CN105678740B - A kind of camera geometric calibration processing method and processing device - Google Patents
A kind of camera geometric calibration processing method and processing device Download PDFInfo
- Publication number
- CN105678740B CN105678740B CN201511015634.5A CN201511015634A CN105678740B CN 105678740 B CN105678740 B CN 105678740B CN 201511015634 A CN201511015634 A CN 201511015634A CN 105678740 B CN105678740 B CN 105678740B
- Authority
- CN
- China
- Prior art keywords
- value
- cosine distance
- camera
- parameter
- optimized
- 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
- 238000012545 processing Methods 0.000 title claims abstract description 17
- 238000003672 processing method Methods 0.000 title claims abstract description 9
- 238000000034 method Methods 0.000 claims abstract description 38
- 238000004364 calculation method Methods 0.000 claims abstract description 18
- 238000010801 machine learning Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 abstract description 17
- 238000005457 optimization Methods 0.000 description 17
- 230000000694 effects Effects 0.000 description 8
- 238000013507 mapping Methods 0.000 description 8
- 238000012360 testing method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 238000012897 Levenberg–Marquardt algorithm Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Studio Devices (AREA)
Abstract
The embodiment of the present invention provides a kind of camera geometric calibration processing method and processing device, the described method includes: determining camera parameter initial value, and the first COS distance of the point pair gone out based on the calculation of initial value is obtained, the point is to the coordinate to correspond to two pixels of space coordinate same position in two adjacent pictures；Judge whether first COS distance is consistent with preset value, if it is not, then obtaining the first iterative parameter, and adjusts the initial value using first iterative parameter, obtain the first optimal value；Obtain the second COS distance based on the calculated point pair of first optimal value；Judge whether second COS distance is consistent with the preset value, if it is, first optimal value is determined as camera parameter calibration value.Such scheme can substantially reduce calculation amount involved in treatment process, help to realize real-time online processing.
Description
Technical Field
The invention relates to the field of image processing, in particular to a camera geometric calibration processing method and device.
Background
Panoramic photography generally refers to a photographing and picture splicing method in which a certain point is taken as a center to perform horizontal 360-degree and vertical 180-degree photographing, and a plurality of photographed pictures are spliced into one panoramic picture. Generally, panoramic photography may include at least two forms of panoramic images and panoramic video.
Generally, when a plurality of original pictures taken are used to be spliced into a panoramic picture, two parts of mapping and splicing are involved. The mapping can be understood as projecting pixel points on the original pictures to positions corresponding to the panoramic pictures, and the splicing can be understood as performing fusion transition on overlapping regions of two adjacent original pictures.
Generally, the camera parameters can include external parameters (such as roll, yaw, pitch, Tx, Ty, Tz) of the camera and internal parameters (such as u, v, w, α, gamma) of the camera, wherein (Tx, Ty, Tz) represents a translation vector, (roll, yaw, pitch) represents a rotation matrix, respectively represents a rotation angle gamma around the z-axis, a rotation angle β around the y-axis, and a rotation angle α around the x-axis, (u, v, w) represents a deflection distortion, and (α, gamma) represents a fisheye lens imaging model parameter.
At present, most of Levenberg-Marquardt algorithms (L-M algorithms for short) are used for iterative calculation to realize camera geometric calibration. In this way, each iteration process needs to solve a second-order partial derivative for each parameter to be estimated, so as to obtain a black plug (Hessian) matrix and a Jacobi (Jacobi) matrix. When the N cameras are used for panoramic shooting, 12 × N parameters to be estimated are involved in the calibration process, the calculation amount is huge, a large amount of calculation time needs to be consumed, the calibration method is commonly used for server off-line (offline) calibration, and real-time on-line processing cannot be realized at present.
In addition, when geometric calibration is performed based on the L-M algorithm, if the determinant of two matrixes is zero in the iteration process, namely the transformation matrix is singular, the local optimal value of the camera parameter can be obtained. Usually, the local optimal value is affected by initial values of camera parameters, and different initial values may cause singularity of a change matrix in different iteration processes, so as to obtain different local optimal values, and therefore, the method has a high requirement on selection of the initial values.
