CN108416811B - Camera self-calibration method and device - Google Patents
Camera self-calibration method and device Download PDFInfo
- Publication number
- CN108416811B CN108416811B CN201810189115.8A CN201810189115A CN108416811B CN 108416811 B CN108416811 B CN 108416811B CN 201810189115 A CN201810189115 A CN 201810189115A CN 108416811 B CN108416811 B CN 108416811B
- Authority
- CN
- China
- Prior art keywords
- feature point
- adjacent images
- matching
- matrix
- subunit
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a camera self-calibration method and a device, comprising the following steps: and acquiring a feature point matching pair in two adjacent images. And estimating initial values of the internal parameters by using the property that the first two singular values of the intrinsic matrix are equal according to the acquired feature point matching pairs. And correcting the initial value of the internal parameter according to a Kruppa equation and outputting a result. The method comprises the steps of estimating initial values of internal parameters by utilizing the characteristic that first two singular values of an essential matrix are equal and through a particle swarm algorithm, substituting the estimated initial values of the internal parameters into a target function constructed according to a Kruppa equation, optimizing the target function by adopting an internal point penalty function method and combining a variable-scale method, and finally obtaining the optimized internal parameters. The method combines the self-calibration technology based on the intrinsic matrix and the self-calibration technology based on the Kruppa equation, so that the calibration of the internal parameters is performed by two steps of initial value calculation and correction, and the accuracy and the stability of the calibration of the internal parameters are effectively ensured.
Description
Technical Field
The application relates to the technical field of computer vision, in particular to a camera self-calibration method and device.
Background
One of the basic tasks of computer vision is to calculate the geometric information of an object in three-dimensional space based on the image information acquired by a camera, and to reconstruct and identify the object therefrom, and the correlation between the three-dimensional geometric position of a point on the surface of the object in space and the corresponding point in the image is determined by the geometric model imaged by the camera, and the parameters of the geometric model are the parameters of the camera. In most cases, these parameters must be obtained through experiments and calculations, and this process is called camera calibration (or calibration). The calibration process is to determine the geometric and optical parameters of the camera, i.e. the internal parameters, and the orientation of the camera with respect to the world coordinate system, i.e. the external parameters. In a word, camera calibration is an indispensable step for extracting three-dimensional spatial information from a two-dimensional image in the field of computer vision, and is widely applied to the fields of three-dimensional reconstruction, navigation, visual monitoring and the like. Camera calibration can be broadly divided into three categories: the method comprises a traditional calibration method, a calibration method based on active vision and a self-calibration method. The traditional calibration method needs to use a calibration block which is subjected to precision machining, and calculates internal and external parameters of the camera by establishing a corresponding relation between a point with a known three-dimensional coordinate on the calibration block and an image point of the calibration block; the method has the advantages that higher precision can be obtained, but the calibration process is time-consuming and labor-consuming, and is not suitable for on-line calibration and occasions where the use of a calibration block is impossible. The calibration method based on active vision needs to control the camera to do some special motions, such as rotation around the optical center or pure translation, and the internal parameters can be calculated by utilizing the particularity of the motions; the method has the advantages that the algorithm is simple, linear solution can be obtained frequently, and the method is not suitable for occasions where the motion trail of the camera is unknown or cannot be controlled. The two calibration methods all use information of special scenes or camera motion, and cannot be used in the most common situation that the scenes are arbitrary and the motion trail of the camera is unknown.
In order to obtain the intrinsic parameters of the camera under the condition that neither scene information nor camera motion information is known, a self-calibration method is proposed. Some have demonstrated that two quadratic nonlinear constraints in the form of Kruppa equations exist between every two images from the perspective of projective geometry, and the internal parameters can be solved by directly solving the Kruppa equations. In view of the difficulty of directly solving the Kruppa equation, people also provide the idea of layered gradual calibration, namely firstly performing projective reconstruction on an image sequence, and then performing affine calibration and Euclidean calibration on the basis. The self-calibration method utilizes the constraints of the parameters in the camera to carry out calibration, and the constraints are irrelevant to the scene and the motion of the camera, which is the reason that the self-calibration method is more flexible than the former two calibration methods.
