CN108335333A - A kind of linear camera scaling method - Google Patents
A kind of linear camera scaling method Download PDFInfo
- Publication number
- CN108335333A CN108335333A CN201810294884.4A CN201810294884A CN108335333A CN 108335333 A CN108335333 A CN 108335333A CN 201810294884 A CN201810294884 A CN 201810294884A CN 108335333 A CN108335333 A CN 108335333A
- Authority
- CN
- China
- Prior art keywords
- point
- neural network
- camera
- matrix
- coordinate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The invention belongs to optical measurement and field of visual inspection, and in particular to a kind of linear camera scaling method；The space coordinate that this method passes through the several points of known calibration block；Using manual angle point grid, the pixel coordinate of each point is found out；According to the correspondence and Cross ration invariability of corresponding vertex in harmonic conjugates theory in projective geometry and projective transformation, the vanishing point of orthogonal parallel line is solved；Camera intrinsic parameter matrix is solved according to constraint equation；Using the product of camera intrinsic parameter inverse of a matrix and each picpointed coordinate as the input layer of Bp neural network algorithms, using each space point coordinates as the output layer of neural network algorithm, linear approximation is carried out to its outer parameter using Bp neural network algorithms and solves external parameters of cameras matrix.
Description
Technical field
The invention belongs to optical measurement and field of visual inspection, and in particular to a kind of linear camera scaling method.
Background technology
One of basic task of computer vision is exactly the image obtained from video camera, calculates the object in three dimensions
Body geological information, and thus rebuild and identify object, and certain three-dimensional geometry positions on space object surface with its in the picture
Corresponding points between relationship be to determine that these geometrical model parameters are exactly video camera by the imaging geometry model of video camera
Parameter, the process for calculating these parameters is exactly the calibration of video camera.
The calibration of video camera can be taken under certain conditions, based on object of reference, such calibration known to shape, size
Method has following several：Using the scaling method of optimization algorithm, typically there is conventional method and linear in photogrammetry
Transform method；Utilize the scaling method of video camera transformation matrix；Consider the two-step method of distortion compensation；Improved Zhang Zhengyou methods.
These methods obtain the mapping relations of real space object and corresponding points on image using specific object of reference, and algorithm is simple, but
Need certain test requirements document.
Camera calibration process may be considered as an optimization process, that is, introduces optimization algorithm and solved, such as introduce
Gradient descent method, Newton iteration method, existing introducing optimization method the problem is that calculate time-consuming, unstable, precision is not
Height is easily disturbed.
Invention content
In view of the above-mentioned problems, the invention discloses a kind of linear camera scaling method, considers algorithm and calibration is joined
Accuracy, the robustness of number solution, and then improve stated accuracy.
It is an object of the present invention to what is be achieved：
A kind of linear camera scaling method, including following steps：
Step a, world coordinate system O-XYZ is established according to calibrating block, measures the space coordinate of A, B, C, D, E, F, G, H, L.
Step b, to prevent the influence of noise, manual extraction angle point from obtaining the picture point A of each pointm、Bm、 Cm、Dm、Em、Fm、
Gm、Hm、LmPixel coordinate (uA, vA)、(uB, vB)、 (uC, vC)、(uD, vD)、(uE, vE)、(uF, vF)、(uG, vG)、(uH, vH)、
(uL, vL)。
Step c, according to harmonic conjugates theory and projective transformation corresponding vertex correspondence and cross ratio invariability in projective geometry
Property, solve the vanishing point p of orthogonal parallel line1、p2、p3、p4Pixel coordinate, wherein p1、p2、p3、p4Respectively AmCm、GmHm、
DmBm、FmEmVanishing point.
Step d, camera intrinsic parameter matrix is solved according to constraint equation
Step e, using the product of inverse and each picture point of camera intrinsic parameter matrix K as the input of Bp neural network algorithms
Layer.
Step f, using each space point coordinates as the output layer of Bp neural network algorithms.
Step g, linear approximation is carried out to its outer parameter using Bp neural network algorithms and solves external parameters of cameras matrix M.
A kind of above-mentioned linear camera standardization, the step a are specially：
Using one point O of calibrating block as origin, world coordinate system O-XYZ is established, measures and obtains A, B, C, D, E, F, G, H, L
Space coordinate, the wherein midpoint of C, E, F, H AD, AB, DC, BC, L AC, DB, EF and GH midpoint.
The step b is specially：
To prevent the influence of noise, manual extraction angle point from obtaining the picture point A of each pointm、Bm、Cm、 Dm、Em、Fm、Gm、Hm、Lm
Pixel coordinate (uA, vA)、(uB, vB)、(uC, vC)、(uD, vD)、(uE, vE)、(uF, vF)、(uG, vG)、(uH, vH)、(uL, vL)。
The step c is specially：
The infinite point in the direction AC, GH, DB, FE is P respectively1∞、P2∞、P3∞、P4∞, AmCm、GmHm、DmBm、FmEmBlanking
Point is respectively p1、p2、p3、p4, can be obtained according to harmonic conjugates theory in photography geometry：
According to the correspondence and cross ratio invariability shape of photography transformation respective point, can obtain：
Finally obtain the pixel coordinate expression formula of each vanishing point：
The step d is specially：
According to optical center on the ball that two vanishing points that orthogonal parallel line projection obtains are diameter, sphere equation can be obtained：
By optical center (0,0,0)TSubstitution can obtain：
Vanishing point is substituted into equation, finally finds out the f in camera intrinsic parameter matrix Kx、fy、 Cx、CyFour parameters.
Described step e, f, g are specially：
Using the product of inverse and each picpointed coordinate of camera intrinsic parameter matrix K as the input layer of Bp neural network algorithms,
Using each space point coordinates as the output layer of neural network algorithm, its outer parameter is linearly forced using Bp neural network algorithms
It is close to solve external parameters of cameras matrix M.
Advantageous effect：The present invention has considered the influence of noise on image, the time-consuming of algorithm, complexity, and then improves
The precision and stability of calibration.
Description of the drawings
Fig. 1 camera calibration flow charts.
Fig. 2 camera calibration block schematic diagrames.
Fig. 3 square templates figure and image.
Tri- layers of Bp neural network diagrams of Fig. 4.
Specific implementation mode：
The specific implementation mode of the present invention is described in further detail below in conjunction with the accompanying drawings.
The present invention provides a kind of linear camera scaling method, and flow chart is as shown in Figure 1, this approach includes the following steps：
Step a, as shown in Figure 2 with one point O of calibrating block, establish world coordinate system O-XYZ, measure obtain A, B, C, D, E, F,
G, the space coordinate of H, L, the wherein midpoint of C, E, F, H AD, AB, DC, BC, L AC, DB, EF and GH midpoint.
Step b, to prevent the influence of noise, manual extraction angle point from obtaining the picture point A of each pointm、Bm、 Cm、Dm、Em、Fm、
Gm、Hm、LmPixel coordinate (uA, vA)、(uB, vB)、 (uC, vC)、(uD, vD)、(uE, vE)、(uF, vF)、(uG, vG)、(uH, vH)、
(uL, vL)。
Step c, as shown in Figure 3 according to harmonic conjugates theory and projective transformation corresponding vertex correspondence in projective geometry
And Cross ration invariability, solve the vanishing point p of orthogonal parallel line1、p2、p3、 p4Pixel coordinate, wherein p1、p2、p3、p4Respectively
AmCm、GmHm、DmBm、 FmEmVanishing point.
Step d, the f in camera intrinsic parameter matrix K is solved according to constraint equationx、fy、Cx、CyFour parameters.
Step e, as shown in figure 4, using the coordinate product of inverse and each picture point of camera intrinsic parameter matrix K as Bp nerve nets
The input layer of network algorithm, using each space point coordinates as the output layer of neural network algorithm, using Bp neural network algorithms to it
Outer parameter carries out linear approximation and solves external parameters of cameras matrix M.
Claims (5)
1. a kind of linear camera scaling method, which is characterized in that include the following steps：
Step a, world coordinate system O-XYZ is established according to calibrating block, measures the spatial point seat for obtaining A, B, C, D, E, F, G, H, L
Mark；
Step b, to prevent the influence of noise, manual extraction angular coordinate from obtaining each picture point Am、Bm、Cm、Dm、Em、Fm、Gm、Hm、Lm
Pixel coordinate (uA, vA)、(uB, vB)、(uC, vC)、(uD, vD)、(uE, vE)、(uF, vF)、(uG, vG)、(uH, vH)、(uL, vL)；
Step c, according to harmonic conjugates in projective geometry are theoretical and projective transformation corresponding vertex correspondence and Cross ration invariability,
Solve the vanishing point p of orthogonal parallel line1、p2、p3、p4Pixel coordinate, wherein p1、p2、p3、p4Respectively AmCm、GmHm、DmBm、
FmEmVanishing point；
Step d, camera intrinsic parameter battle array is solved according to constraint equationWherein fx=f/dx, fy=f/
dyIt is the scale factor in camera image plane horizontal direction x-axis and vertical direction y-axis respectively, f is the effective focal length of video camera,
dx, dyRespectively each pixel is in the physical size of x-axis and y-axis direction, Cx、CyFor main point coordinates；
Step e, using the product of inverse and each picture point of camera intrinsic parameter matrix K as the input layer of Bp neural network algorithms；
Step f, using the space coordinate of each point as the output layer of Bp neural network algorithms；
Step g, using Bp neural network algorithms to its outer parameter carry out linear approximation solve external parameters of cameras matrix M=(R |
T), wherein R includes corresponding three vectors of three dimensions XYZ axis deflection angles, T indicate camera coordinate system and world coordinate system it
Between offset.
2. according to each point space coordinates way is sought in claims 1, which is characterized in that the step a is specially：
Using one point O of calibrating block as origin, world coordinate system O-XYZ is established, measures the space seat for obtaining A, B, C, D, E, F, G, H, L
Mark, the wherein midpoint of G, E, F, H AD, AB, DC, BC, L AC, DB, EF and GH midpoint.
3. the method for seeking vanishing point according to claims 1, which is characterized in that the step c is specially：
The infinite point in the direction AC, GH, DB, FE is P respectively1∞、P2∞、P3∞、P4∞, AmCm、GmHm、DmBm、FmEmVanishing point point
It Wei not p1、p2、p3、p4, can be obtained according to harmonic conjugates theory in photography geometry：
According to the correspondence and Cross ration invariability of photography transformation respective point, can obtain：
Finally obtain the pixel coordinate expression formula of each vanishing point：
4. the method that the constraint equation according to claims 1 solves camera intrinsic parameter matrix K, which is characterized in that
The step d is specially：
According to optical center on the ball that two vanishing points that orthogonal parallel line projection obtains are diameter, sphere equation can be obtained：
By optical center (0,0,0)TSubstitution obtains constraint equation：
Vanishing point is substituted into constraint equation, finds out the f in camera intrinsic parameter matrix Kx、fy、Cx、CyFour parameters.
5. the method for solving external parameters of cameras matrix M using Bp neural network algorithms according to claims 1,
It is characterized in that, described step e, f, g's is specially：
Using three layers of Bp neural network algorithms, using the product of inverse and each picpointed coordinate of camera intrinsic parameter matrix K as Bp god
Input layer through network algorithm utilizes Bp neural network algorithms using each space point coordinates as the output layer of neural network algorithm
Linear approximation is carried out to its outer parameter and solves external parameters of cameras matrix M.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810294884.4A CN108335333A (en) | 2018-04-03 | 2018-04-03 | A kind of linear camera scaling method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810294884.4A CN108335333A (en) | 2018-04-03 | 2018-04-03 | A kind of linear camera scaling method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108335333A true CN108335333A (en) | 2018-07-27 |
Family
ID=62932661
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810294884.4A Pending CN108335333A (en) | 2018-04-03 | 2018-04-03 | A kind of linear camera scaling method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108335333A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109583495A (en) * | 2018-11-28 | 2019-04-05 | 深圳爱莫科技有限公司 | Display image treating method and apparatus |
| CN110009696A (en) * | 2019-04-10 | 2019-07-12 | 哈尔滨理工大学 | It is demarcated based on ant colony algorithm Optimized BP Neural Network trinocular vision |
| CN110969657A (en) * | 2018-09-29 | 2020-04-07 | 杭州海康威视数字技术股份有限公司 | Gun and ball coordinate association method and device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102982550A (en) * | 2012-12-14 | 2013-03-20 | 云南大学 | Method of solving intrinsic parameters of camera with regular pentagonal prismatic table |
| CN106097322A (en) * | 2016-06-03 | 2016-11-09 | 江苏大学 | A kind of vision system calibration method based on neutral net |
-
2018
- 2018-04-03 CN CN201810294884.4A patent/CN108335333A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102982550A (en) * | 2012-12-14 | 2013-03-20 | 云南大学 | Method of solving intrinsic parameters of camera with regular pentagonal prismatic table |
| CN106097322A (en) * | 2016-06-03 | 2016-11-09 | 江苏大学 | A kind of vision system calibration method based on neutral net |
Non-Patent Citations (2)
| Title |
|---|
| 隋树林 等: "基于BP神经网络的摄像机参数标定的研究", 《2010亚太地区信息论学术会议论文集》 * |
| 霍炬 等: "基于消隐点几何特性的摄像机自标定方法", 《光学学报》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110969657A (en) * | 2018-09-29 | 2020-04-07 | 杭州海康威视数字技术股份有限公司 | Gun and ball coordinate association method and device, electronic equipment and storage medium |
| CN110969657B (en) * | 2018-09-29 | 2023-11-03 | 杭州海康威视数字技术股份有限公司 | Gun ball coordinate association method and device, electronic equipment and storage medium |
| CN109583495A (en) * | 2018-11-28 | 2019-04-05 | 深圳爱莫科技有限公司 | Display image treating method and apparatus |
| CN109583495B (en) * | 2018-11-28 | 2019-10-22 | 深圳爱莫科技有限公司 | Display image treating method and apparatus |
| CN110009696A (en) * | 2019-04-10 | 2019-07-12 | 哈尔滨理工大学 | It is demarcated based on ant colony algorithm Optimized BP Neural Network trinocular vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106683173B (en) | A method of Three-dimensional Gravity is improved based on neighborhood Block- matching and is laid foundations the dense degree of cloud | |
| CN107507235B (en) | Registration method of color image and depth image acquired based on RGB-D equipment | |
| CN104537707B (en) | Image space type stereoscopic vision moves real-time measurement system online | |
| JP5392415B2 (en) | Stereo image generation apparatus, stereo image generation method, and computer program for stereo image generation | |
| CN106780388B (en) | Optical distortion correction method for linear array camera | |
| CN108510551B (en) | Method and system for calibrating camera parameters under long-distance large-field-of-view condition | |
| CN111192235B (en) | Image measurement method based on monocular vision model and perspective transformation | |
| CN109544628B (en) | Accurate reading identification system and method for pointer instrument | |
| JPWO2018235163A1 (en) | Calibration apparatus, calibration chart, chart pattern generation apparatus, and calibration method | |
| US20170223338A1 (en) | Three dimensional scanning system and framework | |
| JP2005308553A (en) | Three-dimensional image measuring device and method | |
| CN109163657A (en) | A kind of circular target position and posture detection method rebuild based on binocular vision 3 D | |
| CN111667536A (en) | Parameter calibration method based on zoom camera depth estimation | |
| CN110378969A (en) | A kind of convergence type binocular camera scaling method based on 3D geometrical constraint | |
| CN109272555B (en) | External parameter obtaining and calibrating method for RGB-D camera | |
| CN108335333A (en) | A kind of linear camera scaling method | |
| CN110738608B (en) | Plane image correction method and system | |
| Li et al. | Cross-ratio invariant based line scan camera geometric calibration with static linear data | |
| CN113450416B (en) | TCSC method applied to three-dimensional calibration of three-dimensional camera | |
| Liu et al. | Epipolar rectification method for a stereovision system with telecentric cameras | |
| CN111105467B (en) | Image calibration method and device and electronic equipment | |
| CN113012234A (en) | High-precision camera calibration method based on plane transformation | |
| Nie et al. | Method for calibration accuracy improvement of projector-camera-based structured light system | |
| CN114998448A (en) | Method for calibrating multi-constraint binocular fisheye camera and positioning space point | |
| Portalés et al. | An efficient projector calibration method for projecting virtual reality on cylindrical surfaces |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180727 |
|
| WD01 | Invention patent application deemed withdrawn after publication |