Disclosure of Invention
The embodiment of the invention provides a camera geometric calibration processing method and device, which can greatly reduce the calculated amount involved in the processing process and is beneficial to realizing real-time online processing.
A camera geometry calibration processing method, the method comprising:
determining an initial value of a camera parameter, and obtaining a first cosine distance of a point pair calculated based on the initial value, wherein the point pair is coordinates of two pixel points corresponding to the same position of a space coordinate in two adjacent pictures;
judging whether the first cosine distance is consistent with a preset value or not, if not, acquiring a first iteration parameter, and adjusting the initial value by using the first iteration parameter to acquire a first optimized value;
obtaining a second cosine distance of the point pair calculated based on the first optimized value;
and judging whether the second cosine distance accords with the preset value, and if so, determining the first optimized value as a camera parameter calibration value.
Preferably, if the second cosine distance does not conform to the preset value, the method further comprises:
acquiring a second iteration parameter, and adjusting the first optimized value by using the second iteration parameter to obtain a second optimized value;
obtaining a third cosine distance of the point pair calculated based on the second optimized value;
and judging whether the third cosine distance accords with the preset value, and if so, determining the second optimized value as a camera parameter calibration value.
Preferably, the determining the initial value of the camera parameter includes:
determining a priori values of the camera parameters as the initial values; or,
and adding random disturbance on the basis of the prior value of the camera to obtain the initial value.
Preferably, the manner of obtaining the iteration parameter is as follows:
and performing machine learning on a preset sample to obtain the iteration parameter.
Preferably, machine learning is performed on a preset sample, and the first iterative parameter is obtained in a manner that:
wherein (M)0，N0) Representing a first iteration parameter, CjA number representing the identity of the camera,representing the camera parameter calibration for the jth camera,represents the initial values of the camera parameters of the jth camera,the representation is based on (M)0，N0) And calculating a first cosine distance of the point pair.
Preferably, the manner of adjusting the initial value by using the first iteration parameter to obtain a first optimized valueComprises the following steps:
a camera geometry calibration processing apparatus, the apparatus comprising:
the system comprises a cosine distance calculation unit, a first image acquisition unit and a second image acquisition unit, wherein the cosine distance calculation unit is used for determining an initial value of a camera parameter and obtaining a first cosine distance of a point pair calculated based on the initial value, and the point pair is coordinates of two pixel points corresponding to the same position of a space coordinate in two adjacent images;
the optimized value adjusting unit is used for judging whether the first cosine distance accords with a preset value, if not, acquiring a first iteration parameter, and adjusting the initial value by using the first iteration parameter to acquire a first optimized value;
the cosine distance calculation unit is further configured to obtain a second cosine distance of the point pair calculated based on the first optimized value;
and the calibration value determining unit is used for judging whether the second cosine distance accords with the preset value, and if so, determining the first optimized value as a camera parameter calibration value.
Preferably, the apparatus further comprises:
the optimized value adjusting unit is further configured to obtain a second iteration parameter when the second cosine distance does not match the preset value, and adjust the first optimized value by using the second iteration parameter to obtain a second optimized value;
the cosine distance calculation unit is further configured to obtain a third cosine distance of the point pair calculated based on the second optimized value;
the calibration value determining unit is further configured to determine whether the third cosine distance matches the preset value, and if so, determine the second optimized value as a camera parameter calibration value.
Preferably, the apparatus further comprises:
an iteration parameter obtaining unit, configured to perform machine learning on a preset sample to obtain the first iteration parameter:
wherein (M)0，N0) Representing a first iteration parameter, CjA number representing the identity of the camera,representing the camera parameter calibration for the jth camera,represents the initial values of the camera parameters of the jth camera,the representation is based on (M)0，N0) And calculating a first cosine distance of the point pair.