Although the self-calibration method based on the Kruppa equation establishes an equation between every two images, all the images are unified to a consistent shooting frame, so that the calibration of the camera is more convenient. However, in the method, the parameter of the infinity plane supporting the absolute quadratic curve needs to be eliminated in the column equation process, so that when more images are input, the consistency of the infinity plane cannot be ensured, and thus when the image sequence is longer, the self-calibration method based on the Kruppa equation is unstable and has poor accuracy and robustness.
Disclosure of Invention
The application provides a camera self-calibration method, a device, equipment and a storage medium, which are used for solving the problems of instability, poor accuracy and poor robustness of the existing self-calibration method based on the Kruppa equation.
In order to solve the technical problem, the embodiment of the application discloses the following technical scheme:
in a first aspect, an embodiment of the present application discloses a camera self-calibration method, including: and acquiring a feature point matching pair in two adjacent images. And estimating initial values of the internal parameters by using the property that the first two singular values of the intrinsic matrix are equal according to the acquired feature point matching pairs. And correcting the initial value of the internal parameter according to a Kruppa equation and outputting a result.
Further, obtaining the matching pairs of feature points in two adjacent images includes: and detecting two adjacent images by adopting an SIFT algorithm, and acquiring feature point descriptors in the two images. And matching the obtained feature point descriptors in the two adjacent images to obtain a feature point descriptor matching pair. And searching the feature points respectively corresponding to the feature point descriptors in the two adjacent images to obtain the feature points in the two adjacent images respectively corresponding to the feature point descriptor matching pairs, namely the feature point matching pairs.
Matching the obtained feature point descriptors in the two adjacent images comprises the following steps: and selecting the feature point descriptor pair with the shortest Euclidean distance in the two adjacent images as a feature point descriptor matching pair.
Further, estimating initial values of the internal parameters by using the property that the first two singular values of the essential matrix are equal according to the obtained feature point matching pairs comprises: and estimating a basic matrix F of two adjacent images by using a RANSAC algorithm according to the obtained feature point matching pairs. And setting an internal parameter K according to the assumed image principal point coordinates, the scale factor, the focal length of the camera along the direction of the u axis of the image and the focal length along the direction of the v axis of the image. Obtaining an intrinsic matrix E ═ K according to the internal parameter K and the basic matrix FTFK, wherein KTIs the transposed matrix of K. And constructing a function according to the property that the first two singular values of the essential matrix are equal, enabling the function to approach zero by adopting a particle swarm algorithm, and estimating the initial value of the internal parameter.
Further, the correcting the initial value of the internal parameter according to the Kruppa equation and outputting the result comprises: and estimating the image pole e' of the two adjacent images by using a RANSAC algorithm according to the obtained feature point matching pair. And constructing a Kruppa equation according to the fundamental matrix F and the image poles e' between two adjacent images. And constructing an objective function according to a Kruppa equation. And substituting the initial value of the internal parameter into the objective function, optimizing the objective function by using an internal penalty function method and combining a variable scale method, and obtaining and outputting the optimized internal parameter value as an output result.
In a second aspect, an embodiment of the present application discloses a camera self-calibration apparatus, including: and the acquisition unit is configured to acquire the feature point matching pairs in the two adjacent images. And the estimation unit is configured to estimate the initial value of the internal parameter by using the property that the first two singular values of the intrinsic matrix are equal according to the acquired feature point matching pair. And the correcting unit is configured and used for correcting the initial value of the internal parameter according to a Kruppa equation and outputting a result.
Further, the acquisition unit includes: and the obtaining subunit is configured to detect two adjacent images by adopting an SIFT algorithm, and obtain the special point descriptors in the two images. And the matching subunit is used for matching the feature point descriptors in the two adjacent images to obtain a feature point descriptor matching pair. And the searching subunit is configured to search the feature points in the two adjacent images respectively corresponding to the feature point descriptors to obtain the feature points in the two adjacent images respectively corresponding to the feature point descriptor matching pairs, namely the feature point matching pairs.
The matching subunit includes: and the matching module is configured to select the feature point descriptor pair with the shortest Euclidean distance in the two adjacent images as the feature point descriptor matching pair.
Further, the estimation unit includes: and the estimation operator unit is configured and used for estimating a basic matrix F of two adjacent images by adopting an RANSAC algorithm according to the acquired feature point matching pairs. And a setting subunit, configured to set the internal parameter K according to the assumed image principal point coordinates, the scale factor, and the focal length of the camera along the image u-axis direction and the focal length along the image v-axis direction. A substitution subunit configured to obtain an intrinsic matrix E ═ K according to the intrinsic parameter K and the basic matrix FTFK, wherein KTIs the transposed matrix of K. And the computing subunit is configured to construct a function according to the property that the first two singular values of the intrinsic matrix are equal, make the function approach zero by adopting a particle swarm algorithm, and estimate the initial value of the internal parameter.
