CN117197231A - Feature point uncertainty-based monocular camera pose estimation method - Google Patents
Feature point uncertainty-based monocular camera pose estimation method Download PDFInfo
- Publication number
- CN117197231A CN117197231A CN202310666923.XA CN202310666923A CN117197231A CN 117197231 A CN117197231 A CN 117197231A CN 202310666923 A CN202310666923 A CN 202310666923A CN 117197231 A CN117197231 A CN 117197231A
- Authority
- CN
- China
- Prior art keywords
- matrix
- uncertainty
- feature points
- feature point
- equation model
- 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
- 238000000034 method Methods 0.000 title claims abstract description 42
- 239000011159 matrix material Substances 0.000 claims abstract description 95
- 238000005259 measurement Methods 0.000 claims abstract description 33
- 230000009466 transformation Effects 0.000 claims abstract description 29
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims abstract description 27
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 18
- 238000009826 distribution Methods 0.000 claims abstract description 17
- 238000013519 translation Methods 0.000 claims abstract description 14
- 238000004590 computer program Methods 0.000 claims description 14
- 238000003384 imaging method Methods 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 10
- 241001270131 Agaricus moelleri Species 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 description 22
- 238000010586 diagram Methods 0.000 description 7
- 238000012545 processing Methods 0.000 description 4
- 230000006978 adaptation Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The application discloses a monocular camera pose estimation method based on feature point uncertainty, which comprises the following steps: constructing a covariance inverse matrix of the measurement error of the feature points according to the pixel gray level distribution information of the feature point data space; singular value decomposition is carried out on the covariance inverse matrix of the characteristic points, and an affine transformation matrix of the characteristic points is obtained; carrying out uncertainty weighting, constructing a weighted algebraic error function, obtaining a linear equation model according to the weighted algebraic error function, and rewriting the linear equation model into a nonlinear equation model; and obtaining consistent noise variance, solving the nonlinear equation model based on the consistent noise variance to obtain a closed solution with consistent deviation, obtaining an estimated value of the rotation matrix and an estimated value of the translation vector according to the closed solution, and outputting a camera pose estimated result. The uncertainty of the feature points is considered, asymptotic deviation is eliminated, and the precision and efficiency of pose estimation are improved.
Description
Technical Field
The application relates to a monocular camera pose estimation method based on feature point uncertainty, and belongs to the technical field of vision measurement.
Background
Camera pose estimation has wide application in three-dimensional reconstruction, autopilot, camera calibration, augmented reality, photogrammetry. The pose of a camera may be described by a rotation matrix and a translation vector, which are also referred to as the camera's extrinsic parameters. The external parameters of the camera are solved by establishing a 3D-2D matching relation of the feature points through a world coordinate system and a plane pixel coordinate system, and the process is called PnP problem.
Most of the existing PnP algorithms are used for obtaining the optimal solution of the pose through establishing algebraic error functions and iteration algorithms. However, these algorithms ignore the observation errors of the feature points and do not take the uncertainty of the feature points into account. In the PnP algorithm considering uncertainty of feature points, the observation errors are generally isotropic and distributed independently and uniformly. In actual feature point extraction, since the gray distribution pattern around the feature point is greatly different, the error is often anisotropic and is not independently distributed.
Disclosure of Invention
The purpose is as follows: in order to overcome the defects in the prior art, the application provides a monocular camera pose estimation method based on feature point uncertainty.
The technical scheme is as follows: in order to solve the technical problems, the application adopts the following technical scheme:
in a first aspect, the present application provides a method for estimating a pose of a camera based on feature point uncertainty, including:
extracting uncertain information of the characteristic points according to pixel gray level distribution information of the characteristic point data space, and constructing a covariance inverse matrix Q of measurement errors of the characteristic points -1 ；
Covariance inverse matrix Q for feature points -1 Performing singular value decomposition to obtain an affine transformation matrix F of the characteristic points;
carrying out uncertainty weighting on the feature points based on an affine transformation matrix F of the feature points, and constructing a weighted algebraic error function;
obtaining a linear equation model according to the weighted algebraic error function;
rewriting the linear equation model into a nonlinear equation model;
based on nonlinear equation modeObtaining consistent noise variance, and solving a nonlinear equation model based on the consistent noise variance to obtain a closed solution with consistent deviation
According to the closed form solutionObtaining an estimate of the rotation matrix +.>And an estimate of the translation vector +.>And outputting a camera pose estimation result.
