix, obtains known junction point quantity；
S4, is optimized the pose parameter obtained and known junction point, and repeats S3, until obtaining all of image
Relative pose parameter and junction point relative position coordinates；
S5, carries out compensating computation to every image, obtains final image pose parameter and junction point coordinate.
Optionally, and according to described two images, obtain essential matrix, according to described essential matrix, obtain described two
The relative pose parameter that image is corresponding, including:
According to two image plane space geometry restriction relations, utilize 5 methods to solve essential matrix, and utilization is adopted at random
The essential matrix obtained of every time sampling is evaluated by sample consistency algorithm, and obtains stable essential matrix；
Stable essential matrix is carried out singular value decomposition, according to point different solution of the eliminating of principle before camera, obtains two
The relative pose parameter that image is corresponding.
Optionally, the number of the point that the unknown image of statistics residue is overlapping with junction point, select the number of overlapping point
More than the image of predetermined threshold value, and according to described image, obtain camera matrix, by consistency algorithm, described camera matrix is entered
Row is evaluated, and obtains stable camera matrix, according to stable camera matrix, obtains known junction point quantity, including:
The number of the point that the unknown image of statistics residue is overlapping with junction point, selects the number of overlapping point more than presetting
The image of threshold value, utilizes 6 method straight linear conversion to ask and obtains camera matrix, utilize stochastical sampling consistency algorithm to adopting every time
The camera matrix that sample obtains is evaluated, and obtains stable camera matrix；
Stable camera matrix decomposition is obtained the inside and outside parameter of camera, utilizes known image to carry out forward intersection,
To known junction point quantity.
Optionally, the pose parameter obtained and known junction point are optimized, and repeat S3, until obtaining all of
The relative pose parameter of image and junction point relative position coordinates, including:
Optimized algorithm is utilized to be iterated image pose parameter and known junction point optimizing；
Repeat step S3, and the principle optimized according to calculating limit, limit, until the relative pose parameter of all of image and company
Contact relative coordinate is obtained and is optimized.
Optionally, every image is carried out compensating computation, obtain final image pose parameter and junction point coordinate
Afterwards, described method also includes:
The imaging model tried to achieve is converted into co-colouration effect by matrix operations；
Use bundle block adjustment method that 6 elements of exterior orientation of every image and the position of each junction point are sat
Mark carries out compensating computation, eliminates error of coordinate.
A kind of measurement system based on multi-vision aviation image, including:
Pose parameter acquisition module, for selecting two images in all images, and according to described two images, obtains
Essential matrix, according to described essential matrix, obtains the relative pose parameter that described two images are corresponding；
First optimizes module, for according to presetting optimized algorithm, described relative pose parameter being optimized computing, rejects
Rough error point in described relative pose parameter and error, obtain stable relative pose parameter；
Matrix operations module, for adding up the number of the unknown image of the residue point overlapping with junction point, selects weight
The number of folded point is more than the image of predetermined threshold value, and according to described image, obtains camera matrix, by consistency algorithm to described
Camera matrix is evaluated, and obtains stable camera matrix, according to stable camera matrix, obtains known junction point quantity；
Second optimizes module, for being optimized the pose parameter obtained and known junction point, until obtaining all
The relative pose parameter of image and junction point relative position coordinates；
Modular converter, for every image is carried out compensating computation, obtains final image pose parameter and junction point
Coordinate.
Optionally, described pose parameter acquisition module, specifically for according to two image plane space geometry restriction relations,
Utilize 5 methods to solve essential matrix, and utilize stochastical sampling consistency algorithm that the essential matrix obtained of every time sampling is carried out
Evaluate, and obtain stable essential matrix；Stable essential matrix is carried out singular value decomposition, according to some principle before camera
Get rid of different solution, obtain two relative pose parameters corresponding to image.
Optionally, described matrix operations module, the number of the point that the unknown image of statistics residue is overlapping with junction point, choosing
Select out the number image more than predetermined threshold value of overlapping point, utilize 6 method straight linear conversion to ask and obtain camera matrix, utilize with
The camera matrix obtained of every time sampling is evaluated by machine sampling consistency algorithm, and obtains stable camera matrix；Will be stable
Camera matrix decomposition obtain the inside and outside parameter of camera, utilize known image to carry out forward intersection, obtain known connection and count
Amount.
