CN104794718A - Single-image CT (computed tomography) machine room camera calibration method - Google Patents

Abstract

The invention discloses a single-image CT (computed tomography) machine room camera calibration method. The method includes: using a camera to be calibrated to photograph a regular solid object with known structure to acquire an image; according to structural characteristics of the regular solid object, acquiring three-dimensional world coordinates of a plurality of angular points and corresponding two-dimensional image coordinates; applying the two-dimensional image coordinates and three-dimensional world coordinates of each angular point to a camera model, performing first iterative optimization solving by means of iterative optimization of set initial values so as to obtain first inside and outside camera parameters; according to the structural characteristics of the regular solid object, acquiring three-dimensional world coordinates of a plurality of equally-divided discrete points and corresponding two-dimensional image coordinates; applying the two-dimensional image coordinates and three-dimensional world coordinates of each equally-divided discrete point to the camera model, performing second iterative optimization solving by means of iterative optimization using the first inside and outside camera parameters as initial values, and thus acquiring precise inside and outside camera parameters.

Description

A kind of method of single image CT center monitoring camera calibration

Technical field

The present invention relates to Computer Vision Detection Technique field, particularly relate to a kind of method of single image CT center monitoring camera calibration.

Background technology

In computer vision application, in order to obtain the corresponding relation between image slices vegetarian refreshments point and actual physics spatial point, camera calibration is absolutely necessary process.Camera calibration, supposition camera model under, utilize shot by camera to image to the object reduced in space.Through the process to image, utilize a series of mathematic(al) manipulation and computing method, obtain geometry and the optical characteristics (being also inner parameter) of video camera inside, and camera coordinate system is relative to the position relationship (being also external parameter) of space coordinates.

1) inner parameter comprises f, (Cx, Cy), k1, sx, wherein:

F: focal length, unit millimeter;

(Cx, Cy): the coordinate of image center or principal point, unit is pixel;

K1: the coefficient of first order of camera lens radial distortion;

Sx: uncertainty scale factor, it is caused by video camera transversal scanning and sample-timing error.

2) external parameter comprises R and T, wherein:

R, T are rotation matrix between world coordinate system and camera coordinate system and translation vector respectively.If the direction of camera coordinate system under world coordinate system: around X-axis rotated counterclockwise by angle (alpha, α), around Y-axis rotated counterclockwise by angle (beta, β), around Z axis rotated counterclockwise by angle (gamma, γ), then rotation matrix is:

$R=R_{\mathrm{\α}}{R}_{\mathrm{\β}}{R}_{\mathrm{\γ}}=\left[\begin{array}{ccc}{r}_{11}& r_{12}& r_{13}\\ r_{21}& r_{22}& r_{23}\\ r_{31}& r_{32}& r_{33}\end{array}\right]$

Wherein,

$R_{\mathrm{\α}}=\left[\begin{array}{ccc}1& 0& 0\\ 0& \cos \mathrm{\α}& \sin \mathrm{\α}\\ 0& -\sin \mathrm{\α}& \cos \mathrm{\α}\end{array}\right],R_{\mathrm{\β}}=\left[\begin{array}{ccc}\cos \mathrm{\β}& 0& -\sin \mathrm{\β}\\ 0& 1& 0\\ \sin \mathrm{\β}& 0& \cos \mathrm{\β}\end{array}\right],R_{\mathrm{\γ}}=\left[\begin{array}{ccc}\cos \mathrm{\γ}& \sin \mathrm{\γ}& 0\\ -\sin \mathrm{\γ}& \cos \mathrm{\γ}& 0\\ 0& 0& 1\end{array}\right]$

And T=[Tx, Ty, Tz] ', wherein Tx, Ty, Tz: from world coordinates be tied to camera coordinate system conversion along three translation of axes amounts.

In computer vision, what camera model solved is the problem that point in three-dimensional scenic is corresponding with the point on the plane of delineation.Camera model is the simplification of optical imagery geometric relationship, and the most the most frequently used camera model is pin-hole model (pinhole model).As shown in Figure 1, pinhole camera model is derived according to lens imaging principle, and it is linear, therefore also becomes Camera Linear Model.Pinhole camera model does not consider lens distortion, but can provide actual camera one well approximate, the research of a lot of camera marking method be all be based upon pinhole camera model basis on.

With reference to figure 1 ~ Fig. 2, we consider that spatial point is to the central projection in a sheet of planar, and make projection centre be positioned at the initial point of an European coordinate system, and plane z=f is called as the plane of delineation, wherein f is the focal length of video camera.Under pinhole camera modeling, volume coordinate is M=(X, Y, Z) ^tsome M be mapped on the plane of delineation a bit, this point is the straight line of tie point M and projection centre and the intersection point of the plane of delineation.According to similar triangles, point (X, Y, Z) can be calculated very soon ^tbe mapped to point (fX/Z, fY/Z, f) on the plane of delineation ^t.After omitting last image coordinate, the central projection from world coordinates to image coordinate is:

This is from 3 dimension theorem in Euclid space IR ³to 2 dimension theorem in Euclid space IR ²a mapping.

Projection centre is called as video camera center, and video camera center is called the main shaft of video camera to the vertical line of the plane of delineation, and the intersection point of main shaft and the plane of delineation is called principal point.Cross video camera center parallel is called video camera principal plane in the plane of the plane of delineation.

If by the homogeneous vector representation world and picture point, so central projection can be expressed as the linear mapping between homogeneous coordinate system very simply.Specifically, above-mentioned formula can be write as following matrix product form:

Suppose our the world point M 4 homogeneous vector of dimensions (X, Y, Z, 1) ^trepresent; Picture point m is expressed as the form of the homogeneous vector of 3 dimension; P represents 3 × 4 homogeneous video camera projection matrixes.So above-mentioned formula can be write as m=PM compactly.The camera matrix it defining the pin-hole model of central projection is:

$P=\mathrm{diag}(f,f,1)\left[I\right|0]=\left[\begin{array}{cccc}f& & & 0\\ & f& & 0\\ & & 1& 0\end{array}\right].$

Common camera marking method has following several:

1) traditional cameras standardization:

When utilizing traditional standardization to carry out camera calibration, need to use calibrated reference.Tradition standardization can comprise direct linear transformation's standardization (DLT), radial arrangement restraint (RAC) standardization, active vision standardization and plane standardization etc.Below for direct linear transformation's standardization and radial arrangement restraint standardization, the ultimate principle of traditional standardization is once described.

Direct linear transformation's standardization (DLT): direct linear transformation's standardization is proposed the beginning of the seventies by Abdal-Aziz and Karara.First the method needs to set up camera imaging model system of linear equations, and measure the world coordinates of one group of point in scene and its respective coordinates on imaging plane, then these coordinate figures are substituted into the unknowm coefficient obtaining this system of linear equations in this system of linear equations, the coefficient of this system of linear equations contains the inner parameter of video camera.

Tsai two step camera calibration method: the method first step utilizes perspective matrix to convert set up linear equation and solve, and obtains the exact solution of most of external parameter; Second step, the parameter of trying to achieve is initial value, brings into nonlinear factor such as all the other external parameter and distortion factors etc. and carries out iterative.Because RAC method considers radial distortion, therefore compare DLT method, its precision is higher.

2) based on the scaling method of active vision:

Control video camera and do some peculair motion, as rotated around the wide heart or pure flatly moving, utilize the singularity of this motion to calculate inner parameter.

3) self-calibrating method:

Usual supposition is when taking different images, the inner parameter of video camera does not change, and the corresponding relation between picture point is determined, according to the special restriction relation existed between imaging point in multiple image, by the corresponding relation between picture point, video camera is demarcated.

Because traditional standardization needs to use calibrated reference in shooting and calibration process always, bring very large inconvenience thus to the use of shooting operation and scaling method.And need to make to demarcate between object and video camera to produce relative motion, to make the multiple image taken have large otherness, make operation very inconvenient.Video camera changes angle at every turn, or shift position all needs to reuse and demarcates object and carry out camera calibration.

When using the scaling method based on active vision, video camera must be controlled and do some peculair motion, but be in most cases difficult to know that the motion state of video camera or the motion state of video camera cannot be known at all, in these cases, the scaling method based on active vision can not just be used.

When using self-calibrating method: owing to being do not change at supposition inner parameter, and demarcate under the prerequisite determined of the corresponding relation between picture point, stated accuracy is poor, and stability is not so good, and can not use when inner parameter changes.

Therefore, the camera marking method that a kind of improvement is provided is needed, to solve the problem.

Summary of the invention

The object of the embodiment of the present invention is to provide a kind of method of single image CT center monitoring camera calibration, utilize the familiar object of an inconspicuous rule as scaling reference, only need take and once can demarcate video camera, whole operating process is simple, quick, and precision is high.

Embodiments provide a kind of method of single image CT center monitoring camera calibration, comprise step:

Utilize video camera to be calibrated to take the regular stereoscopic article that structure is known, obtain piece image;

With described regular stereoscopic article for benchmark builds world coordinate system to it, thus obtain the three-dimensional world coordinate value of some angle points of described regular stereoscopic article;

Image procossing is carried out to described image, obtains the two dimensional image coordinate figure corresponding to described some angle points;

The two dimensional image coordinate figure of angle point described in each, three-dimensional world coordinate value are substituted in the camera model of mapping relations between Description Image coordinate system and world coordinate system respectively, and utilize the iteration method for optimizing of setting initial value to carry out first iteration optimization and solve, thus obtain first camera interior and exterior parameter;

According to the described world coordinate system set up, the some deciles obtained on the boundary line of described regular stereoscopic article cut the three-dimensional world coordinate value of discrete point;

Image procossing is carried out to described image, obtains described some deciles and cut two dimensional image coordinate figure corresponding to discrete point;

Decile described in each is cut the two dimensional image coordinate figure of discrete point, three-dimensional world coordinate value substitutes in the camera model of mapping relations between described Description Image coordinate system and world coordinate system respectively, and utilize described first camera interior and exterior parameter to carry out second iteration Optimization Solution as the iteration method for optimizing of initial value, thus obtain accurate camera interior and exterior parameter.

As the improvement of such scheme, between described Description Image coordinate system and world coordinate system, the camera model of mapping relations is:

$s\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=K\left(\begin{array}{cc}R& t\end{array}\right)\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]=P_{3\×4}\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]$

Wherein, s is a scale factor, (X _w, Y _w, Z _w, 1) and represent that the coordinate figure of any point P in world coordinate system, (u, v, 1) represent that the coordinate figure that this image coordinate corresponding to P point is fastened, R, t are respectively rotation matrix in external parameters of cameras and translation vector, P _{3 × 4}for video camera projection matrix;

K is camera intrinsic parameter, and meets:

$K=\left[\begin{array}{ccc}{f}_u& 0& x_0\\ 0& f_v& y_0\\ 0& 0& 1\end{array}\right]$

Wherein, (x ₀, y ₀) be the center point coordinate value of described image, f _ufor scale factor on described plane of delineation transverse axis, f _vfor scale factor on the described plane of delineation longitudinal axis, and f _u=f/dx, f _v=f/dy, f are the focal length of described video camera to be calibrated.

As the improvement of such scheme, described regular stereoscopic article includes but not limited to right cylinder, spheroid, cube or spheroid.

As the improvement of such scheme, the central point using the mid point between the arbitrary described angle point of described regular stereoscopic article or angle point as the described world coordinate system built.

As the improvement of such scheme, described decile cuts the number of number more than described angle point of discrete point.

As the improvement of such scheme, the method for described image procossing comprises SIFT feature point detection method or color segmentation method.

As the improvement of such scheme, described iteration method for optimizing is least square method.

As the improvement of such scheme, described setting initial value comprises: the initial value of focal length is set to the half of described figure image width high sum; Rotation matrix initial value is set to 0; Translation vector initial value is set to 0.

As the improvement of such scheme, carry out first iteration optimization by the iteration optimization in Zhang Zhengyou gridiron pattern standardization, Tsai gridiron pattern standardization or Hartley and solve and second iteration Optimization Solution, and it is minimum to make to solve error.

Compared with prior art, the method of single image CT center monitoring camera calibration disclosed by the invention utilizes video camera to be calibrated once to take acquisition image to the regular stereoscopic article that structure is known, the architectural feature of the regular stereoscopic article utilizing structure known, obtain the three-dimensional world coordinate value of object angle point and corresponding two dimensional image coordinate figure, then carry out first time iteration optimization, thus obtain the first inside and outside parameter (coarse parameter) of video camera; And the architectural feature of the regular stereoscopic article utilizing structure known, the some deciles obtained on object boundary line cut the three-dimensional world coordinate value of discrete point (segmentation) and corresponding two dimensional image coordinate figure, and with first inside and outside parameter (coarse parameter) for initial value, carry out second time iteration optimization, obtain final camera interior and exterior parameter (accurate parameters), whole operating process is simple, quick, and precision is high.