Preferably, the optimization value adjusting unit is specifically used for adjusting the optimization value according toCalculating to obtain a first optimized value
Compared with the prior art, the method and the device have the advantages that the Taylor polynomial is used for fitting the mapping relation in the picture splicing process, and the optimization problem of the image model parameters is converted into the nonlinear convex quadratic optimization problem of the error between the polynomial and the mapping function. Specifically, pixel point projection can be performed based on current parameters of the camera, cosine distances between the projected point pairs are obtained, whether camera parameter optimization is needed or not is judged according to the cosine distances, and if not, the current parameters can be used as camera parameter calibration values; if necessary, machine learning can be used for obtaining iterative parameters to optimize the current parameters, and iterative calculation is carried out by using the optimized parameters until camera parameter calibration values are obtained. By the scheme, the calculation amount involved in the processing process can be greatly reduced, and real-time online processing is facilitated.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive labor.
FIG. 1 is a flow chart of a camera geometry calibration processing method of the present invention;
FIG. 2 is an effect display diagram of picture stitching implemented based on the scheme of the present invention;
FIG. 3 is an effect display diagram of picture stitching implemented based on prior art solutions;
fig. 4 is a schematic structural diagram of the camera geometric calibration processing apparatus of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of 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 invention.
Before the specific scheme of the invention is introduced, the design idea of the invention is briefly introduced.
When panoramic shooting is carried out, partial overlapping areas exist in original pictures shot by any two adjacent cameras, and the partial overlapping areas are subjected to fusion transition to obtain panoramic pictures. In this respect, we can understand that for the overlapping area, the following scenario exists: two adjacent original pictures A and B, wherein a pixel point a exists in the picture A, a pixel point B exists in the picture B, and both a and B correspond to the same position in a space coordinate. Correspondingly, the pixel points a and b can be referred to as a point pair.
In the stitching process, based on the current camera parameters, the pixel points a and b on the original picture can be respectively mapped to the Equirectangular (equidistant spherical) projection image, and the point pairs a 'and b' of the a and b corresponding to the Equirectangular image are obtained, and usually, if the a 'and the b' can be overlapped, the best stitching effect can be achieved. As an example, a cosine distance between vectors a 'and b' in the three-dimensional space may be calculated, and the cosine distance may reflect a difference between a 'and b', so that it may be known whether the current camera parameters are suitable for stitching a panoramic picture.
That is, whether the current camera parameters are suitable or not can be judged through the cosine distance, and if so, the panoramic picture can be spliced by using the camera parameters; if not, camera parameter optimization may be performed to determine camera parameter calibration values that can be used for panorama picture stitching. In the scheme of the invention, whether the camera parameters are proper or not is judged, and whether the cosine distance accords with the preset value or not can be understood, namely whether the cosine distance falls into the range allowed by the preset value or not. For example, the preset value may be 0.1, and generally, the smaller the preset value is, that is, the closer the cosine distance is to 0, the better the effect of the panoramic picture spliced based on the calibrated value obtained by optimization is, the specific value of the preset value is not limited in the embodiment of the present invention and may be determined by practical application.
In conclusion, the geometric calibration process of the panoramic camera can be converted into the following optimization problem by finding a gradient direction step length Δ x:
wherein,
aiand biAn ith point pair representing the overlapping area of the adjacent original pictures;
h (—) represents a function formed by camera parameters to perform coordinate transformation from points on the original picture to points on the equidistant spherical image based on the camera parameters, e.g., aiConversion to ai’，biConversion to bi’(ii) a H (—) can be generally embodied in various forms according to practical applications, and the present invention is not particularly limited thereto as long as coordinate conversion can be performed based on camera parameters;
c (,) represents the cosine distance of two points on the solved equidistant spherical image, such as point pair ai’And bi’The cosine distance of (d).
The following illustrates the invention.
Referring to fig. 1, a flowchart of a camera geometric calibration processing method according to an embodiment of the present invention is shown, which may include the following steps:
s101, determining an initial value of a camera parameter, and obtaining a first cosine distance of a point pair calculated based on the initial value, wherein the point pair is coordinates of two pixel points corresponding to the same position of a space coordinate in two adjacent pictures.