Optionally, described second optimizes module, specifically for utilizing optimized algorithm to image pose parameter and known connection
Point is iterated optimizing；And according to the principle of calculating limit, limit optimization, until the relative pose parameter of all of image and junction point
Relative coordinate is obtained and is optimized.
Optionally, also include:
Processing module, for being converted into co-colouration effect by the imaging model tried to achieve by matrix operations；Use light
Bundle method block adjustment method carries out adjustment meter to 6 elements of exterior orientation of every image and the position coordinates of each junction point
Calculate, eliminate error of coordinate.
The technical scheme provided by the embodiment of the present invention, is solved owing to tilting image inclination angle excessive, causes tradition
The problem that in photogrammetry, relative orientation iteration does not restrains.Use side computer vision and photogrammetry combined
Method, recovers true pose parameter and the junction point true coordinate of all images exactly.
Accompanying drawing explanation
Fig. 1 is the flow chart of a kind of measuring method based on multi-vision aviation image in the embodiment of the present invention；
Fig. 2 is the structural representation of a kind of measurement system based on multi-vision aviation image in the embodiment of the present invention.
Detailed description of the invention
Embodiments providing a kind of measuring method based on multi-vision aviation image, the method includes: obtain two
The essential matrix of image, and essential matrix is obtained to the relative pose parameter of two images, according to default optimized algorithm, to phase
Pose parameter is optimized computing, rejects the rough error point in described relative pose parameter and error, obtain stable phase para-position
Appearance parameter；The number of the point that the unknown image of statistics residue is overlapping with junction point, selects the number of overlapping point more than presetting
The image of threshold value, and according to described image, obtain camera matrix, by consistency algorithm, described camera matrix is evaluated,
Obtain stable camera matrix, according to stable camera matrix, obtain known junction point quantity；To the pose parameter obtained and
Known junction point is optimized, until obtaining relative pose parameter and the junction point relative position coordinates of all of image；To often
Open image and carry out compensating computation, obtain final image pose parameter and junction point coordinate.
The technical scheme provided by the embodiment of the present invention, is solved owing to tilting image inclination angle excessive, causes tradition
The problem that in photogrammetry, relative orientation iteration does not restrains.Use side computer vision and photogrammetry combined
Method, recovers true pose parameter and the junction point true coordinate of all images exactly.
Below by accompanying drawing and specific embodiment, technical solution of the present invention is described in detail, it will be appreciated that this
Concrete technical characteristic in bright embodiment and embodiment is the explanation to technical solution of the present invention rather than restriction, not
In the case of conflict, the concrete technical characteristic in the embodiment of the present invention and embodiment can be mutually combined.
It is illustrated in figure 1 the flow chart of a kind of measuring method based on multi-vision aviation image in the embodiment of the present invention, should
Method includes:
S1, selects two images in all images, and according to described two images, obtains essential matrix, according to described
Essential matrix, obtains the relative pose parameter that described two images are corresponding；
S2, according to default optimized algorithm, is optimized computing to described relative pose parameter, rejects described relative pose ginseng
Rough error point in number and error, obtain stable relative pose parameter；
S3, the number of the point that the unknown image of statistics residue is overlapping with junction point, the number selecting overlapping point is more than
The image of predetermined threshold value, and according to described image, obtain camera matrix, by consistency algorithm, described camera matrix is commented
Valency, obtains stable camera matrix, according to stable camera matrix, obtains known junction point quantity；
S4, is optimized the pose parameter obtained and known junction point, and repeats S3, until obtaining all of image
Relative pose parameter and junction point relative position coordinates；
S5, carries out compensating computation to every image, obtains final image pose parameter and junction point coordinate.