In some embodiments, extracting uncertainty information of the feature points according to pixel gray scale distribution information of the feature point data space, constructing a covariance inverse matrix of measurement errors of the feature points, including:
covariance inverse matrix Q of measurement error through feature points -1 Describing uncertainty of feature points, Q -1 The modeling of (2) is as follows:
wherein Q is a covariance matrix of the measurement error, W is an elliptical region centered on the feature point, W (u, v) is a sum of pixel grayscales of the elliptical region,and->The gradients of the image in the u and v directions, respectively.
In some embodiments, the covariance inverse matrix Q for the feature points -1 Performing singular value decomposition to obtain an affine transformation matrix F of the feature points, wherein the affine transformation matrix F comprises the following steps:
for Q -1 Singular value decomposition is carried out to obtain Q -1 ＝U∑ -1 U T Wherein, the method comprises the steps of, wherein,
affine transformation matrix
Wherein sigma 1 Sum sigma 2 Representing uncertainty of feature points, Σ -1/2 Transforming uncertainty ellipses into unit circles, U T Is a rotation matrix that rotates the tilted uncertainty ellipse to an ellipse that is the same as the direction of the image planes u and v;
the feature point data space is projected into the weighted covariance space through an affine transformation matrix F, and the measurement error of the feature points is converted from anisotropic distribution to isotropic distribution.
In some embodiments, uncertainty weighting the feature points based on an affine transformation matrix F of the feature points, constructing a weighted algebraic error function, comprising:
the camera aperture imaging model is:
wherein,the unit is pixel, K is known quantity; q i ＝[u i v i ] T For the i-th pixel plane feature point, +.>Is a reference point in the world coordinate system; d, d i Representing depth information; r and t are external parameter information to be solved, and respectively represent a rotation matrix and a translation vector;
based on an affine transformation matrix F of the characteristic points, the camera aperture imaging model is rewritten into the following by taking the measurement errors of the characteristic points into consideration:
[u 0 v 0 ] T it is known to simplify the above equation and weight the feature points to obtain the observation equation:
the weighted algebraic error function is:
wherein f x For the focal length f of the pixel in the x direction y For the focal length of the pixel in the y direction, u 0 Is the center in the u direction of the pixel plane, v 0 Epsilon is the center in the v direction of the pixel plane i Measurement error, W-finger matrixE indicates matrix-> Measurement error representing characteristic point of ith pixel plane,/->Representing the i-th weighted pixel plane feature point, F i An affine transformation matrix representing the i-th pixel plane feature point.
In some embodiments, deriving the linear equation model from the weighted algebraic error function includes:
on both sides of the weighted algebraic error functionAt the same time multiply d i According to the camera aperture imaging model,intermediate parameter e= [0 0 0 1] T Obtaining a measurement equation:
and (3) rewriting the measurement equation into a matrix form to obtain a linear equation model:
wherein M is a coefficient matrix, x is composed of external parameters R and t,is an error term;wherein->Representing tensor product of matrix, I 3 Representing a 3 x 3 identity matrix.
In some embodiments, the overwriting of the linear equation model to the nonlinear equation model includes:
introducing constraintsAnd->To eliminate scale ambiguity; the rotation matrix R is denoted as [ R1R 2R 3 ]] T Translation vector t= [ t ] 1 t 2 t 3 ] T Get->The linear equation model is carried in, and the method comprises the following steps:
and (3) rewriting the model into a matrix form to obtain a nonlinear equation model:
Aθ+η＝b
wherein,is a vector of weighted feature points, +.>Eta is an error term, theta is an unknown quantity to be solved for, < ->A is a coefficient matrix:
wherein alpha is a scale factor, n is the number of feature points,Centroid coordinates representing feature points in world coordinate system, +.>Representing the coordinates of the i-th weighted pixel plane feature point in the u-direction, +.>Representing the coordinates of the i-th weighted pixel plane feature point in the v-direction.