Further, the correction unit includes: and the estimating operator unit is configured to estimate an image pole e' of two adjacent images by using an RANSAC algorithm according to the obtained feature point matching pair. And the equation constructing subunit is configured for constructing a Kruppa equation according to the fundamental matrix F and the image poles e' between two adjacent images. And the function constructing subunit is configured to construct the objective function according to the Kruppa equation. And the optimization subunit is configured to substitute the initial value of the internal parameter into the objective function, optimize the objective function by using an internal penalty function method in combination with a variable-scale method, and obtain and output the optimized value of the internal parameter as an output result.
Compared with the prior art, the beneficial effect of this application is:
the method comprises the steps of estimating initial values of internal parameters by utilizing the characteristics that first two singular values of an essential matrix are equal and through a particle swarm algorithm, substituting the estimated initial values of the internal parameters into a target function constructed according to a Kruppa equation, optimizing the target function by adopting an internal point penalty function method and combining a variable-scale method, and finally obtaining the optimized internal parameters. The method combines the self-calibration technology based on the intrinsic matrix and the self-calibration technology based on the Kruppa equation, so that the calibration of the internal parameters is performed by two steps of initial value calculation and correction, and the accuracy and the stability of the calibration of the internal parameters are effectively ensured.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.
Drawings
In order to more clearly explain the technical solution of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious to those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic diagram of image acquisition according to an embodiment of the present application;
FIG. 2 is a flow chart of the intrinsic parameter self-calibration provided in the embodiments of the present application;
fig. 3 is an image of the power fitting captured according to the embodiment of the present application.
Detailed Description
In order to make those skilled in the art better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Example 1 illustratively provides a method of calibrating an internal parameter of an X-ray non-destructive inspection system for inspecting electrical equipment. When the three-dimensional information of the power equipment is recovered from the two-dimensional X-ray image, the traditional self-calibration method is not suitable for calibrating the internal parameters of the X-ray nondestructive testing system due to the penetrability of X-rays. Therefore, the internal parameters of the X-ray nondestructive testing system are calibrated by a self-calibration method.
Referring to fig. 2, the specific calibration method includes the following steps:
a) acquiring a feature point matching pair in two adjacent images;
b) estimating initial values of the internal parameters by using the property that the first two singular values of the intrinsic matrix are equal according to the acquired feature point matching pairs;
c) and correcting the initial value of the internal parameter according to a Kruppa equation and outputting a result.
Further, referring to fig. 1, step a) obtains a matching pair of feature points in two adjacent images, and includes the following steps:
1) and detecting image feature points in two adjacent views by using an SIFT algorithm, and obtaining feature point descriptors (128-dimensional feature point descriptors) of the feature points.
2) And matching the obtained feature point descriptors in the two adjacent images to obtain a feature point descriptor matching pair.
3) And searching the feature points respectively corresponding to the feature point descriptors in the two adjacent images to obtain the feature points in the two adjacent images respectively corresponding to the feature point descriptor matching pairs, namely the feature point matching pairs.
Further, according to the one-to-one relationship of the matching pairs of the feature points, the RANSAC algorithm is adopted to calculate the fundamental matrix F between two adjacent images, so that 3 fundamental matrices can be obtained, and the image poles e' are calculated according to the fundamental matrices.
The specific method for calculating the fundamental matrix F and the image pole e' between two adjacent images by adopting the RANSAC algorithm comprises the following steps:
and (4) randomly taking 8 pairs from all the matched pairs of the image, and solving a basic matrix F by using an 8-point method. The 8-point method comprises the following specific steps:
(1) and recording one pair of matching point pairs in a homogeneous coordinate form as: x ═ u, v,1)TAnd x ═ u ', v',1)TThen, the following relationship exists between the two points: x is the number ofTFx' is 0, F is the basis matrix,
the specific equation is
After the formula is expanded, each element of the F matrix is extracted as follows: (u 'u u' v u 'v' u v 'v v' u v 1) (f)11f12f13f21f22f23f31f32f33) 0 and f (f)11f12f13f21f22f23f31f32f33) Then, given a set of 8 sets of points, we have the following equation:
singular Value Decomposition (SVD) of A, with A ═ UDVTWherein the vector form of V is:
V＝(v1v2v3v4v5v6v7v8v9) And f ═ v9Thus, the basis matrix F is obtained. But since the base matrix rank is 2, the F matrix solved as described above does not usually satisfy this constraint, where we force the rank-2 constraint by SVD decomposition. If F is UDVTIf the diagonal matrix D is diag (r, s, t) and r ≧ s ≧ t, F is Udiag (r, s,0) VTI.e. the solution of the basis matrix.