From the point of view of step S1, utilize sparse matching result, from the beginning of two images that same place quantity is most, according to two
Open image plane space geometry restriction relation, utilize 5 methods to solve essential matrix E, and utilize stochastical sampling consistency algorithm
The E obtained that every time samples is evaluated by (RANSAC iterative algorithm), until obtaining stable E.Then essential matrix is carried out
Singular value decomposition, gets rid of different solution according to some principle before camera, finally tries to achieve the relative pose parameter of two images, pass through two panels
Front friendship method solves the relative position coordinates of all same places.
From the point of view of step S2, utilize LM (Levenberg-Marquardt) optimized algorithm to image pose parameter and
Know that junction point is iterated optimizing, excluding gross error point, eliminate incidental error.
From the point of view of step S3, the number of the point that the unknown image of statistics residue is overlapping with junction point, and with maximum weight
The 75% of folded number is threshold value, picks out the overlapping number image more than this threshold value, utilizes 6 method straight linear conversion (DLT) to try to achieve
Camera matrix P, utilizes stochastical sampling consistency algorithm (RANSAC iterative algorithm) to be evaluated the P obtained that every time samples, directly
To obtaining stable P.And then P decomposition is obtained the inside and outside parameter of camera.Forward intersection is carried out also with known image,
To new known junction point, increase known junction point quantity with this.
From the point of view of step S4, also with LM (Levenberg-Marquardt) optimized algorithm to image pose parameter
It is iterated optimizing with known junction point.By that analogy, repeat step 3, and the principle optimized according to calculating limit, limit, until institute
Relative pose parameter and the junction point relative coordinate of some images are obtained and are optimized.In embodiments of the present invention, have employed solely
Special optimization method.Being different from traditional first calculating, the method for rear iteration optimization, this technology uses the side that calculating limit, limit optimizes
Method, from the beginning of two images of initial selected, when after the relative pose information calculating two images and identical point coordinates, just
Carry out LM iteration optimization by these two, often add calculate new image time, will be by the image one of all known posture information
Rise and carry out LM optimization, until all images participate in optimizing, it is thus possible to excluding gross error point effectively, and improve image posture information and
The precision of junction point coordinate.
From the point of view of step S5, if having GCP or gps data, utilize the absolute orientation method in photogrammetry, by image
Relative pose parameter and junction point relative coordinate be transformed under true coordinate system.
In embodiments of the present invention, computer vision and the combination of photogrammetry.By the derivation of equation, complete and will count
The camera imaging model conversion of calculation machine visual field is to the co-colouration effect in photogrammetry, and utilizes in photogrammetry
Absolute orientation and bundle block adjustment method, complete the true pose parameter of all images and the accurate of junction point coordinate
Solve.
A kind of measuring method based on multi-vision aviation image in the corresponding embodiment of the present invention, also carries in the embodiment of the present invention
Supply a kind of measurement system based on multi-vision aviation image, be illustrated in figure 2 in the embodiment of the present invention a kind of based on various visual angles
The structural representation of the measurement system of aviation image, this system includes:
Pose parameter acquisition module 201, for selecting two images in all images, and according to described two images,
Obtain essential matrix, according to described essential matrix, obtain the relative pose parameter that described two images are corresponding；
First optimizes module 202, for according to presetting optimized algorithm, described relative pose parameter being optimized computing,
Reject the rough error point in described relative pose parameter and error, obtain stable relative pose parameter；
Matrix operations module 203, for adding up the number of the unknown image of the residue point overlapping with junction point, selects
The number of overlapping point is more than the image of predetermined threshold value, and according to described image, obtains camera matrix, by consistency algorithm to institute
State camera matrix to be evaluated, obtain stable camera matrix, according to stable camera matrix, obtain known junction point quantity；
Second optimizes module 204, for being optimized the pose parameter obtained and known junction point, until obtaining institute
The relative pose parameter of some images and junction point relative position coordinates；
Modular converter 205, for every image is carried out compensating computation, obtains final image pose parameter and connection
Point coordinates.
Further, in embodiments of the present invention, described pose parameter acquisition module 201, specifically for according to two images
Plane space geometrical-restriction relation, utilizes 5 methods to solve essential matrix, and utilizes stochastical sampling consistency algorithm to adopting every time
The essential matrix that sample obtains is evaluated, and obtains stable essential matrix；Stable essential matrix is carried out singular value decomposition,
According to point different solution of the eliminating of principle before camera, obtain two relative pose parameters corresponding to image.