In some embodiments, a consistent noise variance is obtained based on the nonlinear equation model, and the nonlinear equation model is solved based on the consistent noise variance to obtain a closed-form solution with consistent biasComprising the following steps:
construction of matrixDefinition matrix->And
the barycenter coordinates of the feature points in the world coordinate system are represented, A is a coefficient matrix, and b is a vector formed by weighted feature points;
constructing a function H (lambda) =phi-lambda delta, wherein lambda is a generalized eigenvalue, and solving to obtain a consistent noise variance
Solving the nonlinear equation model based on the consistent noise variance to obtain a closed solution with consistent deviation:
in some embodiments, according to the closed form solutionObtaining an estimate of the rotation matrix +.>And an estimate of the translation vector +.>Comprising the following steps:
further, the method further comprises the following steps: by singular value decompositionProjection into the space of the prune group, the formula is as follows:
wherein the method comprises the steps ofThe representation will->Projection formula projected into the lie space, diag diagonal matrix, det determinant, assuming ∈ ->The singular value decomposition result of +.>Wherein->And->Are parameters in the singular value decomposition result.
In a second aspect, the application provides a camera pose estimation device based on feature point uncertainty, which comprises a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the method according to the first aspect.
In a third aspect, the application provides an apparatus comprising,
a memory;
a processor;
and
a computer program;
wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method of the first aspect described above.
In a fourth aspect, the present application provides a storage medium having stored thereon a computer program which, when executed by a processor, implements the method of the first aspect.
The beneficial effects are that: the monocular camera pose estimation method based on feature point uncertainty provided by the application has the following advantages: according to the application, anisotropic noise of the feature points is considered, and an algebraic error function based on the feature point uncertainty weighting coefficient is used to establish a monocular camera pose solving model based on the feature point uncertainty weighting coefficient; and deducing a solution algorithm with consistent deviation; an uncertainty model is built through image gray level distribution information around the feature points, the uncertainty model is integrated into an objective function for pose solving, a least square solution for eliminating deviation is introduced, asymptotic deviation of the feature points is eliminated, accuracy of an algorithm is improved, and the pose of a camera is accurately and reliably solved.
Drawings
FIG. 1 is a flow chart of a method according to an embodiment of the application.
FIG. 2 is a schematic diagram of feature point uncertainty in one embodiment of the present application.
Detailed Description
The application is further described below with reference to the drawings and examples. The following examples are only for more clearly illustrating the technical aspects of the present application, and are not intended to limit the scope of the present application.
In the description of the present application, the meaning of a number is one or more, the meaning of a number is two or more, and greater than, less than, exceeding, etc. are understood to exclude the present number, and the meaning of a number is understood to include the present number. The description of the first and second is for the purpose of distinguishing between technical features only and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of the technical features indicated.
In the description of the present application, the descriptions of the terms "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Example 1
In a first aspect, as shown in fig. 1, the present embodiment provides a method for estimating a pose of a camera based on uncertainty of feature points, including:
s1, extracting uncertain information of the feature points according to pixel gray level distribution information of a feature point data space, and constructing a covariance inverse matrix Q of measurement errors of the feature points -1 ；
S2 covariance inverse matrix Q of characteristic points -1 Performing singular value decomposition to obtain an affine transformation matrix F of the characteristic points;
s3, carrying out uncertainty weighting on the feature points based on an affine transformation matrix F of the feature points, and constructing a weighted algebraic error function;
s4, obtaining a linear equation model according to the weighted algebraic error function;
s5, rewriting the linear equation model into a nonlinear equation model;
s6, obtaining consistent noise variance based on the nonlinear equation model, and solving the nonlinear equation model based on the consistent noise variance to obtain a closed solution with consistent deviation
S7 according to the closed solutionObtaining an estimate of the rotation matrix +.>And an estimate of the translation vector +.>And outputting a camera pose estimation result.