Secondly, judging all the matching points by utilizing epipolar geometric constraint relation to see whether the matching points belong to interior points (namely correct matching point pairs). The method specifically comprises the following steps: knowing from epipolar geometry that the matching point x' of point x on the first image must be epipolar line l on the second image corresponding to xxThe above. However, due to noise interference, the point x' will often deviate from the polar line, i.e. xTFx' is no longer equal to 0. Let the matching point xiAnd xi' the corresponding polar lines are respectively
And
then point xi' and Point xiThe distances to the two polar lines are respectively:
the sum D of the two deviations is calculatedi＝di'+diIf the sum of the deviations is less than a predetermined threshold value t (t ═ 0.1), the matching point x is determinediAnd xi' is called inner point, otherwise is outer point, i.e. is wrong matching point.
And after judging all the matching point pairs, counting the number of the inner points under the basic matrix F.
And thirdly, randomly extracting 8 pairs of matching point pairs again, repeating the steps of the first step and the second step, randomly extracting samples for 20 times, and comparing the number of the inner points in each sample, wherein the matrix with the most inner points is the robust basic matrix F.
Taking the image shown in fig. 3 taken by the X-ray detection system shown in fig. 1 as an example, table 1 shows the fundamental matrix and the image poles of the image shown in fig. 3 obtained by the self-calibration method provided by the present application.
TABLE 1 fundamental matrix and image poles derived from FIG. 3 taken with an X-ray detection system
Further, the specific method for estimating the initial value of the internal parameter by using the property that the first two singular values of the essential matrix are equal according to the obtained feature point matching pair in the step b) comprises the following steps:
1) setting internal parameters of X-ray nondestructive testing system
fuFor the focal length of the X-ray nondestructive testing system along the u-axis direction of the image, fvFor the focal length of the X-ray nondestructive testing system along the v-axis direction of the image, (u)0,v0) Is the principal point of the image and s is the scale factor. Thus, the essential matrix can be obtained as E ═ KTFK, further adopting RANSAC algorithm to solve to obtain 3 essential matrixes.
2) Optimizing an objective function by particle swarm optimization
Making it close to zero to estimate the initial value of the intrinsic parameter. Wherein n represents n pairs of image pairs, and n is more than or equal to 3;1σijand2σijrepresenting the first two singular values of the essential matrix, which should be equal in theory. w is aijThe number of feature points of two adjacent images is shown,
representing the sum of the number of matching points in all adjacent image pairs, and obtaining an initial value of the internal parameter as
Further, the specific method for correcting the initial value of the internal parameter according to the Kruppa equation and outputting the result in the step c) comprises the following steps:
1) constructing a Kruppa equation according to a basic matrix F between adjacent images and an image pole e
Wherein λ is a non-zero constant, [ e']×The antisymmetric matrix defined for the pole e',
is a symmetric matrix, so that
Then the two sides of the Kruppa equation FCFTAnd
can all representIs given by (c)1,c2,c3,c4,c5) Linear function of c1,c2,c3,c4,c5Are the elements of the matrix C and,
so the Kruppa equation can be converted to
2) Constructing an objective function
Where n represents n pairs of images and n ≧ 3. The objective function has the following constraints
Calculating the initial value of the matrix C according to the initial value of the internal parameter obtained in the step b)
The optimization matrix C can be estimated by optimizing the objective function through an internal penalty function method and a variable-scale method, and the specific method comprises the following steps:
(1) defining barrier functions
And converting the objective function into an unconstrained programming problem. Wherein
rkFor penalty factors, for decreasing series, rk+1＝rk10, get r1＝1。
(2) To be provided with
As an initial point, given an error e1＝10-4，ε2＝10-4And solving the unconstrained optimization problem by adopting a variable-scale method. The method comprises the following specific steps:
1) to giveDefinite positive definite matrix H0E, where E is an identity matrix of order 5.
