CN105096317A - Fully automatic calibration method for high performance camera under complicated background - Google Patents

Abstract

With an object of solving problems existing in the prior art, the invention provides a fully automatic calibration method for a high performance camera under a complicated background. The method is combined with a Robust's checkerboard corner detection method, and two groups of checkerboards corners are adopted to serve as filters on the basis of the characteristics of the corners to filter marker images. Eight filters of two types are reduced to four filters of two types, which means that the wave processing amount halves and therefore, the calibration speed increases. Further, the method utilizes a Zhang's camera calibration method to mark the camera calibration parameters. The method utilizes the marked camera calibration parameters to standardize images to be corrected, and a normalized correlation of the standardized images is calculated, and then sub-pixel precision checkerboard corners are obtained. The method then re-projects the coordinate of the checkerboard corners onto image space to obtain a precise coordinate of the corner image. The newly obtained coordinate of the corners is then substituted into the Zhang's calibration method to obtain new computed camera parameters. By repeating the above steps, a performer can obtain highly precise camera parameters. According to the embodiments of the invention, it is possible for a performer to complete automatic corner detection and camera calibration without having to resort to human-machine interactive operations.

Description

High-performance camera full automatic calibration method in a kind of complex background

Technical field

The invention belongs to camera calibration technical field, particularly relate to a kind of high-performance camera full automatic calibration method.

Background technology

Phase camera calibration is one of basic problem in computer vision, comprises and calculates camera inside and outside parameter, have important application in fields such as dimension of object measurement, three-dimensional reconstruction, object identification, robot navigations.Calibration algorithm based on pinhole camera model needs the one-to-one relationship setting up 3D object coordinates and 2D image coordinate, namely.Conventional camera calibration algorithm has the dull and stereotyped standardization of the two-stage calibration method of Tsai and Zhang.ZhangShi standardization is (see non-patent literature 1:ZhengyouZhang:AFlexibleNewTechniqueforCameraCalibratio n.IEEETrans.PatternAnal.Mach.Intell. (PAMI), 22 (11): 1330-1334, 2000.) plane reference plate is utilized to provide one flexibly, the camera parameter computing method of low cost, the method utilizes the cross-hatch pattern picture taken under different attitude and obtains the subpixel coordinates of X-comers in image and known gridiron pattern physical coordinates, the calibration algorithm of camera is set up according to camera imaging model.Adopt the method increasing training points quantity also can obtain good precision, in actual items, obtain good application.Therefore gridiron pattern is able to widespread use as pattern, in machine calibration process, the problem of a core how in complex background, automatically to extract tessellated angular coordinate, especially at outdoor environment as robot navigation, three-dimensional measurement etc., but also very high precision to be had.

It is a problem extensively studied that X-comers detects automatically with camera calibration, conventional Corner Detection Algorithm has Harris method, SUSAN method etc., these methods can only detect the obvious unique point of significant graded in image, can not obtain tessellated structural information from the unique point detected.As the function that OpenCV, Matlab calibration tool case etc. provides X-comers to detect, but all need to adjust compared with multiparameter or manually delimit the scope detected, for manual mode, if the distortion of camera lens is excessive, the distortion parameter valuation also needing to provide camera lens is as initial value to correct image, and this all causes the inconvenience of use.Non-patent literature 2 (J.EHa.AutomaticDetectionofCalibrationMarkersonaChessboar d, OpticalEngineering, 2007,46 (10): 1-8.) the accurate side of geometrical symmetry design judgment of the symmetry of gridiron pattern brightness and the four edges by each angle point is utilized, this algorithm has certain antijamming capability to illumination variation, but the method just obtains tessellated angle point information, do not do the Accurate Calibration of camera.Non-patent literature 3 (MartinRufli, etal.AutomaticDetectionofCheckerboardsonBlurredandDistor tedImages.IEEEIROS2008) propose a kind of method detecting angle point in the cross-hatch pattern picture of fuzzy and Severe distortion, first the method utilizes adaptive threshold method that image is converted into bianry image, then self-adaptation corrosion window is adopted to process to bianry image the aspect ratio retaining grid, recycling self-adaptation quadrilateral link distance criterion carries out heuristic quadrilateral link to the image that previous step process obtains, tessellated four limits and angle point information is obtained finally by polygonal segments, the method has good stability to blurred picture and from the severe-distortion image that omnidirectional camera (Omnidirectionalcamera) obtains, but the method can only obtain the angular coordinate of pixel precision.Non-patent literature 4 (B.Atchesonetal.CALTag:HighPrecisionFiducialMarkersforCam eraCalibration, the15thInternationalWorkshoponVision, Modeling, andVisualization, 2010, 41-48.) a kind of high precision X-comers detection algorithm is proposed, the method is passed through to reach robust in the binary pattern (being similar to the Quick Response Code of simplification) of the special matrix of gridiron pattern indoor design and is detected X-comers, be 0.316 pixel by the average re-projection error of traditional ZhangShi standardization.Although the method can extract X-comers in complex background, need the binary pattern of particular design as supplementary means, versatility is not strong.