In some embodiments, step S1 extracts uncertainty information of the feature points according to pixel gray distribution information of the feature point data space, and constructs a covariance inverse matrix of measurement errors of the feature points, including:
in the actual visual measurement, different feature points have different gray pattern distributions on the imaging plane, and the directionality of the gray pattern distribution is introduced when the image point is extracted and reflected in the u and v directions, so that the uncertainty of the image point error can be used for representing the anisotropy and the non-independent identical distribution of the extraction error. Different image point errors have different magnitudes, anisotropy distributions, and uncertainties. Q (Q) -1 Described as feature point q i ＝[u i v i ] T An uncertainty ellipse with a circle center, and long and short axes a and b of the ellipse represent a characteristic point q i The angles of the major axis a and the minor axis b relative to the directions u and v represent the uncertainty direction. Mainly shown in FIG. 2Three different types of image point measurement error uncertainties. If the image point error is in the case of fig. 2 (II), this means that the uncertainty of the feature point is isotropic and there is no directional uncertainty, only scale uncertainty. If all the pixel errors are the cases of fig. 2 (I) and fig. 2 (III), the uncertainty of the feature points is directional, and the uncertainty of the pixel error direction in the actual situation needs to be considered.
Q -1 The modeling of (2) is as follows:
wherein Q is a covariance matrix of the measurement error, W is an elliptical region centered on the feature point, W (u, v) is a sum of pixel grayscales of the elliptical region,and->The gradients of the image in the u and v directions, respectively.
In some embodiments, step S2 is performed on the covariance inverse matrix Q of the feature points -1 Performing singular value decomposition to obtain an affine transformation matrix F of the feature points, wherein the affine transformation matrix F comprises the following steps:
will Q -1 Singular value decomposition is carried out to obtain Q -1 ＝U∑ -1 U T Wherein, the method comprises the steps of, wherein,defining a transformation matrix->
Wherein sigma 1 Sum sigma 2 Representing uncertainty of feature points, Σ -1/2 Transforming uncertainty ellipses into unit circles, U T Is a rotation matrix that rotates the tilted uncertainty ellipse to the same ellipse as the image planes u and v.
F is a 2 x 2 affine transformation matrix that functions to project the feature point coordinates into the weighted covariance space through transformation F, where the error can be considered isotropic.
In some embodiments, step S3 performs uncertainty weighting on the feature points based on the affine transformation matrix F of the feature points, and constructs a weighted algebraic error function, including:
establishing an algebraic measurement error function according to the camera aperture imaging model; the camera aperture imaging model is as follows:
wherein,is a camera reference matrix, and the unit is a pixel. In the present application, K is a known amount. q i ＝[u i v i ] T Is pixel plane feature point +.>Is a reference point in the world coordinate system. d, d i Representing depth information. R and t are solved extrinsic information, and respectively represent a rotation matrix and a translation vector;
based on an affine transformation matrix F of the characteristic points, the camera aperture imaging model is rewritten into the following by taking the measurement errors of the characteristic points into consideration:
due to [ u ] 0 v 0 ] T It is known that the above equation can be simplified, and the feature points are weighted, so as to obtain the observation equation:
the weighted algebraic error function is:
wherein f x For the focal length f of the pixel in the x direction y For the focal length of the pixel in the y direction, u 0 Is the center in the u direction of the pixel plane, v 0 Epsilon is the center in the v direction of the pixel plane i Measurement error, W-finger matrixE indicates matrix-> Measurement error representing characteristic point of ith pixel plane,/->Representing the i-th weighted pixel plane feature point, F i An affine transformation matrix representing the i-th pixel plane feature point.
The step S4 obtains a linear equation model according to the weighted algebraic error function, and the method comprises the following steps:
multiplying d simultaneously on both sides of a weighted algebraic error function i According to the camera aperture imaging model,intermediate parameter e= [0 0 0 1] T Obtaining a measurement equation:
and (3) rewriting the measurement equation into a matrix form to obtain a linear equation model:
wherein M is a coefficient matrix, x is composed of external parameters R and t,is an error term;wherein->Representing tensor product of matrix, I 3 Representing a 3 x 3 identity matrix.