2) Calculating search direction
Represents the barrier function G (c)k,rk) At ckThe gradient of (a).
3) Let ck+1＝ck+λkPkBy solving for
Can be solved to obtain lambdak。
4) If G (c)k+1)≤ε1Then the optimal solution c*(rk)＝ck+1. Otherwise proceed to step 5).
5) Order to
Δck＝ck+1-ck. According to the correction formula
Calculate Hk+1And starting the calculation from the step 2) again. Finally, the penalty factor r can be obtainedkOptimal solution c of time*(rk)＝ck+1。
(3) If r iskq(ck+1)≤ε2Then c isk+1Is the optimal solution of the objective function, otherwise, r in the barrier functionkIs replaced by rk+1And starting the calculation from the step 2 again until the precision requirement is met.
Finally, the optimal solution can be obtained
Are combined into
Then, the cholesky decomposition is carried out to obtain the internal parameter of
Corresponding to the method for self-calibrating the parameters in the camera provided by the application, the application also provides a device for self-calibrating the parameters in the camera, which comprises the following steps: and the acquisition unit is configured to acquire the feature point matching pairs in the two adjacent images. And the estimation unit is configured to estimate the initial value of the internal parameter by using the property that the first two singular values of the intrinsic matrix are equal according to the acquired feature point matching pair. And the correcting unit is configured and used for correcting the initial value of the internal parameter according to a Kruppa equation and outputting a result.
Further, the acquisition unit includes: and the obtaining subunit is configured to detect two adjacent images by adopting an SIFT algorithm, and obtain the special point descriptors in the two images. And the matching subunit is used for matching the feature point descriptors in the two adjacent images to obtain a feature point descriptor matching pair. And the searching subunit is configured to search the feature points in the two adjacent images respectively corresponding to the feature point descriptors to obtain the feature points in the two adjacent images respectively corresponding to the feature point descriptor matching pairs, namely the feature point matching pairs.
The matching subunit includes: and the matching module is configured to select the feature point descriptor pair with the shortest Euclidean distance in the two adjacent images as the feature point descriptor matching pair.
The estimation unit includes: and the estimation operator unit is configured and used for estimating a basic matrix F of two adjacent images by adopting an RANSAC algorithm according to the acquired feature point matching pairs. And a setting subunit, configured to set the internal parameter K according to the assumed image principal point coordinates, the scale factor, and the focal length of the camera along the image u-axis direction and the focal length along the image v-axis direction. A substitution subunit configured to obtain an intrinsic matrix E ═ K according to the intrinsic parameter K and the basic matrix FTFK, wherein KTIs the transposed matrix of K. And the computing subunit is configured to construct a function according to the property that the first two singular values of the intrinsic matrix are equal, make the function approach zero by adopting a particle swarm algorithm, and estimate the initial value of the internal parameter.
The correction unit includes: and the estimating operator unit is configured to estimate an image pole e' of two adjacent images by using an RANSAC algorithm according to the obtained feature point matching pair. And the equation constructing subunit is configured for constructing a Kruppa equation according to the fundamental matrix F and the image poles e' between two adjacent images. And the function constructing subunit is configured to construct the objective function according to the Kruppa equation. And the optimization subunit is configured to substitute the initial value of the internal parameter into the objective function, optimize the objective function by using an internal penalty function method in combination with a variable-scale method, and obtain and output the optimized value of the internal parameter as an output result.
Since the above embodiments are all described by referring to and combining with other embodiments, the same portions are provided between different embodiments, and the same and similar portions between the various embodiments in this specification may be referred to each other. And will not be described in detail herein.
It is noted that, in this specification, relational terms such as "first" and "second," and the like, are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such circuit structure, article, or apparatus. Without further limitation, the presence of an element identified by the phrase "comprising an … …" does not exclude the presence of other like elements in a circuit structure, article or device comprising the element.
Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.
The above-described embodiments of the present application do not limit the scope of the present application.