Non-patent literature 5 (Yang Xingfang etc. for the checkerboard image Corner Detection new algorithm of camera calibration, Chinese journal of scientific instrument, 2011, 32 (5): 1109-1113.) for the shortcoming of SUSAN Corner Detection Algorithm, the chessboard Corner Detection Algorithm that to propose with the transition times of pixel grey scale in small neighbourhood annular template be criterion, utilize this character, this algorithm is easy to reject non-X-type angle point, the symmetry of recycling checkerboard image, in the local neighborhood of first location angle point, by grey scale centre of gravity method, sub-pixel positioning is carried out to angle point, ensure that stability and the precision of algorithm, but this algorithm can only process pure cross-hatch pattern picture, also namely the method can not process the problem of the X-comers extraction in complex environment.Non-patent literature 6 (storage Jun etc. adopt the gridiron pattern Jiao Jiao of annular template to detect, optical precision engineering, 2013,21 (1): 189-196.) a kind of X-comers detection algorithm based on annular template is proposed to over-exposed and lens distortion image, the nature examination angle point of the symmetry that this algorithm is met by the intensity profile analyzing X-comers annex and gray scale alternately change, and utilize the symmetry of the superfluous She's angle point distribution in local to remove superfluous She's angle point, reach the angular coordinate of sub-pixel precision.If but image can not strictly meet in the symmetry of inhomogeneous illumination or the superfluous She's angle point distribution in local under affecting compared with very noisy, therefore the robustness of this algorithm does not obtain strict checking.

In above-mentioned existing Corner Detection, major part method all needs to adopt pre-service, automatically the Corner Detection problem in any complex background can not be processed, and adopt pre-service that the stability of algorithm is reduced, such as, utilize binaryzation that cross-hatch pattern picture is become bianry image in most of method, but the automatic threshold selection in binarization method itself is also a difficult problem, be usually difficult to automatically select suitable threshold value that cross-hatch pattern picture is converted into bianry image in complex scene.When carrying out denoising and cluster to the unique point detected, length and the angle in usual employing X-comers region carry out true angle point grid as eigenwert, but due to affined transformation and lens distortion, make these features not high in the stability of actual application.Utilize during the characteristic detection methods such as SUSAN or Harris and also need to carry out the operations such as image increase, filtering, binaryzation, morphology, and after obtaining all features in image, the sub-pix characteristic of the structuring angle point of the gridiron pattern M × N how stably to obtain also is a problem.

Camera calibration aspect, adopts OpenCV or Matlab calibration tool case can meet the requirement of general precision demarcation after obtaining gridiron pattern sub-pix angular coordinate, but in high-acruracy survey or three-dimensional reconstruction, needs more high-precision camera calibration parameter.Current published method adopts the caliberating device of redesign mostly, as orthogonal surface plate, wire drawing caliberating device etc., or improve on the basis of traditional peg model, but from disclosed method, the precision of calibration algorithm does not significantly improve.In sum, also not having robustly to detect X-comers up till now again can the algorithm of high-precision calibrating camera.

Summary of the invention

In order to solve the problems of the prior art, the present invention proposes the high-performance camera full automatic calibration method in a kind of complex background, the X-comers coordinate of sub-pixel precision completely automatically can be extracted in any complex background, without the need to any manual operation and intervention, demarcate through Zhang Shi after extracting angular coordinate and obtain initial camera parameter, cross-hatch pattern is as normalization correction, based on the sub-pix Corner Detection of NCC in planningization image, angle point re-projection, carry out the iteration searching process of Zhang Shi demarcation again, until obtain high-precision camera parameter, the method will have important application in the demarcation of high-acruracy survey and outdoor robotic vision system.