In some embodiments, step S5 rewrites the linear equation model to a nonlinear equation model, comprising:
introducing constraintsAnd->To eliminate scale ambiguity; the rotation matrix R is denoted as [ R1R 2R 3 ]] T Translation vector t= [ t ] 1 t 2 t 3 ] T Can get->The linear equation model is carried in, and the method comprises the following steps:
and (3) rewriting the model into a matrix form to obtain a nonlinear equation model:
Aθ+η＝b
wherein,is a vector of weighted feature points, +.>Eta is an error term, theta is an unknown quantity to be solved for, < ->A is a coefficient matrix:
wherein alpha is a scale factor, n is the number of feature points,Centroid coordinates representing feature points in world coordinate system, +.>Representing the coordinates of the i-th weighted pixel plane feature point in the u-direction, +.>Representing the coordinates of the i-th weighted pixel plane feature point in the v-direction.
In some embodiments, step S6 obtains a consistent noise variance based on the nonlinear equation model, and solves the nonlinear equation model based on the consistent noise variance to obtain a closed-form solution with consistent biasComprising the following steps:
construction of matrixDefinition matrix->And the barycenter coordinates of the feature points in the world coordinate system are represented, A is a coefficient matrix, and b is a vector formed by weighted feature points;
constructing a function H (lambda) =phi-lambda delta, wherein lambda is a generalized eigenvalue, and solving the generalized eigenvalue to obtain a consistent noise variance
The closed-form solution of the non-linear system to eliminate bias is as follows:
in some embodiments, step S7 is based on the closed-form solutionObtaining an estimate of the rotation matrix +.>And an estimate of the translation vector +.>Comprising the following steps:
due to the rotation matrixDoes not necessarily satisfy->Is a condition of (2). Therefore, it is necessary to divide +.>Projecting into the space of the plum clusters, wherein the method comprises the following steps:
wherein the method comprises the steps ofThe representation will->Projection formula projected into the lie space, diag diagonal matrix, det determinant, assuming ∈ ->The singular value decomposition result of +.>Wherein->And->Are parameters in the singular value decomposition result.
Example 2
In a second aspect, based on embodiment 1, the present embodiment provides a camera pose estimation device based on feature point uncertainty, including a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the method according to embodiment 1.
Example 3
In a third aspect, based on embodiment 1, the present embodiment provides an apparatus, comprising,
a memory;
a processor;
and
a computer program;
wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method of embodiment 1.
Example 4
In a fourth aspect, based on embodiment 1, the present embodiment provides a storage medium having stored thereon a computer program which, when executed by a processor, implements the method described in embodiment 1.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing is only a preferred embodiment of the application, it being noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present application, and such modifications and adaptations are intended to be comprehended within the scope of the application.
Claims (10)
1. A camera pose estimation method based on feature point uncertainty, the method comprising:
extracting uncertain information of the characteristic points according to pixel gray level distribution information of the characteristic point data space, and constructing a covariance inverse matrix Q of measurement errors of the characteristic points -1 ；
Covariance inverse matrix Q for feature points -1 Performing singular value decomposition to obtain an affine transformation matrix F of the characteristic points;
carrying out uncertainty weighting on the feature points based on an affine transformation matrix F of the feature points, and constructing a weighted algebraic error function;
obtaining a linear equation model according to the weighted algebraic error function;
rewriting the linear equation model into a nonlinear equation model;
obtaining consistent noise variance based on nonlinear equation model, and solving the nonlinear equation model based on consistent noise variance to obtain deviationConsistent closed form solution
According to the closed form solutionObtaining an estimate of the rotation matrix +.>And an estimate of the translation vector +.>And outputting a camera pose estimation result.
2. The camera pose estimation method based on feature point uncertainty according to claim 1, wherein extracting uncertainty information of feature points according to pixel gray distribution information of feature point data space, constructing covariance inverse matrix of measurement error of feature points, comprises:
covariance inverse matrix Q of measurement error through feature points -1 Describing uncertainty of feature points, Q -1 The modeling of (2) is as follows:
wherein Q is a covariance matrix of the measurement error, W is an elliptical region centered on the feature point, W (u, v) is a sum of pixel grayscales of the elliptical region,and->The gradients of the image in the u and v directions, respectively.
3. The camera pose estimation method based on feature point uncertainty as claimed in claim 1, wherein covariance inverse matrix Q for feature points -1 Performing singular value decomposition to obtain an affine transformation matrix F of the feature points, wherein the affine transformation matrix F comprises the following steps:
for Q -1 Singular value decomposition is carried out to obtain Q -1 ＝U∑ -1 U T Wherein, the method comprises the steps of, wherein,
affine transformation matrix
Wherein sigma 1 Sum sigma 2 Representing uncertainty of feature points, Σ -1/2 Transforming uncertainty ellipses into unit circles, U T Is a rotation matrix that rotates the tilted uncertainty ellipse to an ellipse that is the same as the direction of the image planes u and v;
the feature point data space is projected into the weighted covariance space through an affine transformation matrix F, and the measurement error of the feature points is converted from anisotropic distribution to isotropic distribution.
4. The camera pose estimation method based on feature point uncertainty of claim 1, wherein uncertainty weighting is performed on feature points based on an affine transformation matrix F of the feature points, and constructing a weighted algebraic error function comprises:
the camera aperture imaging model is:
wherein,the unit is pixel, K is known quantity; q i ＝[u i v i ] T For the i-th pixel plane feature point, +.>Is a reference point in the world coordinate system; d, d i Representing depth information; r and t are external parameter information to be solved, and respectively represent a rotation matrix and a translation vector;
based on an affine transformation matrix F of the characteristic points, the camera aperture imaging model is rewritten into the following by taking the measurement errors of the characteristic points into consideration:
[u 0 v 0 ] T it is known to simplify the above equation and weight the feature points to obtain the observation equation:
the weighted algebraic error function is:
wherein f x For the focal length f of the pixel in the x direction y For the focal length of the pixel in the y direction, u 0 Is the center in the u direction of the pixel plane, v 0 Epsilon is the center in the v direction of the pixel plane i Measurement error, W-finger matrixE indicates matrix->Measurement error representing characteristic point of ith pixel plane,/->Representing the i-th weighted pixel plane feature point, F i An affine transformation matrix representing the i-th pixel plane feature point.
5. The method for estimating camera pose based on feature point uncertainty as claimed in claim 4, wherein obtaining a linear equation model from a weighted algebraic error function comprises:
multiplying d simultaneously on both sides of a weighted algebraic error function i According to the camera aperture imaging model,intermediate parameter e= [0 0 0 1] T Obtaining a measurement equation:
and (3) rewriting the measurement equation into a matrix form to obtain a linear equation model:
wherein M is a coefficient matrix, x is composed of external parameters R and t,is an error term;wherein->Representing tensor product of matrix, I 3 Representing a 3 x 3 identity matrix.
6. The feature point uncertainty-based camera pose estimation method according to claim 5, wherein rewriting the linear equation model into the nonlinear equation model comprises:
introducing constraintsAnd->To eliminate scale ambiguity; the rotation matrix R is denoted as [ R1R 2R 3 ]] T Translation vector t= [ t ] 1 t 2 t 3 ] T Get->The linear equation model is carried in, and the method comprises the following steps:
and (3) rewriting the model into a matrix form to obtain a nonlinear equation model:
Aθ+η＝b
wherein,is a vector of weighted feature points, +.>Eta is an error term, theta is an unknown quantity to be solved for, < ->A is a coefficient matrix:
wherein alpha is a scale factor, n is the number of feature points,Centroid coordinates representing feature points in world coordinate system, +.>Representing the coordinates of the i-th weighted pixel plane feature point in the u-direction, +.>Representing the coordinates of the i-th weighted pixel plane feature point in the v-direction.
7. The camera pose estimation method based on feature point uncertainty as claimed in claim 1, wherein the consistent noise variance is obtained based on a nonlinear equation model, and the nonlinear equation model is solved based on the consistent noise variance to obtain a closed-form solution with consistent deviationComprising the following steps:
construction of matrixDefinition matrix->Andthe barycenter coordinates of the feature points in the world coordinate system are represented, A is a coefficient matrix, and b is a vector formed by weighted feature points;
constructing a function H (lambda) =phi-lambda delta, wherein lambda is a generalized eigenvalue, and solving to obtain a consistent noise variance
Solving the nonlinear equation model based on the consistent noise variance to obtain a closed solution with consistent deviation:
8. the feature point uncertainty-based camera pose estimation method according to claim 1, wherein the closed-form solution is based onObtaining an estimate of the rotation matrix +.>And an estimate of the translation vector +.>Comprising the following steps:
9. the feature point uncertainty-based camera pose estimation method of claim 8, further comprising: by singular value decompositionProjection into the space of the prune group, the formula is as follows:
wherein the method comprises the steps ofThe representation will->Projection formula projected into the lie space, diag diagonal matrix, det determinant, assuming ∈ ->The singular value decomposition result of +.>Wherein->And->Are parameters in the singular value decomposition result.
10. A storage medium having stored thereon a computer program which, when executed by a processor, implements the method of any of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310666923.XA CN117197231A (en) | 2023-06-07 | 2023-06-07 | Feature point uncertainty-based monocular camera pose estimation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310666923.XA CN117197231A (en) | 2023-06-07 | 2023-06-07 | Feature point uncertainty-based monocular camera pose estimation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117197231A true CN117197231A (en) | 2023-12-08 |
Family
ID=89000439
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310666923.XA Pending CN117197231A (en) | 2023-06-07 | 2023-06-07 | Feature point uncertainty-based monocular camera pose estimation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117197231A (en) |
-
2023
- 2023-06-07 CN CN202310666923.XA patent/CN117197231A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Heimann et al. | 3D active shape models using gradient descent optimization of description length | |
| US10380413B2 (en) | System and method for pose-invariant face alignment | |
| US6434278B1 (en) | Generating three-dimensional models of objects defined by two-dimensional image data | |
| US20180225823A1 (en) | Adversarial and Dual Inverse Deep Learning Networks for Medical Image Analysis | |
| CN111723691B (en) | Three-dimensional face recognition method and device, electronic equipment and storage medium | |
| JP4985516B2 (en) | Information processing apparatus, information processing method, and computer program | |
| CN101377812B (en) | Method for recognizing position and attitude of space plane object | |
| CN110335297A (en) | A kind of point cloud registration method based on feature extraction | |
| CN113888689A (en) | Image rendering model training method, image rendering method and image rendering device | |
| CN112651944A (en) | 3C component high-precision six-dimensional pose estimation method and system based on CAD model | |
| CN117197159A (en) | Efficient and high-quality high-temperature object point cloud segmentation method | |
| CN108830888A (en) | Thick matching process based on improved multiple dimensioned covariance matrix Feature Descriptor | |
| CN112364881B (en) | Advanced sampling consistency image matching method | |
| Heitz et al. | Statistical shape model generation using nonrigid deformation of a template mesh | |
| CN113658194B (en) | Point cloud splicing method and device based on reference object and storage medium | |
| CN111178429A (en) | System and method for providing medical guidance using patient depth images | |
| Liu | Automatic range image registration in the markov chain | |
| CN117197231A (en) | Feature point uncertainty-based monocular camera pose estimation method | |
| CN114022526B (en) | SAC-IA point cloud registration method based on three-dimensional shape context | |
| CN100590658C (en) | Method for matching two dimensional object point and image point with bilateral constraints | |
| Kirschner et al. | Construction of groupwise consistent shape parameterizations by propagation | |
| CN110264562A (en) | Skull model characteristic point automatic calibration method | |
| CN112307971B (en) | Sphere target identification method and device based on three-dimensional point cloud data | |
| CN115511935A (en) | Normal distribution transformation point cloud registration method based on iterative discretization and linear interpolation | |
| CN114419255A (en) | Three-dimensional human head model generation method and device fusing real human faces, electronic equipment and storage medium |
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 |