Claims (6)
1. A method for self-calibration of a camera, the method comprising:
acquiring a feature point matching pair in two adjacent images;
estimating initial values of intrinsic parameters by using the property that the first two singular values of an intrinsic matrix are equal according to the obtained feature point matching pair, namely estimating a fundamental matrix F of two adjacent images by using a RANSAC algorithm according to the obtained feature point matching pair, setting the intrinsic parameters K according to the assumed image principal point coordinates, the scale factors, the focal length of a camera along the image u-axis direction and the focal length along the image v-axis direction, and obtaining the intrinsic matrix E which is K according to the intrinsic parameters K and the fundamental matrix FTFK, wherein KTConstructing a function according to the property that the first two singular values of the essential matrix are equal, adopting a particle swarm algorithm to enable the function to approach zero, and estimating an initial value of an internal parameter;
and correcting the initial values of the internal parameters according to a Kruppa equation and outputting a result, namely estimating the image poles e 'of two adjacent images by adopting a RANSAC algorithm according to the obtained feature point matching pairs, constructing a Kruppa equation according to a basic matrix F and the image poles e' between the two adjacent images, constructing an objective function according to the Kruppa equation, substituting the initial values of the internal parameters into the objective function, optimizing the objective function by using an internal point penalty function method in combination with a variable-scale method, and obtaining and outputting the optimized internal parameter values as an output result.
2. The method according to claim 1, wherein the obtaining of the matching pairs of feature points in two adjacent images comprises:
detecting two adjacent images by adopting an SIFT algorithm, and acquiring feature point descriptors in the two images;
matching the obtained feature point descriptors in the two adjacent images to obtain a feature point descriptor matching pair;
and searching the feature points in the two adjacent images respectively corresponding to the feature point descriptors to obtain the feature points in the two adjacent images respectively corresponding to the feature point descriptor matching pairs, namely the feature point matching pairs.
3. The method according to claim 1, wherein the matching the feature point descriptors in the two adjacent images comprises:
and selecting the feature point descriptor pair with the shortest Euclidean distance in the two adjacent images as a feature point descriptor matching pair.
4. A camera self-calibration apparatus, the apparatus comprising:
the acquisition unit is configured to acquire a feature point matching pair in two adjacent images;
an estimating unit configured to estimate initial values of intrinsic parameters by using the property that the first two singular values of the intrinsic matrix are equal according to the obtained feature point matching pair, the estimating unit comprises an estimating subunit configured to estimate a basic matrix F of two adjacent images by using a RANSAC algorithm according to the obtained feature point matching pair, a setting subunit configured to set an intrinsic parameter K according to the assumed image principal point coordinates, the scale factor, the focal length of the camera along the image u-axis direction and the focal length along the image v-axis direction, and a substituting subunit configured to obtain the intrinsic matrix E-K according to the intrinsic parameter K and the basic matrix FTFK, wherein KTThe calculation subunit is configured to construct a function according to the property that the first two singular values of the essential matrix are equal, and the particle swarm algorithm is adopted to make the function approach zero and estimate the initial value of the internal parameter;
the correction unit is configured for correcting the initial value of the internal parameter according to a Kruppa equation and outputting a result, and comprises an estimation operator unit and a function construction subunit, wherein the estimation operator unit is configured for estimating an image pole e 'of two adjacent images by adopting a RANSAC algorithm according to an acquired feature point matching pair, the equation construction subunit is configured for constructing a Kruppa equation according to a basic matrix F and the image pole e' between the two adjacent images, the function construction subunit is configured for constructing a target function according to the Kruppa equation, the optimization subunit is configured for substituting the initial value of the internal parameter into the target function, the target function is optimized by using an internal point penalty function method and a variable scale method, and the optimized internal parameter value is obtained and output as an output result.
5. The apparatus of claim 4, wherein the obtaining unit comprises:
the acquisition subunit is configured to detect two adjacent images by adopting an SIFT algorithm and acquire special point descriptors in the two images;
the matching subunit is used for matching the feature point descriptors in the two adjacent images to obtain a feature point descriptor matching pair;
and the searching subunit is configured to search the feature points in the two adjacent images respectively corresponding to the feature point descriptors to obtain the feature points in the two adjacent images respectively corresponding to the feature point descriptor matching pairs, namely the feature point matching pairs.