The invention judges whether the camera parameter optimization is needed or not by judging whether the cosine distance of the point pair falls into the allowable range of the preset value or not, so that an initial value of the camera parameter can be selected first, the point pair mapping is carried out based on the initial value, and the first cosine distance of the point pair is obtained when the initial value is corresponded to.
For example, the initial value in the scheme of the invention can be set by combining with the actual operation experience; or, considering that the prior value of the camera parameter is usually a better value, the initial value in the solution of the present invention may also be set according to the prior value, for example, the prior value of the camera parameter may be determined as the initial value; alternatively, random perturbations may be added based on the a priori values of the cameras to obtain the initial values. For example, the random disturbance value may be set according to actual operation experience, or may also be set to ± (a priori value/100), which may not be specifically limited by the embodiment of the present invention.
For example, CjRepresents the identity number of the jth camera at the time of panorama shooting,initial value of camera parameter, a, representing jth cameraj，iAnd bj，iRepresents the ith point pair on the original picture taken by the jth camera, and can beSubstituting the formula f (x) to obtain a first cosine distance d of the point pair calculated based on the initial value1。
S102, judging whether the first cosine distance is consistent with a preset value or not, if not, obtaining a first iteration parameter, and adjusting the initial value by using the first iteration parameter to obtain a first optimized value.
To obtain d1Thereafter, a determination of whether to optimize the camera parameters, i.e., determination d, may be performed once1Whether or not to cooperate withThe preset values are matched. Still taking the default value of 0.1 as an example, the first cosine distance is consistent with the default value and can be understood as d1And (4) less than or equal to 0.1, otherwise, the first cosine distance is not consistent with the preset value.
If the camera parameters are judged to be needed to be optimized, a first iteration parameter can be obtained from a predetermined iteration parameter set, and the initial value in the step S101 is adjusted by the first iteration parameter to obtain a first optimized value of the camera parameters.
For example, the camera parameters may be optimized by the following formula:wherein (M)k-1，Nk-1) The parameter of the k-th iteration is represented,indicating that the jth camera needs to have optimized camera parameters,representing the optimized camera parameters of the j-th camera.
In this step, the first optimized value(M0，N0) The parameters of the first iteration are represented,represents the initial values of the camera parameters of the jth camera,the representation is based on (M)0，N0) The first cosine distance d of the point pair calculated1。
In the scheme of the invention, an iterative parameter set can be obtained by a machine learning mode on a preset sample, and usually, a group of iterative parameters is needed for each iteration. The manner of obtaining the iterative parameters according to the present invention can be described as follows, and will not be described in detail here.
S103, obtaining a second cosine distance of the point pair calculated based on the first optimized value.
And S104, judging whether the second cosine distance accords with the preset value, and if so, determining the first optimized value as a camera parameter calibration value.
Similar to S101, a first optimized value is obtainedThen, can be used forSubstituting into formula f (x) to obtain a second cosine distance d corresponding to the first optimized value2And further reuse d2A determination is performed once whether to optimize camera parameters.
Specifically, if the judgment result indicates d2If the first optimized value is consistent with the preset value, the iterative process can be ended, and the first optimized value is obtainedAnd determining a parameter calibration value of the jth camera, and enabling the splicing effect of the panoramic picture to be optimal when pixel point mapping is carried out based on the calibration value.
In summary, the method uses Taylor polynomials to fit the mapping relation in the picture splicing process, and converts the optimization problem of the image model parameters into the nonlinear convex quadratic optimization problem of the error between the polynomials and the mapping function. The camera geometric calibration realized based on the scheme of the invention does not need to solve the second-order partial derivative of the parameters to be estimated as the existing scheme, can greatly reduce the calculated amount related to the processing process, and is beneficial to realizing real-time online processing. In addition, the gradient descending direction in the optimization problem is fitted based on the algorithm of machine learning, so that the processing flow is simplified, the convergence speed is accelerated, and the convergence efficiency of the optimization problem is improved. In addition, the camera parameter calibration value determined based on the scheme of the invention is not easy to fall into local optimum, so that the optimization precision of the scheme of the invention is higher.