Accompanying drawing explanation

Fig. 1 is the principle schematic of pinhole camera model in prior art.

Fig. 2 is the principle schematic of pinhole camera model in prior art.

Fig. 3 is the process flow diagram of the method for a kind of single image CT center monitoring camera calibration in the embodiment of the present invention.

Fig. 4 is that in the method for embodiment of the present invention camera calibration, iteration optimization solves the schematic diagram of exact solution for the first time.

Fig. 5 is the schematic diagram that in the method for embodiment of the present invention camera calibration, second time iteration optimization solves exact solution.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, be clearly and completely described the technical scheme in the embodiment of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.

The method of a kind of single image CT center monitoring camera calibration of the present invention is demarcated mainly for the high-end video camera in medical system, the type video camera often can ignore distortion parameter in intrinsic parameters of the camera and uncertainty scale factor, therefore, camera marking method of the present invention is without the need to solving this two camera intrinsic parameters.

Fig. 3 is the schematic flow sheet of the method for a kind of single image CT center monitoring camera calibration that the embodiment of the present invention provides.As shown in Figure 3, according to the method for the camera calibration of the present embodiment, comprise the following steps:

Step S101: utilize video camera to be calibrated to take the regular stereoscopic article that structure is known, obtains piece image.

Concrete, described regular stereoscopic article includes but not limited to right cylinder, spheroid, cube or spheroid.And known structure can arrange as the case may be or measure.Then, utilize the video camera of inside and outside parameter to be calibrated to take the regular stereoscopic article that this structure is known, thus obtain a width picture (image).

Step S102: with described regular stereoscopic article for benchmark builds world coordinate system to it, thus obtain the three-dimensional world coordinate value of some angle points of described regular stereoscopic article.

Concrete, world coordinate system can be built using the mid point between the arbitrary described angle point of described regular stereoscopic article or angle point as the central point of world coordinate system.Because the structure of described regular stereoscopic article is known, therefore, when central point (initial point) using the mid point between one of them angle point or angle point as world coordinate system, the three-dimensional world coordinate value of other angle points can correspondingly calculate.

Step S103: carry out image procossing to described image, obtains the two dimensional image coordinate figure corresponding to described some angle points.

Wherein, the image procossing such as SIFT feature point detection method or color segmentation method can be taked, calculate the two dimensional image coordinate figure that described some angle points are corresponding.

Step S104: the two dimensional image coordinate figure of angle point described in each, three-dimensional world coordinate value are substituted in the camera model of mapping relations between Description Image coordinate system and world coordinate system respectively, and utilize the iteration method for optimizing of setting initial value to carry out first iteration optimization and solve, thus obtain first camera interior and exterior parameter.

Concrete, between Description Image coordinate system and world coordinate system, the camera model of mapping relations is:

$s\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=K\left(\begin{array}{cc}R& t\end{array}\right)\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]=P_{3\×4}\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]---\left(1\right)$

Wherein, s is a scale factor, (X _w, Y _w, Z _w, 1) and represent that the coordinate figure of any point P in world coordinate system, (u, v, 1) represent that the coordinate figure that this image coordinate corresponding to P point is fastened, R, t are respectively rotation matrix in external parameters of cameras and translation vector, P _{3 × 4}for video camera projection matrix;

K is camera intrinsic parameter, and meets:

$K=\left[\begin{array}{ccc}{f}_u& 0& x_0\\ 0& f_v& y_0\\ 0& 0& 1\end{array}\right]$

Wherein, (x ₀, y ₀) be the center point coordinate value of described image, f _ufor scale factor on described plane of delineation transverse axis, f _vfor scale factor on the described plane of delineation longitudinal axis, and f _u=f/dx, f _v=f/dy, f are the focal length of described video camera to be calibrated.

Successively the two dimensional image coordinate figure of angle point described in each, three-dimensional world coordinate value are substituted into after in this camera model respectively, carry out first iteration optimization by the iteration optimization in Zhang Zhengyou gridiron pattern standardization, Tsai gridiron pattern standardization or Hartley and solve this overdetermined linear system, and it is minimum to make to solve error.

Preferably, least square method can be taked to solve this overdetermined linear system, obtain first camera interior and exterior parameter (coarse camera interior and exterior parameter).

Understandable, when carrying out first iteration optimization, needing (camera interior and exterior parameter) initial value pre-setting this iteration optimization, comprising focal length, rotation matrix, translation vector.Such as, the initial value of setting can be: the initial value of focal length is set to the half of described figure image width high sum; Rotation matrix initial value is set to 0; Translation vector initial value is set to 0.

Step S105: according to the described world coordinate system set up, the some deciles obtained on the boundary line of described regular stereoscopic article cut the three-dimensional world coordinate value of discrete point.

Wherein, described decile cuts the number of number more than described angle point of discrete point.In addition, described decile cuts discrete point and can comprise above-mentioned angle point.In like manner, because the structure of described regular stereoscopic article is known, therefore, when central point (initial point) using the mid point between one of them angle point or angle point as world coordinate system, the three-dimensional world coordinate value that described decile cuts discrete point can correspondingly calculate.

Step S106: carry out image procossing to described image, obtains described some deciles and cuts two dimensional image coordinate figure corresponding to discrete point.

Concrete, the image processing methods such as SIFT feature point detection method or color segmentation method can be taked boundary line on image to be detected, then wait component curve by approximate, or divide curve according to the ratio that successively decreases and calculate described decile and cut two dimensional image coordinate figure corresponding to discrete point.

Step S107: decile described in each is cut the two dimensional image coordinate figure of discrete point, three-dimensional world coordinate value substitutes in the camera model of mapping relations between described Description Image coordinate system and world coordinate system respectively, and utilize described first camera interior and exterior parameter to carry out second iteration Optimization Solution as the iteration method for optimizing of initial value, thus obtain accurate camera interior and exterior parameter.

Concrete, between Description Image coordinate system and world coordinate system, the camera model of mapping relations is as shown in above-mentioned formula (1), successively decile described in each is cut the two dimensional image coordinate figure of discrete point, three-dimensional world coordinate value substitutes into after in this camera model respectively, by Zhang Zhengyou gridiron pattern standardization, iteration optimization in Tsai gridiron pattern standardization or Hartley or least square method carry out this overdetermined linear system of second iteration Optimization Solution, and, when carrying out second iteration Optimization Solution, the initial value that the first camera interior and exterior parameter (coarse camera interior and exterior parameter) that obtains solves as this iteration optimization is solved using first iteration optimization, thus obtain final camera interior and exterior parameter (accurate camera interior and exterior parameter).

Below, composition graphs 4 ~ Fig. 5, by a specific embodiment, is described further the method for a kind of single image CT center monitoring camera calibration of the present invention.

In the present embodiment, the method for this single image camera calibration mainly comprises following key step:

1, the image of a given reference substance is taken;

Given reference substance is: the regular stereoscopic article of the right cylinder of a known structure, spheroid, band broken line, known structure can arrange as the case may be or measure.

As shown in Fig. 4 ~ Fig. 5, utilize the object (right cylinder) of video camera to be calibrated to the known rule of structure to take, thus obtain an amplitude object image.

2, coarse estimation: the architectural feature utilizing object, obtains two dimension, the three-dimensional coordinate of object angle point, and carries out first time iteration optimization, obtains the inside and outside parameter of video camera and preserves.Specifically comprise:

2.1 camera models use pinhole camera modeling.

The video camera mapping equation that three-dimensional world coordinate system is mapped to two dimensional image coordinate system is:

$s\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=K\left(\begin{array}{cc}R& t\end{array}\right)\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]=P_{3\×4}\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]$

Wherein,

1) $K=\left[\begin{array}{ccc}{f}_u& 0& x_0\\ 0& f_v& y_0\\ 0& 0& 1\end{array}\right],$ Inner parameter can simplify, and the focal length of two axis is equal is f; And without radial, tangential distortion.

2) f: focal length, unit millimeter;

3) (x0, y0): the coordinate of image center or principal point, unit is pixel, and the central point of simplified image is just the central point of captured image, for known;

2.2 calculate three-dimensional world coordinate value

With reference to figure 4, this cylindrical angle point comprises ABCD tetra-points, and the step calculating the three-dimensional world coordinate value of these four points comprises:

1) can central point any point in cylindrical ABCD point of hypothetical world coordinate system, also can be the mid point of AB, or the mid point of CD;

2) length of hypothesis AB is a linear module further, and ruler can certainly be taken to measure how many millimeters;

3) in units of AB, then the three-dimensional world coordinate point coordinate of A, B, C, D is easy to derive, because be a known right cylinder.

2.3 calculate corresponding two dimensional image coordinate figure

Calculate A, B, C, D three-dimensional world coordinate point corresponding picture point A ', B ', C ', D ' coordinate.

Such as, by image procossing, the method detected as SIFT feature point can obtain.Also can obtain curved surface by the method for color segmentation, then find the two dimensional image coordinate figure of these four angle points.

2.4 initial values arranging first iteration optimization, as focal length, rotation matrix, translation vector.

Focal length initial value is set to the half of figure image width high sum;

Rotation matrix initial value is set to 0;

Translation vector initial value is set to 0

2.5 first time iteration optimization solve

According to the method that the iteration optimization in the multiple view geometry of Zhang Zhengyou gridiron pattern standardization, Tsai gridiron pattern standardization or Hartley solves, solve least error:

$s\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=K\left(\begin{array}{cc}R& t\end{array}\right)\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]=P_{3\×4}\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]$

For in above-mentioned formula, we have been aware of value (u, the v of this two-dimensional image coordinate points now, 1) value (Xw, the Yw of three-dimensional coordinate point corresponding to this two-dimensional image coordinate points, is also aware of, Zx, 1), and that the unknown is exactly video camera internal reference K and outer parameter R and t.Solve described overdetermined linear system by least square method, solve first coarse camera interior and exterior parameter, as the initial value of second iteration Optimization Solution.

3, accurately estimate: the architectural feature utilizing object, generate the three-dimensional coordinate point of segmentation and corresponding two-dimensional coordinate point; Utilize the two and three dimensions coordinate obtained, and with the inside and outside parameter of coarse estimation for initial value, carry out second time iteration optimization, obtain accurate camera interior and exterior parameter.Specifically comprise:

3.1 calculate three-dimensional world coordinate value

As shown in Figure 5, boundary line on right cylinder comprises parallel AB curve, CD curve, it is also parallel that the decile of its correspondence cuts discrete point, and the point on curve A B and the point on curve C D differ from a direction, length is the translation relation of AC or BD, so can or the three-dimensional world coordinate value of these discrete points;

3.2 calculate two dimensional image coordinate figure

1) ' B ' and the curve C ' D ' that as shown in Figure 5, can curve A be detected by the method for image procossing;

2) same, be similar to and wait component curve, or divide curve according to the ratio that successively decreases;

3) thus obtain three-dimensional decile and cut two dimensional image coordinate figure corresponding to discrete point.

3.3 second time iteration optimization solve

In like manner first time, can first time iteration optimization error equation to merge with secondary error equation together with carry out second time iteration optimization (namely, first time iteration optimization is solved the coarse camera interior and exterior parameter obtained, the initial value as second time iteration optimization solves).

$s\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=K\left(\begin{array}{cc}R& t\end{array}\right)\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]=P_{3\×4}\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]$

Equally, for in above-mentioned formula, we have been aware of the value (u of this two-dimensional image coordinate points now, v, 1), be also aware of the value (Xw of three-dimensional coordinate point corresponding to this two-dimensional image coordinate points, Yw, Zx, 1), and that the unknown is exactly video camera internal reference K and outer parameter R and t.Solve described overdetermined linear system (initial value is that iteration optimization solves the coarse camera interior and exterior parameter obtained for the first time) by least square method, solve the accurate camera interior and exterior parameter of second step.

In sum, the method of single image CT center monitoring camera calibration disclosed by the invention utilizes video camera to be calibrated once to take acquisition image to the regular stereoscopic article that structure is known, the architectural feature of the regular stereoscopic article utilizing structure known, obtain the three-dimensional world coordinate value of object angle point and corresponding two dimensional image coordinate figure, then carry out first time iteration optimization, thus obtain the first inside and outside parameter (coarse parameter) of video camera; And the architectural feature of the regular stereoscopic article utilizing structure known, the some deciles obtained on object boundary line cut the three-dimensional world coordinate value of discrete point (segmentation) and corresponding two dimensional image coordinate figure, and with first inside and outside parameter (coarse parameter) for initial value, carry out second time iteration optimization, obtain final camera interior and exterior parameter (accurate parameters), whole operating process is simple, quick, and precision is high.

The above is the preferred embodiment of the present invention; it should be pointed out that for those skilled in the art, under the premise without departing from the principles of the invention; can also make some improvements and modifications, these improvements and modifications are also considered as protection scope of the present invention.

Claims (9)

1. a method for single image CT center monitoring camera calibration, is characterized in that, comprises step:

Utilize video camera to be calibrated to take the regular stereoscopic article that structure is known, obtain piece image;

With described regular stereoscopic article for benchmark builds world coordinate system to it, thus obtain the three-dimensional world coordinate value of some angle points of described regular stereoscopic article;

Image procossing is carried out to described image, obtains the two dimensional image coordinate figure corresponding to described some angle points;

The two dimensional image coordinate figure of angle point described in each, three-dimensional world coordinate value are substituted in the camera model of mapping relations between Description Image coordinate system and world coordinate system respectively, and utilize the iteration method for optimizing of setting initial value to carry out first iteration optimization and solve, thus obtain first camera interior and exterior parameter;

According to the described world coordinate system set up, the some deciles obtained on the boundary line of described regular stereoscopic article cut the three-dimensional world coordinate value of discrete point;

Image procossing is carried out to described image, obtains described some deciles and cut two dimensional image coordinate figure corresponding to discrete point;

Decile described in each is cut the two dimensional image coordinate figure of discrete point, three-dimensional world coordinate value substitutes in the camera model of mapping relations between described Description Image coordinate system and world coordinate system respectively, and utilize described first camera interior and exterior parameter to carry out second iteration Optimization Solution as the iteration method for optimizing of initial value, thus obtain accurate camera interior and exterior parameter.

2. the method for single image CT center monitoring camera calibration as claimed in claim 1, it is characterized in that, between described Description Image coordinate system and world coordinate system, the camera model of mapping relations is:

$s\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=K\left(\begin{array}{cc}R& t\end{array}\right)\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]=P_{3\×4}\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]$

Wherein, s is a scale factor, (X _w, Y _w, Z _w, 1) and represent that the coordinate figure of any point P in world coordinate system, (u, v, 1) represent that the coordinate figure that this image coordinate corresponding to P point is fastened, R, t are respectively rotation matrix in external parameters of cameras and translation vector, P _{3 × 4}for video camera projection matrix;

K is camera intrinsic parameter, and meets:

$K=\left[\begin{array}{ccc}{f}_u& 0& x_0\\ 0& f_v& y_0\\ 0& 0& 1\end{array}\right]$

Wherein, (x ₀, y ₀) be the center point coordinate value of described image, f _ufor scale factor on described plane of delineation transverse axis, f _vfor scale factor on the described plane of delineation longitudinal axis, and f _u=f/dx, f _v=f/dy, f are the focal length of described video camera to be calibrated.

3. the method for single image CT center monitoring camera calibration as claimed in claim 1, it is characterized in that, described regular stereoscopic article includes but not limited to right cylinder, spheroid, cube or spheroid.

4. the method for single image CT center monitoring camera calibration as claimed in claim 1, is characterized in that, the central point using the mid point between the arbitrary described angle point of described regular stereoscopic article or angle point as the described world coordinate system built.

5. the method for single image CT center monitoring camera calibration as claimed in claim 1, it is characterized in that, described decile cuts the number of number more than described angle point of discrete point.

6. the method for single image CT center monitoring camera calibration as claimed in claim 1, it is characterized in that, the method for described image procossing comprises SIFT feature point detection method or color segmentation method.

7. the method for single image CT center monitoring camera calibration as claimed in claim 1, it is characterized in that, described iteration method for optimizing is least square method.

8. the method for single image CT center monitoring camera calibration as claimed in claim 1 or 2, it is characterized in that, described setting initial value comprises: the initial value of focal length is set to the half of described figure image width high sum; Rotation matrix initial value is set to 0; Translation vector initial value is set to 0.

9. the method for single image CT center monitoring camera calibration as claimed in claim 1 or 2, it is characterized in that, carry out first iteration optimization by the iteration optimization in Zhang Zhengyou gridiron pattern standardization, Tsai gridiron pattern standardization or Hartley to solve and second iteration Optimization Solution, and it is minimum to make to solve error.