6. The apparatus of claim 5, wherein the matching subunit comprises:
and the matching module is configured to select the feature point descriptor pair with the shortest Euclidean distance in the two adjacent images as the feature point descriptor matching pair.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810189115.8A CN108416811B (en) | 2018-03-08 | 2018-03-08 | Camera self-calibration method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810189115.8A CN108416811B (en) | 2018-03-08 | 2018-03-08 | Camera self-calibration method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108416811A CN108416811A (en) | 2018-08-17 |
| CN108416811B true CN108416811B (en) | 2022-04-12 |
Family
ID=63130295
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810189115.8A Active CN108416811B (en) | 2018-03-08 | 2018-03-08 | Camera self-calibration method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108416811B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091144B (en) * | 2019-11-27 | 2023-06-27 | 云南电网有限责任公司电力科学研究院 | Image feature point matching method and device based on depth pseudo-twin network |
| CN113256696B (en) * | 2021-06-28 | 2021-09-24 | 中国人民解放军国防科技大学 | External parameter calibration method of laser radar and camera based on natural scene |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0537021A3 (en) * | 1991-10-11 | 1995-01-18 | Marconi Gec Ltd | Camera-calibration in a computer vision system |
| CN101373538A (en) * | 2008-08-25 | 2009-02-25 | 同济大学 | Self-calibrating method of translational initial value operating |
| CN102750704A (en) * | 2012-06-29 | 2012-10-24 | 吉林大学 | Step-by-step video camera self-calibration method |
| CN103208122A (en) * | 2013-04-18 | 2013-07-17 | 湖南大学 | Multi-camera calibration method based on one-dimensional calibration rod design |
-
2018
- 2018-03-08 CN CN201810189115.8A patent/CN108416811B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0537021A3 (en) * | 1991-10-11 | 1995-01-18 | Marconi Gec Ltd | Camera-calibration in a computer vision system |
| CN101373538A (en) * | 2008-08-25 | 2009-02-25 | 同济大学 | Self-calibrating method of translational initial value operating |
| CN102750704A (en) * | 2012-06-29 | 2012-10-24 | 吉林大学 | Step-by-step video camera self-calibration method |
| CN103208122A (en) * | 2013-04-18 | 2013-07-17 | 湖南大学 | Multi-camera calibration method based on one-dimensional calibration rod design |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108416811A (en) | 2018-08-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Urban et al. | Improved wide-angle, fisheye and omnidirectional camera calibration | |
| Haralick et al. | Pose estimation from corresponding point data | |
| CN110264528B (en) | Rapid self-calibration method for binocular camera with fish-eye lens | |
| CN108510551B (en) | Method and system for calibrating camera parameters under long-distance large-field-of-view condition | |
| KR101791590B1 (en) | Object pose recognition apparatus and method using the same | |
| KR100855657B1 (en) | System for estimating self-position of the mobile robot using monocular zoom-camara and method therefor | |
| Bartoli et al. | Nonlinear estimation of the fundamental matrix with minimal parameters | |
| CN112465877B (en) | Kalman filtering visual tracking stabilization method based on motion state estimation | |
| CN111144349A (en) | Indoor visual relocation method and system | |
| CN116433737A (en) | Method and device for registering laser radar point cloud and image and intelligent terminal | |
| CN108416811B (en) | Camera self-calibration method and device | |
| El Akkad et al. | Reconstruction of 3D scenes by camera self-calibration and using genetic algorithms | |
| CN113706635B (en) | Long-focus camera calibration method based on point feature and line feature fusion | |
| JP7044016B2 (en) | Line sensor camera calibration device and method | |
| CN108447084B (en) | Stereo matching compensation method based on ORB characteristics | |
| Ge et al. | Binocular vision calibration and 3D re-construction with an orthogonal learning neural network | |
| CN112288813B (en) | Pose estimation method based on multi-view vision measurement and laser point cloud map matching | |
| Gaspar et al. | New depth from focus filters in active monocular vision systems for indoor 3-D tracking | |
| Li et al. | Least squares image matching: A comparison of the performance of robust estimators | |
| Zhou et al. | Method for fundamental matrix estimation combined with feature lines | |
| CN114877826B (en) | Binocular stereo matching three-dimensional measurement method, system and storage medium | |
| CN113048985B (en) | Camera relative motion estimation method under known relative rotation angle condition | |
| WO2019058487A1 (en) | Three-dimensional reconstructed image processing device, three-dimensional reconstructed image processing method, and computer-readable storage medium having three-dimensional reconstructed image processing program stored thereon | |
| Taraglio et al. | Evolutionary approach to epipolar geometry estimation | |
| CN115100535B (en) | Satellite remote sensing image rapid reconstruction method and device based on affine camera model |
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 |