Alternatively, if the judgment result of S104 represents d2If the value is not consistent with the preset value, the camera parameter calibration value needs to be obtained through a further iteration process. Specifically, a second iteration parameter may be obtained, and the first optimized value may be adjusted by using the second iteration parameter to obtain a second optimized value; obtaining a third cosine distance of the point pair calculated based on the second optimized value; and judging whether the third cosine distance accords with the preset value, and if so, determining the second optimized value as a camera parameter calibration value. In this example, the manner of obtaining the second optimized value, the manner of calculating the third cosine distance by using the second optimized value, the manner of determining whether the third cosine distance matches the preset value, the manner of performing the subsequent processing according to the determination result, and the like may refer to the descriptions of S102 to S104 above, and are not described herein again.
The manner in which the iterative parameter set is obtained by the present invention is briefly described below.
In a first mode, the iterative parameter set can be obtained through a machine learning mode.
Specifically, through the idea of supervised learning in machine learning, the iteration parameter set { M required by each iteration is learned by using a training sample0，M1，…，Mk-1，MkAnd { N }0，N1，…，Nk-1，Nk}。
The parameters involved in the training samples may be embodied as follows:
identity codes C of a series of cameras for panoramic photographyj}；
Calibration value of camera parametersCamera for representing j cameraThe parameter calibration value is a 12 x n-dimensional vector;
a series of point pairs { a }j，i，bj，i}，aj，iAnd bj，iRepresenting an ith point pair on an original picture taken by a jth camera;
initial value of camera parametersDenotes the initial value of the camera parameter of the jth camera.
In combination with the sample parameters described above, the first iteration parameter (M) may be obtained by solving the following linear optimization problem0，N0)：
Obtaining (M)0，N0) Then can be based onCalculating to obtain a first optimized valueAt the same time, the utility model can also be used forSubstituting into the formula f (x), and calculating to obtainThe second iteration parameter (M) in the scheme of the invention is further obtained according to the following formula1，N1)：
By continuously calculating according to the mode, the scheme required by the invention can be determinedAnd (5) iterating the parameter set. For example, if the above learning process finds that k is 5, a better result can be obtained, and after the learning is finished, the following iterative parameter set { M }can be obtained0，M1，M2，M3，M4，M5And { N }0，N1，N2，N3，N4，N5}. The scheme of the invention does not specifically limit the value of k, the value of iterative parameters and the like, and can be determined by practical application.
And in the second mode, the iterative parameter set can be obtained in a mode of combining machine learning and parameter verification.
Specifically, learning may be performed according to the procedure shown in the first embodiment, and an iterative parameter set obtained by learning is referred to as a test iterative parameter set. The test specimen is then used, e.g., a series of point pairs { a } for a test camerai，biAnd adopting the scheme of the invention, starting from the initial value of the camera parameter, using the test iteration parameter set to optimize the camera parameter, and acquiring the point pair { a ] after k iterationsi，biProjection on equidistant spherical image { a }i’，bi’And obtaining the error parameter of the test camera. For example, the error parameters may be an average pixel error and a highest pixel error.
Generally, if the error parameter is consistent with a preset threshold value, the test iteration parameter set obtained in the learning stage is considered to be available, and is determined as the iteration parameter set in the scheme of the invention; if the error parameter is not in accordance with the preset threshold value, the testing iterative parameter set obtained in the learning stage is considered to be unavailable, and the initial value of the camera parameter in the learning stage can be adjusted correspondinglyAnd performing machine learning based on the adjusted initial value of the camera parameter until the error parameter obtained in the parameter verification stage is consistent with the preset threshold value. As an example, initial values of camera parameters for a learning phase are adjustedIt may be that a random perturbation is added on the basis of the initial value.
In order to better verify the beneficial effects brought by the scheme of the invention, experimental comparison results shown in the following table are also provided. It should be noted that the test environment involved in the verification is the intel i5 processor (3.3GHz), Win7 Pro system.
| Average pixel error | Highest pixel error | Number of iterations | Operation time(s) | |
| Scheme of the invention | 1.85 | 5.20 | 5 | 3 |
| Prior Art | 3.25 | 8.75 | 20 | 16 |
TABLE 1900 ten thousand pixels (2048 × 1536 × 3) three-head panoramic camera calibration results
In addition to the above experimental comparison results, the scheme of the present invention is also superior to the prior art in view of the picture splicing effect, and specifically, see the picture splicing effect display diagram implemented based on the scheme of the present invention shown in fig. 2 and the picture splicing effect display diagram implemented based on the prior art shown in fig. 3, especially the circled area in the figure.