Further, in embodiments of the present invention, described matrix operations module 203, statistics remains unknown image and has connected
The number of the point that contact is overlapping, selects the number image more than predetermined threshold value of overlapping point, utilizes 6 method straight linear conversion
Ask and obtain camera matrix, utilize stochastical sampling consistency algorithm that the camera matrix obtained of every time sampling is evaluated, and obtain
Stable camera matrix；Stable camera matrix decomposition is obtained the inside and outside parameter of camera, utilizes known image to carry out front
Intersection, obtains known junction point quantity.
Further, in embodiments of the present invention, described second optimizes module 204, specifically for utilizing optimized algorithm to shadow
As pose parameter and known junction point are iterated optimizing；And according to the principle of calculating limit, limit optimization, until all of image
Relative pose parameter and junction point relative coordinate are obtained and are optimized.
Further, in embodiments of the present invention, this system also includes:
Processing module, for being converted into co-colouration effect by the imaging model tried to achieve by matrix operations；Use light
Bundle method block adjustment method carries out adjustment meter to 6 elements of exterior orientation of every image and the position coordinates of each junction point
Calculate, eliminate error of coordinate.
Although having been described for the preferred embodiment of the application, but one of ordinary skilled in the art once knowing substantially
Creative concept, then can make other change and amendment to these embodiments.So, claims are intended to be construed to bag
Include preferred embodiment and fall into all changes and the amendment of the application scope.
Obviously, those skilled in the art can carry out various change and the modification essence without deviating from the application to the application
God and scope.So, if these amendments of the application and modification belong to the scope of the application claim and equivalent technologies thereof
Within, then the application is also intended to comprise these change and modification.
Claims (10)
1. a measuring method based on multi-vision aviation image, it is characterised in that described method includes:
S1, selects two images in all images, and according to described two images, obtains essential matrix, according to described essence
Matrix, obtains the relative pose parameter that described two images are corresponding；
S2, according to default optimized algorithm, is optimized computing to described relative pose parameter, rejects in described relative pose parameter
Rough error point and error, obtain stable relative pose parameter；
S3, the number of the point that the unknown image of statistics residue is overlapping with junction point, select the number of overlapping point more than presetting
The image of threshold value, and according to described image, obtain camera matrix, by consistency algorithm, described camera matrix is evaluated,
Obtain stable camera matrix, according to stable camera matrix, obtain known junction point quantity；
S4, is optimized the pose parameter obtained and known junction point, and repeats S3, until obtaining the phase of all of image
To pose parameter and junction point relative position coordinates；
S5, carries out compensating computation to every image, obtains final image pose parameter and junction point coordinate.
2. the method for claim 1, it is characterised in that and according to described two images, obtain essential matrix, according to institute
State essential matrix, obtain the relative pose parameter that described two images are corresponding, including:
According to two image plane space geometry restriction relations, utilize 5 methods to solve essential matrix, and utilize stochastical sampling one
The essential matrix obtained of every time sampling is evaluated by cause property algorithm, and obtains stable essential matrix；
Stable essential matrix is carried out singular value decomposition, according to point different solution of the eliminating of principle before camera, obtains two images
Corresponding relative pose parameter.
3. the method for claim 1, it is characterised in that the point that the unknown image of statistics residue is overlapping with junction point
Number, selects the number image more than predetermined threshold value of overlapping point, and according to described image, obtains camera matrix, by one
Described camera matrix is evaluated by cause property algorithm, obtains stable camera matrix, according to stable camera matrix, obtains known
Junction point quantity, including:
The number of the point that the unknown image of statistics residue is overlapping with junction point, selects the number of overlapping point more than predetermined threshold value
Image, utilize the conversion of 6 method straight linear to ask and obtain camera matrix, utilize stochastical sampling consistency algorithm to sampling every time
To camera matrix be evaluated, and obtain stable camera matrix；
Stable camera matrix decomposition is obtained the inside and outside parameter of camera, utilizes known image to carry out forward intersection, obtain
Know junction point quantity.
4. the method for claim 1, it is characterised in that the pose parameter obtained and known junction point are carried out excellent
Change, and repeat S3, until obtaining relative pose parameter and the junction point relative position coordinates of all of image, including:
Optimized algorithm is utilized to be iterated image pose parameter and known junction point optimizing；
Repeat step S3, and the principle optimized according to calculating limit, limit, until the relative pose parameter of all of image and junction point
Relative coordinate is obtained and is optimized.
5. the method for claim 1, it is characterised in that every image is being carried out compensating computation, is obtaining final shadow
After pose parameter and junction point coordinate, described method also includes:
The imaging model tried to achieve is converted into co-colouration effect by matrix operations；
Use bundle block adjustment method that 6 elements of exterior orientation of every image and the position coordinates of each junction point are entered
Row compensating computation, eliminates error of coordinate.
6. a measurement system based on multi-vision aviation image, it is characterised in that including:
Pose parameter acquisition module, for selecting two images in all images, and according to described two images, obtains essence
Matrix, according to described essential matrix, obtains the relative pose parameter that described two images are corresponding；
First optimizes module, for according to presetting optimized algorithm, described relative pose parameter being optimized computing, rejects described
Rough error point in relative pose parameter and error, obtain stable relative pose parameter；
Matrix operations module, for adding up the number of the unknown image of the residue point overlapping with junction point, selects overlapping point
Number more than the image of predetermined threshold value, and according to described image, obtain camera matrix, by consistency algorithm to described camera
Matrix is evaluated, and obtains stable camera matrix, according to stable camera matrix, obtains known junction point quantity；
Second optimizes module, for being optimized the pose parameter obtained and known junction point, until obtaining all of shadow
The relative pose parameter of picture and junction point relative position coordinates；
Modular converter, for every image is carried out compensating computation, obtains final image pose parameter and junction point coordinate.
7. system as claimed in claim 6, it is characterised in that described pose parameter acquisition module, specifically for according to two
Image plane space geometry restriction relation, utilizes 5 methods to solve essential matrix, and utilizes stochastical sampling consistency algorithm to often
The essential matrix that secondary sampling obtains is evaluated, and obtains stable essential matrix；Stable essential matrix is carried out singular value
Decompose, according to point different solution of the eliminating of principle before camera, obtain two relative pose parameters corresponding to image.
8. system as claimed in claim 6, it is characterised in that described matrix operations module, the unknown image of statistics residue with
Through the number of the point of junction point overlap, the number selecting overlapping point is more than the image of predetermined threshold value, utilizes 6 method straight linear
Conversion is asked and is obtained camera matrix, utilizes stochastical sampling consistency algorithm to be evaluated the camera matrix obtained of every time sampling, and
Obtain stable camera matrix；Stable camera matrix decomposition is obtained the inside and outside parameter of camera, utilizes known image to carry out
Forward intersection, obtains known junction point quantity.
9. system as claimed in claim 6, it is characterised in that described second optimizes module, specifically for utilizing optimized algorithm
It is iterated image pose parameter and known junction point optimizing；And according to the principle of calculating limit, limit optimization, until all of shadow
Relative pose parameter and the junction point relative coordinate of picture are obtained and are optimized.
10. system as claimed in claim 6, it is characterised in that also include:
Processing module, for being converted into co-colouration effect by the imaging model tried to achieve by matrix operations；Use flux of light method
Block adjustment method carries out compensating computation to 6 elements of exterior orientation of every image and the position coordinates of each junction point, disappears
Except error of coordinate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610639829.5A CN106289188A (en) | 2016-08-05 | 2016-08-05 | A kind of measuring method based on multi-vision aviation image and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610639829.5A CN106289188A (en) | 2016-08-05 | 2016-08-05 | A kind of measuring method based on multi-vision aviation image and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106289188A true CN106289188A (en) | 2017-01-04 |
Family
ID=57666027
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610639829.5A Pending CN106289188A (en) | 2016-08-05 | 2016-08-05 | A kind of measuring method based on multi-vision aviation image and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106289188A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106949880A (en) * | 2017-03-10 | 2017-07-14 | 中国电建集团昆明勘测设计研究院有限公司 | The method that unmanned plane image local degree of overlapping in area's crosses high disposal is surveyed in height above sea level big rise and fall |
| CN107507277A (en) * | 2017-07-31 | 2017-12-22 | 北京康邦科技有限公司 | Three-dimensional point cloud reconstructing method and device, server and readable storage medium storing program for executing |
| CN107885701A (en) * | 2017-11-07 | 2018-04-06 | 中国科学院电子学研究所 | Conjugate gradient method bundle block adjustment method based on the calculus of variations |
| CN109631850A (en) * | 2019-01-03 | 2019-04-16 | 甘肃大禹九洲测绘地理信息有限公司 | A kind of canted shot relative positioning method based on deep learning |
| CN109949357A (en) * | 2019-02-27 | 2019-06-28 | 武汉大学 | A kind of stereopsis is to relative attitude restoration methods |
| CN110068829A (en) * | 2019-04-11 | 2019-07-30 | 上海快仓智能科技有限公司 | The simplified laser positioning method based on Fast Shapes Matching |
| CN110296686A (en) * | 2019-05-21 | 2019-10-01 | 北京百度网讯科技有限公司 | Localization method, device and the equipment of view-based access control model |
| CN110954068A (en) * | 2019-12-09 | 2020-04-03 | 南京搜天数据科技有限公司 | Urban green land automatic extraction method based on unmanned aerial vehicle data |
| CN111222586A (en) * | 2020-04-20 | 2020-06-02 | 广州都市圈网络科技有限公司 | Inclined image matching method and device based on three-dimensional inclined model visual angle |
| WO2021114434A1 (en) * | 2019-12-11 | 2021-06-17 | 上海交通大学 | Pure pose solution method and system for multi-view camera pose and scene |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102721410A (en) * | 2012-06-20 | 2012-10-10 | 唐粮 | Island-air triangular measuring method based on GPS/IMU positioning and orientating technology |
| CN105783881A (en) * | 2016-04-13 | 2016-07-20 | 西安航天天绘数据技术有限公司 | Aerial triangulation method and device |
-
2016
- 2016-08-05 CN CN201610639829.5A patent/CN106289188A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102721410A (en) * | 2012-06-20 | 2012-10-10 | 唐粮 | Island-air triangular measuring method based on GPS/IMU positioning and orientating technology |
| CN105783881A (en) * | 2016-04-13 | 2016-07-20 | 西安航天天绘数据技术有限公司 | Aerial triangulation method and device |
Non-Patent Citations (2)
| Title |
|---|
| 周春艳，等: "基于五点算法估计基础矩阵的研究", 《计算机技术与发展》 * |
| 胡凌山: "计算机视觉中基本矩阵的估计方法", 《应用科技》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106949880B (en) * | 2017-03-10 | 2019-03-19 | 中国电建集团昆明勘测设计研究院有限公司 | The method that area's unmanned plane image part degree of overlapping crosses high disposal is surveyed in height above sea level big rise and fall |
| CN106949880A (en) * | 2017-03-10 | 2017-07-14 | 中国电建集团昆明勘测设计研究院有限公司 | The method that unmanned plane image local degree of overlapping in area's crosses high disposal is surveyed in height above sea level big rise and fall |
| CN107507277A (en) * | 2017-07-31 | 2017-12-22 | 北京康邦科技有限公司 | Three-dimensional point cloud reconstructing method and device, server and readable storage medium storing program for executing |
| CN107507277B (en) * | 2017-07-31 | 2021-04-06 | 北京康邦科技有限公司 | Three-dimensional point cloud reconstruction method and device, server and readable storage medium |
| CN107885701B (en) * | 2017-11-07 | 2021-02-09 | 中国科学院电子学研究所 | Conjugate gradient method light beam method block adjustment method based on variational method |
| CN107885701A (en) * | 2017-11-07 | 2018-04-06 | 中国科学院电子学研究所 | Conjugate gradient method bundle block adjustment method based on the calculus of variations |
| CN109631850A (en) * | 2019-01-03 | 2019-04-16 | 甘肃大禹九洲测绘地理信息有限公司 | A kind of canted shot relative positioning method based on deep learning |
| CN109949357A (en) * | 2019-02-27 | 2019-06-28 | 武汉大学 | A kind of stereopsis is to relative attitude restoration methods |
| CN109949357B (en) * | 2019-02-27 | 2022-07-05 | 武汉大学 | Method for recovering relative posture of stereo image pair |
| CN110068829A (en) * | 2019-04-11 | 2019-07-30 | 上海快仓智能科技有限公司 | The simplified laser positioning method based on Fast Shapes Matching |
| CN110296686A (en) * | 2019-05-21 | 2019-10-01 | 北京百度网讯科技有限公司 | Localization method, device and the equipment of view-based access control model |
| CN110954068A (en) * | 2019-12-09 | 2020-04-03 | 南京搜天数据科技有限公司 | Urban green land automatic extraction method based on unmanned aerial vehicle data |
| WO2021114434A1 (en) * | 2019-12-11 | 2021-06-17 | 上海交通大学 | Pure pose solution method and system for multi-view camera pose and scene |
| US20230041433A1 (en) * | 2019-12-11 | 2023-02-09 | Shanghai Jiao Tong University | Pure pose solution method and system for multi-view camera pose and scene |
| CN111222586A (en) * | 2020-04-20 | 2020-06-02 | 广州都市圈网络科技有限公司 | Inclined image matching method and device based on three-dimensional inclined model visual angle |
| CN111222586B (en) * | 2020-04-20 | 2020-09-18 | 广州都市圈网络科技有限公司 | Inclined image matching method and device based on three-dimensional inclined model visual angle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106289188A (en) | A kind of measuring method based on multi-vision aviation image and system | |
| CN109816703B (en) | Point cloud registration method based on camera calibration and ICP algorithm | |
| CN112785702B (en) | SLAM method based on tight coupling of 2D laser radar and binocular camera | |
| Koutsoudis et al. | Multi-image 3D reconstruction data evaluation | |
| CN105976312B (en) | Point cloud autoegistration method based on point feature histogram | |
| Merrell et al. | Real-time visibility-based fusion of depth maps | |
| CN106485690A (en) | Cloud data based on a feature and the autoregistration fusion method of optical image | |
| CN107358633A (en) | Join scaling method inside and outside a kind of polyphaser based on 3 points of demarcation things | |
| CN104677277B (en) | A kind of method and system for measuring object geometric attribute or distance | |
| CN104240289A (en) | Three-dimensional digitalization reconstruction method and system based on single camera | |
| WO2023046211A1 (en) | Photogrammetry method, apparatus and device, and storage medium | |
| CN105701787B (en) | Depth map fusion method based on confidence level | |
| CN108399631B (en) | Scale invariance oblique image multi-view dense matching method | |
| Gadasin et al. | Reconstruction of a Three-Dimensional Scene from its Projections in Computer Vision Systems | |
| Ahmadabadian et al. | Image selection in photogrammetric multi-view stereo methods for metric and complete 3D reconstruction | |
| KR100623666B1 (en) | Method for manufacturing three-dimensional geographic information system digital image drawing using digital photommetry and non prism total station | |
| Ylimäki et al. | Accurate 3-d reconstruction with rgb-d cameras using depth map fusion and pose refinement | |
| GB2569609A (en) | Method and device for digital 3D reconstruction | |
| CN107194334B (en) | Video satellite image dense Stereo Matching method and system based on optical flow estimation | |
| Cavegn et al. | Evaluation of Matching Strategies for Image-Based Mobile Mapping | |
| Wang et al. | What can we learn about the scene structure from three orthogonal vanishing points in images | |
| Thanusutiyabhorn et al. | Image-based 3D laser scanner | |
| Previtali et al. | An automatic multi-image procedure for accurate 3D object reconstruction | |
| Jiang et al. | A multi-view structured light measurement method based on pose estimation using deep learning | |
| Pollefeys et al. | Acquisition of detailed models for virtual reality |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170104 |