The present invention is achieved through the following technical solutions:

A high-performance camera full automatic calibration method in complex background, said method comprising the steps of:

Step 1: the true origin of world coordinates M and image coordinate m is unified, obtains the X-comers coordinate of image space by the Corner Detection Algorithm of robust;

Step 2: after obtaining corresponding point set, utilizes Zhang Shi calibration algorithm to obtain camera calibration parameter as initial value;

Step 3: utilize the calibrating parameters of described step 2 to carry out distortion correction and normalization correction to image;

Step 4: image after calibration sets up region of interest ROI, then calculates normalized-cross-correlation function NCC in each ROI, calculates the X-comers (u of sub-pixel precision on normalization correction image _p, v _p)

Step 5: the angular coordinate re-projection in normalized images is obtained accurate angle point image coordinate to image space;

Step 6: the new angular coordinate obtained is substituted in Zhang Shi calibration algorithm calculate new camera parameter again;

Step 7: repeat the process of step 3-6 until convergence can obtain high-precision camera calibration parameter.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the high-performance camera full automatic calibration method in complex background of the present invention;

Fig. 2 is the filter schematic of two types of the present invention;

Fig. 3 is the scaling board gridiron pattern schematic diagram of symmetric shape;

Fig. 4 is that the initial point of world coordinates and image coordinate unifies schematic diagram;

Fig. 5 is image normalization correction and extracts schematic diagram based on the angular coordinate of NCC, and wherein, Fig. 5 (a) is original image, and Fig. 5 (b) is normalization correction image, and Fig. 5 (c) is the NCC template of amplifying.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.

In computer vision, for determining three-dimensional geometry position and its mutual relationship in the picture between corresponding point of certain point of space object surface, must set up the geometric model of camera imaging, these geometric model parameters are exactly camera parameter.The precision of camera calibration result and the stability of algorithm directly affect the accuracy of result of calculation.

The present invention proposes the camera high-precision calibrating method in a kind of complex environment, the X-comers coordinate of sub-pixel precision completely automatically can be extracted in any complex background, without the need to any manual operation and intervention, extract after angular coordinate through Zhang Shi demarcate obtain initial camera parameter, cross-hatch pattern as normalization correction, planningization image in based on the sub-pix Corner Detection of NCC, angle point re-projection, carry out the iteration searching process of Zhang Shi demarcation again, until obtain high-precision camera parameter, computation process as shown in Figure 1.

First, according to different application, gridiron pattern mark is carried out to the image of more than image acquisition at least 3 width, the black and white lattice quantity of gridiron pattern mark is M × N.Identical with traditional Zhang Shi scaling method to the requirement of image acquisition.If the image of camera shooting is coloured image, need image to be converted into gray level image.

Then, according to the feature of X-comers image, design bank of filters as shown in Figure 2, wherein Typel:AB, CD are axle alignment angle point wave filter; Type2:AB, CD are the angle point wave filter of rotation 45 °, and this two classes wave filter can detect the angle point of the distortion chessboard table images caused due to perspective projection.Like this, two classes, 8 wave filters are become two class 4, reduce half filtering computational processing, improve the speed of algorithm.

Adopt the bank of filters of two types shown in Fig. 1 can improve the speed of algorithm, the stability of algorithm also increases.First adopt four wave filters of two types shown in Fig. 1 to carry out filtering operation to cross-hatch pattern picture, then calculate the Similarity value of angle point wherein with represent the likelihood score of two kinds may reverse (AB and CD) of filter type i (i=1,2), for pixel is for the response of the X wave filter of filter type i, with be defined as follows:

$s_1^i=min(f_{AR}^i-\mathrm{\μ},\mathrm{\μ}-f_{CD}^i)$

$s_2^i=\min (\mathrm{\μ}-f_{AB}^i,f_{CD}^i-\mathrm{\μ})$

$\mathrm{\μ}=0.5(f_{AB}^i+f_{CD}^i)$

Utilize the similarity c of pixel to form likelihood figure C, on C, then utilize non-maximum restraining method to obtain the candidate point detected, utilize Sobel filter result to obtain weighting direction histogram, and utilize average drifting to obtain two principal direction α ₁and α ₂, finally from edge direction structure template T, by calculating the normalized-cross-correlation function (NCC) of T and expectation gradient obtain the score value of X-comers, and utilize threshold value T _cornercalculate the candidate point of X-comers.By optimized image gradient g _p∈ R ²the angular coordinate of sub-pixel precision is obtained with the orthogonality of vector p-c:

$c={\left(\underset{p\∈N_I}{\Σ}{g}_p{g}_p^T\right)}^{-1}\underset{p\∈N_I}{\Σ}\left(g_p{g}_p^T\right)p,$

Wherein N _ifor angle point 11 × 11 neighborhood territory pixel.Tessellated direction vector e ₁∈ R ²and e ₂∈ R ²obtained by minimization method vector sum gradient direction:

$e_i=\arg \underset{e_i^{\′}}{\min }\underset{p\∈M_i}{\Σ}{\left(g_p^T{e}_i^{\′}\right)}^2,$

s.t.e′ _i ^Te′ _i＝1。

Now, the point of approximate Corner Feature on the X-comers of all sub-pixel precisions and image can be obtained comprising.Employing minimizes based on energy function the algorithm expanded with Seed Points and extracts X-comers E (X, Y)=min{E _corners(Y)+E _struct(X, Y) }, wherein X={c ₁, c ₂..., c _nbe unique points detected all in image, Y={y ₁, y ₂..., y _nfor comprising the point set of X-comers and noise, for noise point set, E _struct(X, Y) is gridiron pattern structuring Angle function.

Camera parameter comprises intrinsic parameter and outer parameter, is set up the some M=(x, y, z) in world coordinate system by the inside and outside parameter of camera ^twith some m=(u, v) of image space ^tbetween transformation relation, homogeneous coordinates are expressed as $\tilde{M}={(x,y,z,1)}^T$ With $\tilde{m}={(u,v,1)}^T,$ Pinhole camera model representation is in formula, s is scale factor, and (Rt) is Camera extrinsic number, and A is camera internal reference matrix.By obtaining n to the corresponding point in world coordinate system and image coordinate with set up matrix equation, least square method or Levenberg-Marquardt algorithm is utilized to calculate camera projection matrix and camera inside and outside parameter, adopt these algorithms usually can not reach very high precision, enough accuracy requirements cannot be obtained in high-acruracy survey and three-dimensional reconstruction.

The present invention proposes based on the accurate camera calibration method of Iterative Optimal Control point methods in conjunction with the X-comers detection algorithm of previously described robust.

The initial point of world coordinates and image coordinate is unified: the X-comers coordinate first being obtained image space by the Corner Detection Algorithm of robust, if scaling board gridiron pattern is symmetric shape, as shown in Figure 3, A, B point and C, D point symmetry, for the true origin of world coordinates M and image coordinate m is unified, we stick the circle marker of inverse in the lower left corner gridiron pattern of scaling board, as shown in Figure 4.A ', B ', C ', D ' point is respectively tessellated four angle points in image, when calculating angle point, has calculated the apex coordinate of each angle point and the direction v on limit ₁, v ₂, wherein.Then four the diagonal angle direction vectors calculating each angle point are respectively: (v _x, v _y), (-v _x,-v _y) (in accompanying drawing 4 black gray expandable vector), (v _x,-v _y), (-v _xv _y) (in accompanying drawing 4 white vector), wherein (v _xv _y)=(v ₁+ v ₂) * 0.5, A ' four diagonal angle direction vectors putting as shown in Figure 4.Then the minimum length of side r of grid on scaling board image is calculated _s, with r _sfor radius does four circles respectively on four angle points, then calculate diagonal angle direction vector and the intersection point of circle, calculate the average of pixel in point of intersection 2 × 2 neighborhood, that is:

Wherein i is four angle points, and j is each corner point four direction vector and the intersection point of circle, supposes (black gray expandable vector), the pixel average that (white vector) is diagoned vector intersection point, if or time, i point is the initial point of world coordinates, i.e. A ' point.Corresponding point M ~ the m of unified initial point can be set up by above-mentioned algorithm.

After obtaining corresponding point set, utilize Zhang Shi calibration algorithm to obtain camera calibration parameter as initial value (concrete steps repeat no more, can see non-patent literature 1), utilize calibrating parameters can carry out distortion correction and normalization correction to image.