In correspondence with the method described above, an embodiment of the present invention further provides a camera geometry calibration processing apparatus, and referring to fig. 4, the apparatus may include:
a cosine distance calculation unit 201, configured to determine an initial value of a camera parameter, and obtain a first cosine distance of a point pair calculated based on the initial value, where the point pair is coordinates of two pixel points corresponding to a same position of a spatial coordinate in two adjacent pictures;
an optimized value adjusting unit 202, configured to determine whether the first cosine distance matches a preset value, if not, obtain a first iteration parameter, and adjust the initial value by using the first iteration parameter to obtain a first optimized value;
the cosine distance calculation unit 201 is further configured to obtain a second cosine distance of the point pair calculated based on the first optimized value;
a calibration value determining unit 203, configured to determine whether the second cosine distance matches the preset value, and if so, determine the first optimized value as a camera parameter calibration value.
Optionally, the apparatus further comprises:
the optimized value adjusting unit is further configured to obtain a second iteration parameter when the second cosine distance does not match the preset value, and adjust the first optimized value by using the second iteration parameter to obtain a second optimized value;
the cosine distance calculation unit is further configured to obtain a third cosine distance of the point pair calculated based on the second optimized value;
the calibration value determining unit is further configured to determine whether the third cosine distance matches the preset value, and if so, determine the second optimized value as a camera parameter calibration value.
Optionally, the apparatus further comprises:
an iteration parameter obtaining unit, configured to perform machine learning on a preset sample to obtain the first iteration parameter:
wherein (M)0，N0) Representing a first iteration parameter, CjA number representing the identity of the camera,representing the camera parameter calibration for the jth camera,represents the initial values of the camera parameters of the jth camera,the representation is based on (M)0，N0) And calculating a first cosine distance of the point pair.
Optionally, the optimization value adjusting unit is specifically configured to adjust the optimization value according toCalculating to obtain a first optimized value
It should be noted that, in the present specification, the embodiments are all described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments may be referred to each other. For the device-like embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
Finally, it should also be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, 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 process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The above-mentioned detailed descriptions of the solutions provided by the present invention apply specific examples to explain the principles and embodiments of the present invention, and the descriptions of the above-mentioned examples are only used to help understanding the method of the present invention and its core ideas; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (5)
1. A camera geometric calibration processing method, the method comprising:
determining an initial value of a camera parameter, and obtaining a first cosine distance of a point pair calculated based on the initial value, wherein the point pair is coordinates of two pixel points corresponding to the same position of a space coordinate in two adjacent pictures in original pictures shot by any two adjacent cameras;
judging whether the first cosine distance accords with a preset value, if not, acquiring a first iteration parameter from a predetermined iteration parameter set, and adjusting the initial value by using the first iteration parameter to acquire a first optimized value;
obtaining a second cosine distance of the point pair calculated based on the first optimized value;
judging whether the second cosine distance accords with the preset value, if so, determining the first optimized value as a camera parameter calibration value;
the method for acquiring the iteration parameter set comprises the following steps:
through supervised learning in machine learning, a preset sample is used for learning an iteration parameter set required by each iteration.
2. The method of claim 1, wherein if the second cosine distance does not match the predetermined value, the method further comprises:
acquiring a second iteration parameter from the iteration parameter set, and adjusting the first optimized value by using the second iteration parameter to obtain a second optimized value;
obtaining a third cosine distance of the point pair calculated based on the second optimized value;
and judging whether the third cosine distance accords with the preset value, and if so, determining the second optimized value as a camera parameter calibration value.
3. The method of claim 1, wherein determining initial values for camera parameters comprises:
determining a priori values of the camera parameters as the initial values; or,
and adding random disturbance on the basis of the prior value of the camera to obtain the initial value.
4. A camera geometry calibration processing apparatus, the apparatus comprising:
the system comprises a cosine distance calculation unit, a first image acquisition unit and a second image acquisition unit, wherein the cosine distance calculation unit is used for determining a camera parameter initial value and obtaining a first cosine distance of a point pair calculated based on the initial value, and the point pair is the coordinates of two pixel points corresponding to the same position of a space coordinate in two adjacent images in original images shot by any two adjacent cameras;
an optimized value adjusting unit, configured to determine whether the first cosine distance matches a preset value, if not, obtain a first iteration parameter from a predetermined iteration parameter set, and adjust the initial value by using the first iteration parameter to obtain a first optimized value, where the manner of obtaining the iteration parameter set is: learning an iteration parameter set required by each iteration by using a preset sample through supervised learning in machine learning;
the cosine distance calculation unit is further configured to obtain a second cosine distance of the point pair calculated based on the first optimized value;
and the calibration value determining unit is used for judging whether the second cosine distance accords with the preset value, and if so, determining the first optimized value as a camera parameter calibration value.
5. The apparatus of claim 4, further comprising:
the optimized value adjusting unit is further configured to obtain a second iteration parameter from the iteration parameter set when the second cosine distance does not match the preset value, and adjust the first optimized value by using the second iteration parameter to obtain a second optimized value;
the cosine distance calculation unit is further configured to obtain a third cosine distance of the point pair calculated based on the second optimized value;
the calibration value determining unit is further configured to determine whether the third cosine distance matches the preset value, and if so, determine the second optimized value as a camera parameter calibration value.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201511015634.5A CN105678740B (en) | 2015-12-30 | 2015-12-30 | A kind of camera geometric calibration processing method and processing device |
| PCT/CN2016/078908 WO2017113535A1 (en) | 2015-12-30 | 2016-04-08 | Method and apparatus for geometric calibration of camera |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201511015634.5A CN105678740B (en) | 2015-12-30 | 2015-12-30 | A kind of camera geometric calibration processing method and processing device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105678740A CN105678740A (en) | 2016-06-15 |
| CN105678740B true CN105678740B (en) | 2019-06-07 |
Family
ID=56298011
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201511015634.5A Active CN105678740B (en) | 2015-12-30 | 2015-12-30 | A kind of camera geometric calibration processing method and processing device |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN105678740B (en) |
| WO (1) | WO2017113535A1 (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106875451B (en) * | 2017-02-27 | 2020-09-08 | 安徽华米智能科技有限公司 | Camera calibration method and device and electronic equipment |
| CN108320026B (en) * | 2017-05-16 | 2022-02-11 | 腾讯科技（深圳）有限公司 | Machine learning model training method and device |
| CN109166152A (en) * | 2018-07-27 | 2019-01-08 | 深圳六滴科技有限公司 | Bearing calibration, system, computer equipment and the storage medium of panorama camera calibration |
| CN110378963A (en) * | 2018-12-04 | 2019-10-25 | 北京京东振世信息技术有限公司 | Camera parameter scaling method and device |
| CN109598715B (en) * | 2018-12-05 | 2023-03-24 | 山西镭谱光电科技有限公司 | Material granularity online detection method based on machine vision |
| CN109785395A (en) * | 2019-01-16 | 2019-05-21 | 深圳市卓简科技有限公司 | A kind of panorama model parameter optimization method and system |
| CN110910458B (en) * | 2019-12-19 | 2022-12-02 | 长光禹辰信息技术与装备(青岛)有限公司 | Gray board calibration method, device, equipment and medium |
| CN113393529B (en) * | 2020-03-12 | 2024-05-10 | 浙江宇视科技有限公司 | Method, device, equipment and medium for calibrating camera |
| CN112102417B (en) * | 2020-09-15 | 2024-04-19 | 阿波罗智联(北京)科技有限公司 | Method and device for determining world coordinates |