First image is standardized wherein for normalized images coordinate, for standardization projection matrix. for the list of scaling board and image Direct Transform is answered, the coordinate of normalized images coordinate after homograph is u ' _n=W (1)/W (3), v ' _n=W (2)/W (3), normalization correction image is calculated as follows:

$\left\{\begin{array}{c}{u}_d^{\′}=u_n^{\′}-c\left(1\right)/f\left(1\right)\\ v_d^{\′}=v_n^{\′}-c\left(2\right)/f\left(2\right)\end{array}\right.---\left(1\right)$

$\left\{\begin{array}{c}{u}_d=u_d^{\′}(1+kc(1)r_d+kc\left(2\right)r_d^2)+2kc\left(3\right)u_d^{\′}{v}_d^{\′}+kc\left(4\right)(r_d+2{u_d^{\′}}^2)\\ v_d=v_d^{\′}(1+kc(1)r_d+kc\left(2\right)r_d^2)+2kc\left(4\right)u_d^{\′}{v}_d^{\′}+kc\left(3\right)(r_d+2{v_d^{\′}}^2)\end{array}\right.---\left(2\right)$

$\left\{\begin{array}{c}{u}_n=u_{d}f\left(1\right)+c\left(1\right)\\ v_n=v_{d}f\left(2\right)+c\left(2\right)\end{array}\right.---\left(3\right)$

Wherein (u _n, v _n) for distortion after pixel coordinate, (u ' _n, v ' _n) be distortionless pixel coordinate, (u _d, v _d) be the pixel normalization coordinate after distortion, (u ' _d, v ' _d) be the normalization coordinate of undistorted pixel, f (), c (), k () is respectively the distortion factor of focal length, principal point coordinate and camera lens in camera internal reference.Original image is as shown in accompanying drawing 5 (a), and the image after normalization correction is as shown in accompanying drawing 5 (b), and tessellated average grid is of a size of d _g.Then the template image of X-comers black/white and white/black m × m (m gets 21 or 23) is set up, as shown in accompanying drawing 5 (c).With (d on image after calibration _g/ 2, d _g/ 2) starting point, with d _gfor the length of side sets up region of interest ROI (RegionofInterest), then in each ROI, the formwork calculation normalized-cross-correlation function (NCC shown in accompanying drawing 5 (c) is used, Normalizedcrosscorrelation), the X-comers (u of sub-pixel precision on normalization correction image is calculated _p, v _p), finally the angular coordinate re-projection in normalized images is obtained accurate angle point image coordinate to image space.Again the new angular coordinate obtained is substituted in Zhang Shi calibration algorithm and calculate new camera parameter, repeat said process until convergence can obtain high-precision camera calibration parameter.

The process calculating normalized-cross-correlation function is as follows:

$NCC(u,v)=\frac{{\mathrm{\Σ}}_{x,y}\[f(x,y)-{\stackrel{\‾}{f}}_{u,v}\]\[t(x-u,y-v)-t\]}{\sqrt{{\mathrm{\Σ}}_{x,y}{\[f(x,y)-{\stackrel{\‾}{f}}_{u,v}\]}^2{\mathrm{\Σ}}_{x,y}{\[t(x-u,y-v)-\stackrel{\‾}{t}\]}^2}}---\left(4\right)$

Wherein f (x, y) is image pixel, and t (u, v) is template pixel, for the average of image inner formword area pixel, for the average of template image.If it is very consuming time for directly adopting formula (1) each point calculates normalized correlation coefficient in the picture.Be the image of M × M for a width size, template size is n × n, and computation complexity is O (M ²× n ²), this computing method is difficult to reach the requirement of real-time in industrial machine.A kind of algorithm of the quick calculating normalized correlation coefficient based on Fast Fourier Transform (FFT) and pixel and table is adopted to carry out X-comers location.First formula (4) molecule can be represented by such as formula (5):

$\underset{x,y}{\Σ}\[f\left(x,y\right)-{\stackrel{\‾}{f}}_{u,v})\[t(x-u,y-v)-\stackrel{\‾}{t}\]---\left(5\right)$

Above formula can regard the convolution between image and template as, so can be calculated by Fast Fourier Transform (FFT) and inverse transformation, formula is as follows:

F ^-1{F(f)F ^*(T)}(6)

The denominator of normalized correlation coefficient is:

$\underset{x,y}{\Σ}{\[f(x,y)-{\stackrel{\‾}{f}}_{u,v}\]}^2\underset{x,y}{\Σ}{\[t(x-u,y-v)-\stackrel{\‾}{t}\]}^2---\left(7\right)$

Above formula is calculated by the algorithm of look-up table.Look-up table can be obtained by following formulae discovery in advance:

S(u，v)＝f(u，v)+S(u-1，v)+S(u，v-1)-S(u-1，v-1)(8)

S ²(u，v)＝f ²(u，v)+S ²(u-1，v)+S ²(u，v-1)-S ²(u-1，v-1)(9)

Then formula (7) can be combined by the value in look-up table, and expression formula is:

E _f(u, v)=S ²(u+N-1, v+N-1)-S ²(u-1, v+N-1)-S ²(u+N-1, v-1)+S ²the computation complexity of (u-1, v-1) (10) formula (10) is O (3M ²), and the computation complexity of formula (7) is O (3N ²(M-N+1) ²), greatly can reduce the complexity of algorithm.

Prove by experiment, method of the present invention is consuming time less, without the need to setting area-of-interest, can automatically complete Corner Detection, camera calibration, operating without the need to man-machine interactively.

Above content is in conjunction with concrete preferred implementation further description made for the present invention, can not assert that specific embodiment of the invention is confined to these explanations.For general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, some simple deduction or replace can also be made, all should be considered as belonging to protection scope of the present invention.

Claims (3)

1. the high-performance camera full automatic calibration method in complex background, said method comprising the steps of:

Step 1: the true origin of world coordinates M and image coordinate m is unified, obtains the X-comers coordinate of image space by the Corner Detection Algorithm of robust; Wherein, the Corner Detection Algorithm of described robust adopts two groups of gridiron patterns to carry out filtering as wave filter to scaling board image to corner characteristics, and wherein Type1:AB, CD are axle alignment angle point wave filter; Type2:AB, CD are the angle point wave filter of rotation 45 °;

Step 2: after obtaining corresponding point set, utilizes Zhang Shi calibration algorithm to obtain camera calibration parameter as initial value;

Step 3: utilize the calibrating parameters of described step 2 to carry out distortion correction and normalization correction to image;

Step 4: image after calibration sets up region of interest ROI, then calculates normalized-cross-correlation function NCC in each ROI, calculates the X-comers (u of sub-pixel precision on normalization correction image _p, v _p);

Step 5: the angular coordinate re-projection in normalized images is obtained accurate angle point image coordinate to image space;

Step 6: the new angular coordinate described step 5 obtained again substitutes in Zhang Shi calibration algorithm and calculates new camera calibration parameter;

Step 7: repeat the process of step 3-6 until convergence can obtain high-precision camera calibration parameter.

2. high-performance camera full automatic calibration method according to claim 1, is characterized in that: described step 3 is specially: standardize to image wherein for normalized images coordinate, for standardization projection matrix, for the list of scaling board and image Direct Transform is answered, the coordinate of normalized images coordinate after homograph is u ' _n=W (1)/W (3), v ' _n=W (2)/W (3), normalization correction image is calculated as follows:

$\left\{\begin{array}{c}{u}_d^{\′}=u_n^{\′}-c\left(1\right)/f\left(1\right)\\ v_d^{\′}=v_n^{\′}-c\left(2\right)/f\left(2\right)\end{array}\right.,$

$\left\{\begin{array}{c}{u}_d=u_d^{\′}(1+kc(1)r_d+kc(2\left)r_d^2\right)+2kc\left(3\right)u_d^{\′}{v}_d^{\′}+kc\left(4\right)(r_d+2u_d^{\′2})\\ v_d=v_d^{\′}(1+kc(1)r_d+kc(2\left)r_d^2\right)+2kc\left(4\right)u_d^{\′}{v}_d^{\′}+kc\left(3\right)(r_d+2v_d^{\′2})\end{array}\right.,$

$\left\{\begin{array}{c}{u}_n=u_{d}f\left(1\right)+c\left(1\right)\\ v_n=v_{d}f\left(2\right)+c\left(2\right)\end{array}\right.,$

Wherein, f (), c (), kc () is respectively the distortion factor of focal length, principal point coordinate and camera lens in camera internal reference, (u _n, v _n) for distortion after pixel coordinate, (u ' _n, v ' _n) be distortionless pixel coordinate, (u _d, v _d) be the pixel normalization coordinate after distortion, (u ' _d, v ' _d) be the normalization coordinate of undistorted pixel.

3. high-performance camera full automatic calibration method according to claim 1, is characterized in that: described area-of-interest is with (d _g/ 2, d _g/ 2) starting point, with d _gfor the length of side, wherein, d _gfor tessellated average grid size.