| CN113405455B (en) * | 2021-05-17 | 2023-03-07 | 中创智维科技有限公司 | Method for automatically calibrating pixel ratio by calculating geometric parameters of contact network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101794442A (en) * | 2010-01-25 | 2010-08-04 | 哈尔滨工业大学 | Calibration method for extracting illumination-insensitive information from visible images |
| CN101957990A (en) * | 2010-08-13 | 2011-01-26 | 武汉大学 | Camera calibration method, image processing equipment and motor vehicle |
| CN102881016A (en) * | 2012-09-19 | 2013-01-16 | 中科院微电子研究所昆山分所 | Vehicle 360-degree surrounding reconstruction method based on internet of vehicles |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102208108B (en) * | 2011-04-01 | 2013-01-09 | 哈尔滨工业大学 | Large-field high-precision quick field overall calibration method for camera |
| US9955142B2 (en) * | 2013-07-05 | 2018-04-24 | Mediatek Inc. | On-line stereo camera calibration device and method for generating stereo camera parameters |
| CN103745458B (en) * | 2013-12-26 | 2015-07-29 | 华中科技大学 | A kind of space target rotating axle based on binocular light flow of robust and mass center estimation method |
| CN104751486B (en) * | 2015-03-20 | 2017-07-11 | 安徽大学 | A kind of moving target relay tracking algorithm of many ptz cameras |
-
2015
- 2015-12-30 CN CN201511015634.5A patent/CN105678740B/en active Active
-
2016
- 2016-04-08 WO PCT/CN2016/078908 patent/WO2017113535A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101794442A (en) * | 2010-01-25 | 2010-08-04 | 哈尔滨工业大学 | Calibration method for extracting illumination-insensitive information from visible images |
| CN101957990A (en) * | 2010-08-13 | 2011-01-26 | 武汉大学 | Camera calibration method, image processing equipment and motor vehicle |
| CN102881016A (en) * | 2012-09-19 | 2013-01-16 | 中科院微电子研究所昆山分所 | Vehicle 360-degree surrounding reconstruction method based on internet of vehicles |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2017113535A1 (en) | 2017-07-06 |
| CN105678740A (en) | 2016-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105678740B (en) | A kind of camera geometric calibration processing method and processing device | |
| CN108564617B (en) | Three-dimensional reconstruction method and device for multi-view camera, VR camera and panoramic camera | |
| CN108257183B (en) | Camera lens optical axis calibration method and device | |
| CN105096329B (en) | Method for accurately correcting image distortion of ultra-wide-angle camera | |
| WO2017161608A1 (en) | Geometric calibration processing method and device for camera | |
| US20150093042A1 (en) | Parameter calibration method and apparatus | |
| CN104994367B (en) | A kind of image correction method and camera | |
| WO2015039067A1 (en) | Method for combining multiple image fields | |
| CN106875451A (en) | Camera calibration method, device and electronic equipment | |
| CN111553939B (en) | Image registration algorithm of multi-view camera | |
| CN111340737B (en) | Image correction method, device and electronic system | |
| Aghayari et al. | Geometric calibration of full spherical panoramic Ricoh-Theta camera | |
| CN110738608B (en) | Plane image correction method and system | |
| CN110225321B (en) | Training sample data acquisition system and method for trapezoidal correction | |
| US20190197734A1 (en) | Calibration for multi-camera systems | |
| CN106157246A (en) | A kind of full automatic quick cylinder panoramic image joining method | |
| CN111445537A (en) | Calibration method and system of camera | |
| Wakai et al. | Deep single fisheye image camera calibration for over 180-degree projection of field of view | |
| CN116309844A (en) | Three-dimensional measurement method based on single aviation picture of unmanned aerial vehicle | |
| CN111432117B (en) | Image rectification method, device and electronic system | |
| CN108520541A (en) | A kind of scaling method of wide angle cameras | |
| CN111353945B (en) | Fisheye image correction method, device and storage medium | |
| CN111899158A (en) | Image splicing method considering geometric distortion | |
| CN105096296A (en) | Stereo camera imaging correction device, method and electronic equipment | |
| CN116109681A (en) | Image fusion method, device, electronic equipment and readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |